Pieter Abbeel

EDBT 2026 Demo / reviewers in the wild / expert

dblp:a/PieterAbbeel · DBLP profile ↗

← Back

330ranked-venue papers

10as first author

127since 2021 · last 2026

0009-0007-0628-9932ORCID · verified

Domains — the database's venue-derived domains; a paper can count in several

Artificial intelligence and machine learning · 314 · 10 first-author · 122 since 2021Systems, architecture and hardware · 104 · 1 first-author · 23 since 2021Graphics, computer vision, multimedia, augmented reality and games · 22 · 14 since 2021Applied, interdisciplinary, general and emerging computing · 7 · 1 since 2021Databases, data management, data science and information retrieval · 2 · 1 first-author

Year	Publication	Venue	Position
2026	Cliqueformer: Model-Based Optimization with Structured Transformers abstract Large neural networks excel at prediction tasks, but their application to design problems, such as protein engineering or materials discovery, requires solving offline model-based optimization (MBO) problems. While predictive models may not directly translate to effective design, recent MBO algorithms incorporate reinforcement learning and generative modeling approaches. Meanwhile, theoretical work suggests that exploiting the target function’s structure can enhance MBO performance. We present Cliqueformer, a transformer- based architecture that learns the black-box function’s structure through functional graphical models (FGM), addressing distribution shift without relying on explicit conservative approaches. Across various domains, including chemical and genetic design tasks, Cliqueformer demonstrates superior performance compared to existing methods. Jakub Grudzien Kuba, Pieter Abbeel, Sergey Levine	AAAI	2
2026	Deep Sensorimotor Control by Imitating Predictive Models of Human Motion: Leveraging Visual Understanding of Motion for Policy Learning Himanshu Gaurav Singh, Pieter Abbeel, Jitendra Malik, Antonio Loquercio	ICPRAM	2
2025	Efficient Long Video Tokenization via Coordinate-based Patch Reconstruction abstract Efficient tokenization of videos remains a challenge in training vision models that can process long videos. One promising direction is to develop a tokenizer that can encode long video clips, as it would enable the tokenizer to leverage the temporal coherence of videos better for tokenization. However, training existing tokenizers on long videos often incurs a huge training cost as they are trained to reconstruct all the frames at once. In this paper, we introduce CoordTok, a video tokenizer that learns a mapping from coordinate-based representations to the corresponding patches of input videos, inspired by recent advances in 3D generative models. In particular, CoordTok encodes a video into factorized triplane representations and reconstructs patches that correspond to randomly sampled (x, y, t) coordinates. This allows for training large tokenizer models directly on long videos without requiring excessive training resources. Our experiments show that CoordTok can drastically reduce the number of tokens for encoding long video clips. For instance, CoordTok can encode a 128-frame video with 128 × 128 resolution into 1280 tokens, while baselines need 6144 or 8192 tokens to achieve similar reconstruction quality. We further show that this efficient video tokenization enables memory-efficient training of a diffusion transformer that can generate 128 frames at once. Huiwon Jang, Sihyun Yu, Jinwoo Shin, Pieter Abbeel, Younggyo Seo	CVPR	4
2025	SEMDICE: Off-policy State Entropy Maximization via Stationary Distribution Correction Estimation abstract In the unsupervised pre-training for reinforcement learning, the agent aims to learn a prior policy for downstream tasks without relying on task-specific reward functions. We focus on state entropy maximization (SEM), where the goal is to learn a policy that maximizes the entropy of the state's stationary distribution. In this paper, we introduce SEMDICE, a principled off-policy algorithm that computes an SEM policy from an arbitrary off-policy dataset, which optimizes the policy directly within the space of stationary distributions. SEMDICE computes a single, stationary Markov state-entropy-maximizing policy from an arbitrary off-policy dataset. Experimental results demonstrate that SEMDICE outperforms baseline algorithms in maximizing state entropy while achieving the best adaptation efficiency for downstream tasks among SEM-based unsupervised RL pre-training methods. Jongmin Lee 0004, Meiqi Sun, Pieter Abbeel	ICLR	3
2025	World Model on Million-Length Video And Language With Blockwise RingAttention abstract Enabling long-context understanding remains a key challenge in scaling existing sequence models -- a crucial component in developing generally intelligent models that can process and operate over long temporal horizons that potentially consist of millions of tokens. In this paper, we aim to address these challenges by providing a comprehensive exploration of the full development process for producing 1M context language models and video-language models, setting new benchmarks in language retrieval and new capabilities in long video understanding. We detail our long context data curation process, progressive context extension from 4K to 1M tokens, and present an efficient open-source implementation for scalable training on long sequences. Additionally, we open-source a family of 7B parameter models capable of processing long text documents and videos exceeding 1M tokens. Hao Liu 0055, Wilson Yan, Matei Zaharia, Pieter Abbeel	ICLR	4
2025	One Step Diffusion via Shortcut Models abstract Diffusion models and flow matching models have enabled generating diverse and realistic images by learning to transfer noise to data. However, sampling from these models involves iterative denoising over many neural network passes, making generation slow and expensive. Previous approaches for speeding up sampling require complex training regimes, such as multiple training phases, multiple networks, or fragile scheduling. We introduce Shortcut Models, a family of generative models that use a single network and training phase to produce high-quality samples in a single or multiple sampling steps. Shortcut models condition the network not only on the current noise level but also on the desired step size, allowing the model to skip ahead in the generation process. Across a wide range of sampling step budgets, shortcut models consistently produce higher quality samples than previous approaches, such as consistency models and reflow. Compared to distillation, shortcut models reduce complexity to a single network and training phase and additionally allow varying step budgets at inference time. Kevin Frans, Danijar Hafner, Sergey Levine, Pieter Abbeel	ICLR	4
2025	Protein Language Model Fitness is a Matter of Preference abstract Leveraging billions of years of evolution, scientists have trained protein language models (pLMs) to understand the sequence and structure space of proteins aiding in the design of more functional proteins. Although they have shown ability to improve efficiency in engineering, it remains unclear under what conditions they will succeed or fail. We aim to predict the circumstances in which pLMs can successfully perform zero-shot fitness estimation. Our work demonstrates the trends observed over hundreds of deep mutational scans across multiple different fitness objectives. We find that the likelihood, or abstractly, implicit preference of a certain protein sequence imbued during pretraining is predictive fitness prediction capabilities. Both over-preferred and under-preferred wild type sequences harm performance. Generating a causal link between training data and likelihood, we show a power law tail over what data increases protein likelihood which is tied to training sequence homology. Lastly, proteins of low likelihood can be remedied by unsupervised finetuning. In sum, the zero-shot fitness estimation abilities of pLMs can be predicted by the likelihood of the engineered sequence, thus suggesting when pLMs should be deployed in protein maturation campaigns and a way to improve their performance under circumstances of low likelihood. Cade W. Gordon, Amy X. Lu, Pieter Abbeel	ICLR	3
2025	MaxInfoRL: Boosting exploration in reinforcement learning through information gain maximization abstract Reinforcement learning (RL) algorithms aim to balance exploiting the current best strategy with exploring new options that could lead to higher rewards. Most common RL algorithms use undirected exploration, i.e., select random sequences of actions. Exploration can also be directed using intrinsic rewards, such as curiosity or model epistemic uncertainty. However, effectively balancing task and intrinsic rewards is challenging and often task-dependent. In this work, we introduce a framework, MaxInfoRL, for balancing intrinsic and extrinsic exploration. MaxInfoRL steers exploration towards informative transitions, by maximizing intrinsic rewards such as the information gain about the underlying task. When combined with Boltzmann exploration, this approach naturally trades off maximization of the value function with that of the entropy over states, rewards, and actions. We show that our approach achieves sublinear regret in the simplified setting of multi-armed bandits. We then apply this general formulation to a variety of off-policy model-free RL methods for continuous state-action spaces, yielding novel algorithms that achieve superior performance across hard exploration problems and complex scenarios such as visual control tasks. Bhavya Sukhija, Stelian Coros, Andreas Krause 0001, Pieter Abbeel, Carmelo Sferrazza	ICLR	4
2025	Prioritized Generative Replay abstract Sample-efficient online reinforcement learning often uses replay buffers to store experience for reuse when updating the value function. However, uniform replay is inefficient, since certain classes of transitions can be more relevant to learning. While prioritization of more useful samples is helpful, this strategy can also lead to overfitting, as useful samples are likely to be more rare. In this work, we instead propose a prioritized, parametric version of an agent's memory, using generative models to capture online experience. This paradigm enables (1) densification of past experience, with new generations that benefit from the generative model's generalization capacity and (2) guidance via a family of "relevance functions" that push these generations towards more useful parts of an agent's acquired history. We show this recipe can be instantiated using conditional diffusion models and simple relevance functions such as curiosity- or value-based metrics. Our approach consistently improves performance and sample efficiency in both state- and pixel-based domains. We expose the mechanisms underlying these gains, showing how guidance promotes diversity in our generated transitions and reduces overfitting. We also showcase how our approach can train policies with even higher update-to-data ratios than before, opening up avenues to better scale online RL agents. Project page available at: https://pgenreplay.github.io Renhao Wang, Kevin Frans, Pieter Abbeel, Sergey Levine, Alexei A. Efros	ICLR	3
2025	ElasticTok: Adaptive Tokenization for Image and Video abstract Efficient video tokenization remains a key bottleneck in learning general purpose vision models that are capable of processing long video sequences. Prevailing approaches are restricted to encoding videos to a fixed number of tokens, where too few tokens will result in overly lossy encodings, and too many tokens will result in prohibitively long sequence lengths. In this work, we introduce ElasticTok, a method that conditions on prior frames to adaptively encode a frame into a variable number of tokens. To enable this in a computationally scalable way, we propose a masking technique that drops a random number of tokens at the end of each frames's token encoding. During inference, ElasticTok can dynamically allocate tokens when needed -- more complex data can leverage more tokens, while simpler data only needs a few tokens. Our empirical evaluations on images and video demonstrate the effectiveness of our approach in efficient token usage, paving the way for future development of more powerful multimodal models, world models, and agents. Video examples of using ElasticTok can be found on our website: http://largeworldmodel.github.io/elastictok Wilson Yan, Volodymyr Mnih, Aleksandra Faust, Matei Zaharia, Pieter Abbeel, Hao Liu 0055	ICLR	5
2025	OTTER: A Vision-Language-Action Model with Text-Aware Visual Feature Extraction abstract Vision-Language-Action (VLA) models aim to predict robotic actions based on visual observations and language instructions. Existing approaches require fine-tuning pre-trained vision-language models (VLMs) as visual and language features are independently fed into downstream policies, degrading the pre-trained semantic alignments. We propose OTTER, a novel VLA architecture that leverages these existing alignments through explicit, text-aware visual feature extraction. Instead of processing all visual features, OTTER selectively extracts and passes only task-relevant visual features that are semantically aligned with the language instruction to the policy transformer. This allows OTTER to keep the pre-trained vision-language encoders frozen. Thereby, OTTER preserves and utilizes the rich semantic understanding learned from large-scale pre-training, enabling strong zero-shot generalization capabilities. In simulation and real-world experiments, OTTER significantly outperforms existing VLA models, demonstrating strong zero-shot generalization to novel objects and environments. Video, code, checkpoints, and dataset: https://ottervla.github.io/. Fangchen Liu, Letian Fu, Tingfan Wu, Mustafa Mukadam, Jitendra Malik, Kenneth Y. Goldberg, Pieter Abbeel	ICML	8
2025	Chip Placement with Diffusion Models abstract Macro placement is a vital step in digital circuit design that defines the physical location of large collections of components, known as macros, on a 2D chip. Because key performance metrics of the chip are determined by the placement, optimizing it is crucial. Existing learning-based methods typically fall short because of their reliance on reinforcement learning (RL), which is slow and struggles to generalize, requiring online training on each new circuit. Instead, we train a diffusion model capable of placing new circuits zero-shot, using guided sampling in lieu of RL to optimize placement quality. To enable such models to train at scale, we designed a capable yet efficient architecture for the denoising model, and propose a novel algorithm to generate large synthetic datasets for pre-training. To allow zero-shot transfer to real circuits, we empirically study the design decisions of our dataset generation algorithm, and identify several key factors enabling generalization. When trained on our synthetic data, our models generate high-quality placements on unseen, realistic circuits, achieving competitive performance on placement benchmarks compared to state-of-the-art methods. Vint Lee, Leena Elzeiny, Chun Deng, Pieter Abbeel, John Wawrzynek	ICML	5
2025	Value-Based Deep RL Scales Predictably abstract Scaling data and compute is critical in modern machine learning. However, scaling also demands _predictability_: we want methods to not only perform well with more compute or data, but also have their performance be predictable from low compute or low data runs, without ever running the large-scale experiment. In this paper, we show predictability of value-based off-policy deep RL. First, we show that data and compute requirements to reach a given performance level lie on a _Pareto frontier_, controlled by the updates-to-data (UTD) ratio. By estimating this frontier, we can extrapolate data requirements into a higher compute regime, and compute requirements into a higher data regime. Second, we determine the optimal allocation of total _budget_ across data and compute to obtain given performance and use it to determine hyperparameters that maximize performance for a given budget. Third, this scaling behavior is enabled by first estimating predictable relationships between different _hyperparameters_, which is used to counteract effects of overfitting and plasticity loss unique to RL. We validate our approach using three algorithms: SAC, BRO, and PQL on DeepMind Control, OpenAI gym, and IsaacGym, when extrapolating to higher levels of data, compute, budget, or performance. Oleh Rybkin, Michal Nauman, Preston Fu, Charlie Snell, Pieter Abbeel, Sergey Levine, Aviral Kumar	ICML	5
2025	Beyond Sight: Finetuning Generalist Robot Policies with Heterogeneous Sensors via Language Grounding abstract Interacting with the world is a multi-sensory experience: achieving effective general-purpose interaction requires making use of all available modalities - including vision, touch, and audio - to fill in gaps from partial observation. For example, when vision is occluded reaching into a bag, a robot should rely on its senses of touch and sound. However, state-of-the-art generalist robot policies are typically trained on large datasets to predict robot actions solely from visual and proprioceptive observations. In this work, we propose FuSe, a novel approach that enables finetuning visuomotor generalist policies on heterogeneous sensor modalities for which large datasets are not readily available by leveraging natural language as a common cross-modal grounding. We combine a multimodal contrastive loss with a sensory-grounded language generation loss to encode high-level semantics. In the context of robot manipulation, we show that FuSe enables performing challenging tasks that require reasoning jointly over modalities such as vision, touch, and sound in a zero-shot setting, such as multimodal prompting, compositional cross-modal prompting, and descriptions of objects it interacts with. We show that the same recipe is applicable to widely different generalist policies, including both diffusion-based generalist policies and large vision-language-action (VLA) models. Extensive experiments in the real world show that FuSe is able to increase success rates by over 20% compared to all considered baselines. Joshua Jones, Oier Mees, Carmelo Sferrazza, Kyle Stachowicz, Pieter Abbeel, Sergey Levine	ICRA	5
2025	Hand-Object Interaction Pretraining from Videos abstract We present an approach to learn general robot manipulation priors from 3D hand-object interaction trajectories. We build a framework to use in-the-wild videos to generate sensorimotor robot trajectories. We do so by lifting both the human hand and the manipulated object in a shared 3D space and retargeting human motions to robot actions. Generative modeling on this data gives us a task-agnostic base policy. This policy captures a general yet flexible manipulation prior. We empirically demonstrate that finetuning this policy, with both reinforcement learning (RL) and behavior cloning (BC), enables sample-efficient adaptation to downstream tasks and simultaneously improves robustness and generalizability compared to prior approaches. Qualitative experiments are available at: https://hgaurav2k.github.io/hop/. Himanshu Gaurav Singh, Antonio Loquercio, Carmelo Sferrazza, Jane Wu, Haozhi Qi, Pieter Abbeel, Jitendra Malik	ICRA	6
2025	Geometric Retargeting: A Principled, Ultrafast Neural Hand Retargeting Algorithm abstract We introduce Geometric Retargeting (GeoRT), an ultrafast, and principled neural hand retargeting algorithm for teleoperation, developed as part of our recent Dexterity Gen (DexGen) system [1]. GeoRT converts human finger keypoints to robot hand keypoints at 1KHz, achieving state-of-the-art speed and accuracy with significantly fewer hyperparameters. This high-speed capability enables flexible postprocessing, such as leveraging a foundational controller for action correction like DexGen. GeoRT is trained in an unsupervised manner, eliminating the need for manual annotation of hand pairs. The core of GeoRT lies in novel geometric objective functions that capture the essence of retargeting: preserving motion fidelity, ensuring configuration space (C-space) coverage, maintaining uniform response through high flatness, pinch correspondence and preventing self-collisions. This approach is free from intensive test-time optimization, offering a more scalable and practical solution for real-time hand retargeting. Zhao-Heng Yin, Luis Pineda, Chaithanya Krishna Bodduluri, Tingfan Wu, Pieter Abbeel, Mustafa Mukadam	IROS	6
2025	A Stable Whitening Optimizer for Efficient Neural Network Training abstract In this work, we take an experimentally grounded look at neural network optimization. Building on the Shampoo family of algorithms, we identify and alleviate three key issues, resulting in the proposed SPlus method. First, we find that naive Shampoo is prone to divergence when matrix-inverses are cached for long periods. We introduce an alternate bounded update combining a historical eigenbasis with instantaneous normalization, resulting in across-the-board stability and significantly lower computational requirements. Second, we adapt a shape-aware scaling to enable learning rate transfer across network width. Third, we find that high learning rates result in large parameter noise, and propose a simple iterate-averaging scheme which unblocks faster learning. To properly confirm these findings, we introduce a pointed Transformer training benchmark, considering three objectives (language modelling, image classification, and diffusion modelling) across different stages of training. On average, SPlus is able to reach the validation performance of Adam within 44-58% of the gradient steps and 62-83% of the wallclock time. Kevin Frans, Sergey Levine, Pieter Abbeel	NeurIPS	3
2025	Compute-Optimal Scaling for Value-Based Deep RL abstract As models grow larger and training them becomes expensive, it becomes increasingly important to scale training recipes not just to larger models and more data, but to do so in a compute-optimal manner that extracts maximal performance per unit of compute. While such scaling has been well studied for language modeling, reinforcement learning (RL) has received less attention in this regard. In this paper, we investigate compute scaling for online, value-based deep RL. These methods present two primary axes for compute allocation: model capacity and the update-to-data (UTD) ratio. Given a fixed compute budget, we ask: how should resources be partitioned across these axes to maximize data efficiency? Our analysis reveals a nuanced interplay between model size, batch size, and UTD. In particular, we identify a phenomenon we call TD-overfitting: increasing the batch quickly harms Q-function accuracy for small models, but this effect is absent in large models, enabling effective use of large batch size at scale. We provide a mental model for understanding this phenomenon and build guidelines for choosing batch size and UTD to optimize compute usage. Our findings provide a grounded starting point for compute-optimal scaling in deep RL, mirroring studies in supervised learning but adapted to TD learning. Project page: https://value-scaling.github.io/. Preston Fu, Oleh Rybkin, Michal Nauman, Pieter Abbeel, Sergey Levine, Aviral Kumar	NeurIPS	5
2025	Bigger, Regularized, Categorical: High-Capacity Value Functions are Efficient Multi-Task Learners abstract Recent advances in language modeling and vision stem from training large models on diverse, multi‑task data. This paradigm has had limited impact in value-based reinforcement learning (RL), where improvements are often driven by small models trained in a single-task context. This is because in multi-task RL sparse rewards and gradient conflicts make optimization of temporal difference brittle. Practical workflows for generalist policies therefore avoid online training, instead cloning expert trajectories or distilling collections of single‑task policies into one agent. In this work, we show that the use of high-capacity value models trained via cross-entropy and conditioned on learnable task embeddings addresses the problem of task interference in online RL, allowing for robust and scalable multi‑task training. We test our approach on 7 multi-task benchmarks with over 280 unique tasks, spanning high degree-of-freedom humanoid control and discrete vision-based RL. We find that, despite its simplicity, the proposed approach leads to state-of-the-art single and multi-task performance, as well as sample-efficient transfer to new tasks. Michal Nauman, Marek Cygan, Carmelo Sferrazza, Aviral Kumar, Pieter Abbeel	NeurIPS	5
2025	Coarse-to-fine Q-Network with Action Sequence for Data-Efficient Reinforcement Learning abstract Predicting a sequence of actions has been crucial in the success of recent behavior cloning algorithms in robotics. Can similar ideas improve reinforcement learning (RL)? We answer affirmatively by observing that incorporating action sequences when predicting ground-truth return-to-go leads to lower validation loss. Motivated by this, we introduce Coarse-to-fine Q-Network with Action Sequence (CQN-AS), a novel value-based RL algorithm that learns a critic network that outputs Q-values over a sequence of actions, i.e., explicitly training the value function to learn the consequence of executing action sequences. Our experiments show that CQN-AS outperforms several baselines on a variety of sparse-reward humanoid control and tabletop manipulation tasks from BiGym and RLBench. Younggyo Seo, Pieter Abbeel	NeurIPS	2
2025	SOMBRL: Scalable and Optimistic Model-Based RL abstract We address the challenge of efficient exploration in model-based reinforcement learning (MBRL), where the system dynamics are unknown and the RL agent must learn directly from online interactions. We propose Scalable and Optimistic MBRL (SOMBRL), an approach based on the principle of optimism in the face of uncertainty. SOMBRL learns an uncertainty-aware dynamics model and greedily maximizes a weighted sum of the extrinsic reward and the agent's epistemic uncertainty. SOMBRL is compatible with any policy optimizers or planners, and under common regularity assumptions on the system, we show that SOMBRL has sublinear regret for nonlinear dynamics in the (i) finite-horizon, (ii) discounted infinite-horizon, and (iii) non-episodic setting. Additionally, SOMBRL offers a flexible and scalable solution for principled exploration. We evaluate SOMBRL on state-based and visual-control environments, where it displays strong performance across all tasks and baselines. We also evaluate SOMBRL on a dynamic RC car hardware and show SOMBRL outperforms the state-of-the-art, illustrating the benefits of principled exploration for MBRL. Bhavya Sukhija, Lenart Treven, Carmelo Sferrazza, Florian Dörfler, Pieter Abbeel, Andreas Krause 0001	NeurIPS	5
2025	DexGarmentLab: Dexterous Garment Manipulation Environment with Generalizable Policy abstract Garment manipulation is a critical challenge due to the diversity in garment categories, geometries, and deformations. Despite this, humans can effortlessly handle garments, thanks to the dexterity of our hands. However, existing research in the field has struggled to replicate this level of dexterity, primarily hindered by the lack of realistic simulations of dexterous garment manipulation. Therefore, we propose DexGarmentLab, the first environment specifically designed for dexterous (especially bimanual) garment manipulation, which features large-scale high-quality 3D assets for 15 task scenarios, and refines simulation techniques tailored for garment modeling to reduce the sim-to-real gap. Previous data collection typically relies on teleoperation or training expert reinforcement learning (RL) policies, which are labor-intensive and inefficient. In this paper, we leverage garment structural correspondence to automatically generate a dataset with diverse trajectories using only a single expert demonstration, significantly reducing manual intervention. However, even extensive demonstrations cannot cover the infinite states of garments, which necessitates the exploration of new algorithms. To improve generalization across diverse garment shapes and deformations, we propose a Hierarchical gArment-manipuLation pOlicy (HALO). It first identifies transferable affordance points to accurately locate the manipulation area, then generates generalizable trajectories to complete the task. Through extensive experiments and detailed analysis of our method and baseline, we demonstrate that HALO consistently outperforms existing methods, successfully generalizing to previously unseen instances even with significant variations in shape and deformation where others fail. Our project page is available at: https://wayrise.github.io/DexGarmentLab/. Ruihai Wu, Yue Chen 0023, Jitendra Malik, Pieter Abbeel, Hao Dong 0003	NeurIPS	9
2025	Object-centric 3D Motion Field for Robot Learning from Human Videos abstract Learning robot control policies from human videos is a promising direction for scaling up robot learning. However, how to extract action knowledge (or action representations) from videos for policy learning remains a key challenge. Existing action representations such as video frames, pixelflow, and pointcloud flow have inherent limitations such as modeling complexity or loss of information. In this paper, we propose to use object-centric 3D motion field to represent actions for robot learning from human videos, and present a novel framework for extracting this representation from videos for zero-shot control. We introduce two novel components. First, a novel training pipeline for training a ``denoising'' 3D motion field estimator to extract fine object 3D motions from human videos with noisy depth robustly. Second, a dense object-centric 3D motion field prediction architecture that favors both cross-embodiment transfer and policy generalization to background. We evaluate the system in real world setups. Experiments show that our method reduces 3D motion estimation error by over 50% compared to the latest method, achieve 55% average success rate in diverse tasks where prior approaches fail ($\lesssim 10$\%), and can even acquire fine-grained manipulation skills like insertion. Zhao-Heng Yin, Sherry Yang 0001, Pieter Abbeel	NeurIPS	3
2025	A Dexterous and Compliant (DexCo) Hand Based on Soft Hydraulic Actuation for Human-Inspired Fine In-Hand Manipulation abstract Human beings possess a remarkable skill for fine in-hand manipulation, utilizing both intrafinger interactions (in-finger) and finger–environment interactions across a wide range of daily tasks. These tasks range from skilled activities like screwing light bulbs, picking and sorting pills, and in-hand rotation, to more complex tasks such as opening plastic bags, cluttered bin picking, and counting cards. Despite its prevalence in human activities, replicating these fine motor skills in robotics remains a substantial challenge. This study tackles the challenge of fine in-hand manipulation by introducing the dexterous and compliant (DexCo) hand system. The DexCo hand mimics human dexterity, replicating the intricate interaction between the thumb, index, and middle fingers, with a contractable palm. The key to maneuverable fine in-hand manipulation lies in its innovative soft hydraulic actuation, which strikes a balance between control complexity, dexterity, compliance, and motion accuracy within a compact structure, enhancing the overall performance of the system. The model of soft hydraulic actuation, based on hydrostatic force analysis, reveals the compliance of hand joints, which is also further extended to a dedicated robot operating system (ROS) package for DexCo hand simulation, considering both motion and stiffness aspects. Dedicated velocity and position teleoperation controllers are designed for implementing real physical manipulation tasks. The benchmark results show that the fingertip achieves a maximum repeatable finger strength of 34.4 N, a grasp cycle time of less than 2.04 s, and a maximum repeatability accuracy of 0.03 mm. Experimental results demonstrate the DexCo hand successfully performs complex fine in-hand manipulation tasks, providing a promising solution for advancing robotic manipulation capabilities toward the human level. Jianshu Zhou, Junda Huang, Qi Dou 0001, Pieter Abbeel, Yun-Hui Liu 0001	IEEE Trans. Robotics	4
2024	Functional Graphical Models: Structure Enables Offline Data-Driven Optimization abstract While machine learning models are typically trained to solve prediction problems, we might often want to use them for optimization problems. For example, given a dataset of proteins and their corresponding fluorescence levels, we might want to optimize for a new protein with the highest possible fluorescence. This kind of data-driven optimization (DDO) presents a range of challenges beyond those in standard prediction problems, since we need models that successfully predict the performance of new designs that are better than the best designs seen in the training set. It is not clear theoretically when existing approaches can even perform better than the na ïve approach that simply selects the best design in the dataset. In this paper, we study how structure can enable sample-efficient data-driven optimization. To formalize the notion of structure, we introduce functional graphical models (FGMs) and show theoretically how they can provide for principled data-driven optimization by decomposing the original high-dimensional optimization problem into smaller sub-problems. This allows us to derive much more practical regret bounds for DDO, and the result implies that DDO with FGMs can achieve nearly optimal designs in situations where naïve approaches fail due to insufficient coverage of the offline data. We further present a data-driven optimization algorithm that inferes the FGM structure itself, either over the original input variables or a latent variable representation of the inputs. Kuba Grudzien Kuba, Masatoshi Uehara, Sergey Levine, Pieter Abbeel	AISTATS	4
2024	Chain of Hindsight aligns Language Models with Feedback abstract Learning from human preferences is important for language models to match human needs and to align with human and social values. Prior works have achieved remarkable successes by learning from human feedback to understand and follow instructions. Nonetheless, these methods are either founded on hand-picked model generations that are favored by human annotators, rendering them inefficient in terms of data utilization and challenging to apply in general, or they depend on reinforcement learning, which often suffers from imperfect reward functions and relies on extremely challenging optimizations. In this work, we propose a novel technique, Chain of Hindsight, that is easy to optimize and can learn from any form of feedback, regardless of its polarity. Our idea is inspired by how humans learn from extensive feedback presented in the form of languages. We convert all types of feedback into sequences of sentences, which are then used to fine-tune the model, allowing us to take advantage of the language comprehension capabilities of language models. We condition the model on a sequence of model generations paired with feedback. By doing so, the model is trained to generate outputs based on feedback, while learning to identify and correct negative attributes or errors. Applying our method to large language models, we observed that Chain of Hindsight significantly surpasses previous methods in aligning language models with human preferences. We report significant improvements on summarization and dialogue benchmarks, with our approach markedly preferred in human evaluations. Hao Liu 0055, Carmelo Sferrazza, Pieter Abbeel	ICLR	3
2024	RingAttention with Blockwise Transformers for Near-Infinite Context abstract Transformers have emerged as the architecture of choice for many state-of-the-art AI models, showcasing exceptional performance across a wide range of AI applications. However, the memory demands imposed by Transformers limit their ability to handle long sequences, thereby posing challenges in utilizing videos, actions, and other long-form sequences and modalities in complex environments. We present a novel approach, Blockwise RingAttention, which leverages blockwise computation of self-attention and feedforward to distribute long sequences across multiple devices while fully overlapping the communication of key-value blocks with the computation of blockwise attention. Our approach enables training and inference of sequences that are up to device count times longer than those achievable by prior memory-efficient Transformers, without resorting to approximations or incurring additional communication and computation overheads. Extensive experiments on language modeling and reinforcement learning tasks demonstrate the effectiveness of our approach in allowing millions of tokens context size and improving performance. Hao Liu 0055, Matei Zaharia, Pieter Abbeel	ICLR	3
2024	Video Language Planning abstract We are interested in enabling visual planning for complex long-horizon tasks in the space of generated videos and language, leveraging recent advances in large generative models pretrained on Internet-scale data. To this end, we present video language planning (VLP), an algorithm that consists of a tree search procedure, where we train (i) vision-language models to serve as both policies and value functions, and (ii) text-to-video models as dynamics models. VLP takes as input a long-horizon task instruction and current image observation, and outputs a long video plan that provides detailed multimodal (video and language) specifications that describe how to complete the final task. VLP scales with increasing computation budget where more computation time results in improved video plans, and is able to synthesize long-horizon video plans across different robotics domains -- from multi-object rearrangement, to multi-camera bi-arm dexterous manipulation. Generated video plans can be translated into real robot actions via goal-conditioned policies, conditioned on each intermediate frame of the generated video. Experiments show that VLP substantially improves long-horizon task success rates compared to prior methods on both simulated and real robots (across 3 hardware platforms). Yilun Du, Sherry Yang 0001, Peter R. Florence, Fei Xia 0002, Ayzaan Wahid, Brian Ichter, Pierre Sermanet, Tianhe Yu, Pieter Abbeel, Josh Tenenbaum, Leslie Pack Kaelbling, Andy Zeng 0001, Jonathan Tompson	ICLR	9
2024	The False Promise of Imitating Proprietary Language Models abstract An emerging method to cheaply improve a weaker language model is to finetune it on outputs from a stronger model, such as a proprietary system like ChatGPT (e.g., Alpaca, Self-Instruct, and others). In this work, we critically analyze this approach of imitating language models. We first finetune a series of LMs that imitate ChatGPT using varying base model sizes (1.5B--13B), data sources, and imitation data amounts (0.3M--150M tokens). We then evaluate the models using crowd raters and canonical NLP benchmarks. Initially, we were surprised by the output quality of our imitation models---they appear far better at following instructions, and crowd workers rate their outputs as competitive with ChatGPT. However, when conducting more targeted automatic evaluations, we find that imitation models close little to none of the gap from the base LM to ChatGPT on tasks that are not heavily supported in the imitation data. We show that these performance discrepancies may slip past human raters because imitation models are adept at mimicking ChatGPT’s style but not its factuality. Overall, we conclude that while model imitation can be useful for training models to follow instructions and avoid toxic outputs, it falls short its full promise in many ways. In particular, there exists a substantial capabilities gap between open and closed LMs that we find cannot be bridged merely by adding more imitation data. Instead, we find that fine-tuning more capable base LMs has a significantly more substantial effect on closing this gap. In turn, we argue that the higher leverage action for improving open-source models is to tackle the difficult challenge of developing better base LMs, rather than taking the shortcut of imitating proprietary systems. Arnav Gudibande, Eric Wallace, Charlie Snell, Xinyang Geng, Hao Liu 0055, Pieter Abbeel, Sergey Levine, Dawn Song	ICLR	6
2024	DreamSmooth: Improving Model-based Reinforcement Learning via Reward Smoothing abstract Model-based reinforcement learning (MBRL) has gained much attention for its ability to learn complex behaviors in a sample-efficient way: planning actions by generating imaginary trajectories with predicted rewards. Despite its success, we found that surprisingly, reward prediction is often a bottleneck of MBRL, especially for sparse rewards that are challenging (or even ambiguous) to predict. Motivated by the intuition that humans can learn from rough reward estimates, we propose a simple yet effective reward smoothing approach, DreamSmooth, which learns to predict a temporally-smoothed reward, instead of the exact reward at the given timestep. We empirically show that DreamSmooth achieves state-of-the-art performance on long-horizon sparse-reward tasks both in sample efficiency and final performance without losing performance on common benchmarks, such as Deepmind Control Suite and Atari benchmarks. Vint Lee, Pieter Abbeel, Youngwoon Lee	ICLR	2
2024	Tensor Trust: Interpretable Prompt Injection Attacks from an Online Game abstract While Large Language Models (LLMs) are increasingly being used in real-world applications, they remain vulnerable to prompt injection attacks: malicious third party prompts that subvert the intent of the system designer. To help researchers study this problem, we present a dataset of over 563,000 prompt injection attacks and 118,000 prompt-based "defenses" against prompt injection, all created by players of an online game called Tensor Trust. To the best of our knowledge, this is the first dataset that includes both human-generated attacks and defenses for instruction-following LLMs. The attacks in our dataset have easily interpretable structure, and shed light on the weaknesses of LLMs. We also use the dataset to create a benchmark for resistance to two types of prompt injection, which we refer to as prompt extraction and prompt hijacking. Our benchmark results show that many models are vulnerable to the attack strategies in the Tensor Trust dataset. Furthermore, we show that some attack strategies from the dataset generalize to deployed LLM-based applications, even though they have a very different set of constraints to the game. We release data and code at [tensortrust.ai/paper](https://tensortrust.ai/paper) Sam Toyer, Olivia Watkins, Ethan Mendes, Justin Svegliato, Luke Bailey, Tiffany Wang, Isaac Ong, Karim Elmaaroufi, Pieter Abbeel, Trevor Darrell, Alan Ritter, Stuart Russell 0001	ICLR	9
2024	Scalable Diffusion for Materials Generation abstract Generative models trained on internet-scale data are capable of generating novel and realistic texts, images, and videos. A natural next question is whether these models can advance science, for example by generating novel stable materials. Traditionally, models with explicit structures (e.g., graphs) have been used in modeling structural relationships in scientific data (e.g., atoms and bonds in crystals), but generating structures can be difficult to scale to large and complex systems. Another challenge in generating materials is the mismatch between standard generative modeling metrics and downstream applications. For instance, common metrics such as the reconstruction error do not correlate well with the downstream goal of discovering novel stable materials. In this work, we tackle the scalability challenge by developing a unified crystal representation that can represent any crystal structure (UniMat), followed by training a diffusion probabilistic model on these UniMat representations. Our empirical results suggest that despite the lack of explicit structure modeling, UniMat can generate high fidelity crystal structures from larger and more complex chemical systems, outperforming previous graph-based approaches under various generative modeling metrics. To better connect the generation quality of materials to downstream applications, such as discovering novel stable materials, we propose additional metrics for evaluating generative models of materials, including per-composition formation energy and stability with respect to convex hulls through decomposition energy from Density Function Theory (DFT). Lastly, we show that conditional generation with UniMat can scale to previously established crystal datasets with up to millions of crystals structures, outperforming random structure search (the current leading method for structure discovery) in discovering new stable materials. Sherry Yang 0001, KwangHwan Cho, Amil Merchant, Pieter Abbeel, Dale Schuurmans, Igor Mordatch, Ekin Dogus Cubuk	ICLR	4
2024	Probabilistic Adaptation of Black-Box Text-to-Video Models abstract Large text-to-video models trained on internet-scale data have demonstrated exceptional capabilities in generating high-fidelity videos from arbitrary textual descriptions. However, similar to proprietary language models, large text-to-video models are often black boxes whose weight parameters are not publicly available, posing a significant challenge to adapting these models to specific domains such as robotics, animation, and personalized stylization. Inspired by how a large language model can be prompted to perform new tasks without access to the model weights, we investigate how to adapt a black-box pretrained text-to-video model to a variety of downstream domains without weight access to the pretrained model. In answering this question, we propose \emph{\methodname}, which leverages the score function of a large pretrained video diffusion model as a probabilistic prior to guide the generation of a task-specific small video model. Our experiments show that, by incorporating broad knowledge and fidelity of the pretrained model probabilistically, a small model with as few as 1.25% parameters of the pretrained model can generate high-quality yet domain-specific videos for a variety of downstream domains such as animation, egocentric modeling, and modeling of simulated and real-world robotics data. As large text-to-video models starting to become available as a service similar to large language models, we advocate for private institutions to expose scores of video diffusion models as outputs in addition to generated videos to allow flexible adaptation of large pretrained text-to-video models by the general public. Sherry Yang 0001, Yilun Du, Bo Dai 0001, Dale Schuurmans, Josh Tenenbaum, Pieter Abbeel	ICLR	6
2024	Learning Interactive Real-World Simulators abstract Generative models trained on internet data have revolutionized how text, image, and video content can be created. Perhaps the next milestone for generative models is to simulate realistic experience in response to actions taken by humans, robots, and other interactive agents. Applications of a real-world simulator range from controllable content creation in games and movies, to training embodied agents purely in simulation that can be directly deployed in the real world. We explore the possibility of learning a universal simulator (UniSim) of real-world interaction through generative modeling. We first make the important observation that natural datasets available for learning a real-world simulator are often rich along different axes (e.g., abundant objects in image data, densely sampled actions in robotics data, and diverse movements in navigation data). With careful orchestration of diverse datasets, each providing a different aspect of the overall experience, UniSim can emulate how humans and agents interact with the world by simulating the visual outcome of both high-level instructions such as “open the drawer” and low-level controls such as “move by x,y” from otherwise static scenes and objects. There are numerous use cases for such a real-world simulator. As an example, we use UniSim to train both high-level vision-language planners and low-level reinforcement learning policies, each of which exhibit zero-shot real-world transfer after training purely in a learned real-world simulator. We also show that other types of intelligence such as video captioning models can benefit from training with simulated experience in UniSim, opening up even wider applications. Sherry Yang 0001, Yilun Du, Seyed Kamyar Seyed Ghasemipour, Jonathan Tompson, Leslie Pack Kaelbling, Dale Schuurmans, Pieter Abbeel	ICLR	7
2024	Unsupervised Zero-Shot Reinforcement Learning via Functional Reward Encodings abstract Can we pre-train a generalist agent from a large amount of unlabeled offline trajectories such that it can be immediately adapted to any new downstream tasks in a zero-shot manner? In this work, we present a functional reward encoding (FRE) as a general, scalable solution to this zero-shot RL problem. Our main idea is to learn functional representations of any arbitrary tasks by encoding their state-reward samples using a transformer-based variational auto-encoder. This functional encoding not only enables the pre-training of an agent from a wide diversity of general unsupervised reward functions, but also provides a way to solve any new downstream tasks in a zero-shot manner, given a small number of reward-annotated samples. We empirically show that FRE agents trained on diverse random unsupervised reward functions can generalize to solve novel tasks in a range of simulated robotic benchmarks, often outperforming previous zero-shot RL and offline RL methods. Kevin Frans, Seohong Park, Pieter Abbeel, Sergey Levine	ICML	3
2024	Visual Representation Learning with Stochastic Frame Prediction abstract Self-supervised learning of image representations by predicting future frames is a promising direction but still remains a challenge. This is because of the under-determined nature of frame prediction; multiple potential futures can arise from a single current frame. To tackle this challenge, in this paper, we revisit the idea of stochastic video generation that learns to capture uncertainty in frame prediction and explore its effectiveness for representation learning. Specifically, we design a framework that trains a stochastic frame prediction model to learn temporal information between frames. Moreover, to learn dense information within each frame, we introduce an auxiliary masked image modeling objective along with a shared decoder architecture. We find this architecture allows for combining both objectives in a synergistic and compute-efficient manner. We demonstrate the effectiveness of our framework on a variety of tasks from video label propagation and vision-based robot learning domains, such as video segmentation, pose tracking, vision-based robotic locomotion, and manipulation tasks. Code is available on the project webpage: https://sites.google.com/view/2024rsp. Huiwon Jang, Dongyoung Kim, Jinwoo Shin, Pieter Abbeel, Younggyo Seo	ICML	5
2024	Learning to Model the World With Language abstract To interact with humans and act in the world, agents need to understand the range of language that people use and relate it to the visual world. While current agents can learn to execute simple language instructions, we aim to build agents that leverage diverse language---language like "this button turns on the TV" or "I put the bowls away"---that conveys general knowledge, describes the state of the world, provides interactive feedback, and more. Our key idea is that agents should interpret such diverse language as a signal that helps them predict the future: what they will observe, how the world will behave, and which situations will be rewarded. This perspective unifies language understanding with future prediction as a powerful self-supervised learning objective. We instantiate this in Dynalang, an agent that learns a multimodal world model to predict future text and image representations, and learns to act from imagined model rollouts. While current methods that learn language-conditioned policies degrade in performance with more diverse types of language, we show that Dynalang learns to leverage environment descriptions, game rules, and instructions to excel on tasks ranging from game-playing to navigating photorealistic home scans. Finally, we show that our method enables additional capabilities due to learning a generative model: Dynalang can be pretrained on text-only data, enabling learning from offline datasets, and generate language grounded in an environment. Jessy Lin, Olivia Watkins, Danijar Hafner, Pieter Abbeel, Daniel Klein 0001, Anca D. Dragan	ICML	5
2024	Learning a Diffusion Model Policy from Rewards via Q-Score Matching abstract Diffusion models have become a popular choice for representing actor policies in behavior cloning and offline reinforcement learning. This is due to their natural ability to optimize an expressive class of distributions over a continuous space. However, previous works fail to exploit the score-based structure of diffusion models, and instead utilize a simple behavior cloning term to train the actor, limiting their ability in the actor-critic setting. In this paper, we present a theoretical framework linking the structure of diffusion model policies to a learned Q-function, by linking the structure between the score of the policy to the action gradient of the Q-function. We focus on off-policy reinforcement learning and propose a new policy update method from this theory, which we denote Q-score matching. Notably, this algorithm only needs to differentiate through the denoising model rather than the entire diffusion model evaluation, and converged policies through Q-score matching are implicitly multi-modal and explorative in continuous domains. We conduct experiments in simulated environments to demonstrate the viability of our proposed method and compare to popular baselines. Source code is available from the project website: https://www.michaelpsenka.io/qsm/. Michael Psenka, Alejandro Escontrela, Pieter Abbeel, Yi Ma 0001	ICML	3
2024	Position: Video as the New Language for Real-World Decision Making abstract Both text and video data are abundant on the internet and support large-scale self-supervised learning through next token or frame prediction. However, they have not been equally leveraged: language models have had significant real-world impact, whereas video generation has remained largely limited to media entertainment. Yet video data captures important information about the physical world that is difficult to express in language. To address this gap, we discuss an under-appreciated opportunity to extend video generation to solve tasks in the real world. We observe how, akin to language, video can serve as a unified interface that can absorb internet knowledge and represent diverse tasks. Moreover, we demonstrate how, like language models, video generation can serve as planners, agents, compute engines, and environment simulators through techniques such as in-context learning, planning and reinforcement learning. We identify major impact opportunities in domains such as robotics, self-driving, and science, supported by recent work that demonstrates how such advanced capabilities in video generation are plausibly within reach. Lastly, we identify key challenges in video generation that mitigate progress. Addressing these challenges will enable video generation models to demonstrate unique value alongside language models in a wider array of AI applications. Sherry Yang 0001, Jacob C. Walker, Jack Parker-Holder, Yilun Du, Jake Bruce, André Barreto 0001, Pieter Abbeel, Dale Schuurmans	ICML	7
2024	SpawnNet: Learning Generalizable Visuomotor Skills from Pre-trained Network abstract The existing internet-scale image and video datasets cover a wide range of everyday objects and tasks, bringing the potential of learning policies that generalize in diverse scenarios. Prior works have explored visual pre-training with different self-supervised objectives. Still, the generalization capabilities of the learned policies and the advantages over well-tuned baselines remain unclear from prior studies. In this work, we present a focused study of the generalization capabilities of the pre-trained visual representations at the categorical level. We identify the key bottleneck in using a frozen pre-trained visual backbone for policy learning and then propose SpawnNet, a novel two-stream architecture that learns to fuse pre-trained multi-layer representations into a separate network to learn a robust policy. Through extensive simulated and real experiments, we show significantly better categorical generalization compared to prior approaches in imitation learning settings. Open-sourced code and videos can be found on our website: https://xingyu-lin.github.io/spawnnet/. John So, Sashwat Mahalingam, Fangchen Liu, Pieter Abbeel	ICRA	5
2024	Closing the Visual Sim-to-Real Gap with Object-Composable NeRFs abstract Deep learning methods for perception are the cornerstone of many robotic systems. Despite their potential for impressive performance, obtaining real-world training data is expensive, and can be impractically difficult for some tasks. Sim-to-real transfer with domain randomization offers a potential workaround, but often requires extensive manual tuning and results in models that are brittle to distribution shift between sim and real. In this work, we introduce Composable Object Volume NeRF (COV-NeRF), an object-composable NeRF model that is the centerpiece of a real-to-sim pipeline for synthesizing training data targeted to scenes and objects from the real world. COV-NeRF extracts objects from real images and composes them into new scenes, generating photorealistic renderings and many types of 2D and 3D supervision, including depth maps, segmentation masks, and meshes. We show that COV-NeRF matches the rendering quality of modern NeRF methods, and can be used to rapidly close the sim-to-real gap across a variety of perceptual modalities. Nikhil Mishra, Maximilian Sieb, Pieter Abbeel, Xi Chen 0022	ICRA	3
2024	Open X-Embodiment: Robotic Learning Datasets and RT-X Models : Open X-Embodiment Collaboration abstract Large, high-capacity models trained on diverse datasets have shown remarkable successes on efficiently tackling downstream applications. In domains from NLP to Computer Vision, this has led to a consolidation of pretrained models, with general pretrained backbones serving as a starting point for many applications. Can such a consolidation happen in robotics? Conventionally, robotic learning methods train a separate model for every application, every robot, and even every environment. Can we instead train "generalist" X-robot policy that can be adapted efficiently to new robots, tasks, and environments? In this paper, we provide datasets in standardized data formats and models to make it possible to explore this possibility in the context of robotic manipulation, alongside experimental results that provide an example of effective X-robot policies. We assemble a dataset from 22 different robots collected through a collaboration between 21 institutions, demonstrating 527 skills (160266 tasks). We show that a high-capacity model trained on this data, which we call RT-X, exhibits positive transfer and improves the capabilities of multiple robots by leveraging experience from other platforms. The project website is robotics-transformer-x.github.io. Abigail O'Neill, Abhiram Maddukuri, Abhishek Gupta 0004, Abhishek Padalkar, Abraham Lee, Acorn Pooley, Agrim Gupta, Ajay Mandlekar, Ajinkya Jain, Albert Tung, Alex Bewley, Alex Irpan, Alexander Khazatsky, Anant Rai, Anchit Gupta, Andrew E. Wang, Anikait Singh, Animesh Garg, Aniruddha Kembhavi, Annie Xie, Anthony Brohan, Antonin Raffin, Archit Sharma, Arefeh Yavary, Arhan Jain, Ashwin Balakrishna, Ayzaan Wahid, Ben Burgess-Limerick, Bernhard Schölkopf, Blake Wulfe, Brian Ichter, Cewu Lu, Charles Xu 0003, Charlotte Le, Chelsea Finn, Chen Wang 0053, Chenfeng Xu, Cheng Chi 0001, Chenguang Huang, Christine Chan, Christopher Agia, Chuer Pan, Chuyuan Fu, Coline Devin, Danfei Xu, Daniel Morton, Danny Drieß, Daphne Chen, Deepak Pathak, Dhruv Shah, Dieter Büchler, Dinesh Jayaraman, Dmitry Kalashnikov, Dorsa Sadigh, Edward Johns, Ethan Paul Foster, Fangchen Liu, Federico Ceola, Fei Xia 0002, Feiyu Zhao, Freek Stulp, Gaoyue Zhou, Gaurav S. Sukhatme, Gautam Salhotra, Gilbert Feng, Giulio Schiavi, Glen Berseth, Gregory Kahn, Guanzhi Wang, Hao Su 0001, Haoshu Fang, Henghui Bao, Heni Ben Amor, Henrik I. Christensen, Hiroki Furuta, Homer Walke, Hongjie Fang, Huy Ha, Igor Mordatch, Ilija Radosavovic, Isabel Leal, Jacky Liang, Jad Abou-Chakra, Jaehyung Kim 0001, Jaimyn Drake, Jan Peters 0001, Jan Schneider 0007, Jasmine Hsu, Jeannette Bohg, Jeffrey T. Bingham, Jensen Gao, Jiaheng Hu, Jiajun Wu 0001, Jiankai Sun, Jianlan Luo, Jiayuan Gu, Jie Tan 0001, Jihoon Oh, Jimmy Wu, Jingpei Lu, Jitendra Malik, João Silvério, Joey Hejna, Jonathan Booher, Jonathan Tompson, Jonathan Yang, Jordi Salvador, Joseph J. Lim, Junhyek Han, Kanishka Rao, Karl Pertsch, Karol Hausman, Keegan Go, Keerthana Gopalakrishnan, Kenneth Y. Goldberg, Kendra Byrne, Kenneth Oslund, Kento Kawaharazuka, Kevin Black, Kevin Zhang 0002, Kiana Ehsani, Kiran Lekkala, Kirsty Ellis, Krishan Rana, Krishnan Srinivasan, Kuan Fang, Kunal Pratap Singh, Kuo-Hao Zeng, Kyle Hatch, Kyle Hsu, Laurent Itti, Yunliang Chen 0001, Lerrel Pinto, Li Fei-Fei 0001, Liam Tan, Linxi Fan, Lionel Ott, Lisa Lee, Luca Weihs, Magnum Chen, Marion Lepert, Marius Memmel, Masayoshi Tomizuka, Masha Itkina, Mateo Guaman Castro, Max Spero, Maximilian Du, Michael Ahn, Michael C. Yip, Mingtong Zhang 0003, Mingyu Ding, Minho Heo, Mohan Kumar Srirama, Mohit Sharma 0001, Moo Jin Kim, Naoaki Kanazawa, Nicklas Hansen 0001, Nicolas Heess, Nikhil J. Joshi, Niko Sünderhauf, Norman Di Palo, Nur Muhammad Shafiullah, Oier Mees, Oliver Kroemer, Osbert Bastani, Pannag R. Sanketi, Patrick Tree Miller, Patrick Yin, Paul Wohlhart, Peng Xu 0010, Peter David Fagan, Peter Mitrano, Pierre Sermanet, Pieter Abbeel, Priya Sundaresan, Qiuyu Chen, Rafael Rafailov, Ria Doshi, Roberto Martin Martin, Rohan Baijal, Rosario Scalise, Rose Hendrix, Roy Lin, Runjia Qian, Russell Mendonca, Rutav Shah, Ryan Hoque, Ryan Julian, Samuel Bustamante-Gomez, Sean Kirmani, Sergey Levine, Sherry Moore, Shikhar Bahl, Shivin Dass, Shubham D. Sonawani, Shuran Song, Sichun Xu, Siddhant Haldar, Siddharth Karamcheti, Simeon Adebola, Simon Guist, Soroush Nasiriany, Stefan Schaal, Stefan Welker, Stephen Tian, Subramanian Ramamoorthy, Sudeep Dasari, Suneel Belkhale, Sungjae Park, Suraj Nair 0003, Suvir Mirchandani, Takayuki Osa, Tanmay Gupta, Tatsuya Harada, Tatsuya Matsushima, Ted Xiao, Thomas Kollar, Tianhe Yu, Tianli Ding, Todor Davchev, Tony Z. Zhao, Travis Armstrong, Trevor Darrell, Trinity Chung, Vidhi Jain, Vincent Vanhoucke, Wolfram Burgard, Xiaolong Wang 0004, Xinghao Zhu, Xinyang Geng, Liangwei Xu, Yecheng Jason Ma 0001, Yejin Kim 0003, Yevgen Chebotar, Yilin Wu 0003, Yonatan Bisk, Yoonyoung Cho, Youngwoon Lee, Yuchen Cui, Yueh-Hua Wu, Yujin Tang, Yuke Zhu, Yunchu Zhang, Yunfan Jiang 0001, Yunshuang Li, Yunzhu Li, Yusuke Iwasawa, Yutaka Matsuo, Zehan Ma, Zichen Jeff Cui, Zichen Zhang 0016, Zipeng Lin	ICRA	185
2024	The Power of the Senses: Generalizable Manipulation from Vision and Touch through Masked Multimodal Learning abstract Humans rely on the synergy of their senses for most essential tasks. For tasks requiring object manipulation, we seamlessly and effectively exploit the complementarity of our senses of vision and touch. This paper draws inspiration from such capabilities and aims to find a systematic approach to fuse visual and tactile information in a reinforcement learning setting. We propose Masked Multimodal Learning (M3L), which jointly learns a policy and visual-tactile representations based on masked autoencoding. The representations jointly learned from vision and touch improve sample efficiency, and unlock generalization capabilities beyond those achievable through each of the senses separately. Remarkably, representations learned in a multimodal setting also benefit vision-only policies at test time. We evaluate M3L on three simulated environments with both visual and tactile observations: robotic insertion, door opening, and dexterous in-hand manipulation, demonstrating the benefits of learning a multimodal policy. Videos of the experiments and the open-source code are available at https://sferrazza.cc/m3l_site. Carmelo Sferrazza, Younggyo Seo, Hao Liu 0055, Youngwoon Lee, Pieter Abbeel	IROS	5
2024	From LLMs to Actions: Latent Codes as Bridges in Hierarchical Robot Control abstract Hierarchical control for robotics has long been plagued by the need to have a well defined interface layer to communicate between high-level task planners and low-level policies. With the advent of LLMs, language has been emerging as a prospective interface layer. However, this has several limitations. Not all tasks can be decomposed into steps that are easily expressible in natural language (e.g. performing a dance routine). Further, it makes end-to-end finetuning on embodied data challenging due to domain shift and catastrophic forgetting. We introduce our method – Latent Codes as Bridges (LCB) – as an alternate architecture to overcome these limitations. LCB uses a learnable latent code to act as a bridge between LLMs and low-level policies. This enables LLMs to flexibly communicate goals in the task plan without being entirely constrained by language limitations. Additionally, it enables end-to-end finetuning without destroying the embedding space of word tokens learned during pre-training. Through experiments on Language Table and Calvin, two common language based benchmarks for embodied agents, we find that LCB outperforms baselines (including those w/ GPT-4V) that leverage pure language as the interface layer on tasks that require reasoning and multi-step behaviors. Yide Shentu, Philipp Wu, Aravind Rajeswaran, Pieter Abbeel	IROS	4
2024	GELLO: A General, Low-Cost, and Intuitive Teleoperation Framework for Robot Manipulators abstract Humans can teleoperate robots to accomplish complex manipulation tasks. Imitation learning has emerged as a powerful framework that leverages human teleoperated demonstrations to teach robots new skills. However, the performance of the learned policies is bottlenecked by the quality, scale, and variety of the demonstration data. In this paper, we aim to lower the barrier to collecting large and high-quality human demonstration data by proposing a GEneraL framework for building LOw-cost and intuitive teleoperation systems for robotic manipulation (GELLO). Given a target robot arm, we build a GELLO controller device that has the same kinematic structure as the target arm, leveraging 3D-printed parts and economical off-the-shelf motors. GELLO is easy to build and intuitive to use. Through an extensive user study, we show that GELLO enables more reliable and efficient demonstration collection compared to other cost efficient teleoperation devices commonly used in the imitation learning literature such as virtual reality controllers and 3D spacemouses. We further demonstrate the capabilities of GELLO for performing complex bi-manual and contact-rich manipulation tasks. To make GELLO accessible to everyone, we have designed and built GELLO systems for 3 commonly used robotic arms: Franka, UR5, and xArm. All software and hardware are open-sourced and can be found on our website: https://wuphilipp.github.io/gello/. Philipp Wu, Yide Shentu, Zhongke Yi, Pieter Abbeel	IROS	5
2024	Vision Foundation Model Enables Generalizable Object Pose Estimation abstract Object pose estimation plays a crucial role in robotic manipulation, however, its practical applicability still suffers from limited generalizability. This paper addresses the challenge of generalizable object pose estimation, particularly focusing on category-level object pose estimation for unseen object categories. Current methods either require impractical instance-level training or are confined to predefined categories, limiting their applicability. We propose VFM-6D, a novel framework that explores harnessing existing vision and language models, to elaborate object pose estimation into two stages: category-level object viewpoint estimation and object coordinate map estimation. Based on the two-stage framework, we introduce a 2D-to-3D feature lifting module and a shape-matching module, both of which leverage pre-trained vision foundation models to improve object representation and matching accuracy. VFM-6D is trained on cost-effective synthetic data and exhibits superior generalization capabilities. It can be applied to both instance-level unseen object pose estimation and category-level object pose estimation for novel categories. Evaluations on benchmark datasets demonstrate the effectiveness and versatility of VFM-6D in various real-world scenarios. Kai Chen 0028, Yiyao Ma, Stephen James, Jianshu Zhou, Yun-Hui Liu 0001, Pieter Abbeel, Qi Dou 0001	NeurIPS	7
2024	A StrongREJECT for Empty Jailbreaks abstract Most jailbreak papers claim the jailbreaks they propose are highly effective, often boasting near-100% attack success rates. However, it is perhaps more common than not for jailbreak developers to substantially exaggerate the effectiveness of their jailbreaks. We suggest this problem arises because jailbreak researchers lack a standard, high-quality benchmark for evaluating jailbreak performance, leaving researchers to create their own. To create a benchmark, researchers must choose a dataset of forbidden prompts to which a victim model will respond, along with an evaluation method that scores the harmfulness of the victim model’s responses. We show that existing benchmarks suffer from significant shortcomings and introduce the StrongREJECT benchmark to address these issues. StrongREJECT's dataset contains prompts that victim models must answer with specific, harmful information, while its automated evaluator measures the extent to which a response gives useful information to forbidden prompts. In doing so, the StrongREJECT evaluator achieves state-of-the-art agreement with human judgments of jailbreak effectiveness. Notably, we find that existing evaluation methods significantly overstate jailbreak effectiveness compared to human judgments and the StrongREJECT evaluator. We describe a surprising and novel phenomenon that explains this discrepancy: jailbreaks bypassing a victim model’s safety fine-tuning tend to reduce its capabilities. Together, our findings underscore the need for researchers to use a high-quality benchmark, such as StrongREJECT, when developing new jailbreak attacks. We release the StrongREJECT code and data at https://strong-reject.readthedocs.io/. Alexandra Souly, Qingyuan Lu, Dillon Bowen, Tu Trinh, Elvis Hsieh, Sana Pandey, Pieter Abbeel, Justin Svegliato, Scott Emmons, Olivia Watkins, Sam Toyer	NeurIPS	7
2023	Improving Long-Horizon Imitation through Instruction Prediction abstract Complex, long-horizon planning and its combinatorial nature pose steep challenges for learning-based agents. Difficulties in such settings are exacerbated in low data regimes where over-fitting stifles generalization and compounding errors hurt accuracy. In this work, we explore the use of an often unused source of auxiliary supervision: language. Inspired by recent advances in transformer-based models, we train agents with an instruction prediction loss that encourages learning temporally extended representations that operate at a high level of abstraction. Concretely, we demonstrate that instruction modeling significantly improves performance in planning environments when training with a limited number of demonstrations on the BabyAI and Crafter benchmarks. In further analysis we find that instruction modeling is most important for tasks that require complex reasoning, while understandably offering smaller gains in environments that require simple plans. More details and code can be found at \url{https://github.com/jhejna/instruction-prediction}. Joey Hejna, Pieter Abbeel, Lerrel Pinto	AAAI	2
2023	VectorFusion: Text-to-SVG by Abstracting Pixel-Based Diffusion Models abstract Diffusion models have shown impressive results in text-to-image synthesis. Using massive datasets of captioned images, diffusion models learn to generate raster images of highly diverse objects and scenes. However, designers frequently use vector representations of images like Scalable Vector Graphics (SVGs) for digital icons or art. Vector graphics can be scaled to any size, and are compact. We show that a text-conditioned diffusion model trained on pixel representations of images can be used to generate SVG-exportable vector graphics. We do so without access to large datasets of captioned SVGs. By optimizing a differentiable vector graphics rasterizer, our method, VectorFusion, distills abstract semantic knowledge out of a pretrained diffusion model. Inspired by recent text-to-3D work, we learn an SVG consistent with a caption using Score Distillation Sampling. To accelerate generation and improve fidelity, VectorFusion also initializes from an image sample. Experiments show greater quality than prior work, and demonstrate a range of styles including pixel art and sketches. Ajay Jain, Amber Xie, Pieter Abbeel	CVPR	3
2023	Preference Transformer: Modeling Human Preferences using Transformers for RL Changyeon Kim, Jongjin Park, Jinwoo Shin, Honglak Lee, Pieter Abbeel, Kimin Lee	ICLR	5
2023	Dichotomy of Control: Separating What You Can Control from What You Cannot Sherry Yang 0001, Dale Schuurmans, Pieter Abbeel, Ofir Nachum	ICLR	3
2023	Become a Proficient Player with Limited Data through Watching Pure Videos Weirui Ye, Pieter Abbeel, Yang Gao 0029	ICLR	3
2023	CLUTR: Curriculum Learning via Unsupervised Task Representation Learning abstract Reinforcement Learning (RL) algorithms are often known for sample inefficiency and difficult generalization. Recently, Unsupervised Environment Design (UED) emerged as a new paradigm for zero-shot generalization by simultaneously learning a task distribution and agent policies on the generated tasks. This is a non-stationary process where the task distribution evolves along with agent policies; creating an instability over time. While past works demonstrated the potential of such approaches, sampling effectively from the task space remains an open challenge, bottlenecking these approaches. To this end, we introduce CLUTR: a novel unsupervised curriculum learning algorithm that decouples task representation and curriculum learning into a two-stage optimization. It first trains a recurrent variational autoencoder on randomly generated tasks to learn a latent task manifold. Next, a teacher agent creates a curriculum by maximizing a minimax REGRET-based objective on a set of latent tasks sampled from this manifold. Using the fixed-pretrained task manifold, we show that CLUTR successfully overcomes the non-stationarity problem and improves stability. Our experimental results show CLUTR outperforms PAIRED, a principled and popular UED method, in the challenging CarRacing and navigation environments: achieving 10.6X and 45% improvement in zero-shot generalization, respectively. CLUTR also performs comparably to the non-UED state-of-the-art for CarRacing, while requiring 500X fewer environment interactions. We open source our code at https://github.com/clutr/clutr. Abdus Salam Azad, Izzeddin Gur, Jasper Emhoff, Nathaniel Alexis, Aleksandra Faust, Pieter Abbeel, Ion Stoica	ICML	6
2023	Guiding Pretraining in Reinforcement Learning with Large Language Models abstract Reinforcement learning algorithms typically struggle in the absence of a dense, well-shaped reward function. Intrinsically motivated exploration methods address this limitation by rewarding agents for visiting novel states or transitions, but these methods offer limited benefits in large environments where most discovered novelty is irrelevant for downstream tasks. We describe a method that uses background knowledge from text corpora to shape exploration. This method, called ELLM (Exploring with LLMs) rewards an agent for achieving goals suggested by a language model prompted with a description of the agent's current state. By leveraging large-scale language model pretraining, ELLM guides agents toward human-meaningful and plausibly useful behaviors without requiring a human in the loop. We evaluate ELLM in the Crafter game environment and the Housekeep robotic simulator, showing that ELLM-trained agents have better coverage of common-sense behaviors during pretraining and usually match or improve performance on a range of downstream tasks. Olivia Watkins, Cédric Colas, Trevor Darrell, Pieter Abbeel, Abhishek Gupta 0004, Jacob Andreas	ICML	6
2023	Emergent Agentic Transformer from Chain of Hindsight Experience abstract Large transformer models powered by diverse data and model scale have dominated natural language modeling and computer vision and pushed the frontier of multiple AI areas. In reinforcement learning (RL), despite many efforts into transformer-based policies, a key limitation, however, is that current transformer-based policies cannot learn by directly combining information from multiple sub-optimal trials. In this work, we address this issue using recently proposed chain of hindsight to relabel experience, where we train a transformer on a sequence of trajectory experience ascending sorted according to their total rewards. Our method consists of relabelling target return of each trajectory to the maximum total reward among in sequence of trajectories and training an autoregressive model to predict actions conditioning on past states, actions, rewards, target returns, and task completion tokens, the resulting model, Agentic Transformer (AT), can learn to improve upon itself both at training and test time. As we show on D4RL and ExoRL benchmarks, to the best our knowledge, this is the first time that a simple transformer-based model performs competitively with both temporal-difference and imitation-learning-based approaches, even from sub-optimal data. Our Agentic Transformer also shows a promising scaling trend that bigger models consistently improve results. Hao Liu 0055, Pieter Abbeel	ICML	2
2023	Controllability-Aware Unsupervised Skill Discovery abstract One of the key capabilities of intelligent agents is the ability to discover useful skills without external supervision. However, the current unsupervised skill discovery methods are often limited to acquiring simple, easy-to-learn skills due to the lack of incentives to discover more complex, challenging behaviors. We introduce a novel unsupervised skill discovery method, Controllability-aware Skill Discovery (CSD), which actively seeks complex, hard-to-control skills without supervision. The key component of CSD is a controllability-aware distance function, which assigns larger values to state transitions that are harder to achieve with the current skills. Combined with distance-maximizing skill discovery, CSD progressively learns more challenging skills over the course of training as our jointly trained distance function reduces rewards for easy-to-achieve skills. Our experimental results in six robotic manipulation and locomotion environments demonstrate that CSD can discover diverse complex skills including object manipulation and locomotion skills with no supervision, significantly outperforming prior unsupervised skill discovery methods. Videos and code are available at https://seohong.me/projects/csd/ Seohong Park, Kimin Lee, Youngwoon Lee, Pieter Abbeel	ICML	4
2023	Multi-View Masked World Models for Visual Robotic Manipulation abstract Visual robotic manipulation research and applications often use multiple cameras, or views, to better perceive the world. How else can we utilize the richness of multi-view data? In this paper, we investigate how to learn good representations with multi-view data and utilize them for visual robotic manipulation. Specifically, we train a multi-view masked autoencoder which reconstructs pixels of randomly masked viewpoints and then learn a world model operating on the representations from the autoencoder. We demonstrate the effectiveness of our method in a range of scenarios, including multi-view control and single-view control with auxiliary cameras for representation learning. We also show that the multi-view masked autoencoder trained with multiple randomized viewpoints enables training a policy with strong viewpoint randomization and transferring the policy to solve real-robot tasks without camera calibration and an adaptation procedure. Video demonstrations are available at: https://sites.google.com/view/mv-mwm. Younggyo Seo, Stephen James, Kimin Lee, Jinwoo Shin, Pieter Abbeel	ICML	6
2023	Multi-Environment Pretraining Enables Transfer to Action Limited Datasets abstract Using massive datasets to train large-scale models has emerged as a dominant approach for broad generalization in natural language and vision applications. In reinforcement learning, however, a key challenge is that available data of sequential decision making is often not annotated with actions - for example, videos of game-play are much more available than sequences of frames paired with their logged game controls. We propose to circumvent this challenge by combining large but sparsely-annotated datasets from a target environment of interest with fully-annotated datasets from various other source environments. Our method, Action Limited PreTraining (ALPT), leverages the generalization capabilities of inverse dynamics modelling (IDM) to label missing action data in the target environment. We show that utilizing even one additional environment dataset of labelled data during IDM pretraining gives rise to substantial improvements in generating action labels for unannotated sequences. We evaluate our method on benchmark game-playing environments and show that we can significantly improve game performance and generalization capability compared to other approaches, using annotated datasets equivalent to only $12$ minutes of gameplay. Highlighting the power of IDM, we show that these benefits remain even when target and source environments share no common actions. David Venuto, Sherry Yang 0001, Pieter Abbeel, Doina Precup, Igor Mordatch, Ofir Nachum	ICML	3
2023	Masked Trajectory Models for Prediction, Representation, and Control abstract We introduce Masked Trajectory Models (MTM) as a generic abstraction for sequential decision making. MTM takes a trajectory, such as a state-action sequence, and aims to reconstruct the trajectory conditioned on random subsets of the same trajectory. By training with a highly randomized masking pattern, MTM learns versatile networks that can take on different roles or capabilities, by simply choosing appropriate masks at inference time. For example, the same MTM network can be used as a forward dynamics model, inverse dynamics model, or even an offline RL agent. Through extensive experiments in several continuous control tasks, we show that the same MTM network -- i.e. same weights -- can match or outperform specialized networks trained for the aforementioned capabilities. Additionally, we find that state representations learned by MTM can significantly accelerate the learning speed of traditional RL algorithms. Finally, in offline RL benchmarks, we find that MTM is competitive with specialized offline RL algorithms, despite MTM being a generic self-supervised learning method without any explicit RL components. Code is available at https://github.com/facebookresearch/mtm. Philipp Wu, Arjun Majumdar, Kevin Stone, Igor Mordatch, Pieter Abbeel, Aravind Rajeswaran	ICML	6
2023	Temporally Consistent Transformers for Video Generation abstract To generate accurate videos, algorithms have to understand the spatial and temporal dependencies in the world. Current algorithms enable accurate predictions over short horizons but tend to suffer from temporal inconsistencies. When generated content goes out of view and is later revisited, the model invents different content instead. Despite this severe limitation, no established benchmarks exist for video generation with long temporal dependencies. In this paper, we curate 3 challenging video datasets with long-range dependencies by rendering walks through 3D scenes of procedural mazes, Minecraft worlds, and indoor scans. We perform a comprehensive evaluation of current models and observe their limitations in temporal consistency. Moreover, we introduce the Temporally Consistent Transformer (TECO), a generative model that substantially improves long-term consistency while also reducing sampling time. By compressing its input sequence into fewer embeddings, applying a temporal transformer, and expanding back using a spatial MaskGit, TECO outperforms existing models across many metrics. Videos are available on the website: https://wilson1yan.github.io/teco Wilson Yan, Danijar Hafner, Stephen James, Pieter Abbeel	ICML	4
2023	The Wisdom of Hindsight Makes Language Models Better Instruction Followers abstract Reinforcement learning has seen wide success in finetuning large language models to better align with instructions via human feedback. The so-called algorithm, Reinforcement Learning with Human Feedback (RLHF) demonstrates impressive performance on the GPT series models. However, the underlying reinforcement learning algorithm is complex and requires additional training for reward and value networks. In this paper, we consider an alternative approach: converting feedback to instruction by relabeling the original one and training the model for better alignment in a supervised manner. Such an algorithm doesn't require any additional parameters except for the original language model and maximally reuses the pretraining pipeline. To achieve this, we formulate instruction alignment problem for language models as a goal-reaching problem in decision making. We propose Hindsight Instruction Relabeling (HIR), a novel algorithm for aligning language models with instructions. The resulting two-stage algorithm shed light to a family of reward-free approaches that utilize the hindsightly relabeled instructions based on feedback. We evaluate the performance of HIR extensively on 12 challenging BigBench reasoning tasks and show that HIR outperforms the baseline algorithms and is comparable to or even surpasses supervised fine-tuning. The implementation of HIR is available at https://github.com/tianjunz/HIR. Tianjun Zhang, Fangchen Liu, Justin Wong, Pieter Abbeel, Joseph Gonzalez 0001	ICML	4
2023	StereoPose: Category-Level 6D Transparent Object Pose Estimation from Stereo Images via Back-View NOCS abstract Most existing methods for category-level pose estimation rely on object point clouds. However, when considering transparent objects, depth cameras are usually not able to capture high-quality data, resulting in point clouds with severe artifacts. Without a complete point cloud, existing methods are not applicable to challenging transparent objects. To tackle this problem, we present StereoPose, a novel stereo image based framework for category-level object pose estimation, ideally suited for transparent objects. For a robust estimation from pure stereo images, we develop a pipeline that decouples category-level pose estimation into object size estimation, initial pose estimation, and pose refinement. StereoPose then estimates object pose based on representation in the normalized object coordinate space (NOCS). To address the issue of image content aliasing, we further define a back-view NOCS map for the transparent object. The back-view NOCS aims to reduce the network learning ambiguity caused by content aliasing, and leverage informative cues on the back of the transparent object for more accurate pose estimation. To further improve the performance of the stereo framework, StereoPose is equipped with a parallax attention module for stereo feature fusion and an epipolar loss for improving the stereo-view consistency of network predictions. Extensive experiments on the public TOD dataset demonstrate the superiority of the proposed StereoPose framework for category-level 6D transparent object pose estimation. Code and demos will be available on the project homepage: www.cse.cuhk.edu.hk/~kaichen/stereopose.html. Kai Chen 0028, Stephen James, Congying Sui, Yun-Hui Liu 0001, Pieter Abbeel, Qi Dou 0001	ICRA	5
2023	Distributional Instance Segmentation: Modeling Uncertainty and High Confidence Predictions with Latent-MaskRCNN abstract Object recognition and instance segmentation are fundamental skills in any robotic or autonomous system. Existing state-of-the-art methods are often unable to capture meaningful uncertainty in challenging or ambiguous scenes, and as such can cause critical errors in high-performance applications. In this paper, we explore a class of distributional instance segmentation models using latent codes that can model uncertainty over plausible hypotheses of object masks. For robotic picking applications, we propose a confidence mask method to achieve the high precision necessary in industrial use cases. We show that our method can significantly reduce critical errors in robotic systems, including our newly released dataset of ambiguous scenes in a robotic application. On a real-world apparel-picking robot, our method significantly reduces double pick errors while maintaining high performance. Yuxuan Liu 0001, Nikhil Mishra, Pieter Abbeel, Xi Chen 0022	ICRA	3
2023	Train Offline, Test Online: A Real Robot Learning Benchmark abstract Three challenges limit the progress of robot learning research: robots are expensive (few labs can participate), everyone uses different robots (findings do not generalize across labs), and we lack internet-scale robotics data. We take on these challenges via a new benchmark: Train Offline, Test Online (TOTO). TOTO provides remote users with access to shared robots for evaluating methods on common tasks and an open-source dataset of these tasks for offline training. Its manipulation task suite requires challenging generalization to unseen objects, positions, and lighting. We present initial results on TOTO comparing five pretrained visual representations and four offline policy learning baselines, remotely contributed by five institutions. The real promise of TOTO, however, lies in the future: we release the benchmark for additional submissions from any user, enabling easy, direct comparison to several methods without the need to obtain hardware or collect data. Gaoyue Zhou, Victoria Dean, Mohan Kumar Srirama, Aravind Rajeswaran, Jyothish Pari, Kyle Hatch, Aryan Jain, Tianhe Yu, Pieter Abbeel, Lerrel Pinto, Chelsea Finn, Abhinav Gupta 0001	ICRA	9
2023	Self-Supervised Instance Segmentation by Grasping abstract Instance segmentation is a fundamental skill for many robotic applications. We propose a self-supervised method that uses grasp interactions to collect segmentation supervision for an instance segmentation model. When a robot grasps an item, the mask of that grasped item can be inferred from the images of the scene before and after the grasp. Leveraging this insight, we learn a grasp segmentation model from a small dataset of labelled images to segment the grasped object from before and after grasp images. Such a model can segment grasped objects from thousands of grasp interactions without costly human annotation. Using the segmented grasped objects, we can “cut” objects from their original scenes and “paste” them into new scenes to generate instance supervision. We show that our grasp segmentation model provides a 5x error reduction when segmenting grasped objects compared with traditional image subtraction approaches. Combined with our “cut-and-paste” generation method, instance segmentation models trained with our method achieve better performance than a model trained with 10x the amount of labeled data. On a real robotic grasping system, our instance segmentation model reduces the rate of grasp errors by over 3x compared to an image subtraction baseline. Yuxuan Liu 0001, Xi Chen 0022, Pieter Abbeel	IROS	3
2023	Convolutional Occupancy Models for Dense Packing of Complex, Novel Objects abstract Dense packing in pick-and-place systems is an important feature in many warehouse and logistics applications. Prior work in this space has largely focused on planning algorithms in simulation, but real-world packing performance is often bottlenecked by the difficulty of perceiving 3D object geometry in highly occluded, partially observed scenes. In this work, we present a fully-convolutional shape completion model, F-CON, which can be easily combined with off-the-shelf planning methods for dense packing in the real world. We also release a simulated dataset, COB-3D-v2, that can be used to train shape completion models for real-word robotics applications, and use it to demonstrate that F-CON outperforms other state-of-the-art shape completion methods. Finally, we equip a real-world pick-and-place system with F-CON, and demonstrate dense packing of complex, unseen objects in cluttered scenes. Across multiple planning methods, F-CON enables substantially better dense packing than other shape completion methods. Nikhil Mishra, Pieter Abbeel, Xi Chen 0022, Maximilian Sieb	IROS	2
2023	The Impact of Overall Optimization on Warehouse Automation abstract In this study, we propose a novel approach for investigating optimization performance by flexible robot co-ordination in automated warehouses with multi-agent rein-forcement learning (MARL)-based control. Automated systems using robots are expected to achieve efficient operations compared with manual systems in terms of overall optimization performance. However, the impact of overall optimization on performance remains unclear in most automated systems due to a lack of suitable control methods. Thus, we proposed a centralized training-and-decentralized execution MARL frame-work as a practical overall optimization control method. In the proposed framework, we also proposed a single shared critic, trained with global states and rewards, applicable to a case in which heterogeneous agents make decisions asynchronously. Our proposed MARL framework was applied to the task selection of material handling equipment through automated order picking simulation, and its performance was evaluated to determine how far overall optimization outperforms partial optimization by comparing it with other MARL frameworks and rule-based control methods. Hiroshi Yoshitake, Pieter Abbeel	IROS	2
2023	Learning Universal Policies via Text-Guided Video Generation abstract A goal of artificial intelligence is to construct an agent that can solve a wide variety of tasks. Recent progress in text-guided image synthesis has yielded models with an impressive ability to generate complex novel images, exhibiting combinatorial generalization across domains. Motivated by this success, we investigate whether such tools can be used to construct more general-purpose agents. Specifically, we cast the sequential decision making problem as a text-conditioned video generation problem, where, given a text-encoded specification of a desired goal, a planner synthesizes a set of future frames depicting its planned actions in the future, after which control actions are extracted from the generated video. By leveraging text as the underlying goal specification, we are able to naturally and combinatorially generalize to novel goals. The proposed policy-as-video formulation can further represent environments with different state and action spaces in a unified space of images, which, for example, enables learning and generalization across a variety of robot manipulation tasks. Finally, by leveraging pretrained language embeddings and widely available videos from the internet, the approach enables knowledge transfer through predicting highly realistic video plans for real robots. Yilun Du, Sherry Yang 0001, Bo Dai 0001, Hanjun Dai, Ofir Nachum, Josh Tenenbaum, Dale Schuurmans, Pieter Abbeel	NeurIPS	8
2023	Video Prediction Models as Rewards for Reinforcement Learning abstract Specifying reward signals that allow agents to learn complex behaviors is a long-standing challenge in reinforcement learning. A promising approach is to extract preferences for behaviors from unlabeled videos, which are widely available on the internet. We present Video Prediction Rewards (VIPER), an algorithm that leverages pretrained video prediction models as action-free reward signals for reinforcement learning. Specifically, we first train an autoregressive transformer on expert videos and then use the video prediction likelihoods as reward signals for a reinforcement learning agent. VIPER enables expert-level control without programmatic task rewards across a wide range of DMC, Atari, and RLBench tasks. Moreover, generalization of the video prediction model allows us to derive rewards for an out-of-distribution environment where no expert data is available, enabling cross-embodiment generalization for tabletop manipulation. We see our work as starting point for scalable reward specification from unlabeled videos that will benefit from the rapid advances in generative modeling. Source code and datasets are available on the project website: https://ViperRL.com Alejandro Escontrela, Ademi Adeniji, Wilson Yan, Ajay Jain, Xue Bin Peng, Kenneth Y. Goldberg, Youngwoon Lee, Danijar Hafner, Pieter Abbeel	NeurIPS	9
2023	Reinforcement Learning for Fine-tuning Text-to-Image Diffusion Models Olivia Watkins, Hao Liu 0055, Moonkyung Ryu, Craig Boutilier, Pieter Abbeel, Mohammad Ghavamzadeh, Kangwook Lee 0001, Kimin Lee	NeurIPS	7
2023	Accelerating Reinforcement Learning with Value-Conditional State Entropy Exploration abstract A promising technique for exploration is to maximize the entropy of visited state distribution, i.e., state entropy, by encouraging uniform coverage of visited state space. While it has been effective for an unsupervised setup, it tends to struggle in a supervised setup with a task reward, where an agent prefers to visit high-value states to exploit the task reward. Such a preference can cause an imbalance between the distributions of high-value states and low-value states, which biases exploration towards low-value state regions as a result of the state entropy increasing when the distribution becomes more uniform. This issue is exacerbated when high-value states are narrowly distributed within the state space, making it difficult for the agent to complete the tasks. In this paper, we present a novel exploration technique that maximizes the value-conditional state entropy, which separately estimates the state entropies that are conditioned on the value estimates of each state, then maximizes their average. By only considering the visited states with similar value estimates for computing the intrinsic bonus, our method prevents the distribution of low-value states from affecting exploration around high-value states, and vice versa. We demonstrate that the proposed alternative to the state entropy baseline significantly accelerates various reinforcement learning algorithms across a variety of tasks within MiniGrid, DeepMind Control Suite, and Meta-World benchmarks. Source code is available at https://sites.google.com/view/rl-vcse. Dongyoung Kim, Jinwoo Shin, Pieter Abbeel, Younggyo Seo	NeurIPS	3
2023	Blockwise Parallel Transformers for Large Context Models abstract Transformers have emerged as the cornerstone of state-of-the-art natural language processing models, showcasing exceptional performance across a wide range of AI applications. However, the memory demands posed by the self-attention mechanism and the large feedforward network in Transformers limit their ability to handle long sequences, thereby creating challenges for tasks involving multiple long sequences or long-term dependencies. We present a distinct approach, Blockwise Parallel Transformer (BPT), that leverages blockwise computation of self-attention and feedforward network fusion to minimize memory costs. By processing longer input sequences while maintaining memory efficiency, BPT enables training sequences 32 times longer than vanilla Transformers and up to 4 times longer than previous memory-efficient methods. Extensive experiments on language modeling and reinforcement learning tasks demonstrate the effectiveness of BPT in reducing memory requirements and improving performance. Hao Liu 0055, Pieter Abbeel	NeurIPS	2
2023	Language Quantized AutoEncoders: Towards Unsupervised Text-Image Alignment abstract Recent progress in scaling up large language models has shown impressive capabilities in performing few-shot learning across a wide range of natural language tasks. However, a key limitation is that these language models fundamentally lack grounding to visual perception - a crucial attribute needed to extend to real world tasks such as in visual-question answering and robotics. While prior works have largely connected image to text through pretraining or fine-tuning, learning such alignments are generally costly due to a combination of curating massive datasets and large computational burdens. In order to resolve these limitations, we propose a simple yet effective approach called Language-Quantized AutoEncoder (LQAE), a modification of VQ-VAE that learns to align text-image data in an unsupervised manner by leveraging pretrained language model denoisers (e.g., BERT). Our main idea is to encode images as sequences of text tokens by directly quantizing image embeddings using a pretrained language codebook. We then feed a masked version of the quantized embeddings into a BERT to reconstruct the original input. By doing so, LQAE learns to represent similar images with similar clusters of text tokens, thereby aligning these two modalities without the use of aligned text-image pairs. We show LQAE learns text-aligned image tokens that enable few-shot multi-modal learning with large language models, outperforming baseline methods in tasks such as image classification and VQA while requiring as few as 1-10 image-text pairs. Hao Liu 0055, Wilson Yan, Pieter Abbeel	NeurIPS	3
2023	Where are we in the search for an Artificial Visual Cortex for Embodied Intelligence? abstract We present the largest and most comprehensive empirical study of pre-trained visual representations (PVRs) or visual ‘foundation models’ for Embodied AI. First, we curate CortexBench, consisting of 17 different tasks spanning locomotion, navigation, dexterous, and mobile manipulation. Next, we systematically evaluate existing PVRs and find that none are universally dominant. To study the effect of pre-training data size and diversity, we combine over 4,000 hours of egocentric videos from 7 different sources (over 4.3M images) and ImageNet to train different-sized vision transformers using Masked Auto-Encoding (MAE) on slices of this data. Contrary to inferences from prior work, we find that scaling dataset size and diversity does not improve performance universally (but does so on average). Our largest model, named VC-1, outperforms all prior PVRs on average but does not universally dominate either. Next, we show that task- or domain-specific adaptation of VC-1 leads to substantial gains, with VC-1 (adapted) achieving competitive or superior performance than the best known results on all of the benchmarks in CortexBench. Finally, we present real-world hardware experiments, in which VC-1 and VC-1 (adapted) outperform the strongest pre-existing PVR. Overall, this paper presents no new techniques but a rigorous systematic evaluation, a broad set of findings about PVRs (that in some cases, refute those made in narrow domains in prior work), and open-sourced code and models (that required over 10,000 GPU-hours to train) for the benefit of the research community. Arjun Majumdar, Karmesh Yadav, Sergio Arnaud, Yecheng Jason Ma 0001, Sneha Silwal, Aryan Jain, Vincent-Pierre Berges, Tingfan Wu, Jay Vakil, Pieter Abbeel, Jitendra Malik, Dhruv Batra, Oleksandr Maksymets, Aravind Rajeswaran, Franziska Meier	NeurIPS	11
2023	AlberDICE: Addressing Out-Of-Distribution Joint Actions in Offline Multi-Agent RL via Alternating Stationary Distribution Correction Estimation abstract One of the main challenges in offline Reinforcement Learning (RL) is the distribution shift that arises from the learned policy deviating from the data collection policy. This is often addressed by avoiding out-of-distribution (OOD) actions during policy improvement as their presence can lead to substantial performance degradation. This challenge is amplified in the offline Multi-Agent RL (MARL) setting since the joint action space grows exponentially with the number of agents. To avoid this curse of dimensionality, existing MARL methods adopt either value decomposition methods or fully decentralized training of individual agents. However, even when combined with standard conservatism principles, these methods can still result in the selection of OOD joint actions in offline MARL. To this end, we introduce AlberDICE, an offline MARL algorithm that alternatively performs centralized training of individual agents based on stationary distribution optimization. AlberDICE circumvents the exponential complexity of MARL by computing the best response of one agent at a time while effectively avoiding OOD joint action selection. Theoretically, we show that the alternating optimization procedure converges to Nash policies. In the experiments, we demonstrate that AlberDICE significantly outperforms baseline algorithms on a standard suite of MARL benchmarks. Daiki E. Matsunaga, Jongmin Lee 0004, Jaeseok Yoon, Stefanos Leonardos, Pieter Abbeel, Kee-Eung Kim	NeurIPS	5
2023	Pretraining Graph Neural Networks for Few-Shot Analog Circuit Modeling and Design abstract Being able to predict the performance of circuits without running expensive simulations is a desired capability that can catalyze automated design. In this article, we present a supervised pretraining approach to learn circuit representations that can be adapted to new circuit topologies or unseen prediction tasks. We hypothesize that if we train a neural network (NN) that can predict the output direct current (dc) voltages of a wide range of circuit instances it will be forced to learn generalizable knowledge about the role of each circuit element and how they interact with each other. The dataset for this supervised learning objective can be easily collected at scale since the required dc simulation to get ground truth labels is relatively cheap. This representation would then be helpful for few-shot generalization to unseen circuit metrics that require more time-consuming simulations for obtaining the ground-truth labels. To cope with the variable topological structure of different circuits we describe each circuit as a graph and use graph NNs (GNNs) to learn node embeddings. We show that pretraining GNNs on prediction of output node voltages can encourage learning representations that can be adapted to new unseen topologies or prediction of new circuit-level properties with up to 10x more sample efficiency compared to a randomly initialized model. We further show that we can improve the sample efficiency of prior SoTA model-based optimization methods by$2\times $(almost as good as using an oracle model) via fintuning pretrained GNNs as the feature extractor of the learned models. Kourosh Hakhamaneshi, Marcel Nassar, Mariano Phielipp, Pieter Abbeel, Vladimir Stojanovic	IEEE Trans. Comput. Aided Des. Integr. Circuits Syst.	4
2022	Programmatic Modeling and Generation of Real-Time Strategic Soccer Environments for Reinforcement Learning abstract The capability of a reinforcement learning (RL) agent heavily depends on the diversity of the learning scenarios generated by the environment. Generation of diverse realistic scenarios is challenging for real-time strategy (RTS) environments. The RTS environments are characterized by intelligent entities/non-RL agents cooperating and competing with the RL agents with large state and action spaces over a long period of time, resulting in an infinite space of feasible, but not necessarily realistic, scenarios involving complex interaction among different RL and non-RL agents. Yet, most of the existing simulators rely on randomly generating the environments based on predefined settings/layouts and offer limited flexibility and control over the environment dynamics for researchers to generate diverse, realistic scenarios as per their demand. To address this issue, for the first time, we formally introduce the benefits of adopting an existing formal scenario specification language, SCENIC, to assist researchers to model and generate diverse scenarios in an RTS environment in a flexible, systematic, and programmatic manner. To showcase the benefits, we interfaced SCENIC to an existing RTS environment Google Research Football (GRF) simulator and introduced a benchmark consisting of 32 realistic scenarios, encoded in SCENIC, to train RL agents and testing their generalization capabilities. We also show how researchers/RL practitioners can incorporate their domain knowledge to expedite the training process by intuitively modeling stochastic programmatic policies with SCENIC. Abdus Salam Azad, Edward Kim 0005, Qiancheng Wu, Kimin Lee, Ion Stoica, Pieter Abbeel, Alberto L. Sangiovanni-Vincentelli, Sanjit A. Seshia	AAAI	6
2022	Frozen Pretrained Transformers as Universal Computation Engines abstract We investigate the capability of a transformer pretrained on natural language to generalize to other modalities with minimal finetuning -- in particular, without finetuning of the self-attention and feedforward layers of the residual blocks. We consider such a model, which we call a Frozen Pretrained Transformer (FPT), and study finetuning it on a variety of sequence classification tasks spanning numerical computation, vision, and protein fold prediction. In contrast to prior works which investigate finetuning on the same modality as the pretraining dataset, we show that pretraining on natural language can improve performance and compute efficiency on non-language downstream tasks. Additionally, we perform an analysis of the architecture, comparing the performance of a random initialized transformer to a random LSTM. Combining the two insights, we find language-pretrained transformers can obtain strong performance on a variety of non-language tasks. Aditya Grover, Pieter Abbeel, Igor Mordatch	AAAI	3
2022	Zero-Shot Text-Guided Object Generation with Dream Fields abstract We combine neural rendering with multi-modal image and text representations to synthesize diverse 3D objects solely from natural language descriptions. Our method, Dream Fields, can generate the geometry and color of a wide range of objects without 3D supervision. Due to the scarcity of diverse, captioned 3D data, prior methods only generate objectsfrom a handful of categories, such as ShapeNet. Instead, we guide generation with image-text models pre-trained on large datasets of captioned images from the web. Our method optimizes a Neural Radiance Field from many camera views so that rendered images score highly with a target caption according to a pre-trained CLIP model. To improve fidelity and visual quality, we introduce simple geometric priors, including sparsity-inducing transmittance regularization, scene bounds, and new MLP architectures. In experiments, Dream Fields produce realistic, multi-view consistent object geometry and color from a variety of natural language captions. Ajay Jain, Ben Mildenhall, Jonathan T. Barron, Pieter Abbeel, Ben Poole	CVPR	4
2022	Autoregressive Uncertainty Modeling for 3D Bounding Box Prediction Yuxuan Liu 0001, Nikhil Mishra, Maximilian Sieb, Yide Shentu, Pieter Abbeel, Xi Chen 0022	ECCV (10)	5
2022	Sim-to-Real 6D Object Pose Estimation via Iterative Self-training for Robotic Bin Picking Kai Chen 0028, Stephen James, Yichuan Li 0002, Yun-Hui Liu 0001, Pieter Abbeel, Qi Dou 0001	ECCV (39)	6
2022	HARP: Autoregressive Latent Video Prediction with High-Fidelity Image Generator abstract Video prediction is an important yet challenging problem; burdened with the tasks of generating future frames and learning environment dynamics. Recently, autoregressive latent video models have proved to be a powerful video prediction tool, by separating the video prediction into two sub-problems: pre-training an image generator model, followed by learning an autoregressive prediction model in the latent space of the image generator. However, successfully generating high-fidelity and high-resolution videos has yet to be seen. In this work, we investigate how to train an autoregressive latent video prediction model capable of predicting high-fidelity future frames with minimal modification to existing models, and produce high-resolution (256x256) videos. Specifically, we scale up prior models by employing a high-fidelity image generator (VQ-GAN) with a causal transformer model, and introduce additional techniques of top-k sampling and data augmentation to further improve video prediction quality. Despite the simplicity, the proposed method achieves competitive performance to state-of-the-art approaches on standard video prediction benchmarks with fewer parameters, and enables high-resolution video prediction on complex and large-scale datasets. Younggyo Seo, Kimin Lee, Fangchen Liu, Stephen James, Pieter Abbeel	ICIP	5
2022	It Takes Four to Tango: Multiagent Self Play for Automatic Curriculum Generation Pieter Abbeel, Aditya Grover	ICLR	2
2022	Hierarchical Few-Shot Imitation with Skill Transition Models Kourosh Hakhamaneshi, Ruihan Zhao 0001, Albert Zhan, Pieter Abbeel, Michael Laskin	ICLR	4
2022	Reward Uncertainty for Exploration in Preference-based Reinforcement Learning Xinran Liang, Katherine Shu, Kimin Lee, Pieter Abbeel	ICLR	4
2022	SURF: Semi-supervised Reward Learning with Data Augmentation for Feedback-efficient Preference-based Reinforcement Learning Jongjin Park, Younggyo Seo, Jinwoo Shin, Honglak Lee, Pieter Abbeel, Kimin Lee	ICLR	5
2022	Language Models as Zero-Shot Planners: Extracting Actionable Knowledge for Embodied Agents abstract Can world knowledge learned by large language models (LLMs) be used to act in interactive environments? In this paper, we investigate the possibility of grounding high-level tasks, expressed in natural language (e.g. “make breakfast”), to a chosen set of actionable steps (e.g. “open fridge”). While prior work focused on learning from explicit step-by-step examples of how to act, we surprisingly find that if pre-trained LMs are large enough and prompted appropriately, they can effectively decompose high-level tasks into mid-level plans without any further training. However, the plans produced naively by LLMs often cannot map precisely to admissible actions. We propose a procedure that conditions on existing demonstrations and semantically translates the plans to admissible actions. Our evaluation in the recent VirtualHome environment shows that the resulting method substantially improves executability over the LLM baseline. The conducted human evaluation reveals a trade-off between executability and correctness but shows a promising sign towards extracting actionable knowledge from language models. Wenlong Huang, Pieter Abbeel, Deepak Pathak, Igor Mordatch	ICML	2
2022	Reducing Variance in Temporal-Difference Value Estimation via Ensemble of Deep Networks abstract In temporal-difference reinforcement learning algorithms, variance in value estimation can cause instability and overestimation of the maximal target value. Many algorithms have been proposed to reduce overestimation, including several recent ensemble methods, however none have shown success in sample-efficient learning through addressing estimation variance as the root cause of overestimation. In this paper, we propose MeanQ, a simple ensemble method that estimates target values as ensemble means. Despite its simplicity, MeanQ shows remarkable sample efficiency in experiments on the Atari Learning Environment benchmark. Importantly, we find that an ensemble of size 5 sufficiently reduces estimation variance to obviate the lagging target network, eliminating it as a source of bias and further gaining sample efficiency. We justify intuitively and empirically the design choices in MeanQ, including the necessity of independent experience sampling. On a set of 26 benchmark Atari environments, MeanQ outperforms all tested baselines, including the best available baseline, SUNRISE, at 100K interaction steps in 16/26 environments, and by 68% on average. MeanQ also outperforms Rainbow DQN at 500K steps in 21/26 environments, and by 49% on average, and achieves average human-level performance using 200K ($\pm$100K) interaction steps. Our implementation is available at https://github.com/indylab/MeanQ. Litian Liang, Yaosheng Xu, Stephen McAleer, Dailin Hu, Alexander Ihler, Pieter Abbeel, Roy Fox	ICML	6
2022	Reinforcement Learning with Action-Free Pre-Training from Videos abstract Recent unsupervised pre-training methods have shown to be effective on language and vision domains by learning useful representations for multiple downstream tasks. In this paper, we investigate if such unsupervised pre-training methods can also be effective for vision-based reinforcement learning (RL). To this end, we introduce a framework that learns representations useful for understanding the dynamics via generative pre-training on videos. Our framework consists of two phases: we pre-train an action-free latent video prediction model, and then utilize the pre-trained representations for efficiently learning action-conditional world models on unseen environments. To incorporate additional action inputs during fine-tuning, we introduce a new architecture that stacks an action-conditional latent prediction model on top of the pre-trained action-free prediction model. Moreover, for better exploration, we propose a video-based intrinsic bonus that leverages pre-trained representations. We demonstrate that our framework significantly improves both final performances and sample-efficiency of vision-based RL in a variety of manipulation and locomotion tasks. Code is available at \url{https://github.com/younggyoseo/apv}. Younggyo Seo, Kimin Lee, Stephen James, Pieter Abbeel	ICML	4
2022	Towards More Generalizable One-shot Visual Imitation Learning abstract A general-purpose robot should be able to master a wide range of tasks and quickly learn a novel one by leveraging past experiences. One-shot imitation learning (OSIL) approaches this goal by training an agent with (pairs of) expert demonstrations, such that at test time, it can directly execute a new task from just one demonstration. However, so far this framework has been limited to training on many variations of one task, and testing on other unseen but similar variations of the same task. In this work, we push for a higher level of generalization ability by investigating a more ambitious multi-task setup. We introduce a diverse suite of vision-based robot manipulation tasks, consisting of 7 tasks, a total of 61 variations, and a continuum of instances within each variation. For consistency and comparison purposes, we first train and evaluate single-task agents (as done in prior few-shot imitation work). We then study the multi-task setting, where multi-task training is followed by (i) one-shot imitation on variations within the training tasks, (ii) one-shot imitation on new tasks, and (iii) fine-tuning on new tasks. Prior state-of-the-art, while performing well within some single tasks, struggles in these harder multi-task settings. To address these limitations, we propose MOSAIC (Multi-task One-Shot Imitation with self-Attention and Contrastive learning), which integrates a self-attention model architecture and a temporal contrastive module to enable better task disambiguation and more robust representation learning. Our experiments show that MOSAIC outperforms prior state of the art in learning efficiency, final performance, and learns a multi-task policy with promising generalization ability via fine-tuning on novel tasks. Mandi Zhao, Fangchen Liu, Kimin Lee, Pieter Abbeel	ICRA	4
2022	Adversarial Motion Priors Make Good Substitutes for Complex Reward Functions abstract Training a high-dimensional simulated agent with an under-specified reward function often leads the agent to learn physically infeasible strategies that are ineffective when deployed in the real world. To mitigate these unnatural behaviors, reinforcement learning practitioners often utilize complex reward functions that encourage physically plausible behaviors. However, a tedious labor-intensive tuning process is often required to create hand-designed rewards which might not easily generalize across platforms and tasks. We propose substituting complex reward functions with “style rewards” learned from a dataset of motion capture demonstrations. A learned style reward can be combined with an arbitrary task reward to train policies that perform tasks using naturalistic strategies. These natural strategies can also facilitate transfer to the real world. We build upon Adversarial Motion Priors - an approach from the computer graphics domain that encodes a style reward from a dataset of reference motions - to demonstrate that an adversarial approach to training policies can produce behaviors that transfer to a real quadrupedal robot without requiring complex reward functions. We also demonstrate that an effective style reward can be learned from a few seconds of motion capture data gathered from a German Shepherd and leads to energy-efficient locomotion strategies with natural gait transitions. Alejandro Escontrela, Xue Bin Peng, Wenhao Yu 0003, Tingnan Zhang, Atil Iscen, Kenneth Y. Goldberg, Pieter Abbeel	IROS	7
2022	Multi-Objective Policy Gradients with Topological Constraints abstract Multi-objective optimization models that encode ordered sequential constraints provide a solution to model various challenging problems including encoding preferences, modeling a curriculum, and enforcing measures of safety. A recently developed theory of topological Markov decision processes (TMDPs) captures this range of problems for the case of discrete states and actions. In this work, we extend TMDPs towards continuous spaces and unknown transition dynamics by formulating, proving, and implementing the policy gradient theorem for TMDPs. This theoretical result enables the creation of TMDP learning algorithms that use function approximators, and can generalize existing deep reinforcement learning (DRL) approaches. Specifically, we present a new algorithm for a policy gradient in TMDPs by a simple extension of the proximal policy optimization (PPO) algorithm. We demonstrate this on a real-world multiple-objective navigation problem with an arbitrary ordering of objectives both in simulation and on a real robot. Kyle Hollins Wray, Stas Tiomkin, Mykel J. Kochenderfer, Pieter Abbeel	IROS	4
2022	Playful Interactions for Representation Learning abstract Ahstract- One of the key challenges in visual imitation learning is collecting large amounts of expert demonstrations for a given task. While methods for collecting human demonstrations are becoming easier with teleoperation methods and the use of low-cost assistive tools, we often still require 100–1000 demonstrations for every task to learn a visual representation and policy. To address this, we turn to an alternate form of data that does not require task-specific demonstrations - play. Playing is a fundamental method children use to learn a set of skills and behaviors and visual representations in early learning. Importantly, play data is diverse, task-agnostic, and relatively cheap to obtain. In this work, we propose to use playful interactions in a self-supervised manner to learn visual representations for downstream tasks. We collect 2 hours of playful data in 19 diverse environments and use self-predictive learning to extract visual representations. Given these representations, we train policies using imitation learning for two downstream tasks: Pushing and Stacking. We demonstrate that our visual representations generalize better than standard behavior cloning and can achieve similar performance with only half the number of required demonstrations. Our representations, which are trained from scratch, compare favorably against ImageNet pretrained representations. Finally, we provide an experimental analysis on the effects of different pretraining modes on downstream task learning. Sarah Young, Jyothish Pari, Pieter Abbeel, Lerrel Pinto	IROS	3
2022	Learning Visual Robotic Control Efficiently with Contrastive Pre-training and Data Augmentation abstract Recent advances in unsupervised representation learning significantly improved the sample efficiency of training Reinforcement Learning policies in simulated environments. However, similar gains have not yet been seen for real-robot reinforcement learning. In this work, we focus on enabling data-efficient real-robot learning from pixels. We present Contrastive Pre-training and Data Augmentation for Efficient Robotic Learning (CoDER), a method that utilizes data augmentation and unsupervised learning to achieve sample-efficient training of real-robot arm policies from sparse rewards. While contrastive pre-training, data augmentation, demonstrations, and reinforcement learning are alone insufficient for efficient learning, our main contribution is showing that the combination of these disparate techniques results in a simple yet data-efficient method. We show that, given only 10 demonstrations, a single robotic arm can learn sparse-reward manipulation policies from pixels, such as reaching, picking, moving, pulling a large object, flipping a switch, and opening a drawer in just 30 minutes of mean real-world training time. We include videos and code on the project website: https://sites.google.com/view/efficient-robotic-manipulation/home. Albert Zhan, Ruihan Zhao 0001, Lerrel Pinto, Pieter Abbeel, Michael Laskin	IROS	4
2022	Deep Hierarchical Planning from Pixels abstract Intelligent agents need to select long sequences of actions to solve complex tasks. While humans easily break down tasks into subgoals and reach them through millions of muscle commands, current artificial intelligence is limited to tasks with horizons of a few hundred decisions, despite large compute budgets. Research on hierarchical reinforcement learning aims to overcome this limitation but has proven to be challenging, current methods rely on manually specified goal spaces or subtasks, and no general solution exists. We introduce Director, a practical method for learning hierarchical behaviors directly from pixels by planning inside the latent space of a learned world model. The high-level policy maximizes task and exploration rewards by selecting latent goals and the low-level policy learns to achieve the goals. Despite operating in latent space, the decisions are interpretable because the world model can decode goals into images for visualization. Director learns successful behaviors across a wide range of environments, including visual control, Atari games, and DMLab levels and outperforms exploration methods on tasks with very sparse rewards, including 3D maze traversal with a quadruped robot from an egocentric camera and proprioception, without access to the global position or top-down view used by prior work. Danijar Hafner, Kuang-Huei Lee, Ian Fischer, Pieter Abbeel	NeurIPS	4
2022	Unsupervised Reinforcement Learning with Contrastive Intrinsic Control abstract We introduce Contrastive Intrinsic Control (CIC), an unsupervised reinforcement learning (RL) algorithm that maximizes the mutual information between state-transitions and latent skill vectors. CIC utilizes contrastive learning between state-transitions and skills vectors to learn behaviour embeddings and maximizes the entropy of these embeddings as an intrinsic reward to encourage behavioural diversity. We evaluate our algorithm on the Unsupervised RL Benchmark (URLB) in the asymptotic state-based setting, which consists of a long reward-free pre-training phase followed by a short adaptation phase to downstream tasks with extrinsic rewards. We find that CIC improves over prior exploration algorithms in terms of adaptation efficiency to downstream tasks on state-based URLB. Michael Laskin, Hao Liu 0055, Xue Bin Peng, Denis Yarats, Aravind Rajeswaran, Pieter Abbeel	NeurIPS	6
2022	Masked Autoencoding for Scalable and Generalizable Decision Making abstract We are interested in learning scalable agents for reinforcement learning that can learn from large-scale, diverse sequential data similar to current large vision and language models. To this end, this paper presents masked decision prediction (MaskDP), a simple and scalable self-supervised pretraining method for reinforcement learning (RL) and behavioral cloning (BC). In our MaskDP approach, we employ a masked autoencoder (MAE) to state-action trajectories, wherein we randomly mask state and action tokens and reconstruct the missing data. By doing so, the model is required to infer masked out states and actions and extract information about dynamics. We find that masking different proportions of the input sequence significantly helps with learning a better model that generalizes well to multiple downstream tasks. In our empirical study we ﬁnd that a MaskDP model gains the capability of zero-shot transfer to new BC tasks, such as single and multiple goal reaching, and it can zero-shot infer skills from a few example transitions. In addition, MaskDP transfers well to offline RL and shows promising scaling behavior w.r.t. to model size. It is amenable to data efficient finetuning, achieving competitive results with prior methods based on autoregressive pretraining. Fangchen Liu, Hao Liu 0055, Aditya Grover, Pieter Abbeel	NeurIPS	4
2022	Chain of Thought Imitation with Procedure Cloning abstract Imitation learning aims to extract high-performance policies from logged demonstrations of expert behavior. It is common to frame imitation learning as a supervised learning problem in which one fits a function approximator to the input-output mapping exhibited by the logged demonstrations (input observations to output actions). While the framing of imitation learning as a supervised input-output learning problem allows for applicability in a wide variety of settings, it is also an overly simplistic view of the problem in situations where the expert demonstrations provide much richer insight into expert behavior. For example, applications such as path navigation, robot manipulation, and strategy games acquire expert demonstrations via planning, search, or some other multi-step algorithm, revealing not just the output action to be imitated but also the procedure for how to determine this action. While these intermediate computations may use tools not available to the agent during inference (e.g., environment simulators), they are nevertheless informative as a way to explain an expert’s mapping of state to actions. To properly leverage expert procedure information without relying on the privileged tools the expert may have used to perform the procedure, we propose procedure cloning, which applies supervised sequence prediction to imitate the complete series of expert computations. This way, procedure cloning learns not only what to do (i.e., the output action), but how and why to do it (i.e., the procedure). Through empirical analysis on navigation, simulated robotic manipulation, and game-playing environments, we show that imitating the intermediate computations of an expert’s behavior enables procedure cloning to learn policies exhibiting significant generalization to unseen environment configurations, including those configurations for which running the expert’s procedure directly is infeasible. Sherry Yang 0001, Dale Schuurmans, Pieter Abbeel, Ofir Nachum	NeurIPS	3
2022	Spending Thinking Time Wisely: Accelerating MCTS with Virtual Expansions abstract One of the most important AI research questions is to trade off computation versus performance since ``perfect rationality" exists in theory but is impossible to achieve in practice. Recently, Monte-Carlo tree search (MCTS) has attracted considerable attention due to the significant performance improvement in various challenging domains. However, the expensive time cost during search severely restricts its scope for applications. This paper proposes the Virtual MCTS (V-MCTS), a variant of MCTS that spends more search time on harder states and less search time on simpler states adaptively. We give theoretical bounds of the proposed method and evaluate the performance and computations on $9 \times 9$ Go board games and Atari games. Experiments show that our method can achieve comparable performances to the original search algorithm while requiring less than $50\%$ search time on average. We believe that this approach is a viable alternative for tasks under limited time and resources. The code is available at \url{https://github.com/YeWR/V-MCTS.git}. Weirui Ye, Pieter Abbeel, Yang Gao 0029	NeurIPS	2
2022	On the Effectiveness of Fine-tuning Versus Meta-reinforcement Learning abstract Intelligent agents should have the ability to leverage knowledge from previously learned tasks in order to learn new ones quickly and efficiently. Meta-learning approaches have emerged as a popular solution to achieve this. However, meta-reinforcement learning (meta-RL) algorithms have thus far been restricted to simple environments with narrow task distributions and have seen limited success. Moreover, the paradigm of pretraining followed by fine-tuning to adapt to new tasks has emerged as a simple yet effective solution in supervised learning. This calls into question the benefits of meta learning approaches also in reinforcement learning, which typically come at the cost of high complexity. We therefore investigate meta-RL approaches in a variety of vision-based benchmarks, including Procgen, RLBench, and Atari, where evaluations are made on completely novel tasks. Our findings show that when meta-learning approaches are evaluated on different tasks (rather than different variations of the same task), multi-task pretraining with fine-tuning on new tasks performs equally as well, or better, than meta-pretraining with meta test-time adaptation. This is encouraging for future research, as multi-task pretraining tends to be simpler and computationally cheaper than meta-RL. From these findings, we advocate for evaluating future meta-RL methods on more challenging tasks and including multi-task pretraining with fine-tuning as a simple, yet strong baseline. Mandi Zhao, Pieter Abbeel, Stephen James	NeurIPS	2
2022	AdaCat: Adaptive categorical discretization for autoregressive models abstract Autoregressive generative models can estimate complex continuous data distributions, like trajectory rollouts in an RL environment, image intensities, and audio. Most state-of-the-art models discretize continuous data into several bins and use categorical distributions over the bins to approximate the continuous data distribution. The advantage is that the categorical distribution can easily express multiple modes and are straightforward to optimize. However, such approximation cannot express sharp changes in density without using significantly more bins, which makes it parameter inefficient. We propose an efficient, expressive, multimodal parameterization called Adaptive Categorical Discretization (AdaCat). AdaCat discretizes each dimension of an autoregressive model adaptively, which allows the model to allocate density to fine intervals of interest, improving parameter efficiency. AdaCat generalizes both categoricals and quantile-based regression. AdaCat is a simple add-on to any discretization-based distribution estimator. In experiments, AdaCat improves density estimation for real-world tabular data, images, audio, and trajectories, and improves planning in model-based offline RL. Qiyang Li, Ajay Jain, Pieter Abbeel	UAI	3
2021	Bottleneck Transformers for Visual Recognition abstract We present BoTNet, a conceptually simple yet powerful backbone architecture that incorporates self-attention for multiple computer vision tasks including image classification, object detection and instance segmentation. By just replacing the spatial convolutions with global self-attention in the final three bottleneck blocks of a ResNet and no other changes, our approach improves upon the baselines significantly on instance segmentation and object detection while also reducing the parameters, with minimal overhead in latency. Through the design of BoTNet, we also point out how ResNet bottleneck blocks with self-attention can be viewed as Transformer blocks. Without any bells and whistles, BoTNet achieves 44.4% Mask AP and 49.7% Box AP on the COCO Instance Segmentation benchmark using the Mask R-CNN framework; surpassing the previous best published single model and single scale results of ResNeSt [67] evaluated on the COCO validation set. Finally, we present a simple adaptation of the BoTNet design for image classification, resulting in models that achieve a strong performance of 84.7% top-1 accuracy on the ImageNet benchmark while being up to 1.64x faster in "compute"1time than the popular EfficientNet models on TPU-v3 hardware. We hope our simple and effective approach will serve as a strong baseline for future research in self-attention models for vision.2 Aravind Srinivas, Tsung-Yi Lin, Niki Parmar, Jonathon Shlens, Pieter Abbeel, Ashish Vaswani	CVPR	5
2021	Contrastive Code Representation Learning abstract Recent work learns contextual representations of source code by reconstructing tokens from their context.For downstream semantic understanding tasks like code clone detection, these representations should ideally capture program functionality.However, we show that the popular reconstruction-based RoBERTa model is sensitive to source code edits, even when the edits preserve semantics.We propose Con-traCode: a contrastive pre-training task that learns code functionality, not form.Con-traCode pre-trains a neural network to identify functionally similar variants of a program among many non-equivalent distractors.We scalably generate these variants using an automated source-to-source compiler as a form of data augmentation.Contrastive pretraining outperforms RoBERTa on an adversarial code clone detection benchmark by 39% AUROC.Surprisingly, improved adversarial robustness translates to better accuracy over natural code; ContraCode improves summarization and TypeScript type inference accuracy by 2 to 13 percentage points over competitive baselines.All source is available at https://github.com/parasj/contracode. Paras Jain 0001, Ajay Jain, Tianjun Zhang, Pieter Abbeel, Joseph Gonzalez 0001, Ion Stoica	EMNLP (1)	4
2021	Putting NeRF on a Diet: Semantically Consistent Few-Shot View Synthesis abstract We present DietNeRF, a 3D neural scene representation estimated from a few images. Neural Radiance Fields (NeRF) learn a continuous volumetric representation of a scene through multi-view consistency, and can be rendered from novel viewpoints by ray casting. While NeRF has an impressive ability to reconstruct geometry and fine details given many images, up to 100 for challenging 360° scenes, it often finds a degenerate solution to its image reconstruction objective when only a few input views are available. To improve few-shot quality, we propose DietNeRF. We introduce an auxiliary semantic consistency loss that encourages realistic renderings at novel poses. DietNeRF is trained on individual scenes to (1) correctly render given input views from the same pose, and (2) match high-level semantic attributes across different, random poses. Our semantic loss allows us to supervise DietNeRF from arbitrary poses. We extract these semantics using a pre-trained visual encoder such as CLIP, a Vision Transformer trained on hundreds of millions of diverse single-view, 2D photographs mined from the web with natural language supervision. In experiments, DietNeRF improves the perceptual quality of few-shot view synthesis when learned from scratch, can render novel views with as few as one observed image when pre-trained on a multi-view dataset, and produces plausible completions of completely unobserved regions. Our project website is available at https://www.ajayj.com/dietnerf. Ajay Jain, Matthew Tancik, Pieter Abbeel	ICCV	3
2021	Efficient Empowerment Estimation for Unsupervised Stabilization Ruihan Zhao 0001, Pieter Abbeel, Stas Tiomkin	ICLR	3
2021	Self-Supervised Policy Adaptation during Deployment Nicklas Hansen 0001, Rishabh Jangir, Yu Sun 0020, Guillem Alenyà, Pieter Abbeel, Alexei A. Efros, Lerrel Pinto, Xiaolong Wang 0004	ICLR	5
2021	Task-Agnostic Morphology Evolution Joey Hejna, Pieter Abbeel, Lerrel Pinto	ICLR	2
2021	Learning What To Do by Simulating the Past David Lindner, Rohin Shah, Pieter Abbeel, Anca D. Dragan	ICLR	3
2021	Reset-Free Lifelong Learning with Skill-Space Planning Aditya Grover, Pieter Abbeel, Igor Mordatch	ICLR	3
2021	Mutual Information State Intrinsic Control Rui Zhao 0011, Yang Gao 0029, Pieter Abbeel, Volker Tresp	ICLR	3
2021	Unsupervised Learning of Visual 3D Keypoints for Control abstract Learning sensorimotor control policies from high-dimensional images crucially relies on the quality of the underlying visual representations. Prior works show that structured latent space such as visual keypoints often outperforms unstructured representations for robotic control. However, most of these representations, whether structured or unstructured are learned in a 2D space even though the control tasks are usually performed in a 3D environment. In this work, we propose a framework to learn such a 3D geometric structure directly from images in an end-to-end unsupervised manner. The input images are embedded into latent 3D keypoints via a differentiable encoder which is trained to optimize both a multi-view consistency loss and downstream task objective. These discovered 3D keypoints tend to meaningfully capture robot joints as well as object movements in a consistent manner across both time and 3D space. The proposed approach outperforms prior state-of-art methods across a variety of reinforcement learning benchmarks. Code and videos at https://buoyancy99.github.io/unsup-3d-keypoints/. Boyuan Chen 0003, Pieter Abbeel, Deepak Pathak	ICML	2
2021	SUNRISE: A Simple Unified Framework for Ensemble Learning in Deep Reinforcement Learning abstract Off-policy deep reinforcement learning (RL) has been successful in a range of challenging domains. However, standard off-policy RL algorithms can suffer from several issues, such as instability in Q-learning and balancing exploration and exploitation. To mitigate these issues, we present SUNRISE, a simple unified ensemble method, which is compatible with various off-policy RL algorithms. SUNRISE integrates two key ingredients: (a) ensemble-based weighted Bellman backups, which re-weight target Q-values based on uncertainty estimates from a Q-ensemble, and (b) an inference method that selects actions using the highest upper-confidence bounds for efficient exploration. By enforcing the diversity between agents using Bootstrap with random initialization, we show that these different ideas are largely orthogonal and can be fruitfully integrated, together further improving the performance of existing off-policy RL algorithms, such as Soft Actor-Critic and Rainbow DQN, for both continuous and discrete control tasks on both low-dimensional and high-dimensional environments. Kimin Lee, Michael Laskin, Aravind Srinivas, Pieter Abbeel	ICML	4
2021	PEBBLE: Feedback-Efficient Interactive Reinforcement Learning via Relabeling Experience and Unsupervised Pre-training abstract Conveying complex objectives to reinforcement learning (RL) agents can often be difficult, involving meticulous design of reward functions that are sufficiently informative yet easy enough to provide. Human-in-the-loop RL methods allow practitioners to instead interactively teach agents through tailored feedback; however, such approaches have been challenging to scale since human feedback is very expensive. In this work, we aim to make this process more sample- and feedback-efficient. We present an off-policy, interactive RL algorithm that capitalizes on the strengths of both feedback and off-policy learning. Specifically, we learn a reward model by actively querying a teacher’s preferences between two clips of behavior and use it to train an agent. To enable off-policy learning, we relabel all the agent’s past experience when its reward model changes. We additionally show that pre-training our agents with unsupervised exploration substantially increases the mileage of its queries. We demonstrate that our approach is capable of learning tasks of higher complexity than previously considered by human-in-the-loop methods, including a variety of locomotion and robotic manipulation skills. We also show that our method is able to utilize real-time human feedback to effectively prevent reward exploitation and learn new behaviors that are difficult to specify with standard reward functions. Kimin Lee, Laura M. Smith, Pieter Abbeel	ICML	3
2021	APS: Active Pretraining with Successor Features abstract We introduce a new unsupervised pretraining objective for reinforcement learning. During the unsupervised reward-free pretraining phase, the agent maximizes mutual information between tasks and states induced by the policy. Our key contribution is a novel lower bound of this intractable quantity. We show that by reinterpreting and combining variational successor features \citep{Hansen2020Fast} with nonparametric entropy maximization \citep{liu2021behavior}, the intractable mutual information can be efficiently optimized. The proposed method Active Pretraining with Successor Feature (APS) explores the environment via nonparametric entropy maximization, and the explored data can be efficiently leveraged to learn behavior by variational successor features. APS addresses the limitations of existing mutual information maximization based and entropy maximization based unsupervised RL, and combines the best of both worlds. When evaluated on the Atari 100k data-efficiency benchmark, our approach significantly outperforms previous methods combining unsupervised pretraining with task-specific finetuning. Hao Liu 0055, Pieter Abbeel	ICML	2
2021	MSA Transformer abstract Unsupervised protein language models trained across millions of diverse sequences learn structure and function of proteins. Protein language models studied to date have been trained to perform inference from individual sequences. The longstanding approach in computational biology has been to make inferences from a family of evolutionarily related sequences by fitting a model to each family independently. In this work we combine the two paradigms. We introduce a protein language model which takes as input a set of sequences in the form of a multiple sequence alignment. The model interleaves row and column attention across the input sequences and is trained with a variant of the masked language modeling objective across many protein families. The performance of the model surpasses current state-of-the-art unsupervised structure learning methods by a wide margin, with far greater parameter efficiency than prior state-of-the-art protein language models. Roshan Rao, Robert Verkuil, Joshua Meier, John F. Canny, Pieter Abbeel, Tom Sercu, Alexander Rives	ICML	6
2021	State Entropy Maximization with Random Encoders for Efficient Exploration abstract Recent exploration methods have proven to be a recipe for improving sample-efficiency in deep reinforcement learning (RL). However, efficient exploration in high-dimensional observation spaces still remains a challenge. This paper presents Random Encoders for Efficient Exploration (RE3), an exploration method that utilizes state entropy as an intrinsic reward. In order to estimate state entropy in environments with high-dimensional observations, we utilize a k-nearest neighbor entropy estimator in the low-dimensional representation space of a convolutional encoder. In particular, we find that the state entropy can be estimated in a stable and compute-efficient manner by utilizing a randomly initialized encoder, which is fixed throughout training. Our experiments show that RE3 significantly improves the sample-efficiency of both model-free and model-based RL methods on locomotion and navigation tasks from DeepMind Control Suite and MiniGrid benchmarks. We also show that RE3 allows learning diverse behaviors without extrinsic rewards, effectively improving sample-efficiency in downstream tasks. Younggyo Seo, Jinwoo Shin, Honglak Lee, Pieter Abbeel, Kimin Lee	ICML	5
2021	Decoupling Representation Learning from Reinforcement Learning abstract In an effort to overcome limitations of reward-driven feature learning in deep reinforcement learning (RL) from images, we propose decoupling representation learning from policy learning. To this end, we introduce a new unsupervised learning (UL) task, called Augmented Temporal Contrast (ATC), which trains a convolutional encoder to associate pairs of observations separated by a short time difference, under image augmentations and using a contrastive loss. In online RL experiments, we show that training the encoder exclusively using ATC matches or outperforms end-to-end RL in most environments. Additionally, we benchmark several leading UL algorithms by pre-training encoders on expert demonstrations and using them, with weights frozen, in RL agents; we find that agents using ATC-trained encoders outperform all others. We also train multi-task encoders on data from multiple environments and show generalization to different downstream RL tasks. Finally, we ablate components of ATC, and introduce a new data augmentation to enable replay of (compressed) latent images from pre-trained encoders when RL requires augmentation. Our experiments span visually diverse RL benchmarks in DeepMind Control, DeepMind Lab, and Atari, and our complete code is available at \url{https://github.com/astooke/rlpyt/tree/master/rlpyt/ul}. Adam Stooke, Kimin Lee, Pieter Abbeel, Michael Laskin	ICML	3
2021	Auto-Tuned Sim-to-Real Transfer abstract Policies trained in simulation often fail when transferred to the real world due to the ‘reality gap’ where the simulator is unable to accurately capture the dynamics and visual properties of the real world. Current approaches to tackle this problem, such as domain randomization, require prior knowledge and engineering to determine how much to randomize system parameters in order to learn a policy that is robust to sim-to-real transfer while also not being too conservative. We propose a method for automatically tuning simulator system parameters to match the real world using only raw RGB images of the real world without the need to define rewards or estimate state. Our key insight is to reframe the auto-tuning of parameters as a search problem where we iteratively shift the simulation system parameters to approach the real world system parameters. We propose a Search Param Model (SPM) that, given a sequence of observations and actions and a set of system parameters, predicts whether the given parameters are higher or lower than the true parameters used to generate the observations. We evaluate our method on multiple robotic control tasks in both sim-to-sim and sim-to-real transfer, demonstrating significant improvement over naive domain randomization. Project videos at https://yuqingd.github.io/autotuned-sim2real/. Olivia Watkins, Trevor Darrell, Pieter Abbeel, Deepak Pathak	ICRA	4
2021	Reinforcement Learning for Robust Parameterized Locomotion Control of Bipedal Robots abstract Developing robust walking controllers for bipedal robots is a challenging endeavor. Traditional model-based locomotion controllers require simplifying assumptions and careful modelling; any small errors can result in unstable control. To address these challenges for bipedal locomotion, we present a model-free reinforcement learning framework for training robust locomotion policies in simulation, which can then be transferred to a real bipedal Cassie robot. To facilitate sim-to-real transfer, domain randomization is used to encourage the policies to learn behaviors that are robust across variations in system dynamics. The learned policies enable Cassie to perform a set of diverse and dynamic behaviors, while also being more robust than traditional controllers and prior learning-based methods that use residual control. We demonstrate this on versatile walking behaviors such as tracking a target walking velocity, walking height, and turning yaw. (Video1) Zhongyu Li 0003, Xuxin Cheng, Xue Bin Peng, Pieter Abbeel, Sergey Levine, Glen Berseth, Koushil Sreenath	ICRA	4
2021	Decision Transformer: Reinforcement Learning via Sequence Modeling abstract We introduce a framework that abstracts Reinforcement Learning (RL) as a sequence modeling problem. This allows us to draw upon the simplicity and scalability of the Transformer architecture, and associated advances in language modeling such as GPT-x and BERT. In particular, we present Decision Transformer, an architecture that casts the problem of RL as conditional sequence modeling. Unlike prior approaches to RL that fit value functions or compute policy gradients, Decision Transformer simply outputs the optimal actions by leveraging a causally masked Transformer. By conditioning an autoregressive model on the desired return (reward), past states, and actions, our Decision Transformer model can generate future actions that achieve the desired return. Despite its simplicity, Decision Transformer matches or exceeds the performance of state-of-the-art model-free offline RL baselines on Atari, OpenAI Gym, and Key-to-Door tasks. Aravind Rajeswaran, Kimin Lee, Aditya Grover, Michael Laskin, Pieter Abbeel, Aravind Srinivas, Igor Mordatch	NeurIPS	7
2021	Improving Computational Efficiency in Visual Reinforcement Learning via Stored Embeddings abstract Recent advances in off-policy deep reinforcement learning (RL) have led to impressive success in complex tasks from visual observations. Experience replay improves sample-efficiency by reusing experiences from the past, and convolutional neural networks (CNNs) process high-dimensional inputs effectively. However, such techniques demand high memory and computational bandwidth. In this paper, we present Stored Embeddings for Efficient Reinforcement Learning (SEER), a simple modification of existing off-policy RL methods, to address these computational and memory requirements. To reduce the computational overhead of gradient updates in CNNs, we freeze the lower layers of CNN encoders early in training due to early convergence of their parameters. Additionally, we reduce memory requirements by storing the low-dimensional latent vectors for experience replay instead of high-dimensional images, enabling an adaptive increase in the replay buffer capacity, a useful technique in constrained-memory settings. In our experiments, we show that SEER does not degrade the performance of RL agents while significantly saving computation and memory across a diverse set of DeepMind Control environments and Atari games. Kimin Lee, Aravind Srinivas, Pieter Abbeel	NeurIPS	4
2021	Behavior From the Void: Unsupervised Active Pre-Training abstract We introduce a new unsupervised pre-training method for reinforcement learning called APT, which stands for Active Pre-Training. APT learns behaviors and representations by actively searching for novel states in reward-free environments. The key novel idea is to explore the environment by maximizing a non-parametric entropy computed in an abstract representation space, which avoids challenging density modeling and consequently allows our approach to scale much better in environments that have high-dimensional observations (e.g., image observations). We empirically evaluate APT by exposing task-specific reward after a long unsupervised pre-training phase. In Atari games, APT achieves human-level performance on 12 games and obtains highly competitive performance compared to canonical fully supervised RL algorithms. On DMControl suite, APT beats all baselines in terms of asymptotic performance and data efficiency and dramatically improves performance on tasks that are extremely difficult to train from scratch. Hao Liu 0055, Pieter Abbeel	NeurIPS	2
2021	Hindsight Task Relabelling: Experience Replay for Sparse Reward Meta-RL abstract Meta-reinforcement learning (meta-RL) has proven to be a successful framework for leveraging experience from prior tasks to rapidly learn new related tasks, however, current meta-RL approaches struggle to learn in sparse reward environments. Although existing meta-RL algorithms can learn strategies for adapting to new sparse reward tasks, the actual adaptation strategies are learned using hand-shaped reward functions, or require simple environments where random exploration is sufficient to encounter sparse reward. In this paper we present a formulation of hindsight relabelling for meta-RL, which relabels experience during meta-training to enable learning to learn entirely using sparse reward. We demonstrate the effectiveness of our approach on a suite of challenging sparse reward environments that previously required dense reward during meta-training to solve. Our approach solves these environments using the true sparse reward function, with performance comparable to training with a proxy dense reward function. Charles Packer, Pieter Abbeel, Joseph Gonzalez 0001	NeurIPS	2
2021	Reinforcement Learning with Latent Flow abstract Temporal information is essential to learning effective policies with Reinforcement Learning (RL). However, current state-of-the-art RL algorithms either assume that such information is given as part of the state space or, when learning from pixels, use the simple heuristic of frame-stacking to implicitly capture temporal information present in the image observations. This heuristic is in contrast to the current paradigm in video classification architectures, which utilize explicit encodings of temporal information through methods such as optical flow and two-stream architectures to achieve state-of-the-art performance. Inspired by leading video classification architectures, we introduce the Flow of Latents for Reinforcement Learning (Flare), a network architecture for RL that explicitly encodes temporal information through latent vector differences. We show that Flare recovers optimal performance in state-based RL without explicit access to the state velocity, solely with positional state information. Flare is the most sample efficient model-free pixel-based RL algorithm on the DeepMind Control suite when evaluated on the 500k and 1M step benchmarks across 5 challenging control tasks, and, when used with Rainbow DQN, outperforms the competitive baseline on Atari games at 100M time step benchmark across 8 challenging games. Wenling Shang, Xiaofei Wang 0008, Aravind Srinivas, Aravind Rajeswaran, Yang Gao 0029, Pieter Abbeel, Michael Laskin	NeurIPS	6
2021	Teachable Reinforcement Learning via Advice Distillation abstract Training automated agents to complete complex tasks in interactive environments is challenging: reinforcement learning requires careful hand-engineering of reward functions, imitation learning requires specialized infrastructure and access to a human expert, and learning from intermediate forms of supervision (like binary preferences) is time-consuming and extracts little information from each human intervention. Can we overcome these challenges by building agents that learn from rich, interactive feedback instead? We propose a new supervision paradigm for interactive learning based on "teachable" decision-making systems that learn from structured advice provided by an external teacher. We begin by formalizing a class of human-in-the-loop decision making problems in which multiple forms of teacher-provided advice are available to a learner. We then describe a simple learning algorithm for these problems that first learns to interpret advice, then learns from advice to complete tasks even in the absence of human supervision. In puzzle-solving, navigation, and locomotion domains, we show that agents that learn from advice can acquire new skills with significantly less human supervision than standard reinforcement learning algorithms and often less than imitation learning. Olivia Watkins, Abhishek Gupta 0004, Trevor Darrell, Pieter Abbeel, Jacob Andreas	NeurIPS	4
2021	Mastering Atari Games with Limited Data abstract Reinforcement learning has achieved great success in many applications. However, sample efficiency remains a key challenge, with prominent methods requiring millions (or even billions) of environment steps to train. Recently, there has been significant progress in sample efficient image-based RL algorithms; however, consistent human-level performance on the Atari game benchmark remains an elusive goal. We propose a sample efficient model-based visual RL algorithm built on MuZero, which we name EfficientZero. Our method achieves 194.3% mean human performance and 109.0% median performance on the Atari 100k benchmark with only two hours of real-time game experience and outperforms the state SAC in some tasks on the DMControl 100k benchmark. This is the first time an algorithm achieves super-human performance on Atari games with such little data. EfficientZero's performance is also close to DQN's performance at 200 million frames while we consume 500 times less data. EfficientZero's low sample complexity and high performance can bring RL closer to real-world applicability. We implement our algorithm in an easy-to-understand manner and it is available at https://github.com/YeWR/EfficientZero. We hope it will accelerate the research of MCTS-based RL algorithms in the wider community. Weirui Ye, Shaohuai Liu, Thanard Kurutach, Pieter Abbeel, Yang Gao 0029	NeurIPS	4
2021	AMP: adversarial motion priors for stylized physics-based character control abstract Synthesizing graceful and life-like behaviors for physically simulated characters has been a fundamental challenge in computer animation. Data-driven methods that leverage motion tracking are a prominent class of techniques for producing high fidelity motions for a wide range of behaviors. However, the effectiveness of these tracking-based methods often hinges on carefully designed objective functions, and when applied to large and diverse motion datasets, these methods require significant additional machinery to select the appropriate motion for the character to track in a given scenario. In this work, we propose to obviate the need to manually design imitation objectives and mechanisms for motion selection by utilizing a fully automated approach based on adversarial imitation learning. High-level task objectives that the character should perform can be specified by relatively simple reward functions, while the low-level style of the character's behaviors can be specified by a dataset of unstructured motion clips, without any explicit clip selection or sequencing. For example, a character traversing an obstacle course might utilize a task-reward that only considers forward progress, while the dataset contains clips of relevant behaviors such as running, jumping, and rolling. These motion clips are used to train an adversarial motion prior, which specifies style-rewards for training the character through reinforcement learning (RL). The adversarial RL procedure automatically selects which motion to perform, dynamically interpolating and generalizing from the dataset. Our system produces high-quality motions that are comparable to those achieved by state-of-the-art tracking-based techniques, while also being able to easily accommodate large datasets of unstructured motion clips. Composition of disparate skills emerges automatically from the motion prior, without requiring a high-level motion planner or other task-specific annotations of the motion clips. We demonstrate the effectiveness of our framework on a diverse cast of complex simulated characters and a challenging suite of motor control tasks. Xue Bin Peng, Ze Ma, Pieter Abbeel, Sergey Levine, Angjoo Kanazawa	ACM Trans. Graph.	3
2020	Model-Augmented Actor-Critic: Backpropagating through Paths Ignasi Clavera, Pieter Abbeel	ICLR	3
2020	Sub-policy Adaptation for Hierarchical Reinforcement Learning Alexander C. Li, Carlos Florensa, Ignasi Clavera, Pieter Abbeel	ICLR	4
2020	Hierarchically Decoupled Imitation For Morphological Transfer abstract Learning long-range behaviors on complex high-dimensional agents is a fundamental problem in robot learning. For such tasks, we argue that transferring learned information from a morphologically simpler agent can massively improve the sample efficiency of a more complex one. To this end, we propose a hierarchical decoupling of policies into two parts: an independently learned low-level policy and a transferable high-level policy. To remedy poor transfer performance due to mismatch in morphologies, we contribute two key ideas. First, we show that incentivizing a complex agent’s low-level to imitate a simpler agent’s low-level significantly improves zero-shot high-level transfer. Second, we show that KL-regularized training of the high level stabilizes learning and prevents mode-collapse. Finally, on a suite of publicly released navigation and manipulation environments, we demonstrate the applicability of hierarchical transfer on long-range tasks across morphologies. Our code and videos can be found at https://sites.google.com/berkeley.edu/morphology-transfer. Joey Hejna, Lerrel Pinto, Pieter Abbeel	ICML	3
2020	CURL: Contrastive Unsupervised Representations for Reinforcement Learning abstract We present CURL: Contrastive Unsupervised Representations for Reinforcement Learning. CURL extracts high-level features from raw pixels using contrastive learning and performs off-policy control on top of the extracted features. CURL outperforms prior pixel-based methods, both model-based and model-free, on complex tasks in the DeepMind Control Suite and Atari Games showing 1.9x and 1.2x performance gains at the 100K environment and interaction steps benchmarks respectively. On the DeepMind Control Suite, CURL is the first image-based algorithm to nearly match the sample-efficiency of methods that use state-based features. Our code is open-sourced and available at https://www.github.com/MishaLaskin/curl. Michael Laskin, Aravind Srinivas, Pieter Abbeel	ICML	3
2020	Variable Skipping for Autoregressive Range Density Estimation abstract Deep autoregressive models compute point likelihood estimates of individual data points. However, many applications (i.e., database cardinality estimation), require estimating range densities, a capability that is under-explored by current neural density estimation literature. In these applications, fast and accurate range density estimates over high-dimensional data directly impact user-perceived performance. In this paper, we explore a technique for accelerating range density estimation over deep autoregressive models. This technique, called variable skipping, exploits the sparse structure of range density queries to avoid sampling unnecessary variables during approximate inference. We show that variable skipping provides 10-100x efficiency improvements when targeting challenging high-quantile error metrics, enables complex applications such as text pattern matching, and can be realized via a simple data augmentation procedure without changing the usual maximum likelihood objective. Eric Liang, Zongheng Yang, Ion Stoica, Pieter Abbeel, Yan Duan, Xi Chen 0022	ICML	4
2020	Hallucinative Topological Memory for Zero-Shot Visual Planning abstract In visual planning (VP), an agent learns to plan goal-directed behavior from observations of a dynamical system obtained offline, e.g., images obtained from self-supervised robot interaction. Most previous works on VP approached the problem by planning in a learned latent space, resulting in low-quality visual plans, and difficult training algorithms. Here, instead, we propose a simple VP method that plans directly in image space and displays competitive performance. We build on the semi-parametric topological memory (SPTM) method: image samples are treated as nodes in a graph, the graph connectivity is learned from image sequence data, and planning can be performed using conventional graph search methods. We propose two modifications on SPTM. First, we train an energy-based graph connectivity function using contrastive predictive coding that admits stable training. Second, to allow zero-shot planning in new domains, we learn a conditional VAE model that generates images given a context describing the domain, and use these hallucinated samples for building the connectivity graph and planning. We show that this simple approach significantly outperform the SOTA VP methods, in terms of both plan interpretability and success rate when using the plan to guide a trajectory-following controller. Interestingly, our method can pick up non-trivial visual properties of objects, such as their geometry, and account for it in the plans. Kara Liu, Thanard Kurutach, Christine Tung, Pieter Abbeel, Aviv Tamar	ICML	4
2020	Planning to Explore via Self-Supervised World Models abstract Reinforcement learning allows solving complex tasks, however, the learning tends to be task-specific and the sample efficiency remains a challenge. We present Plan2Explore, a self-supervised reinforcement learning agent that tackles both these challenges through a new approach to self-supervised exploration and fast adaptation to new tasks, which need not be known during exploration. During exploration, unlike prior methods which retrospectively compute the novelty of observations after the agent has already reached them, our agent acts efficiently by leveraging planning to seek out expected future novelty. After exploration, the agent quickly adapts to multiple downstream tasks in a zero or a few-shot manner. We evaluate on challenging control tasks from high-dimensional image inputs. Without any training supervision or task-specific interaction, Plan2Explore outperforms prior self-supervised exploration methods, and in fact, almost matches the performances oracle which has access to rewards. Videos and code: https://ramanans1.github.io/plan2explore/ Ramanan Sekar, Oleh Rybkin, Kostas Daniilidis, Pieter Abbeel, Danijar Hafner, Deepak Pathak	ICML	4
2020	Responsive Safety in Reinforcement Learning by PID Lagrangian Methods abstract Lagrangian methods are widely used algorithms for constrained optimization problems, but their learning dynamics exhibit oscillations and overshoot which, when applied to safe reinforcement learning, leads to constraint-violating behavior during agent training. We address this shortcoming by proposing a novel Lagrange multiplier update method that utilizes derivatives of the constraint function. We take a controls perspective, wherein the traditional Lagrange multiplier update behaves as \emph{integral} control; our terms introduce \emph{proportional} and \emph{derivative} control, achieving favorable learning dynamics through damping and predictive measures. We apply our PID Lagrangian methods in deep RL, setting a new state of the art in Safety Gym, a safe RL benchmark. Lastly, we introduce a new method to ease controller tuning by providing invariance to the relative numerical scales of reward and cost. Our extensive experiments demonstrate improved performance and hyperparameter robustness, while our algorithms remain nearly as simple to derive and implement as the traditional Lagrangian approach. Adam Stooke, Joshua Achiam, Pieter Abbeel	ICML	3
2020	AvE: Assistance via Empowerment abstract One difficulty in using artificial agents for human-assistive applications lies in the challenge of accurately assisting with a person's goal(s). Existing methods tend to rely on inferring the human's goal, which is challenging when there are many potential goals or when the set of candidate goals is difficult to identify. We propose a new paradigm for assistance by instead increasing the human's ability to control their environment, and formalize this approach by augmenting reinforcement learning with human empowerment. This task-agnostic objective increases the person's autonomy and ability to achieve any eventual state. We test our approach against assistance based on goal inference, highlighting scenarios where our method overcomes failure modes stemming from goal ambiguity or misspecification. As existing methods for estimating empowerment in continuous domains are computationally hard, precluding its use in real time learned assistance, we also propose an efficient empowerment-inspired proxy metric. Using this, we are able to successfully demonstrate our method in a shared autonomy user study for a challenging simulated teleoperation task with human-in-the-loop training. Stas Tiomkin, Emre Kiciman, Daniel Polani, Pieter Abbeel, Anca D. Dragan	NeurIPS	5
2020	Sparse Graphical Memory for Robust Planning abstract To operate effectively in the real world, agents should be able to act from high-dimensional raw sensory input such as images and achieve diverse goals across long time-horizons. Current deep reinforcement and imitation learning methods can learn directly from high-dimensional inputs but do not scale well to long-horizon tasks. In contrast, classical graphical methods like A* search are able to solve long-horizon tasks, but assume that the state space is abstracted away from raw sensory input. Recent works have attempted to combine the strengths of deep learning and classical planning; however, dominant methods in this domain are still quite brittle and scale poorly with the size of the environment. We introduce Sparse Graphical Memory (SGM), a new data structure that stores states and feasible transitions in a sparse memory. SGM aggregates states according to a novel two-way consistency objective, adapting classic state aggregation criteria to goal-conditioned RL: two states are redundant when they are interchangeable both as goals and as starting states. Theoretically, we prove that merging nodes according to two-way consistency leads to an increase in shortest path lengths that scales only linearly with the merging threshold. Experimentally, we show that SGM significantly outperforms current state of the art methods on long horizon, sparse-reward visual navigation tasks. Project video and code are available at https://sites.google.com/view/sparse-graphical-memory. Scott Emmons, Ajay Jain, Michael Laskin, Thanard Kurutach, Pieter Abbeel, Deepak Pathak	NeurIPS	5
2020	Denoising Diffusion Probabilistic Models abstract We present high quality image synthesis results using diffusion probabilistic models, a class of latent variable models inspired by considerations from nonequilibrium thermodynamics. Our best results are obtained by training on a weighted variational bound designed according to a novel connection between diffusion probabilistic models and denoising score matching with Langevin dynamics, and our models naturally admit a progressive lossy decompression scheme that can be interpreted as a generalization of autoregressive decoding. On the unconditional CIFAR10 dataset, we obtain an Inception score of 9.46 and a state-of-the-art FID score of 3.17. On 256x256 LSUN, we obtain sample quality similar to ProgressiveGAN. Jonathan Ho, Ajay Jain, Pieter Abbeel	NeurIPS	3
2020	Reinforcement Learning with Augmented Data abstract Learning from visual observations is a fundamental yet challenging problem in Reinforcement Learning (RL). Although algorithmic advances combined with convolutional neural networks have proved to be a recipe for success, current methods are still lacking on two fronts: (a) data-efficiency of learning and (b) generalization to new environments. To this end, we present Reinforcement Learning with Augmented Data (RAD), a simple plug-and-play module that can enhance most RL algorithms. We perform the first extensive study of general data augmentations for RL on both pixel-based and state-based inputs, and introduce two new data augmentations - random translate and random amplitude scale. We show that augmentations such as random translate, crop, color jitter, patch cutout, random convolutions, and amplitude scale can enable simple RL algorithms to outperform complex state-of-the-art methods across common benchmarks. RAD sets a new state-of-the-art in terms of data-efficiency and final performance on the DeepMind Control Suite benchmark for pixel-based control as well as OpenAI Gym benchmark for state-based control. We further demonstrate that RAD significantly improves test-time generalization over existing methods on several OpenAI ProcGen benchmarks. Michael Laskin, Kimin Lee, Adam Stooke, Lerrel Pinto, Pieter Abbeel, Aravind Srinivas	NeurIPS	5
2020	Stochastic Latent Actor-Critic: Deep Reinforcement Learning with a Latent Variable Model abstract Deep reinforcement learning (RL) algorithms can use high-capacity deep networks to learn directly from image observations. However, these high-dimensional observation spaces present a number of challenges in practice, since the policy must now solve two problems: representation learning and task learning. In this work, we tackle these two problems separately, by explicitly learning latent representations that can accelerate reinforcement learning from images. We propose the stochastic latent actor-critic (SLAC) algorithm: a sample-efficient and high-performing RL algorithm for learning policies for complex continuous control tasks directly from high-dimensional image inputs. SLAC provides a novel and principled approach for unifying stochastic sequential models and RL into a single method, by learning a compact latent representation and then performing RL in the model's learned latent space. Our experimental evaluation demonstrates that our method outperforms both model-free and model-based alternatives in terms of final performance and sample efficiency, on a range of difficult image-based control tasks. Our code and videos of our results are available at our website. Alex X. Lee, Anusha Nagabandi, Pieter Abbeel, Sergey Levine	NeurIPS	3
2020	Generalized Hindsight for Reinforcement Learning abstract One of the key reasons for the high sample complexity in reinforcement learning (RL) is the inability to transfer knowledge from one task to another. In standard multi-task RL settings, low-reward data collected while trying to solve one task provides little to no signal for solving that particular task and is hence effectively wasted. However, we argue that this data, which is uninformative for one task, is likely a rich source of information for other tasks. To leverage this insight and efficiently reuse data, we present Generalized Hindsight: an approximate inverse reinforcement learning technique for relabeling behaviors with the right tasks. Intuitively, given a behavior generated under one task, Generalized Hindsight returns a different task that the behavior is better suited for. Then, the behavior is relabeled with this new task before being used by an off-policy RL optimizer. Compared to standard relabeling techniques, Generalized Hindsight provides a substantially more efficient re-use of samples, which we empirically demonstrate on a suite of multi-task navigation and manipulation tasks. Alexander C. Li, Lerrel Pinto, Pieter Abbeel	NeurIPS	3
2020	Trajectory-wise Multiple Choice Learning for Dynamics Generalization in Reinforcement Learning abstract Model-based reinforcement learning (RL) has shown great potential in various control tasks in terms of both sample-efficiency and final performance. However, learning a generalizable dynamics model robust to changes in dynamics remains a challenge since the target transition dynamics follow a multi-modal distribution. In this paper, we present a new model-based RL algorithm, coined trajectory-wise multiple choice learning, that learns a multi-headed dynamics model for dynamics generalization. The main idea is updating the most accurate prediction head to specialize each head in certain environments with similar dynamics, i.e., clustering environments. Moreover, we incorporate context learning, which encodes dynamics-specific information from past experiences into the context latent vector, enabling the model to perform online adaptation to unseen environments. Finally, to utilize the specialized prediction heads more effectively, we propose an adaptive planning method, which selects the most accurate prediction head over a recent experience. Our method exhibits superior zero-shot generalization performance across a variety of control tasks, compared to state-of-the-art RL methods. Source code and videos are available at https://sites.google.com/view/trajectory-mcl. Younggyo Seo, Kimin Lee, Ignasi Clavera Gilaberte, Thanard Kurutach, Jinwoo Shin, Pieter Abbeel	NeurIPS	6
2020	Automatic Curriculum Learning through Value Disagreement abstract Continually solving new, unsolved tasks is the key to learning diverse behaviors. Through reinforcement learning (RL), we have made massive strides towards solving tasks that have a single goal. However, in the multi-task domain, where an agent needs to reach multiple goals, the choice of training goals can largely affect sample efficiency. When biological agents learn, there is often an organized and meaningful order to which learning happens. Inspired by this, we propose setting up an automatic curriculum for goals that the agent needs to solve. Our key insight is that if we can sample goals at the frontier of the set of goals that an agent is able to reach, it will provide a significantly stronger learning signal compared to randomly sampled goals. To operationalize this idea, we introduce a goal proposal module that prioritizes goals that maximize the epistemic uncertainty of the Q-function of the policy. This simple technique samples goals that are neither too hard nor too easy for the agent to solve, hence enabling continual improvement. We evaluate our method across 13 multi-goal robotic tasks and 5 navigation tasks, and demonstrate performance gains over current state-of-the-art methods. Pieter Abbeel, Lerrel Pinto	NeurIPS	2
2020	Locally Masked Convolution for Autoregressive Models abstract High-dimensional generative models have many applications including image compression, multimedia generation, anomaly detection and data completion. State-of-the-art estimators for natural images are autoregressive, decomposing the joint distribution over pixels into a product of conditionals parameterized by a deep neural network, e.g. a convolutional neural network such as the PixelCNN. However, PixelCNNs only model a single decomposition of the joint, and only a single generation order is efficient. For tasks such as image completion, these models are unable to use much of the observed context. To generate data in arbitrary orders, we introduce LMConv: a simple modification to the standard 2D convolution that allows arbitrary masks to be applied to the weights at each location in the image. Using LMConv, we learn an ensemble of distribution estimators that share parameters but differ in generation order, achieving improved performance on whole-image density estimation (2.89 bpd on unconditional CIFAR10), as well as globally coherent image completions. Code is available at https://ajayjain.github.io/lmconv. Ajay Jain, Pieter Abbeel, Deepak Pathak	UAI	2
2019	Analog Circuit Generator based on Deep Neural Network enhanced Combinatorial Optimization abstract A deep neural network (DNN) based stochastic combinatorial optimization framework is presented that can find the optimal sizing of circuits in a sample-efficient manner. This sample efficiency allows us to unify this framework with generator-based tools like Berkeley Analog Generator (BAG) [1] to directly optimize layout, given the high level circuit specifications. We use this tool to design an optical link receiver layout, satisfying high-level design specifications, using post-layout simulations of only 348 design instances. Compared to an evolutionary algorithm without our DNN-based discriminator, our framework improves the sample efficiency and run time by more than 200x. Kourosh Hakhamaneshi, Nick Werblun, Pieter Abbeel, Vladimir Stojanovic	DAC	3
2019	BagNet: Berkeley Analog Generator with Layout Optimizer Boosted with Deep Neural Networks abstract The discrepancy between post-layout and schematic simulation results continues to widen in analog design due in part to the domination of layout parasitics. This paradigm shift is forcing designers to adopt design methodologies that seamlessly integrate layout effects into the standard design flow. Hence, any simulation-based optimization framework should take into account time-consuming post-layout simulation results. This work presents a learning framework that learns to reduce the number of simulations of evolutionary-based combinatorial optimizers, using a DNN that discriminates against generated samples, before running simulations. Using this approach, the discriminator achieves at least two orders of magnitude improvement on sample efficiency for several large circuit examples including an optical link receiver layout. Kourosh Hakhamaneshi, Nick Werblun, Pieter Abbeel, Vladimir Stojanovic	ICCAD	3
2019	Guiding Policies with Language via Meta-Learning John D. Co-Reyes, Abhishek Gupta 0004, Suvansh Sanjeev, Nick Altieri 0001, Jacob Andreas, John DeNero, Pieter Abbeel, Sergey Levine	ICLR (Poster)	7
2019	Learning to Adapt in Dynamic, Real-World Environments through Meta-Reinforcement Learning Anusha Nagabandi, Ignasi Clavera, Ronald S. Fearing, Pieter Abbeel, Sergey Levine, Chelsea Finn	ICLR (Poster)	5
2019	Variational Discriminator Bottleneck: Improving Imitation Learning, Inverse RL, and GANs by Constraining Information Flow Xue Bin Peng, Angjoo Kanazawa, Sam Toyer, Pieter Abbeel, Sergey Levine	ICLR (Poster)	4
2019	ProMP: Proximal Meta-Policy Search Jonas Rothfuss, Dennis Lee 0003, Ignasi Clavera, Tamim Asfour, Pieter Abbeel	ICLR (Poster)	5
2019	Preferences Implicit in the State of the World Rohin Shah, Dmitrii Krasheninnikov, Jordan Alexander, Pieter Abbeel, Anca D. Dragan	ICLR (Poster)	4
2019	Flow++: Improving Flow-Based Generative Models with Variational Dequantization and Architecture Design abstract Flow-based generative models are powerful exact likelihood models with efficient sampling and inference. Despite their computational efficiency, flow-based models generally have much worse density modeling performance compared to state-of-the-art autoregressive models. In this paper, we investigate and improve upon three limiting design choices employed by flow-based models in prior work: the use of uniform noise for dequantization, the use of inexpressive affine flows, and the use of purely convolutional conditioning networks in coupling layers. Based on our findings, we propose Flow++, a new flow-based model that is now the state-of-the-art non-autoregressive model for unconditional density estimation on standard image benchmarks. Our work has begun to close the significant performance gap that has so far existed between autoregressive models and flow-based models. Jonathan Ho, Xi Chen 0022, Aravind Srinivas, Yan Duan, Pieter Abbeel	ICML	5
2019	Population Based Augmentation: Efficient Learning of Augmentation Policy Schedules abstract A key challenge in leveraging data augmentation for neural network training is choosing an effective augmentation policy from a large search space of candidate operations. Properly chosen augmentation policies can lead to significant generalization improvements; however, state-of-the-art approaches such as AutoAugment are computationally infeasible to run for the ordinary user. In this paper, we introduce a new data augmentation algorithm, Population Based Augmentation (PBA), which generates nonstationary augmentation policy schedules instead of a fixed augmentation policy. We show that PBA can match the performance of AutoAugment on CIFAR-10, CIFAR-100, and SVHN, with three orders of magnitude less overall compute. On CIFAR-10 we achieve a mean test error of 1.46%, which is a slight improvement upon the current state-of-the-art. The code for PBA is open source and is available at https://github.com/arcelien/pba. Daniel Ho, Eric Liang, Xi Chen 0022, Ion Stoica, Pieter Abbeel	ICML	5
2019	Bit-Swap: Recursive Bits-Back Coding for Lossless Compression with Hierarchical Latent Variables abstract The bits-back argument suggests that latent variable models can be turned into lossless compression schemes. Translating the bits-back argument into efficient and practical lossless compression schemes for general latent variable models, however, is still an open problem. Bits-Back with Asymmetric Numeral Systems (BB-ANS), recently proposed by Townsend et al,. 2019, makes bits-back coding practically feasible for latent variable models with one latent layer, but it is inefficient for hierarchical latent variable models. In this paper we propose Bit-Swap, a new compression scheme that generalizes BB-ANS and achieves strictly better compression rates for hierarchical latent variable models with Markov chain structure. Through experiments we verify that Bit-Swap results in lossless compression rates that are empirically superior to existing techniques. Friso H. Kingma, Pieter Abbeel, Jonathan Ho	ICML	2
2019	On the Feasibility of Learning, Rather than Assuming, Human Biases for Reward Inference abstract Our goal is for agents to optimize the right reward function, despite how difficult it is for us to specify what that is. Inverse Reinforcement Learning (IRL) enables us to infer reward functions from demonstrations, but it usually assumes that the expert is noisily optimal. Real people, on the other hand, often have systematic biases: risk-aversion, myopia, etc. One option is to try to characterize these biases and account for them explicitly during learning. But in the era of deep learning, a natural suggestion researchers make is to avoid mathematical models of human behavior that are fraught with specific assumptions, and instead use a purely data-driven approach. We decided to put this to the test – rather than relying on assumptions about which specific bias the demonstrator has when planning, we instead learn the demonstrator’s planning algorithm that they use to generate demonstrations, as a differentiable planner. Our exploration yielded mixed findings: on the one hand, learning the planner can lead to better reward inference than relying on the wrong assumption; on the other hand, this benefit is dwarfed by the loss we incur by going from an exact to a differentiable planner. This suggest that at least for the foreseeable future, agents need a middle ground between the flexibility of data-driven methods and the useful bias of known human biases. Code is available at https://tinyurl.com/learningbiases. Rohin Shah, Noah Gundotra, Pieter Abbeel, Anca D. Dragan	ICML	3
2019	SOLAR: Deep Structured Representations for Model-Based Reinforcement Learning abstract Model-based reinforcement learning (RL) has proven to be a data efficient approach for learning control tasks but is difficult to utilize in domains with complex observations such as images. In this paper, we present a method for learning representations that are suitable for iterative model-based policy improvement, even when the underlying dynamical system has complex dynamics and image observations, in that these representations are optimized for inferring simple dynamics and cost models given data from the current policy. This enables a model-based RL method based on the linear-quadratic regulator (LQR) to be used for systems with image observations. We evaluate our approach on a range of robotics tasks, including manipulation with a real-world robotic arm directly from images. We find that our method produces substantially better final performance than other model-based RL methods while being significantly more efficient than model-free RL. Marvin Zhang, Sharad Vikram, Laura Smith 0001, Pieter Abbeel, Matthew J. Johnson 0002, Sergey Levine	ICML	4
2019	Quasi-Direct Drive for Low-Cost Compliant Robotic Manipulation abstract Robots must cost less and be force-controlled to enable widespread, safe deployment in unconstrained human environments. We propose Quasi-Direct Drive actuation as a capable paradigm for robotic force-controlled manipulation in human environments at low-cost. Our prototype - Blue - is a human scale 7 Degree of Freedom arm with 2kg payload. Blue can cost less than $5000. We show that Blue has dynamic properties that meet or exceed the needs of human operators: the robot has a nominal position-control bandwidth of 7.5Hz and repeatability within 4mm. We demonstrate a Virtual Reality based interface that can be used as a method for telepresence and collecting robot training demonstrations. Manufacturability, scaling, and potential use-cases for the Blue system are also addressed. Videos and additional information can be found online at berkeleyopenarms.github.io. David V. Gealy, Stephen McKinley, Brent Yi, Philipp Wu, Phillip R. Downey, Greg Balke, Allan Zhao, Menglong Guo, Rachel Thomasson, Anthony Sinclair, Peter Cuellar, Zoe McCarthy, Pieter Abbeel	ICRA	13
2019	Generalization through Simulation: Integrating Simulated and Real Data into Deep Reinforcement Learning for Vision-Based Autonomous Flight abstract Deep reinforcement learning provides a promising approach for vision-based control of real-world robots. However, the generalization of such models depends critically on the quantity and variety of data available for training. This data can be difficult to obtain for some types of robotic systems, such as fragile, small-scale quadrotors. Simulated rendering and physics can provide for much larger datasets, but such data is inherently of lower quality: many of the phenomena that make the real-world autonomous flight problem challenging, such as complex physics and air currents, are modeled poorly or not at all, and the systematic differences between simulation and the real world are typically impossible to eliminate. In this work, we investigate how data from both simulation and the real world can be combined in a hybrid deep reinforcement learning algorithm. Our method uses real-world data to learn about the dynamics of the system, and simulated data to learn a generalizable perception system that can enable the robot to avoid collisions using only a monocular camera. We demonstrate our approach on a real-world nano aerial vehicle collision avoidance task, showing that with only an hour of real-world data, the quadrotor can avoid collisions in new environments with various lighting conditions and geometry. Code, instructions for building the aerial vehicles, and videos of the experiments can be found at github.com/gkahn13/GtS. Katie Kang, Suneel Belkhale, Gregory Kahn, Pieter Abbeel, Sergey Levine	ICRA	4
2019	Reinforcement Learning on Variable Impedance Controller for High-Precision Robotic Assembly abstract Precise robotic manipulation skills are desirable in many industrial settings, reinforcement learning (RL) methods hold the promise of acquiring these skills autonomously. In this paper, we explicitly consider incorporating operational space force/torque information into reinforcement learning; this is motivated by humans heuristically mapping perceived forces to control actions, which results in completing high-precision tasks in a fairly easy manner. Our approach combines RL with force/torque information by incorporating a proper operational space force controller; where we also exploit different ablations on processing this information. Moreover, we propose a neural network architecture that generalizes to reasonable variations of the environment. We evaluate our method on the open-source Siemens Robot Learning Challenge, which requires precise and delicate force-controlled behavior to assemble a tight-fit gear wheel set. Jianlan Luo, Eugen Solowjow, Chengtao Wen, Juan Aparicio Ojea, Alice M. Agogino, Aviv Tamar, Pieter Abbeel	ICRA	7
2019	Domain Randomization for Active Pose Estimation abstract Accurate state estimation is a fundamental component of robotic control. In robotic manipulation tasks, as is our focus in this work, state estimation is essential for identifying the positions of objects in the scene, forming the basis of the manipulation plan. However, pose estimation typically requires expensive 3D cameras or additional instrumentation such as fiducial markers to perform accurately. Recently, Tobin et al. introduced an approach to pose estimation based on domain randomization, where a neural network is trained to predict pose directly from a 2D image of the scene. The network is trained on computer generated images with a high variation in textures and lighting, thereby generalizing to real world images. In this work, we investigate how to improve the accuracy of domain randomization based pose estimation. Our main idea is that active perception - moving the robot to get a better estimate of pose- can be trained in simulation and transferred to real using domain randomization. In our approach, the robot trains in a domain-randomized simulation how to estimate pose from a sequence of images. We show that our approach can significantly improve the accuracy of standard pose estimation in several scenarios: when the robot holding an object moves, when reference objects are moved in the scene, or when the camera is moved around the object. Xinyi Ren, Jianlan Luo, Eugen Solowjow, Juan Aparicio Ojea, Abhishek Gupta 0004, Aviv Tamar, Pieter Abbeel	ICRA	7
2019	One-Shot Composition of Vision-Based Skills from Demonstration abstract We consider the problem of learning multi-stage vision-based tasks on a real robot from a single video of a human performing the task, while leveraging demonstration data of subtasks with other objects. This problem presents a number of major challenges. Video demonstrations without teleoperation are easy for humans to provide, but do not provide any direct supervision. Learning policies from raw pixels enables full generality but calls for large function approximators with many parameters to be learned. Finally, compound tasks can require impractical amounts of demonstration data, when treated as a monolithic skill. To address these challenges, we propose a method that learns both how to learn primitive behaviors from video demonstrations and how to dynamically compose these behaviors to perform multi-stage tasks by “watching” a human demonstrator. Our results on a simulated Sawyer robot and real PR2 robot illustrate our method for learning a variety of order fulfillment and kitchen serving tasks with novel objects and raw pixel inputs. Video results are linked at https://sites.google.com/view/one-shot-hil. Tianhe Yu, Pieter Abbeel, Sergey Levine, Chelsea Finn	IROS	2
2019	On the Utility of Learning about Humans for Human-AI Coordination abstract While we would like agents that can coordinate with humans, current algorithms such as self-play and population-based training create agents that can coordinate with themselves. Agents that assume their partner to be optimal or similar to them can converge to coordination protocols that fail to understand and be understood by humans. To demonstrate this, we introduce a simple environment that requires challenging coordination, based on the popular game Overcooked, and learn a simple model that mimics human play. We evaluate the performance of agents trained via self-play and population-based training. These agents perform very well when paired with themselves, but when paired with our human model, they are significantly worse than agents designed to play with the human model. An experiment with a planning algorithm yields the same conclusion, though only when the human-aware planner is given the exact human model that it is playing with. A user study with real humans shows this pattern as well, though less strongly. Qualitatively, we find that the gains come from having the agent adapt to the human's gameplay. Given this result, we suggest several approaches for designing agents that learn about humans in order to better coordinate with them. Code is available at https://github.com/HumanCompatibleAI/overcooked_ai. Micah Carroll, Rohin Shah, Mark K. Ho, Thomas L. Griffiths 0001, Sanjit A. Seshia, Pieter Abbeel, Anca D. Dragan	NeurIPS	6
2019	Compositional Plan Vectors abstract Autonomous agents situated in real-world environments must be able to master large repertoires of skills. While a single short skill can be learned quickly, it would be impractical to learn every task independently. Instead, the agent should share knowledge across behaviors such that each task can be learned efficiently, and such that the resulting model can generalize to new tasks, especially ones that are compositions or subsets of tasks seen previously. A policy conditioned on a goal or demonstration has the potential to share knowledge between tasks if it sees enough diversity of inputs. However, these methods may not generalize to a more complex task at test time. We introduce compositional plan vectors (CPVs) to enable a policy to perform compositions of tasks without additional supervision. CPVs represent trajectories as the sum of the subtasks within them. We show that CPVs can be learned within a one-shot imitation learning framework without any additional supervision or information about task hierarchy, and enable a demonstration-conditioned policy to generalize to tasks that sequence twice as many skills as the tasks seen during training. Analogously to embeddings such as word2vec in NLP, CPVs can also support simple arithmetic operations -- for example, we can add the CPVs for two different tasks to command an agent to compose both tasks, without any additional training. Coline Devin, Daniel Geng, Pieter Abbeel, Trevor Darrell, Sergey Levine	NeurIPS	3
2019	Goal-conditioned Imitation Learning abstract Designing rewards for Reinforcement Learning (RL) is challenging because it needs to convey the desired task, be efficient to optimize, and be easy to compute. The latter is particularly problematic when applying RL to robotics, where detecting whether the desired configuration is reached might require considerable supervision and instrumentation. Furthermore, we are often interested in being able to reach a wide range of configurations, hence setting up a different reward every time might be unpractical. Methods like Hindsight Experience Replay (HER) have recently shown promise to learn policies able to reach many goals, without the need of a reward. Unfortunately, without tricks like resetting to points along the trajectory, HER might require many samples to discover how to reach certain areas of the state-space. In this work we propose a novel algorithm goalGAIL, which incorporates demonstrations to drastically speed up the convergence to a policy able to reach any goal, surpassing the performance of an agent trained with other Imitation Learning algorithms. Furthermore, we show our method can also be used when the available expert trajectories do not contain the actions or when the expert is suboptimal, which makes it applicable when only kinesthetic, third person or noisy demonstration is available. Carlos Florensa, Pieter Abbeel, Mariano Phielipp	NeurIPS	3
2019	Compression with Flows via Local Bits-Back Coding abstract Likelihood-based generative models are the backbones of lossless compression due to the guaranteed existence of codes with lengths close to negative log likelihood. However, there is no guaranteed existence of computationally efficient codes that achieve these lengths, and coding algorithms must be hand-tailored to specific types of generative models to ensure computational efficiency. Such coding algorithms are known for autoregressive models and variational autoencoders, but not for general types of flow models. To fill in this gap, we introduce local bits-back coding, a new compression technique for flow models. We present efficient algorithms that instantiate our technique for many popular types of flows, and we demonstrate that our algorithms closely achieve theoretical codelengths for state-of-the-art flow models on high-dimensional data. Jonathan Ho, Evan Lohn, Pieter Abbeel	NeurIPS	3
2019	Guided Meta-Policy Search abstract Reinforcement learning (RL) algorithms have demonstrated promising results on complex tasks, yet often require impractical numbers of samples because they learn from scratch. Meta-RL aims to address this challenge by leveraging experience from previous tasks so as to more quickly solve new tasks. However, in practice, these algorithms generally also require large amounts of on-policy experience during the \emph{meta-training} process, making them impractical for use in many problems. To this end, we propose to learn a reinforcement learning procedure in a federated way, where individual off-policy learners can solve the individual meta-training tasks, and then consolidate these solutions into a single meta-learner. Since the central meta-learner learns by imitating the solutions to the individual tasks, it can accommodate either the standard meta-RL problem setting, or a hybrid setting where some or all tasks are provided with example demonstrations. The former results in an approach that can leverage policies learned for previous tasks without significant amounts of on-policy data during meta-training, whereas the latter is particularly useful in cases where demonstrations are easy for a person to provide. Across a number of continuous control meta-RL problems, we demonstrate significant improvements in meta-RL sample efficiency in comparison to prior work as well as the ability to scale to domains with visual observations. Russell Mendonca, Abhishek Gupta 0004, Rosen Kralev, Pieter Abbeel, Sergey Levine, Chelsea Finn	NeurIPS	4
2019	MCP: Learning Composable Hierarchical Control with Multiplicative Compositional Policies abstract Humans are able to perform a myriad of sophisticated tasks by drawing upon skills acquired through prior experience. For autonomous agents to have this capability, they must be able to extract reusable skills from past experience that can be recombined in new ways for subsequent tasks. Furthermore, when controlling complex high-dimensional morphologies, such as humanoid bodies, tasks often require coordination of multiple skills simultaneously. Learning discrete primitives for every combination of skills quickly becomes prohibitive. Composable primitives that can be recombined to create a large variety of behaviors can be more suitable for modeling this combinatorial explosion. In this work, we propose multiplicative compositional policies (MCP), a method for learning reusable motor skills that can be composed to produce a range of complex behaviors. Our method factorizes an agent's skills into a collection of primitives, where multiple primitives can be activated simultaneously via multiplicative composition. This flexibility allows the primitives to be transferred and recombined to elicit new behaviors as necessary for novel tasks. We demonstrate that MCP is able to extract composable skills for highly complex simulated characters from pre-training tasks, such as motion imitation, and then reuse these skills to solve challenging continuous control tasks, such as dribbling a soccer ball to a goal, and picking up an object and transporting it to a target location. Xue Bin Peng, Michael Chang 0003, Grace Zhang, Pieter Abbeel, Sergey Levine	NeurIPS	4
2019	Evaluating Protein Transfer Learning with TAPE abstract Protein modeling is an increasingly popular area of machine learning research. Semi-supervised learning has emerged as an important paradigm in protein modeling due to the high cost of acquiring supervised protein labels, but the current literature is fragmented when it comes to datasets and standardized evaluation techniques. To facilitate progress in this field, we introduce the Tasks Assessing Protein Embeddings (TAPE), a set of five biologically relevant semi-supervised learning tasks spread across different domains of protein biology. We curate tasks into specific training, validation, and test splits to ensure that each task tests biologically relevant generalization that transfers to real-life scenarios. We benchmark a range of approaches to semi-supervised protein representation learning, which span recent work as well as canonical sequence learning techniques. We find that self-supervised pretraining is helpful for almost all models on all tasks, more than doubling performance in some cases. Despite this increase, in several cases features learned by self-supervised pretraining still lag behind features extracted by state-of-the-art non-neural techniques. This gap in performance suggests a huge opportunity for innovative architecture design and improved modeling paradigms that better capture the signal in biological sequences. TAPE will help the machine learning community focus effort on scientifically relevant problems. Toward this end, all data and code used to run these experiments is available at https://github.com/songlab-cal/tape Roshan Rao, Nicholas Bhattacharya, Yan Duan, Xi Chen 0022, John F. Canny, Pieter Abbeel, Yun S. Song	NeurIPS	7
2019	Addressing Sample Complexity in Visual Tasks Using HER and Hallucinatory GANs abstract Reinforcement Learning (RL) algorithms typically require millions of environment interactions to learn successful policies in sparse reward settings. Hindsight Experience Replay (HER) was introduced as a technique to increase sample efficiency by reimagining unsuccessful trajectories as successful ones by altering the originally intended goals. However, it cannot be directly applied to visual environments where goal states are often characterized by the presence of distinct visual features. In this work, we show how visual trajectories can be hallucinated to appear successful by altering agent observations using a generative model trained on relatively few snapshots of the goal. We then use this model in combination with HER to train RL agents in visual settings. We validate our approach on 3D navigation tasks and a simulated robotics application and show marked improvement over baselines derived from previous work. Himanshu Sahni, Toby Buckley, Pieter Abbeel, Ilya Kuzovkin	NeurIPS	3
2019	Geometry-Aware Neural Rendering abstract Understanding the 3-dimensional structure of the world is a core challenge in computer vision and robotics. Neural rendering approaches learn an implicit 3D model by predicting what a camera would see from an arbitrary viewpoint. We extend existing neural rendering to more complex, higher dimensional scenes than previously possible. We propose Epipolar Cross Attention (ECA), an attention mechanism that leverages the geometry of the scene to perform efficient non-local operations, requiring only $O(n)$ comparisons per spatial dimension instead of $O(n^2)$. We introduce three new simulated datasets inspired by real-world robotics and demonstrate that ECA significantly improves the quantitative and qualitative performance of Generative Query Networks (GQN). Joshua Tobin, Wojciech Zaremba, Pieter Abbeel	NeurIPS	3
2019	Deep Unsupervised Cardinality Estimation abstract Cardinality estimation has long been grounded in statistical tools for density estimation. To capture the rich multivariate distributions of relational tables, we propose the use of a new type of high-capacity statistical model: deep autoregressive models. However, direct application of these models leads to a limited estimator that is prohibitively expensive to evaluate for range or wildcard predicates. To produce a truly usable estimator, we develop a Monte Carlo integration scheme on top of autoregressive models that can efficiently handle range queries with dozens of dimensions or more. Like classical synopses, our estimator summarizes the data without supervision. Unlike previous solutions, we approximate the joint data distribution without any independence assumptions. Evaluated on real-world datasets and compared against real systems and dominant families of techniques, our estimator achieves single-digit multiplicative error at tail, an up to 90x accuracy improvement over the second best method, and is space- and runtime-efficient. Zongheng Yang, Eric Liang, Amog Kamsetty, Chenggang Wu 0001, Yan Duan, Xi Chen 0022, Pieter Abbeel, Joseph M. Hellerstein, Sanjay Krishnan, Ion Stoica	Proc. VLDB Endow.	7
2018	Emergence of Grounded Compositional Language in Multi-Agent Populations abstract By capturing statistical patterns in large corpora, machine learning has enabled significant advances in natural language processing, including in machine translation, question answering, and sentiment analysis. However, for agents to intelligently interact with humans, simply capturing the statistical patterns is insufficient. In this paper we investigate if, and how, grounded compositional language can emerge as a means to achieve goals in multi-agent populations. Towards this end, we propose a multi-agent learning environment and learning methods that bring about emergence of a basic compositional language. This language is represented as streams of abstract discrete symbols uttered by agents over time, but nonetheless has a coherent structure that possesses a defined vocabulary and syntax. We also observe emergence of non-verbal communication such as pointing and guiding when language communication is unavailable. Igor Mordatch, Pieter Abbeel	AAAI	2
2018	Continuous Adaptation via Meta-Learning in Nonstationary and Competitive Environments Maruan Al-Shedivat, Trapit Bansal, Yuri Burda, Ilya Sutskever, Igor Mordatch, Pieter Abbeel	ICLR	6
2018	Meta Learning Shared Hierarchies Kevin Frans, Jonathan Ho, Xi Chen 0022, Pieter Abbeel, John Schulman	ICLR (Poster)	4
2018	Model-Ensemble Trust-Region Policy Optimization Thanard Kurutach, Ignasi Clavera, Yan Duan, Aviv Tamar, Pieter Abbeel	ICLR (Poster)	5
2018	A Simple Neural Attentive Meta-Learner Nikhil Mishra, Mostafa Rohaninejad, Xi Chen 0022, Pieter Abbeel	ICLR (Poster)	4
2018	Parameter Space Noise for Exploration Matthias Plappert, Rein Houthooft, Prafulla Dhariwal, Szymon Sidor, Richard Y. Chen, Xi Chen 0022, Tamim Asfour, Pieter Abbeel, Marcin Andrychowicz	ICLR (Poster)	8
2018	Variance Reduction for Policy Gradient with Action-Dependent Factorized Baselines Cathy Wu 0002, Aravind Rajeswaran, Yan Duan, Alexandre M. Bayen, Sham M. Kakade, Igor Mordatch, Pieter Abbeel	ICLR	8
2018	PixelSNAIL: An Improved Autoregressive Generative Model abstract Autoregressive generative models achieve the best results in density estimation tasks involving high dimensional data, such as images or audio. They pose density estimation as a sequence modeling task, where a recurrent neural network (RNN) models the conditional distribution over the next element conditioned on all previous elements. In this paradigm, the bottleneck is the extent to which the RNN can model long-range dependencies, and the most successful approaches rely on causal convolutions. Taking inspiration from recent work in meta reinforcement learning, where dealing with long-range dependencies is also essential, we introduce a new generative model architecture that combines causal convolutions with self attention. In this paper, we describe the resulting model and present state-of-the-art log-likelihood results on heavily benchmarked datasets: CIFAR-10, $32 \times 32$ ImageNet and $64 \times 64$ ImageNet. Our implementation will be made available at \url{https://github.com/neocxi/pixelsnail-public}. Xi Chen 0022, Nikhil Mishra, Mostafa Rohaninejad, Pieter Abbeel	ICML	4
2018	Self-Consistent Trajectory Autoencoder: Hierarchical Reinforcement Learning with Trajectory Embeddings abstract In this work, we take a representation learning perspective on hierarchical reinforcement learning, where the problem of learning lower layers in a hierarchy is transformed into the problem of learning trajectory-level generative models. We show that we can learn continuous latent representations of trajectories, which are effective in solving temporally extended and multi-stage problems. Our proposed model, SeCTAR, draws inspiration from variational autoencoders, and learns latent representations of trajectories. A key component of this method is to learn both a latent-conditioned policy and a latent-conditioned model which are consistent with each other. Given the same latent, the policy generates a trajectory which should match the trajectory predicted by the model. This model provides a built-in prediction mechanism, by predicting the outcome of closed loop policy behavior. We propose a novel algorithm for performing hierarchical RL with this model, combining model-based planning in the learned latent space with an unsupervised exploration objective. We show that our model is effective at reasoning over long horizons with sparse rewards for several simulated tasks, outperforming standard reinforcement learning methods and prior methods for hierarchical reasoning, model-based planning, and exploration. This model provides a built-in prediction mechanism, by predicting the outcome of closed loop policy behavior. We propose a novel algorithm for performing hierarchical RL with this model, combining model-based planning in the learned latent space with an unsupervised exploration objective. We show that our model is effective at reasoning over long horizons with sparse rewards for several simulated tasks, outperforming standard reinforcement learning methods and prior methods for hierarchical reasoning, model-based planning, and exploration. John D. Co-Reyes, Yuxuan Liu 0001, Abhishek Gupta 0004, Benjamin Eysenbach, Pieter Abbeel, Sergey Levine	ICML	5
2018	Automatic Goal Generation for Reinforcement Learning Agents abstract Reinforcement learning (RL) is a powerful technique to train an agent to perform a task; however, an agent that is trained using RL is only capable of achieving the single task that is specified via its reward function. Such an approach does not scale well to settings in which an agent needs to perform a diverse set of tasks, such as navigating to varying positions in a room or moving objects to varying locations. Instead, we propose a method that allows an agent to automatically discover the range of tasks that it is capable of performing in its environment. We use a generator network to propose tasks for the agent to try to accomplish, each task being specified as reaching a certain parametrized subset of the state-space. The generator network is optimized using adversarial training to produce tasks that are always at the appropriate level of difficulty for the agent, thus automatically producing a curriculum. We show that, by using this framework, an agent can efficiently and automatically learn to perform a wide set of tasks without requiring any prior knowledge of its environment, even when only sparse rewards are available. Videos and code available at https://sites.google.com/view/goalgeneration4rl. Carlos Florensa, David Held, Xinyang Geng, Pieter Abbeel	ICML	4
2018	Latent Space Policies for Hierarchical Reinforcement Learning abstract We address the problem of learning hierarchical deep neural network policies for reinforcement learning. In contrast to methods that explicitly restrict or cripple lower layers of a hierarchy to force them to use higher-level modulating signals, each layer in our framework is trained to directly solve the task, but acquires a range of diverse strategies via a maximum entropy reinforcement learning objective. Each layer is also augmented with latent random variables, which are sampled from a prior distribution during the training of that layer. The maximum entropy objective causes these latent variables to be incorporated into the layer’s policy, and the higher level layer can directly control the behavior of the lower layer through this latent space. Furthermore, by constraining the mapping from latent variables to actions to be invertible, higher layers retain full expressivity: neither the higher layers nor the lower layers are constrained in their behavior. Our experimental evaluation demonstrates that we can improve on the performance of single-layer policies on standard benchmark tasks simply by adding additional layers, and that our method can solve more complex sparse-reward tasks by learning higher-level policies on top of high-entropy skills optimized for simple low-level objectives. Tuomas Haarnoja, Kristian Hartikainen, Pieter Abbeel, Sergey Levine	ICML	3
2018	Soft Actor-Critic: Off-Policy Maximum Entropy Deep Reinforcement Learning with a Stochastic Actor abstract Model-free deep reinforcement learning (RL) algorithms have been demonstrated on a range of challenging decision making and control tasks. However, these methods typically suffer from two major challenges: very high sample complexity and brittle convergence properties, which necessitate meticulous hyperparameter tuning. Both of these challenges severely limit the applicability of such methods to complex, real-world domains. In this paper, we propose soft actor-critic, an off-policy actor-critic deep RL algorithm based on the maximum entropy reinforcement learning framework. In this framework, the actor aims to maximize expected reward while also maximizing entropy. That is, to succeed at the task while acting as randomly as possible. Prior deep RL methods based on this framework have been formulated as Q-learning methods. By combining off-policy updates with a stable stochastic actor-critic formulation, our method achieves state-of-the-art performance on a range of continuous control benchmark tasks, outperforming prior on-policy and off-policy methods. Furthermore, we demonstrate that, in contrast to other off-policy algorithms, our approach is very stable, achieving very similar performance across different random seeds. Tuomas Haarnoja, Aurick Zhou, Pieter Abbeel, Sergey Levine	ICML	3
2018	Universal Planning Networks: Learning Generalizable Representations for Visuomotor Control abstract A key challenge in complex visuomotor control is learning abstract representations that are effective for specifying goals, planning, and generalization. To this end, we introduce universal planning networks (UPN). UPNs embed differentiable planning within a goal-directed policy. This planning computation unrolls a forward model in a latent space and infers an optimal action plan through gradient descent trajectory optimization. The plan-by-gradient-descent process and its underlying representations are learned end-to-end to directly optimize a supervised imitation learning objective. We find that the representations learned are not only effective for goal-directed visual imitation via gradient-based trajectory optimization, but can also provide a metric for specifying goals using images. The learned representations can be leveraged to specify distance-based rewards to reach new target states for model-free reinforcement learning, resulting in substantially more effective learning when solving new tasks described via image based goals. We were able to achieve successful transfer of visuomotor planning strategies across robots with significantly different morphologies and actuation capabilities. Visit https://sites.google. com/view/upn-public/home for video highlights. Aravind Srinivas, Allan Jabri, Pieter Abbeel, Sergey Levine, Chelsea Finn	ICML	3
2018	Deep Object-Centric Representations for Generalizable Robot Learning abstract Robotic manipulation in complex open-world scenarios requires both reliable physical manipulation skills and effective and generalizable perception. In this paper, we propose using an object-centric prior and a semantic feature space for the perception system of a learned policy. We devise an object-level attentional mechanism that can be used to determine relevant objects from a few trajectories or demonstrations, and then immediately incorporate those objects into a learned policy. A task-independent attention locates possible objects in the scene, and a task-specific attention identifies which objects are predictive of the trajectories. The scope of the task-specific attention is easily adjusted by showing demonstrations with distractor objects or with diverse relevant objects. Our results indicate that this approach exhibits good generalization across object instances using very few samples, and can be used to learn a variety of manipulation tasks using reinforcement learning. Coline Devin, Pieter Abbeel, Trevor Darrell, Sergey Levine	ICRA	2
2018	Composable Deep Reinforcement Learning for Robotic Manipulation abstract Model-free deep reinforcement learning has been shown to exhibit good performance in domains ranging from video games to simulated robotic manipulation and locomotion. However, model-free methods are known to perform poorly when the interaction time with the environment is limited, as is the case for most real-world robotic tasks. In this paper, we study how maximum entropy policies trained using soft Q-learning can be applied to real-world robotic manipulation. The application of this method to real-world manipulation is facilitated by two important features of soft Q-learning. First, soft Q-learning can learn multimodal exploration strategies by learning policies represented by expressive energy-based models. Second, we show that policies learned with soft Q-learning can be composed to create new policies, and that the optimality of the resulting policy can be bounded in terms of the divergence between the composed policies. This compositionality provides an especially valuable tool for real-world manipulation, where constructing new policies by composing existing skills can provide a large gain in efficiency over training from scratch. Our experimental evaluation demonstrates that soft Q-learning is substantially more sample efficient than prior model-free deep reinforcement learning methods, and that compositionality can be performed for both simulated and real-world tasks. Tuomas Haarnoja, Vitchyr Pong, Aurick Zhou, Murtaza Dalal, Pieter Abbeel, Sergey Levine	ICRA	5
2018	Self-Supervised Deep Reinforcement Learning with Generalized Computation Graphs for Robot Navigation abstract Enabling robots to autonomously navigate complex environments is essential for real-world deployment. Prior methods approach this problem by having the robot maintain an internal map of the world, and then use a localization and planning method to navigate through the internal map. However, these approaches often include a variety of assumptions, are computationally intensive, and do not learn from failures. In contrast, learning-based methods improve as the robot acts in the environment, but are difficult to deploy in the real-world due to their high sample complexity. To address the need to learn complex policies with few samples, we propose a generalized computation graph that subsumes value-based model-free methods and model-based methods, with specific instantiations interpolating between model-free and model-based. We then instantiate this graph to form a navigation model that learns from raw images and is sample efficient. Our simulated car experiments explore the design decisions of our navigation model, and show our approach outperforms single-step andN-step double Q-learning. We also evaluate our approach on a real-world RC car and show it can learn to navigate through a complex indoor environment with a few hours of fully autonomous, self-supervised training. Videos of the experiments and code can be found at github.com/gkahn13/gcg. Gregory Kahn, Adam Villaflor, Bosen Ding, Pieter Abbeel, Sergey Levine	ICRA	4
2018	Imitation from Observation: Learning to Imitate Behaviors from Raw Video via Context Translation abstract Imitation learning is an effective approach for autonomous systems to acquire control policies when an explicit reward function is unavailable, using supervision provided as demonstrations from an expert, typically a human operator. However, standard imitation learning methods assume that the agent receives examples of observation-action tuples that could be provided, for instance, to a supervised learning algorithm. This stands in contrast to how humans and animals imitate: we observe another person performing some behavior and then figure out which actions will realize that behavior, compensating for changes in viewpoint, surroundings, object positions and types, and other factors. We term this kind of imitation learning “imitation-from-observation,” and propose an imitation learning method based on video prediction with context translation and deep reinforcement learning. This lifts the assumption in imitation learning that the demonstration should consist of observations in the same environment configuration, and enables a variety of interesting applications, including learning robotic skills that involve tool use simply by observing videos of human tool use. Our experimental results show the effectiveness of our approach in learning a wide range of real-world robotic tasks modeled after common household chores from videos of a human demonstrator, including sweeping, ladling almonds, pushing objects as well as a number of tasks in simulation. Yuxuan Liu 0001, Abhishek Gupta 0004, Pieter Abbeel, Sergey Levine	ICRA	3
2018	Overcoming Exploration in Reinforcement Learning with Demonstrations abstract Exploration in environments with sparse rewards has been a persistent problem in reinforcement learning (RL). Many tasks are natural to specify with a sparse reward, and manually shaping a reward function can result in suboptimal performance. However, finding a non-zero reward is exponentially more difficult with increasing task horizon or action dimensionality. This puts many real-world tasks out of practical reach of RL methods. In this work, we use demonstrations to overcome the exploration problem and successfully learn to perform long-horizon, multi-step robotics tasks with continuous control such as stacking blocks with a robot arm. Our method, which builds on top of Deep Deterministic Policy Gradients and Hindsight Experience Replay, provides an order of magnitude of speedup over RL on simulated robotics tasks. It is simple to implement and makes only the additional assumption that we can collect a small set of demonstrations. Furthermore, our method is able to solve tasks not solvable by either RL or behavior cloning alone, and often ends up outperforming the demonstrator policy. Ashvin Nair, Bob McGrew, Marcin Andrychowicz, Wojciech Zaremba, Pieter Abbeel	ICRA	5
2018	Sim-to-Real Transfer of Robotic Control with Dynamics Randomization abstract Simulations are attractive environments for training agents as they provide an abundant source of data and alleviate certain safety concerns during the training process. But the behaviours developed by agents in simulation are often specific to the characteristics of the simulator. Due to modeling error, strategies that are successful in simulation may not transfer to their real world counterparts. In this paper, we demonstrate a simple method to bridge this “reality gap”. By randomizing the dynamics of the simulator during training, we are able to develop policies that are capable of adapting to very different dynamics, including ones that differ significantly from the dynamics on which the policies were trained. This adaptivity enables the policies to generalize to the dynamics of the real world without any training on the physical system. Our approach is demonstrated on an object pushing task using a robotic arm. Despite being trained exclusively in simulation, our policies are able to maintain a similar level of performance when deployed on a real robot, reliably moving an object to a desired location from random initial configurations. We explore the impact of various design decisions and show that the resulting policies are robust to significant calibration error. Xue Bin Peng, Marcin Andrychowicz, Wojciech Zaremba, Pieter Abbeel	ICRA	4
2018	Learning Robotic Assembly from CAD abstract In this work, motivated by recent manufacturing trends, we investigate autonomous robotic assembly. Industrial assembly tasks require contact-rich manipulation skills, which are challenging to acquire using classical control and motion planning approaches. Consequently, robot controllers for assembly domains are presently engineered to solve a particular task, and cannot easily handle variations in the product or environment. Reinforcement learning (RL) is a promising approach for autonomously acquiring robot skills that involve contact-rich dynamics. However, RL relies on random exploration for learning a control policy, which requires many robot executions, and often gets trapped in locally suboptimal solutions. Instead, we posit that prior knowledge, when available, can improve RL performance. We exploit the fact that in modern assembly domains, geometric information about the task is readily available via the CAD design files. We propose to leverage this prior knowledge by guiding RL along a geometric motion plan, calculated using the CAD data. We show that our approach effectively improves over traditional control approaches for tracking the motion plan, and can solve assembly tasks that require high precision, even without accurate state estimation. In addition, we propose a neural network architecture that can learn to track the motion plan, thereby generalizing the assembly controller to changes in the object positions. Garrett Thomas, Melissa Chien, Aviv Tamar, Juan Aparicio Ojea, Pieter Abbeel	ICRA	5
2018	Deep Imitation Learning for Complex Manipulation Tasks from Virtual Reality Teleoperation abstract Imitation learning is a powerful paradigm for robot skill acquisition. However, obtaining demonstrations suitable for learning a policy that maps from raw pixels to actions can be challenging. In this paper we describe how consumer-grade Virtual Reality headsets and hand tracking hardware can be used to naturally teleoperate robots to perform complex tasks. We also describe how imitation learning can learn deep neural network policies (mapping from pixels to actions) that can acquire the demonstrated skills. Our experiments showcase the effectiveness of our approach for learning visuomotor skills. Tianhao Zhang 0001, Zoe McCarthy, Owen Jow, Dennis Lee 0003, Xi Chen 0022, Kenneth Y. Goldberg, Pieter Abbeel	ICRA	7
2018	Establishing Appropriate Trust via Critical States abstract In order to effectively interact with or supervise a robot, humans need to have an accurate mental model of its capabilities and how it acts. Learned neural network policies make that particularly challenging. We propose an approach for helping end-users build a mental model of such policies. Our key observation is that for most tasks, the essence of the policy is captured in a few critical states: states in which it is very important to take a certain action. Our user studies show that if the robot shows a human what its understanding of the task's critical states is, then the human can make a more informed decision about whether to deploy the policy, and if she does deploy it, when she needs to take control from it at execution time. Sandy H. Huang, Kush Bhatia, Pieter Abbeel, Anca D. Dragan	IROS	3
2018	Domain Randomization and Generative Models for Robotic Grasping abstract Deep learning-based robotic grasping has made significant progress thanks to algorithmic improvements and increased data availability. However, state-of-the-art models are often trained on as few as hundreds or thousands of unique object instances, and as a result generalization can be a challenge. In this work, we explore a novel data generation pipeline for training a deep neural network to perform grasp planning that applies the idea of domain randomization to object synthesis. We generate millions of unique, unrealistic procedurally generated objects, and train a deep neural network to perform grasp planning on these objects. Since the distribution of successful grasps for a given object can be highly multimodal, we propose an autoregressive grasp planning model that maps sensor inputs of a scene to a probability distribution over possible grasps. This model allows us to sample grasps efficiently at test time (or avoid sampling entirely). We evaluate our model architecture and data generation pipeline in simulation and the real world. We find we can achieve a >90% success rate on previously unseen realistic objects at test time in simulation despite having only been trained on random objects. We also demonstrate an 80% success rate on real-world grasp attempts despite having only been trained on random simulated objects. Josh Tobin, Lukas Biewald, Rocky Duan, Marcin Andrychowicz, Ankur Handa, Bob McGrew, Alex Ray, Jonas Schneider 0002, Peter Welinder, Wojciech Zaremba, Pieter Abbeel	IROS	12
2018	Meta-Reinforcement Learning of Structured Exploration Strategies abstract Exploration is a fundamental challenge in reinforcement learning (RL). Many current exploration methods for deep RL use task-agnostic objectives, such as information gain or bonuses based on state visitation. However, many practical applications of RL involve learning more than a single task, and prior tasks can be used to inform how exploration should be performed in new tasks. In this work, we study how prior tasks can inform an agent about how to explore effectively in new situations. We introduce a novel gradient-based fast adaptation algorithm – model agnostic exploration with structured noise (MAESN) – to learn exploration strategies from prior experience. The prior experience is used both to initialize a policy and to acquire a latent exploration space that can inject structured stochasticity into a policy, producing exploration strategies that are informed by prior knowledge and are more effective than random action-space noise. We show that MAESN is more effective at learning exploration strategies when compared to prior meta-RL methods, RL without learned exploration strategies, and task-agnostic exploration methods. We evaluate our method on a variety of simulated tasks: locomotion with a wheeled robot, locomotion with a quadrupedal walker, and object manipulation. Abhishek Gupta 0004, Russell Mendonca, Yuxuan Liu 0001, Pieter Abbeel, Sergey Levine	NeurIPS	4
2018	Evolved Policy Gradients abstract We propose a metalearning approach for learning gradient-based reinforcement learning (RL) algorithms. The idea is to evolve a differentiable loss function, such that an agent, which optimizes its policy to minimize this loss, will achieve high rewards. The loss is parametrized via temporal convolutions over the agent's experience. Because this loss is highly flexible in its ability to take into account the agent's history, it enables fast task learning. Empirical results show that our evolved policy gradient algorithm (EPG) achieves faster learning on several randomized environments compared to an off-the-shelf policy gradient method. We also demonstrate that EPG's learned loss can generalize to out-of-distribution test time tasks, and exhibits qualitatively different behavior from other popular metalearning algorithms. Rein Houthooft, Phillip Isola, Bradly C. Stadie, Filip Wolski, Jonathan Ho, Pieter Abbeel	NeurIPS	7
2018	Learning Plannable Representations with Causal InfoGAN abstract In recent years, deep generative models have been shown to 'imagine' convincing high-dimensional observations such as images, audio, and even video, learning directly from raw data. In this work, we ask how to imagine goal-directed visual plans -- a plausible sequence of observations that transition a dynamical system from its current configuration to a desired goal state, which can later be used as a reference trajectory for control. We focus on systems with high-dimensional observations, such as images, and propose an approach that naturally combines representation learning and planning. Our framework learns a generative model of sequential observations, where the generative process is induced by a transition in a low-dimensional planning model, and an additional noise. By maximizing the mutual information between the generated observations and the transition in the planning model, we obtain a low-dimensional representation that best explains the causal nature of the data. We structure the planning model to be compatible with efficient planning algorithms, and we propose several such models based on either discrete or continuous states. Finally, to generate a visual plan, we project the current and goal observations onto their respective states in the planning model, plan a trajectory, and then use the generative model to transform the trajectory to a sequence of observations. We demonstrate our method on imagining plausible visual plans of rope manipulation. Thanard Kurutach, Aviv Tamar, Ge Yang 0003, Stuart Russell 0001, Pieter Abbeel	NeurIPS	5
2018	The Importance of Sampling inMeta-Reinforcement Learning abstract We interpret meta-reinforcement learning as the problem of learning how to quickly find a good sampling distribution in a new environment. This interpretation leads to the development of two new meta-reinforcement learning algorithms: E-MAML and E-$\text{RL}^2$. Results are presented on a new environment we call `Krazy World': a difficult high-dimensional gridworld which is designed to highlight the importance of correctly differentiating through sampling distributions in meta-reinforcement learning. Further results are presented on a set of maze environments. We show E-MAML and E-$\text{RL}^2$ deliver better performance than baseline algorithms on both tasks. Bradly C. Stadie, Ge Yang 0003, Rein Houthooft, Xi Chen 0022, Yan Duan, Yuhuai Wu, Pieter Abbeel, Ilya Sutskever	NeurIPS	7
2018	DeepMimic: example-guided deep reinforcement learning of physics-based character skills abstract A longstanding goal in character animation is to combine data-driven specification of behavior with a system that can execute a similar behavior in a physical simulation, thus enabling realistic responses to perturbations and environmental variation. We show that well-known reinforcement learning (RL) methods can be adapted to learn robust control policies capable of imitating a broad range of example motion clips, while also learning complex recoveries, adapting to changes in morphology, and accomplishing user-specified goals. Our method handles keyframed motions, highly-dynamic actions such as motion-captured flips and spins, and retargeted motions. By combining a motion-imitation objective with a task objective, we can train characters that react intelligently in interactive settings, e.g., by walking in a desired direction or throwing a ball at a user-specified target. This approach thus combines the convenience and motion quality of using motion clips to define the desired style and appearance, with the flexibility and generality afforded by RL methods and physics-based animation. We further explore a number of methods for integrating multiple clips into the learning process to develop multi-skilled agents capable of performing a rich repertoire of diverse skills. We demonstrate results using multiple characters (human, Atlas robot, bipedal dinosaur, dragon) and a large variety of skills, including locomotion, acrobatics, and martial arts. Xue Bin Peng, Pieter Abbeel, Sergey Levine, Michiel van de Panne	ACM Trans. Graph.	2
2018	SFV: reinforcement learning of physical skills from videos abstract Data-driven character animation based on motion capture can produce highly naturalistic behaviors and, when combined with physics simulation, can provide for natural procedural responses to physical perturbations, environmental changes, and morphological discrepancies. Motion capture remains the most popular source of motion data, but collecting mocap data typically requires heavily instrumented environments and actors. In this paper, we propose a method that enables physically simulated characters to learn skills from videos (SFV). Our approach, based on deep pose estimation and deep reinforcement learning, allows data-driven animation to leverage the abundance of publicly available video clips from the web, such as those from YouTube. This has the potential to enable fast and easy design of character controllers simply by querying for video recordings of the desired behavior. The resulting controllers are robust to perturbations, can be adapted to new settings, can perform basic object interactions, and can be retargeted to new morphologies via reinforcement learning. We further demonstrate that our method can predict potential human motions from still images, by forward simulation of learned controllers initialized from the observed pose. Our framework is able to learn a broad range of dynamic skills, including locomotion, acrobatics, and martial arts. (Video 1 ) Xue Bin Peng, Angjoo Kanazawa, Jitendra Malik, Pieter Abbeel, Sergey Levine	ACM Trans. Graph.	4
2017	Learning Invariant Feature Spaces to Transfer Skills with Reinforcement Learning Abhishek Gupta 0004, Coline Devin, Yuxuan Liu 0001, Pieter Abbeel, Sergey Levine	ICLR (Poster)	4
2017	Variational Lossy Autoencoder Xi Chen 0022, Diederik P. Kingma, Tim Salimans, Yan Duan, Prafulla Dhariwal, John Schulman, Ilya Sutskever, Pieter Abbeel	ICLR (Poster)	8
2017	Generalizing Skills with Semi-Supervised Reinforcement Learning Chelsea Finn, Tianhe Yu, Justin Fu, Pieter Abbeel, Sergey Levine	ICLR (Poster)	4
2017	Stochastic Neural Networks for Hierarchical Reinforcement Learning Carlos Florensa, Yan Duan, Pieter Abbeel	ICLR (Poster)	3
2017	Learning Visual Servoing with Deep Features and Fitted Q-Iteration Alex X. Lee, Sergey Levine, Pieter Abbeel	ICLR (Poster)	3
2017	Third Person Imitation Learning Bradly C. Stadie, Pieter Abbeel, Ilya Sutskever	ICLR (Poster)	2
2017	Constrained Policy Optimization abstract For many applications of reinforcement learning it can be more convenient to specify both a reward function and constraints, rather than trying to design behavior through the reward function. For example, systems that physically interact with or around humans should satisfy safety constraints. Recent advances in policy search algorithms (Mnih et al., 2016, Schulman et al., 2015, Lillicrap et al., 2016, Levine et al., 2016) have enabled new capabilities in high-dimensional control, but do not consider the constrained setting. We propose Constrained Policy Optimization (CPO), the first general-purpose policy search algorithm for constrained reinforcement learning with guarantees for near-constraint satisfaction at each iteration. Our method allows us to train neural network policies for high-dimensional control while making guarantees about policy behavior all throughout training. Our guarantees are based on a new theoretical result, which is of independent interest: we prove a bound relating the expected returns of two policies to an average divergence between them. We demonstrate the effectiveness of our approach on simulated robot locomotion tasks where the agent must satisfy constraints motivated by safety. Joshua Achiam, David Held, Aviv Tamar, Pieter Abbeel	ICML	4
2017	Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks abstract We propose an algorithm for meta-learning that is model-agnostic, in the sense that it is compatible with any model trained with gradient descent and applicable to a variety of different learning problems, including classification, regression, and reinforcement learning. The goal of meta-learning is to train a model on a variety of learning tasks, such that it can solve new learning tasks using only a small number of training samples. In our approach, the parameters of the model are explicitly trained such that a small number of gradient steps with a small amount of training data from a new task will produce good generalization performance on that task. In effect, our method trains the model to be easy to fine-tune. We demonstrate that this approach leads to state-of-the-art performance on two few-shot image classification benchmarks, produces good results on few-shot regression, and accelerates fine-tuning for policy gradient reinforcement learning with neural network policies. Chelsea Finn, Pieter Abbeel, Sergey Levine	ICML	2
2017	Reinforcement Learning with Deep Energy-Based Policies abstract We propose a method for learning expressive energy-based policies for continuous states and actions, which has been feasible only in tabular domains before. We apply our method to learning maximum entropy policies, resulting into a new algorithm, called soft Q-learning, that expresses the optimal policy via a Boltzmann distribution. We use the recently proposed amortized Stein variational gradient descent to learn a stochastic sampling network that approximates samples from this distribution. The benefits of the proposed algorithm include improved exploration and compositionality that allows transferring skills between tasks, which we confirm in simulated experiments with swimming and walking robots. We also draw a connection to actor-critic methods, which can be viewed performing approximate inference on the corresponding energy-based model. Tuomas Haarnoja, Pieter Abbeel, Sergey Levine	ICML	3
2017	Prediction and Control with Temporal Segment Models abstract We introduce a method for learning the dynamics of complex nonlinear systems based on deep generative models over temporal segments of states and actions. Unlike dynamics models that operate over individual discrete timesteps, we learn the distribution over future state trajectories conditioned on past state, past action, and planned future action trajectories, as well as a latent prior over action trajectories. Our approach is based on convolutional autoregressive models and variational autoencoders. It makes stable and accurate predictions over long horizons for complex, stochastic systems, effectively expressing uncertainty and modeling the effects of collisions, sensory noise, and action delays. The learned dynamics model and action prior can be used for end-to-end, fully differentiable trajectory optimization and model-based policy optimization, which we use to evaluate the performance and sample-efficiency of our method. Nikhil Mishra, Pieter Abbeel, Igor Mordatch	ICML	2
2017	Learning modular neural network policies for multi-task and multi-robot transfer abstract Reinforcement learning (RL) can automate a wide variety of robotic skills, but learning each new skill requires considerable real-world data collection and manual representation engineering to design policy classes or features. Using deep reinforcement learning to train general purpose neural network policies alleviates some of the burden of manual representation engineering by using expressive policy classes, but exacerbates the challenge of data collection, since such methods tend to be less efficient than RL with low-dimensional, hand-designed representations. Transfer learning can mitigate this problem by enabling us to transfer information from one skill to another and even from one robot to another. We show that neural network policies can be decomposed into “task-specific” and “robot-specific” modules, where the task-specific modules are shared across robots, and the robot-specific modules are shared across all tasks on that robot. This allows for sharing task information, such as perception, between robots and sharing robot information, such as dynamics and kinematics, between tasks. We exploit this decomposition to train mix-and-match modules that can solve new robot-task combinations that were not seen during training. Using a novel approach to train modular neural networks, we demonstrate the effectiveness of our transfer method for enabling zero-shot generalization with a variety of robots and tasks in simulation for both visual and non-visual tasks. Coline Devin, Abhishek Gupta 0004, Trevor Darrell, Pieter Abbeel, Sergey Levine	ICRA	4
2017	Probabilistically safe policy transfer abstract Although learning-based methods have great potential for robotics, one concern is that a robot that updates its parameters might cause large amounts of damage before it learns the optimal policy. We formalize the idea of safe learning in a probabilistic sense by defining an optimization problem: we desire to maximize the expected return while keeping the expected damage below a given safety limit. We study this optimization for the case of a robot manipulator with safety-based torque limits. We would like to ensure that the damage constraint is maintained at every step of the optimization and not just at convergence. To achieve this aim, we introduce a novel method which predicts how modifying the torque limit, as well as how updating the policy parameters, might affect the robot's safety. We show through a number of experiments that our approach allows the robot to improve its performance while ensuring that the expected damage constraint is not violated during the learning process. David Held, Zoe McCarthy, Fred Shentu, Pieter Abbeel	ICRA	5
2017	PLATO: Policy learning using adaptive trajectory optimization abstract Policy search can in principle acquire complex strategies for control of robots and other autonomous systems. When the policy is trained to process raw sensory inputs, such as images and depth maps, it can also acquire a strategy that combines perception and control. However, effectively processing such complex inputs requires an expressive policy class, such as a large neural network. These high-dimensional policies are difficult to train, especially when learning to control safety-critical systems. We propose PLATO, a continuous, reset-free reinforcement learning algorithm that trains complex control policies with supervised learning, using model-predictive control (MPC) to generate the supervision, hence never in need of running a partially trained and potentially unsafe policy. PLATO uses an adaptive training method to modify the behavior of MPC to gradually match the learned policy in order to generate training samples at states that are likely to be visited by the learned policy. PLATO also maintains the MPC cost as an objective to avoid highly undesirable actions that would result from strictly following the learned policy before it has been fully trained. We prove that this type of adaptive MPC expert produces supervision that leads to good long-horizon performance of the resulting policy. We also empirically demonstrate that MPC can still avoid dangerous on-policy actions in unexpected situations during training. Our empirical results on a set of challenging simulated aerial vehicle tasks demonstrate that, compared to prior methods, PLATO learns faster, experiences substantially fewer catastrophic failures (crashes) during training, and often converges to a better policy. Gregory Kahn, Tianhao Zhang 0001, Sergey Levine, Pieter Abbeel	ICRA	4
2017	Reset-free guided policy search: Efficient deep reinforcement learning with stochastic initial states abstract Autonomous learning of robotic skills can allow general-purpose robots to learn wide behavioral repertoires without extensive manual engineering. However, robotic skill learning must typically make trade-offs to enable practical real-world learning, such as requiring manually designed policy or value function representations, initialization from human demonstrations, instrumentation of the training environment, or extremely long training times. We propose a new reinforcement learning algorithm that can train general-purpose neural network policies with minimal human engineering, while still allowing for fast, efficient learning in stochastic environments. We build on the guided policy search (GPS) algorithm, which transforms the reinforcement learning problem into supervised learning from a computational teacher (without human demonstrations). In contrast to prior GPS methods, which require a consistent set of initial states to which the system must be reset after each episode, our approach can handle random initial states, allowing it to be used even when deterministic resets are impossible. We compare our method to existing policy search algorithms in simulation, showing that it can train high-dimensional neural network policies with the same sample efficiency as prior GPS methods, and can learn policies directly from image pixels. We also present real-world robot results that show that our method can learn manipulation policies with visual features and random initial states. William Montgomery, Anurag Ajay, Chelsea Finn, Pieter Abbeel, Sergey Levine	ICRA	4
2017	Combining self-supervised learning and imitation for vision-based rope manipulation abstract Manipulation of deformable objects, such as ropes and cloth, is an important but challenging problem in robotics. We present a learning-based system where a robot takes as input a sequence of images of a human manipulating a rope from an initial to goal configuration, and outputs a sequence of actions that can reproduce the human demonstration, using only monocular images as input. To perform this task, the robot learns a pixel-level inverse dynamics model of rope manipulation directly from images in a self-supervised manner, using about 60K interactions with the rope collected autonomously by the robot. The human demonstration provides a high-level plan of what to do and the low-level inverse model is used to execute the plan. We show that by combining the high and low-level plans, the robot can successfully manipulate a rope into a variety of target shapes using only a sequence of human-provided images for direction. Ashvin Nair, Dian Chen 0001, Pulkit Agrawal 0001, Phillip Isola, Pieter Abbeel, Jitendra Malik, Sergey Levine	ICRA	5
2017	Learning from the hindsight plan - Episodic MPC improvement abstract Model predictive control (MPC) is a popular control method that has proved effective for robotics, among other fields. MPC performs re-planning at every time step. Re-planning is done with a limited horizon per computational and real-time constraints and often also for robustness to potential model errors. However, the limited horizon leads to suboptimal performance. In this work, we consider the iterative learning setting, where the same task can be repeated several times, and propose a policy improvement scheme for MPC. The main idea is that between executions we can, offline, run MPC with a longer horizon, resulting in a hindsight plan. To bring the next real-world execution closer to the hindsight plan, our approach learns to re-shape the original cost function with the goal of satisfying the following property: short horizon planning (as realistic during real executions) with respect to the shaped cost should result in mimicking the hindsight plan. This effectively consolidates long-term reasoning into the short-horizon planning. We empirically evaluate our approach in contact-rich manipulation tasks both in simulated and real environments, such as peg insertion by a real PR2 robot. Aviv Tamar, Garrett Thomas, Tianhao Zhang 0001, Sergey Levine, Pieter Abbeel	ICRA	5
2017	Deep reinforcement learning for tensegrity robot locomotion abstract Tensegrity robots, composed of rigid rods connected by elastic cables, have a number of unique properties that make them appealing for use as planetary exploration rovers. However, control of tensegrity robots remains a difficult problem due to their unusual structures and complex dynamics. In this work, we show how locomotion gaits can be learned automatically using a novel extension of mirror descent guided policy search (MDGPS) applied to periodic locomotion movements, and we demonstrate the effectiveness of our approach on tensegrity robot locomotion. We evaluate our method with real-world and simulated experiments on the SUPERball tensegrity robot, showing that the learned policies generalize to changes in system parameters, unreliable sensor measurements, and variation in environmental conditions, including varied terrains and a range of different gravities. Our experiments demonstrate that our method not only learns fast, power-efficient feedback policies for rolling gaits, but that these policies can succeed with only the limited onboard sensing provided by SUPERball's accelerometers. We compare the learned feedback policies to learned open-loop policies and hand-engineered controllers, and demonstrate that the learned policy enables the first continuous, reliable locomotion gait for the real SUPERball robot. Our code and supplementary material is available from http://rll.berkeley.edu/drl_tensegrity. Marvin Zhang, Xinyang Geng, Jonathan Bruce, Ken Caluwaerts, Massimo Vespignani, Vytas SunSpiral, Pieter Abbeel, Sergey Levine	ICRA	7
2017	The Off-Switch Game abstract It is clear that one of the primary tools we can use to mitigate the potential risk from a misbehaving AI system is the ability to turn the system off. As the capabilities of AI systems improve, it is important to ensure that such systems do not adopt subgoals that prevent a human from switching the system off. This is a challenge because many formulations of rational agents create strong incentives for self-preservation. This is not caused by a built-in instinct, but because a rational agent will maximize expected utility and cannot achieve whatever objective it has been given if it is dead. Our goal is to study the incentives an agent has to allow itself to be switched off. We analyze a simple game between a human H and a robot R, where H can press R’s off switch but R can disable the off switch. A traditional agent takes its reward function for granted: we show that such agents have an incentive to disable the off switch, except in the special case where H is perfectly rational. Our key insight is that for R to want to preserve its off switch, it needs to be uncertain about the utility associated with the outcome, and to treat H’s actions as important observations about that utility. (R also has no incentive to switch itself off in this setting.) We conclude that giving machines an appropriate level of uncertainty about their objectives leads to safer designs, and we argue that this setting is a useful generalization of the classical AI paradigm of rational agents. Dylan Hadfield-Menell, Anca D. Dragan, Pieter Abbeel, Stuart Russell 0001	IJCAI	3
2017	Value Iteration Networks abstract We introduce the value iteration network (VIN): a fully differentiable neural network with a `planning module' embedded within. VINs can learn to plan, and are suitable for predicting outcomes that involve planning-based reasoning, such as policies for reinforcement learning. Key to our approach is a novel differentiable approximation of the value-iteration algorithm, which can be represented as a convolutional neural network, and trained end-to-end using standard backpropagation.We evaluate VIN based policies on discrete and continuous path-planning domains, and on a natural-language based search task. We show that by learning an explicit planning computation, VIN policies generalize better to new, unseen domains.This paper is a significantly abridged and IJCAI audience targeted version of the original NIPS 2016 paper with the same title, available here: https://arxiv.org/abs/1602.02867 Aviv Tamar, Yi Wu 0013, Garrett Thomas, Sergey Levine, Pieter Abbeel	IJCAI	5
2017	Policy transfer via modularity and reward guiding abstract Non-prehensile manipulation, such as pushing, is an important function for robots to move objects and is sometimes preferred as an alternative to grasping. However, due to unknown frictional forces, pushing has been proven a difficult task for robots. We explore the use of reinforcement learning to train a robot to robustly push an object. In order to deal with the sample complexity of training such a method, we train the pushing policy in simulation and then transfer this policy to the real world. In order to ease the transfer from simulation, we propose to use modularity to separate the learned policy from the raw inputs and outputs; rather than training “end-to-end,” we decompose our system into modules and train only a subset of these modules in simulation. We further demonstrate that we can incorporate prior knowledge about the task into the state space and the reward function to speed up convergence. Finally, we introduce “reward guiding” to modify the reward function and further reduce the training time. We demonstrate, in both simulation and real-world experiments, that such an approach can be used to reliably push an object from many initial positions and orientations. Videos available at https://goo.gl/B7LtY3. Ignasi Clavera, David Held, Pieter Abbeel	IROS	3
2017	Domain randomization for transferring deep neural networks from simulation to the real world abstract Bridging the `reality gap' that separates simulated robotics from experiments on hardware could accelerate robotic research through improved data availability. This paper explores domain randomization, a simple technique for training models on simulated images that transfer to real images by randomizing rendering in the simulator. With enough variability in the simulator, the real world may appear to the model as just another variation. We focus on the task of object localization, which is a stepping stone to general robotic manipulation skills. We find that it is possible to train a real-world object detector that is accurate to 1.5 cm and robust to distractors and partial occlusions using only data from a simulator with non-realistic random textures. To demonstrate the capabilities of our detectors, we show they can be used to perform grasping in a cluttered environment. To our knowledge, this is the first successful transfer of a deep neural network trained only on simulated RGB images (without pre-training on real images) to the real world for the purpose of robotic control. Josh Tobin, Rachel Fong, Alex Ray, Jonas Schneider 0002, Wojciech Zaremba, Pieter Abbeel	IROS	6
2017	Gradescope: A Fast, Flexible, and Fair System for Scalable Assessment of Handwritten Work abstract We present a system for online assessment of handwritten homework assignments and exams. First, either instructors or students scan and upload handwritten work. Instructors then grade the work and distribute the results using a web-based platform. Our system optimizes for three key dimensions: speed, consistency, and flexibility. The primary innovation enabling improvements in all three dimensions is a dynamically evolving rubric for each question on an assessment. We also describe how the system minimizes the overhead incurred in the digitization process. Our system has been in use for four years, with instructors at 200 institutions having graded over 10 million pages of student work. We present results as user-reported data and feedback regarding time saved grading, enjoyment, and student experience. Two-thirds of responders report saving 30% or more time relative to their traditional workflow. We also find that the time spent grading an individual response to a question rapidly decays with the number of responses to that question that the grader has already graded. Sergey Karayev, Kevin Gutowski, Pieter Abbeel	L@S	4
2017	Hindsight Experience Replay abstract Dealing with sparse rewards is one of the biggest challenges in Reinforcement Learning (RL). We present a novel technique called Hindsight Experience Replay which allows sample-efficient learning from rewards which are sparse and binary and therefore avoid the need for complicated reward engineering. It can be combined with an arbitrary off-policy RL algorithm and may be seen as a form of implicit curriculum. We demonstrate our approach on the task of manipulating objects with a robotic arm. In particular, we run experiments on three different tasks: pushing, sliding, and pick-and-place, in each case using only binary rewards indicating whether or not the task is completed. Our ablation studies show that Hindsight Experience Replay is a crucial ingredient which makes training possible in these challenging environments. We show that our policies trained on a physics simulation can be deployed on a physical robot and successfully complete the task. The video presenting our experiments is available at https://goo.gl/SMrQnI. Marcin Andrychowicz, Dwight Crow, Alex Ray, Jonas Schneider 0002, Rachel Fong, Peter Welinder, Bob McGrew, Josh Tobin, Pieter Abbeel, Wojciech Zaremba	NIPS	9
2017	One-Shot Imitation Learning abstract Imitation learning has been commonly applied to solve different tasks in isolation. This usually requires either careful feature engineering, or a significant number of samples. This is far from what we desire: ideally, robots should be able to learn from very few demonstrations of any given task, and instantly generalize to new situations of the same task, without requiring task-specific engineering. In this paper, we propose a meta-learning framework for achieving such capability, which we call one-shot imitation learning. Specifically, we consider the setting where there is a very large (maybe infinite) set of tasks, and each task has many instantiations. For example, a task could be to stack all blocks on a table into a single tower, another task could be to place all blocks on a table into two-block towers, etc. In each case, different instances of the task would consist of different sets of blocks with different initial states. At training time, our algorithm is presented with pairs of demonstrations for a subset of all tasks. A neural net is trained that takes as input one demonstration and the current state (which initially is the initial state of the other demonstration of the pair), and outputs an action with the goal that the resulting sequence of states and actions matches as closely as possible with the second demonstration. At test time, a demonstration of a single instance of a new task is presented, and the neural net is expected to perform well on new instances of this new task. Our experiments show that the use of soft attention allows the model to generalize to conditions and tasks unseen in the training data. We anticipate that by training this model on a much greater variety of tasks and settings, we will obtain a general system that can turn any demonstrations into robust policies that can accomplish an overwhelming variety of tasks. Yan Duan, Marcin Andrychowicz, Bradly C. Stadie, Jonathan Ho, Jonas Schneider 0002, Ilya Sutskever, Pieter Abbeel, Wojciech Zaremba	NIPS	7
2017	Inverse Reward Design abstract Autonomous agents optimize the reward function we give them. What they don't know is how hard it is for us to design a reward function that actually captures what we want. When designing the reward, we might think of some specific training scenarios, and make sure that the reward will lead to the right behavior in those scenarios. Inevitably, agents encounter new scenarios (e.g., new types of terrain) where optimizing that same reward may lead to undesired behavior. Our insight is that reward functions are merely observations about what the designer actually wants, and that they should be interpreted in the context in which they were designed. We introduce inverse reward design (IRD) as the problem of inferring the true objective based on the designed reward and the training MDP. We introduce approximate methods for solving IRD problems, and use their solution to plan risk-averse behavior in test MDPs. Empirical results suggest that this approach can help alleviate negative side effects of misspecified reward functions and mitigate reward hacking. Dylan Hadfield-Menell, Smitha Milli, Pieter Abbeel, Stuart Russell 0001, Anca D. Dragan	NIPS	3
2017	Multi-Agent Actor-Critic for Mixed Cooperative-Competitive Environments abstract We explore deep reinforcement learning methods for multi-agent domains. We begin by analyzing the difficulty of traditional algorithms in the multi-agent case: Q-learning is challenged by an inherent non-stationarity of the environment, while policy gradient suffers from a variance that increases as the number of agents grows. We then present an adaptation of actor-critic methods that considers action policies of other agents and is able to successfully learn policies that require complex multi-agent coordination. Additionally, we introduce a training regimen utilizing an ensemble of policies for each agent that leads to more robust multi-agent policies. We show the strength of our approach compared to existing methods in cooperative as well as competitive scenarios, where agent populations are able to discover various physical and informational coordination strategies. Ryan Lowe, Yi Wu 0013, Aviv Tamar, Jean Harb, Pieter Abbeel, Igor Mordatch	NIPS	5
2017	#Exploration: A Study of Count-Based Exploration for Deep Reinforcement Learning abstract Count-based exploration algorithms are known to perform near-optimally when used in conjunction with tabular reinforcement learning (RL) methods for solving small discrete Markov decision processes (MDPs). It is generally thought that count-based methods cannot be applied in high-dimensional state spaces, since most states will only occur once. Recent deep RL exploration strategies are able to deal with high-dimensional continuous state spaces through complex heuristics, often relying on optimism in the face of uncertainty or intrinsic motivation. In this work, we describe a surprising finding: a simple generalization of the classic count-based approach can reach near state-of-the-art performance on various high-dimensional and/or continuous deep RL benchmarks. States are mapped to hash codes, which allows to count their occurrences with a hash table. These counts are then used to compute a reward bonus according to the classic count-based exploration theory. We find that simple hash functions can achieve surprisingly good results on many challenging tasks. Furthermore, we show that a domain-dependent learned hash code may further improve these results. Detailed analysis reveals important aspects of a good hash function: 1) having appropriate granularity and 2) encoding information relevant to solving the MDP. This exploration strategy achieves near state-of-the-art performance on both continuous control tasks and Atari 2600 games, hence providing a simple yet powerful baseline for solving MDPs that require considerable exploration. Rein Houthooft, Davis Foote, Adam Stooke, Xi Chen 0022, Yan Duan, John Schulman, Filip De Turck, Pieter Abbeel	NIPS	9
2017	Inverse Reinforcement Learning via Deep Gaussian Process Ming Jin 0002, Andreas Damianou, Pieter Abbeel, Costas J. Spanos	UAI	3
2016	Benchmarking Deep Reinforcement Learning for Continuous Control abstract Recently, researchers have made significant progress combining the advances in deep learning for learning feature representations with reinforcement learning. Some notable examples include training agents to play Atari games based on raw pixel data and to acquire advanced manipulation skills using raw sensory inputs. However, it has been difficult to quantify progress in the domain of continuous control due to the lack of a commonly adopted benchmark. In this work, we present a benchmark suite of continuous control tasks, including classic tasks like cart-pole swing-up, tasks with very high state and action dimensionality such as 3D humanoid locomotion, tasks with partial observations, and tasks with hierarchical structure. We report novel findings based on the systematic evaluation of a range of implemented reinforcement learning algorithms. Both the benchmark and reference implementations are released at https://github.com/rllab/rllab in order to facilitate experimental reproducibility and to encourage adoption by other researchers. Yan Duan, Xi Chen 0022, Rein Houthooft, John Schulman, Pieter Abbeel	ICML	5
2016	Guided Cost Learning: Deep Inverse Optimal Control via Policy Optimization abstract Reinforcement learning can acquire complex behaviors from high-level specifications. However, defining a cost function that can be optimized effectively and encodes the correct task is challenging in practice. We explore how inverse optimal control (IOC) can be used to learn behaviors from demonstrations, with applications to torque control of high-dimensional robotic systems. Our method addresses two key challenges in inverse optimal control: first, the need for informative features and effective regularization to impose structure on the cost, and second, the difficulty of learning the cost function under unknown dynamics for high-dimensional continuous systems. To address the former challenge, we present an algorithm capable of learning arbitrary nonlinear cost functions, such as neural networks, without meticulous feature engineering. To address the latter challenge, we formulate an efficient sample-based approximation for MaxEnt IOC. We evaluate our method on a series of simulated tasks and real-world robotic manipulation problems, demonstrating substantial improvement over prior methods both in terms of task complexity and sample efficiency. Chelsea Finn, Sergey Levine, Pieter Abbeel	ICML	3
2016	Guided search for task and motion plans using learned heuristics abstract Tasks in mobile manipulation planning often require thousands of individual motions to complete. Such tasks require reasoning about complex goals as well as the feasibility of movements in configuration space. In discrete representations, planning complexity is exponential in the length of the plan. In mobile manipulation, parameters for an action often draw from a continuous space, so we must also cope with an infinite branching factor. Task and motion planning (TAMP) methods integrate logical search over high-level actions with geometric reasoning to address this challenge. We present an algorithm that searches the space of possible task and motion plans and uses statistical machine learning to guide the search process. Our contributions are as follows: 1) we present a complete algorithm for TAMP; 2) we present a randomized local search algorithm for plan refinement that is easily formulated as a Markov decision process (MDP); 3) we apply reinforcement learning (RL) to learn a policy for this MDP; 4) we learn from expert demonstrations to efficiently search the space of high-level task plans, given options that address different (potential) infeasibilities; and 5) we run experiments to evaluate our system in a variety of simulated domains. We show significant improvements in performance over prior work. Rohan Chitnis, Dylan Hadfield-Menell, Abhishek Gupta 0004, Siddharth Srivastava 0001, Edward Groshev, Christopher Lin, Pieter Abbeel	ICRA	7
2016	Deep spatial autoencoders for visuomotor learning abstract Reinforcement learning provides a powerful and flexible framework for automated acquisition of robotic motion skills. However, applying reinforcement learning requires a sufficiently detailed representation of the state, including the configuration of task-relevant objects. We present an approach that automates state-space construction by learning a state representation directly from camera images. Our method uses a deep spatial autoencoder to acquire a set of feature points that describe the environment for the current task, such as the positions of objects, and then learns a motion skill with these feature points using an efficient reinforcement learning method based on local linear models. The resulting controller reacts continuously to the learned feature points, allowing the robot to dynamically manipulate objects in the world with closed-loop control. We demonstrate our method with a PR2 robot on tasks that include pushing a free-standing toy block, picking up a bag of rice using a spatula, and hanging a loop of rope on a hook at various positions. In each task, our method automatically learns to track task-relevant objects and manipulate their configuration with the robot's arm. Chelsea Finn, Xin Yu Tan, Yan Duan, Trevor Darrell, Sergey Levine, Pieter Abbeel	ICRA	6
2016	Combining model-based policy search with online model learning for control of physical humanoids abstract We present an automatic method for interactive control of physical humanoid robots based on high-level tasks that does not require manual specification of motion trajectories or specially-designed control policies. The method is based on the combination of a model-based policy that is trained off-line in simulation and sends high-level commands to a model-free controller that executes these commands on the physical robot. This low-level controller simultaneously learns and adapts a local model of dynamics on-line and computes optimal controls under the learned model. The high-level policy is trained using a combination of trajectory optimization and neural network learning, while considering physical limitations such as limited sensors and communication delays. The entire system runs in real-time on the robot's computer and uses only on-board sensors. We demonstrate successful policy execution on a range of tasks such as leaning, hand reaching, and robust balancing behaviors atop a tilting base on the physical robot and in simulation. Igor Mordatch, Nikhil Mishra, Clemens Eppner, Pieter Abbeel	ICRA	4
2016	TSC-DL: Unsupervised trajectory segmentation of multi-modal surgical demonstrations with Deep Learning abstract The growth of robot-assisted minimally invasive surgery has led to sizable datasets of fixed-camera video and kinematic recordings of surgical subtasks. Segmentation of these trajectories into locally-similar contiguous sections can facilitate learning from demonstrations, skill assessment, and salvaging good segments from otherwise inconsistent demonstrations. Manual, or supervised, segmentation can be prone to error and impractical for large datasets. We present Transition State Clustering with Deep Learning (TSC-DL), a new unsupervised algorithm that leverages video and kinematic data for task-level segmentation, and finds regions of the visual feature space that correlate with transition events using features constructed from layers of pre-trained image classification Deep Convolutional Neural Networks (CNNs). We report results on three datasets comparing Deep Learning architectures (AlexNet and VGG), choice of convolutional layer, dimensionality reduction techniques, visual encoding, and the use of Scale Invariant Feature Transforms (SIFT). We find that the deep architectures extract features that result in up-to a 30.4% improvement in Silhouette Score (a measure of cluster tightness) over the traditional “shallow” features from SIFT. We also present cases where TSC-DL discovers human annotator omissions. Supplementary material, data and code is available at: http://berkeleyautomation.github.io/tsc-dl/. Adithyavairavan Murali, Animesh Garg, Sanjay Krishnan, Florian T. Pokorny, Pieter Abbeel, Trevor Darrell, Kenneth Y. Goldberg	ICRA	5
2016	Model-based reinforcement learning with parametrized physical models and optimism-driven exploration abstract In this paper, we present a robotic model-based reinforcement learning method that combines ideas from model identification and model predictive control. We use a feature-based representation of the dynamics that allows the dynamics model to be fitted with a simple least squares procedure, and the features are identified from a high-level specification of the robot's morphology, consisting of the number and connectivity structure of its links. Model predictive control is then used to choose the actions under an optimistic model of the dynamics, which produces an efficient and goal-directed exploration strategy. We present real time experimental results on standard benchmark problems involving the pendulum, cartpole, and double pendulum systems. Experiments indicate that our method is able to learn a range of benchmark tasks substantially faster than the previous best methods. To evaluate our approach on a realistic robotic control task, we also demonstrate real time control of a simulated 7 degree of freedom arm. Christopher Xie, Sachin Patil, Teodor Mihai Moldovan, Sergey Levine, Pieter Abbeel	ICRA	5
2016	Learning deep control policies for autonomous aerial vehicles with MPC-guided policy search abstract Model predictive control (MPC) is an effective method for controlling robotic systems, particularly autonomous aerial vehicles such as quadcopters. However, application of MPC can be computationally demanding, and typically requires estimating the state of the system, which can be challenging in complex, unstructured environments. Reinforcement learning can in principle forego the need for explicit state estimation and acquire a policy that directly maps sensor readings to actions, but is difficult to apply to unstable systems that are liable to fail catastrophically during training before an effective policy has been found. We propose to combine MPC with reinforcement learning in the framework of guided policy search, where MPC is used to generate data at training time, under full state observations provided by an instrumented training environment. This data is used to train a deep neural network policy, which is allowed to access only the raw observations from the vehicle's onboard sensors. After training, the neural network policy can successfully control the robot without knowledge of the full state, and at a fraction of the computational cost of MPC. We evaluate our method by learning obstacle avoidance policies for a simulated quadrotor, using simulated onboard sensors and no explicit state estimation at test time. Tianhao Zhang 0001, Gregory Kahn, Sergey Levine, Pieter Abbeel	ICRA	4
2016	Learning deep neural network policies with continuous memory states abstract Policy learning for partially observed control tasks requires policies that can remember salient information from past observations. In this paper, we present a method for learning policies with internal memory for high-dimensional, continuous systems, such as robotic manipulators. Our approach consists of augmenting the state and action space of the system with continuous-valued memory states that the policy can read from and write to. Learning general-purpose policies with this type of memory representation directly is difficult, because the policy must automatically figure out the most salient information to memorize at each time step. We show that, by decomposing this policy search problem into a trajectory optimization phase and a supervised learning phase through a method called guided policy search, we can acquire policies with effective memorization and recall strategies. Intuitively, the trajectory optimization phase chooses the values of the memory states that will make it easier for the policy to produce the right action in future states, while the supervised learning phase encourages the policy to use memorization actions to produce those memory states. We evaluate our method on tasks involving continuous control in manipulation and navigation settings, and show that our method can learn complex policies that successfully complete a range of tasks that require memory. Marvin Zhang, Zoe McCarthy, Chelsea Finn, Sergey Levine, Pieter Abbeel	ICRA	5
2016	One-shot learning of manipulation skills with online dynamics adaptation and neural network priors abstract One of the key challenges in applying reinforcement learning to complex robotic control tasks is the need to gather large amounts of experience in order to find an effective policy for the task at hand. Model-based reinforcement learning can achieve good sample efficiency, but requires the ability to learn a model of the dynamics that is good enough to learn an effective policy. In this work, we develop a model-based reinforcement learning algorithm that combines prior knowledge from previous tasks with online adaptation of the dynamics model. These two ingredients enable highly sample-efficient learning even in regimes where estimating the true dynamics is very difficult, since the online model adaptation allows the method to locally compensate for unmodeled variation in the dynamics. We encode the prior experience into a neural network dynamics model, adapt it online by progressively refitting a local linear model of the dynamics, and use model predictive control to plan under these dynamics. Our experimental results show that this approach can be used to solve a variety of complex robotic manipulation tasks in just a single attempt, using prior data from other manipulation behaviors. Justin Fu, Sergey Levine, Pieter Abbeel	IROS	3
2016	Learning dexterous manipulation for a soft robotic hand from human demonstrations abstract Dexterous multi-fingered hands can accomplish fine manipulation behaviors that are infeasible with simple robotic grippers. However, sophisticated multi-fingered hands are often expensive and fragile. Low-cost soft hands offer an appealing alternative to more conventional devices, but present considerable challenges in sensing and actuation, making them difficult to apply to more complex manipulation tasks. In this paper, we describe an approach to learning from demonstration that can be used to train soft robotic hands to perform dexterous manipulation tasks. Our method uses object-centric demonstrations, where a human demonstrates the desired motion of manipulated objects with their own hands, and the robot autonomously learns to imitate these demonstrations using reinforcement learning. We propose a novel algorithm that allows us to blend and select a subset of the most feasible demonstrations, which we use with an extension of the guided policy search framework that learns generalizable neural network policies. We demonstrate our approach on the RBO Hand 2, with learned motor skills for turning a valve, manipulating an abacus, and grasping. Abhishek Gupta 0004, Clemens Eppner, Sergey Levine, Pieter Abbeel	IROS	4
2016	Sequential quadratic programming for task plan optimization abstract We consider the problem of refining an abstract task plan into a motion trajectory. Task and motion planning is a hard problem that is essential to long-horizon mobile manipulation. Many approaches divide the problem into two steps: a search for a task plan and task plan refinement to find a feasible trajectory. We apply sequential quadratic programming to jointly optimize over the parameters in a task plan (e.g., trajectories, grasps, put down locations). We provide two modifications that make our formulation more suitable to task and motion planning. We show how to use movement primitives to reuse previous solutions (and so save optimization effort) without trapping the algorithm in a poor basin of attraction. We also derive an early convergence criterion that lets us quickly detect unsatisfiable constraints so we can re-initialize their variables. We present experiments in a navigation amongst movable objects domain and show substantial improvement in cost over a backtracking refinement algorithm. Dylan Hadfield-Menell, Christopher Lin, Rohan Chitnis, Stuart Russell 0001, Pieter Abbeel	IROS	5
2016	Occlusion-aware multi-robot 3D tracking abstract We introduce an optimization-based control approach that enables a team of robots to cooperatively track a target using onboard sensing. In this setting, the robots are required to estimate their own positions as well as concurrently track the target. Our probabilistic method generates controls that minimize the expected uncertainty of the target. Additionally, our method efficiently reasons about occlusions between robots and takes them into account for the control generation. We evaluate our approach in a number of experiments in which we simulate a team of quadrotor robots flying in three-dimensional space to track a moving target on the ground. We compare our method to other state-of-the-art approaches represented by the random sampling technique, lattice planning method, and our previous method. Our experimental results indicate that our method achieves up to 8 times smaller maximum tracking error and up to 2 times smaller average tracking error than the next best approach in the presented scenarios. Karol Hausman, Gregory Kahn, Sachin Patil, Jörg Müller 0004, Kenneth Y. Goldberg, Pieter Abbeel, Gaurav S. Sukhatme	IROS	6
2016	InfoGAN: Interpretable Representation Learning by Information Maximizing Generative Adversarial Nets abstract This paper describes InfoGAN, an information-theoretic extension to the Generative Adversarial Network that is able to learn disentangled representations in a completely unsupervised manner. InfoGAN is a generative adversarial network that also maximizes the mutual information between a small subset of the latent variables and the observation. We derive a lower bound to the mutual information objective that can be optimized efficiently, and show that our training procedure can be interpreted as a variation of the Wake-Sleep algorithm. Specifically, InfoGAN successfully disentangles writing styles from digit shapes on the MNIST dataset, pose from lighting of 3D rendered images, and background digits from the central digit on the SVHN dataset. It also discovers visual concepts that include hair styles, presence/absence of eyeglasses, and emotions on the CelebA face dataset. Experiments show that InfoGAN learns interpretable representations that are competitive with representations learned by existing fully supervised methods. Xi Chen 0022, Yan Duan, Rein Houthooft, John Schulman, Ilya Sutskever, Pieter Abbeel	NIPS	6
2016	Backprop KF: Learning Discriminative Deterministic State Estimators abstract Generative state estimators based on probabilistic filters and smoothers are one of the most popular classes of state estimators for robots and autonomous vehicles. However, generative models have limited capacity to handle rich sensory observations, such as camera images, since they must model the entire distribution over sensor readings. Discriminative models do not suffer from this limitation, but are typically more complex to train as latent variable models for state estimation. We present an alternative approach where the parameters of the latent state distribution are directly optimized as a deterministic computation graph, resulting in a simple and effective gradient descent algorithm for training discriminative state estimators. We show that this procedure can be used to train state estimators that use complex input, such as raw camera images, which must be processed using expressive nonlinear function approximators such as convolutional neural networks. Our model can be viewed as a type of recurrent neural network, and the connection to probabilistic filtering allows us to design a network architecture that is particularly well suited for state estimation. We evaluate our approach on synthetic tracking task with raw image inputs and on the visual odometry task in the KITTI dataset. The results show significant improvement over both standard generative approaches and regular recurrent neural networks. Tuomas Haarnoja, Anurag Ajay, Sergey Levine, Pieter Abbeel	NIPS	4
2016	Cooperative Inverse Reinforcement Learning abstract For an autonomous system to be helpful to humans and to pose no unwarranted risks, it needs to align its values with those of the humans in its environment in such a way that its actions contribute to the maximization of value for the humans. We propose a formal definition of the value alignment problem as cooperative inverse reinforcement learning (CIRL). A CIRL problem is a cooperative, partial- information game with two agents, human and robot; both are rewarded according to the human’s reward function, but the robot does not initially know what this is. In contrast to classical IRL, where the human is assumed to act optimally in isolation, optimal CIRL solutions produce behaviors such as active teaching, active learning, and communicative actions that are more effective in achieving value alignment. We show that computing optimal joint policies in CIRL games can be reduced to solving a POMDP, prove that optimality in isolation is suboptimal in CIRL, and derive an approximate CIRL algorithm. Dylan Hadfield-Menell, Stuart Russell 0001, Pieter Abbeel, Anca D. Dragan	NIPS	3
2016	VIME: Variational Information Maximizing Exploration abstract Scalable and effective exploration remains a key challenge in reinforcement learning (RL). While there are methods with optimality guarantees in the setting of discrete state and action spaces, these methods cannot be applied in high-dimensional deep RL scenarios. As such, most contemporary RL relies on simple heuristics such as epsilon-greedy exploration or adding Gaussian noise to the controls. This paper introduces Variational Information Maximizing Exploration (VIME), an exploration strategy based on maximization of information gain about the agent's belief of environment dynamics. We propose a practical implementation, using variational inference in Bayesian neural networks which efficiently handles continuous state and action spaces. VIME modifies the MDP reward function, and can be applied with several different underlying RL algorithms. We demonstrate that VIME achieves significantly better performance compared to heuristic exploration methods across a variety of continuous control tasks and algorithms, including tasks with very sparse rewards. Rein Houthooft, Xi Chen 0022, Yan Duan, John Schulman, Filip De Turck, Pieter Abbeel	NIPS	6
2016	Combinatorial Energy Learning for Image Segmentation abstract We introduce a new machine learning approach for image segmentation that uses a neural network to model the conditional energy of a segmentation given an image. Our approach, combinatorial energy learning for image segmentation (CELIS) places a particular emphasis on modeling the inherent combinatorial nature of dense image segmentation problems. We propose efficient algorithms for learning deep neural networks to model the energy function, and for local optimization of this energy in the space of supervoxel agglomerations. We extensively evaluate our method on a publicly available 3-D microscopy dataset with 25 billion voxels of ground truth data. On an 11 billion voxel test set, we find that our method improves volumetric reconstruction accuracy by more than 20% as compared to two state-of-the-art baseline methods: graph-based segmentation of the output of a 3-D convolutional neural network trained to predict boundaries, as well as a random forest classifier trained to agglomerate supervoxels that were generated by a 3-D convolutional neural network. Jeremy Maitin-Shepard, Viren Jain, Michal Januszewski, Peter Li, Pieter Abbeel	NIPS	5
2016	Value Iteration Networks abstract We introduce the value iteration network (VIN): a fully differentiable neural network with a `planning module' embedded within. VINs can learn to plan, and are suitable for predicting outcomes that involve planning-based reasoning, such as policies for reinforcement learning. Key to our approach is a novel differentiable approximation of the value-iteration algorithm, which can be represented as a convolutional neural network, and trained end-to-end using standard backpropagation. We evaluate VIN based policies on discrete and continuous path-planning domains, and on a natural-language based search task. We show that by learning an explicit planning computation, VIN policies generalize better to new, unseen domains. Aviv Tamar, Sergey Levine, Pieter Abbeel, Yi Wu 0013, Garrett Thomas	NIPS	3
2016	Adapting Deep Visuomotor Representations with Weak Pairwise Constraints Eric Tzeng, Coline Devin, Judy Hoffman, Chelsea Finn, Pieter Abbeel, Sergey Levine, Kate Saenko, Trevor Darrell	WAFR	5
2016	End-to-End Training of Deep Visuomotor Policies abstract Policy search methods can allow robots to learn control policies for a wide range of tasks, but practical applications of policy search often require hand-engineered components for perception, state estimation, and low-level control. In this paper, we aim to answer the following question: does training the perception and control systems jointly end-to-end provide better performance than training each component separately? To this end, we develop a method that can be used to learn policies that map raw image observations directly to torques at the robot's motors. The policies are represented by deep convolutional neural networks (CNNs) with 92,000 parameters, and are trained using a guided policy search method, which transforms policy search into supervised learning, with supervision provided by a simple trajectory-centric reinforcement learning method. We evaluate our method on a range of real-world manipulation tasks that require close coordination between vision and control, such as screwing a cap onto a bottle, and present simulated comparisons to a range of prior policy search methods. Sergey Levine, Chelsea Finn, Trevor Darrell, Pieter Abbeel	J. Mach. Learn. Res.	4
2015	Tractability of Planning with Loops abstract We create a unified framework for analyzing and synthesizing plans with loops for solving problems with non-deterministic numeric effects and a limited form of partial observability. Three different action models---with deterministic, qualitative non-deterministic and Boolean non-deterministic semantics---are handled using a single abstract representation. We establish the conditions under which the correctness and termination of solutions, represented as abstract policies, can be verified. We also examine the feasibility of learning abstract policies from examples. We demonstrate our techniques on several planning problems and show that they apply to challenging real-world tasks such as doing the laundry with a PR2 robot. These results resolve a number of open questions about planning with loops and facilitate the development of new algorithms and applications. Siddharth Srivastava 0001, Shlomo Zilberstein, Abhishek Gupta 0004, Pieter Abbeel, Stuart Russell 0001	AAAI	4
2015	Alpha-Beta Divergences Discover Micro and Macro Structures in Data abstract Although recent work in non-linear dimensionality reduction investigates multiple choices of divergence measure during optimization \citeyang2013icml,bunte2012neuro, little work discusses the direct effects that divergence measures have on visualization. We study this relationship, theoretically and through an empirical analysis over 10 datasets. Our works shows how the αand βparameters of the generalized alpha-beta divergence can be chosen to discover hidden macro-structures (categories, e.g. birds) or micro-structures (fine-grained classes, e.g. toucans). Our method, which generalizes t-SNE \citetsne, allows us to discover such structure without extensive grid searches over (α, β) due to our theoretical analysis: such structure is apparent with particular choices of (α, β) that generalize across datasets. We also discuss efficient parallel CPU and GPU schemes which are non-trivial due to the tree-structures employed in optimization and the large datasets that do not fully fit into GPU memory. Our method runs 20x faster than the fastest published code \citefmm. We conclude with detailed case studies on the following very large datasets: ILSVRC 2012, a standard computer vision dataset with 1.2M images; SUSY, a particle physics dataset with 5M instances; and HIGGS, another particle physics dataset with 11M instances. This represents the largest published visualization attained by SNE methods. We have open-sourced our visualization code: \texttthttp://rll.berkeley.edu/absne/. Karthik S. Narayan, Ali Punjani, Pieter Abbeel	ICML	3
2015	Trust Region Policy Optimization abstract In this article, we describe a method for optimizing control policies, with guaranteed monotonic improvement. By making several approximations to the theoretically-justified scheme, we develop a practical algorithm, called Trust Region Policy Optimization (TRPO). This algorithm is effective for optimizing large nonlinear policies such as neural networks. Our experiments demonstrate its robust performance on a wide variety of tasks: learning simulated robotic swimming, hopping, and walking gaits; and playing Atari games using images of the screen as input. Despite its approximations that deviate from the theory, TRPO tends to give monotonic improvement, with little tuning of hyperparameters. John Schulman, Sergey Levine, Pieter Abbeel, Michael I. Jordan, Philipp Moritz	ICML	3
2015	Beyond lowest-warping cost action selection in trajectory transfer abstract We consider the problem of learning from demonstrations to manipulate deformable objects. Recent work [1], [2], [3] has shown promising results that enable robotic manipulation of deformable objects through learning from demonstrations. Their approach is able to generalize from a single demonstration to new test situations, and suggests a nearest neighbor approach to select a demonstration to adapt to a given test situation. Such a nearest neighbor approach, however, ignores important aspects of the problem: brittleness (versus robustness) of demonstrations when generalized through this process, and the extent to which a demonstration makes progress towards a goal. In this paper, we frame the problem of selecting which demonstration to transfer as an options Markov decision process (MDP). We present max-margin Q-function estimation: an approach to learn a Q-function from expert demonstrations. Our learned policies account for variability in robustness of demonstrations and the sequential nature of our tasks. We developed two knot-tying benchmarks to experimentally validate the effectiveness of our proposed approach. The selection strategy described in [2] achieves success rates of 70% and 54%, respectively. Our approach performs significantly better, with success rates of 88% and 76%, respectively. Dylan Hadfield-Menell, Alex X. Lee, Chelsea Finn, Eric Tzeng, Sandy H. Huang, Pieter Abbeel	ICRA	6
2015	Active exploration using trajectory optimization for robotic grasping in the presence of occlusions abstract We consider the task of actively exploring unstructured environments to facilitate robotic grasping of occluded objects. Typically, the geometry and locations of these objects are not known a priori. We mount an RGB-D sensor on the robot gripper to maintain a 3D voxel map of the environment during exploration. The objective is to plan the motion of the sensor in order to search for feasible grasp handles that lie within occluded regions of the map. In contrast to prior work that generates exploration trajectories by sampling, we directly optimize the exploration trajectory to find grasp handles. Since it is challenging to optimize over the discrete voxel map, we encode the uncertainty of the positions of the occluded grasp handles as a mixture of Gaussians, one per occluded region. Our trajectory optimization approach encourages exploration by penalizing a measure of the uncertainty. We then plan a collision-free trajectory for the robot arm to the detected grasp handle. We evaluated our approach by actively exploring and attempting 300 grasps. Our experiments suggest that compared to the baseline method of sampling 10 trajectories, which successfully grasped 58% of the objects, our active exploration formulation with trajectory optimization successfully grasped 93% of the objects, was 1.3× faster, and had 3.2× fewer failed grasp attempts. Gregory Kahn, Peter Sujan, Sachin Patil, Shaunak Dattaprasad Bopardikar, Julian Ryde, Kenneth Y. Goldberg, Pieter Abbeel	ICRA	7
2015	Physics-based trajectory optimization for grasping in cluttered environments abstract Grasping an object in a cluttered, unorganized environment is challenging because of unavoidable contacts and interactions between the robot and multiple immovable (static) and movable (dynamic) obstacles in the environment. Planning an approach trajectory for grasping in such situations can benefit from physics-based simulations that describe the dynamics of the interaction between the robot manipulator and the environment. In this work, we present a physics-based trajectory optimization approach for planning grasp approach trajectories. We present novel cost objectives and identify failure modes relevant to grasping in cluttered environments. Our approach uses rollouts of physics-based simulations to compute the gradient of the objective and of the dynamics. Our approach naturally generates behaviors such as choosing to push objects that are less likely to topple over, recognizing and avoiding situations which might cause a cascade of objects to fall over, and adjusting the manipulator trajectory to push objects aside in a direction orthogonal to the grasping direction. We present results in simulation for grasping in a variety of cluttered environments with varying levels of density of obstacles in the environment. Our experiments in simulation indicate that our approach outperforms a baseline approach that considers multiple straight-line trajectories modified to account for static obstacles by an aggregate success rate of 14% with varying degrees of object clutter. Nikita Kitaev, Igor Mordatch, Sachin Patil, Pieter Abbeel	ICRA	4
2015	A non-rigid point and normal registration algorithm with applications to learning from demonstrations abstract Recent work [1], [2], [3] has shown promising results in learning from demonstrations for the manipulation of deformable objects. Their approach finds a non-rigid registration between points in the demonstration scene and points in the test scene. This registration is then extrapolated and applied to the gripper motions in the demonstration scene to obtain the gripper motions for the test scene. If more than one demonstration is available, a quality score for the non-rigid registration is used to determine the best matching training scene. For many manipulation tasks, however, the gripper's direction of approach with respect to the objects' surface normals is important in order to succeed at the task. This prior work only registers points across scenes and does not register the surface normals, often leading to warps between scenes that are inappropriate for transfer of manipulation primitives. The main contributions of this paper are (i) An algorithm for non-rigid registration that considers both points and normals, and (ii) An evaluation of this registration approach in the context of learning from demonstrations for robotic manipulation. Our experiments, which consider an insertion task in simulation and also knot-tying and towel-folding executions in a PR2, show that incorporating normals results in improved performance and qualitatively better grasps. Alex X. Lee, Max A. Goldstein, Shane T. Barratt, Pieter Abbeel	ICRA	4
2015	Learning force-based manipulation of deformable objects from multiple demonstrations abstract Manipulation of deformable objects often requires a robot to apply specific forces to bring the object into the desired configuration. For instance, tightening a knot requires pulling on the ends, flattening an article of clothing requires smoothing out wrinkles, and erasing a whiteboard requires applying downward pressure. We present a method for learning force-based manipulation skills from demonstrations. Our approach uses non-rigid registration to compute a warping function that transforms both the end-effector poses and forces in each demonstration into the current scene, based on the configuration of the object. Our method then uses the variation between the demonstrations to extract a single trajectory, along with time-varying feedback gains that determine how much to match poses or forces. This results in a learned variable-impedance control strategy that trades off force and position errors, providing for the right level of compliance that applies the necessary forces at each stage of the motion. We evaluate our approach by tying knots in rope, flattening towels, and erasing a whiteboard. Alex X. Lee, Henry Lu, Abhishek Gupta 0004, Sergey Levine, Pieter Abbeel	ICRA	5
2015	Learning contact-rich manipulation skills with guided policy search abstract Autonomous learning of object manipulation skills can enable robots to acquire rich behavioral repertoires that scale to the variety of objects found in the real world. However, current motion skill learning methods typically restrict the behavior to a compact, low-dimensional representation, limiting its expressiveness and generality. In this paper, we extend a recently developed policy search method [1] and use it to learn a range of dynamic manipulation behaviors with highly general policy representations, without using known models or example demonstrations. Our approach learns a set of trajectories for the desired motion skill by using iteratively refitted time-varying linear models, and then unifies these trajectories into a single control policy that can generalize to new situations. To enable this method to run on a real robot, we introduce several improvements that reduce the sample count and automate parameter selection. We show that our method can acquire fast, fluent behaviors after only minutes of interaction time, and can learn robust controllers for complex tasks, including putting together a toy airplane, stacking tight-fitting lego blocks, placing wooden rings onto tight-fitting pegs, inserting a shoe tree into a shoe, and screwing bottle caps onto bottles. Sergey Levine, Nolan Wagener, Pieter Abbeel	ICRA	3
2015	GP-GPIS-OPT: Grasp planning with shape uncertainty using Gaussian process implicit surfaces and Sequential Convex Programming abstract Computing grasps for an object is challenging when the object geometry is not known precisely. In this paper, we explore the use of Gaussian process implicit surfaces (GPISs) to represent shape uncertainty from RGBD point cloud observations of objects. We study the use of GPIS representations to select grasps on previously unknown objects, measuring grasp quality by the probability of force closure. Our main contribution is GP-GPIS-OPT, an algorithm for computing grasps for parallel-jaw grippers on 2D GPIS object representations. Specifically, our method optimizes an approximation to the probability of force closure subject to antipodal constraints on the parallel jaws using Sequential Convex Programming (SCP). We also introduce GPIS-Blur, a method for visualizing 2D GPIS models based on blending shape samples from a GPIS. We test the algorithm on a set of 8 planar objects with transparency, translucency, and specularity. Our experiments suggest that GP-GPIS-OPT computes grasps with higher probability of force closure than a planner that does not consider shape uncertainty on our test objects and may converge to a grasp plan up to 5.7×faster than using Monte-Carlo integration, a common method for grasp planning under shape uncertainty. Furthermore, initial experiments on the Willow Garage PR2 robot suggest that grasps selected with GP-GPIS-OPT are up to 90% more successful than those planned assuming a deterministic shape. Our dataset, code, and videos of our experiments are available at http://rll.berkeley.edu/icra2015grasping/. Jeffrey Mahler, Sachin Patil, Ben Kehoe, Jur P. van den Berg, Matei T. Ciocarlie, Pieter Abbeel, Kenneth Y. Goldberg	ICRA	6
2015	Optimism-driven exploration for nonlinear systems abstract Tasks with unknown dynamics and costly system interaction time present a serious challenge for reinforcement learning. If a model of the dynamics can be learned quickly, interaction time can be reduced substantially. We show that combining an optimistic exploration strategy with model-predictive control can achieve very good sample complexity for a range of nonlinear systems. Our method learns a Dirichlet process mixture of linear models using an exploration strategy based on optimism in the face of uncertainty. Trajectory optimization is used to plan paths in the learned model that both minimize the cost and perform exploration. Experimental results show that our approach achieves some of the most sample-efficient learning rates on several benchmark problems, and is able to successfully learn to control a simulated helicopter during hover and autorotation with only seconds of interaction time. The computational requirements are substantial. Teodor Mihai Moldovan, Sergey Levine, Michael I. Jordan, Pieter Abbeel	ICRA	4
2015	Learning by observation for surgical subtasks: Multilateral cutting of 3D viscoelastic and 2D Orthotropic Tissue Phantoms abstract Automating repetitive surgical subtasks such as suturing, cutting and debridement can reduce surgeon fatigue and procedure times and facilitate supervised tele-surgery. Programming is difficult because human tissue is deformable and highly specular. Using the da Vinci Research Kit (DVRK) robotic surgical assistant, we explore a “Learning By Observation” (LBO) approach where we identify, segment, and parameterize motion sequences and sensor conditions to build a finite state machine (FSM) for each subtask. The robot then executes the FSM repeatedly to tune parameters and if necessary update the FSM structure. We evaluate the approach on two surgical subtasks: debridement of 3D Viscoelastic Tissue Phantoms (3d-DVTP), in which small target fragments are removed from a 3D viscoelastic tissue phantom; and Pattern Cutting of 2D Orthotropic Tissue Phantoms (2d-PCOTP), a step in the standard Fundamentals of Laparoscopic Surgery training suite, in which a specified circular area must be cut from a sheet of orthotropic tissue phantom. We describe the approach and physical experiments with repeatability of 96% for 50 trials of the 3d-DVTP subtask and 70% for 20 trials of the 2d-PCOTP subtask. A video is available at: http://j.mp/Robot-Surgery-Video-Oct-2014. Adithyavairavan Murali, Siddarth Sen, Ben Kehoe, Animesh Garg, Seth McFarland, Sachin Patil, W. Douglas Boyd, Susan Lim, Pieter Abbeel, Kenneth Y. Goldberg	ICRA	9
2015	Range sensor and silhouette fusion for high-quality 3D Scanning abstract We consider the problem of building high-quality 3D object models from commodity RGB and depth sensors. Applications of such a database include instance and object recognition, robot grasping, virtual reality, graphics, and online shopping. Unfortunately, modern reconstruction approaches have difficulties in reconstructing objects with major transparencies (e.g., KinectFusion [22]) and/or concavities (e.g., visual hull). This paper presents a method to fuse visual hull information from off-the-shelf RGB cameras and KinectFusion cues from commodity depth sensors to produce models that are substantially better than either approach on its own. Extensive experiments on the recently published BigBIRD dataset [25] demonstrate that our reconstructions recover more accurate shape and detail than competing approaches, particularly on challenging objects with transparencies and/or concavities. Quantitative evaluations indicate that our approach consistently outperforms competing methods and achieves under 2 mm RMS error. We plan to release our code after the review process. Karthik S. Narayan, James Sha, Pieter Abbeel	ICRA	4
2015	Deep learning helicopter dynamics models abstract We consider the problem of system identification of helicopter dynamics. Helicopters are complex systems, coupling rigid body dynamics with aerodynamics, engine dynamics, vibration, and other phenomena. Resultantly, they pose a challenging system identification problem, especially when considering non-stationary flight regimes. We pose the dynamics modeling problem as direct high-dimensional regression, and take inspiration from recent results in Deep Learning to represent the helicopter dynamics with a Rectified Linear Unit (ReLU) Network Model, a hierarchical neural network model. We provide a simple method for initializing the parameters of the model, and optimization details for training. We describe three baseline models and show that they are significantly outperformed by the ReLU Network Model in experiments on real data, indicating the power of the model to capture useful structure in system dynamics across a rich array of aerobatic maneuvers. Specifically, the ReLU Network Model improves 58% overall in RMS acceleration prediction over state-of-the-art methods. Predicting acceleration along the helicopter's up-down axis is empirically found to be the most difficult, and the ReLU Network Model improves by 60% over the prior state-of-the-art. We discuss explanations of these performance gains, and also investigate the impact of hyperparameters in the novel model. Ali Punjani, Pieter Abbeel	ICRA	2
2015	Toward asymptotically optimal motion planning for kinodynamic systems using a two-point boundary value problem solver abstract We present an approach for asymptotically optimal motion planning for kinodynamic systems with arbitrary nonlinear dynamics amid obstacles. Optimal sampling-based planners like RRT, FMT, and BIT* when applied to kinodynamic systems require solving a two-point boundary value problem (BVP) to perform exact connections between nodes in the tree. Two-point BVPs are non-trivial to solve, hence the prevalence of alternative approaches that focus on specific instances of kinodynamic systems, use approximate solutions to the two-point BVP, or use random propagation of controls. In this work, we explore the feasibility of exploiting recent advances in numerical optimal control and optimization to solve these two-point BVPs for arbitrary kinodynamic systems and how they can be integrated with existing optimal planning algorithms. We combine BIT* with a two-point BVP solver that uses sequential quadratic programming (SQP). We consider the problem of computing minimum-time trajectories. Since the duration of trajectories is not known a-priori, we include the time-step as part of the optimization to allow SQP to optimize over the duration of the trajectory while keeping the number of discrete steps fixed for every connection attempted. Our experiments indicate that using a two-point BVP solver in the inner-loop of BIT* is competitive with the state-of-the-art in sampling-based optimal planning that explicitly avoids the use of two-point BVP solvers. Christopher Xie, Jur P. van den Berg, Sachin Patil, Pieter Abbeel	ICRA	4
2015	Modular task and motion planning in belief space abstract The execution of long-horizon tasks under uncertainty is a fundamental challenge in robotics. Recent approaches have made headway on these tasks with an integration of task and motion planning. In this paper, we present Interfaced Belief Space Planning (IBSP): a modular approach to task and motion planning in belief space. We use a task-independent interface layer to combine an off-the-shelf classical planner with motion planning and inference. We determinize the problem under the maximum likelihood observation assumption to obtain a deterministic representation where successful plans generate goal-directed observations. We leverage properties of maximum likelihood observation determinizations to obtain a simple representation of (optimistic) belief space dynamics that is well-suited to planning. Our interface is implemented with standard belief state queries, requiring only the ability to sample, compute unnormalized likelihoods, and compute maximum likelihood states. Our contribution is a novel algorithm for task and motion planning in belief space that has minimal dependence on the details of the inference engine used. IBSP can work with a broad class of black box state estimators, with zero changes to the algorithm. We validate our approach in simulated tasks for the PR2 that account for continuous state, different types of initial state distributions, and negative observations. Dylan Hadfield-Menell, Edward Groshev, Rohan Chitnis, Pieter Abbeel	IROS	4
2015	Learning compound multi-step controllers under unknown dynamics abstract Applications of reinforcement learning for robotic manipulation often assume an episodic setting. However, controllers trained with reinforcement learning are often situated in the context of a more complex compound task, where multiple controllers might be invoked in sequence to accomplish a higher-level goal. Furthermore, training such controllers typically requires resetting the environment between episodes, which is typically handled manually. We describe an approach for training chains of controllers with reinforcement learning. This requires taking into account the state distributions induced by preceding controllers in the chain, as well as automatically training reset controllers that can reset the task between episodes. The initial state of each controller is determined by the controller that precedes it, resulting in a non-stationary learning problem. We demonstrate that a recently developed method that optimizes linear-Gaussian controllers under learned local linear models can tackle this sort of non-stationary problem, and that training controllers concurrently with a corresponding reset controller only minimally increases training time. We also demonstrate this method on a complex tool use task that consists of seven stages and requires using a toy wrench to screw in a bolt. This compound task requires grasping and handling complex contact dynamics. After training, the controllers can execute the entire task quickly and efficiently. Finally, we show that this method can be combined with guided policy search to automatically train nonlinear neural network controllers for a grasping task with considerable variation in target position. Weiqiao Han, Sergey Levine, Pieter Abbeel	IROS	3
2015	Leveraging appearance priors in non-rigid registration, with application to manipulation of deformable objects abstract Manipulation of deformable objects is a widely applicable but challenging task in robotics. One promising nonparametric approach for this problem is trajectory transfer, in which a non-rigid registration is computed between the starting scene of the demonstration and the scene at test time. This registration is extrapolated to find a function from ℝ3to ℝ3, which is then used to warp the demonstrated robot trajectory to generate a proposed trajectory to execute in the test scene. In prior work [1] [2], only depth information from the scenes has been used to compute this warp function. This approach ignores appearance information, but there are situations in which using both shape and appearance information is necessary for finding high quality non-rigid warp functions. In this paper, we describe an approach to learn relevant appearance information about deformable objects using deep learning, and use this additional information to improve the quality of non-rigid registration between demonstration and test scenes. Our method better registers areas of interest on deformable objects that are crucial for manipulation, such as rope crossings and towel corners and edges. We experimentally validate our approach in both simulation and in the real world, and show that the utilization of appearance information leads to a significant improvement in both selecting the best matching demonstration scene for a given test scene, and finding a high quality non-rigid registration between those two scenes. Sandy H. Huang, Jia Pan 0001, George Mulcaire, Pieter Abbeel	IROS	4
2015	Learning from multiple demonstrations using trajectory-aware non-rigid registration with applications to deformable object manipulation abstract Learning from demonstration by means of non-rigid point cloud registration is an effective tool for learning to manipulate a wide range of deformable objects. However, most methods that use non-rigid registration to transfer demonstrated trajectories assume that the test and demonstration scene are structurally very similar, with any variation explained by a non-linear transformation. In real-world tasks with clutter and distractor objects, this assumption is unrealistic. In this work, we show that a trajectory-aware non-rigid registration method that uses multiple demonstrations to focus the registration process on points that are relevant to the task can effectively handle significantly greater visual variation than prior methods that are not trajectory-aware. We demonstrate that this approach achieves superior generalization on several challenging tasks, including towel folding and grasping objects in a box containing irrelevant distractors. Alex X. Lee, Abhishek Gupta 0004, Henry Lu, Sergey Levine, Pieter Abbeel	IROS	5
2015	Optimized color models for high-quality 3D scanning abstract We consider the problem of estimating high-quality color models of 3D meshes, given a collection of RGB images of the original object. Applications of a database of high-quality colored meshes include object recognition in robot vision, virtual reality, graphics, and online shopping. Most modern approaches that color a 3D object model from a collection of RGB images face problems in (1) producing realistic colors for non-Lambertian surfaces and (2) seamlessly integrating colors from multiple views. Our approach efficiently solves a non-linear least squares optimization problem to jointly estimate the RGB camera poses and color model. We discover that incorporating 2D texture cues, vertex color smoothing, and texture-adaptive camera viewpoint selection into the optimization problem produces qualitatively more coherent color models than those produced by competing methods. We further introduce practical strategies to accelerate optimization. We provide extensive empirical results on the BigBIRD dataset [15], [21]: results from a user study with 133 participants indicate that on all 16 objects considered, our method outperforms competing approaches. Our code is available for download online at http://rll.berkeley.edu/iros2015colormodels. Karthik S. Narayan, Pieter Abbeel	IROS	2
2015	Transition State Clustering: Unsupervised Surgical Trajectory Segmentation for Robot Learning Sanjay Krishnan, Animesh Garg, Sachin Patil, Colin Lea, Gregory D. Hager, Pieter Abbeel, Kenneth Y. Goldberg	ISRR (2)	6
2015	Gradient Estimation Using Stochastic Computation Graphs abstract In a variety of problems originating in supervised, unsupervised, and reinforcement learning, the loss function is defined by an expectation over a collection of random variables, which might be part of a probabilistic model or the external world. Estimating the gradient of this loss function, using samples, lies at the core of gradient-based learning algorithms for these problems. We introduce the formalism of stochastic computation graphs--directed acyclic graphs that include both deterministic functions and conditional probability distributions and describe how to easily and automatically derive an unbiased estimator of the loss function's gradient. The resulting algorithm for computing the gradient estimator is a simple modification of the standard backpropagation algorithm. The generic scheme we propose unifies estimators derived in variety of prior work, along with variance-reduction techniques therein. It could assist researchers in developing intricate models involving a combination of stochastic and deterministic operations, enabling, for example, attention, memory, and control actions. John Schulman, Nicolas Heess, Theophane Weber, Pieter Abbeel	NIPS	4
2015	A Survey of Research on Cloud Robotics and Automation abstract The Cloud infrastructure and its extensive set of Internet-accessible resources has potential to provide significant benefits to robots and automation systems. We consider robots and automation systems that rely on data or code from a network to support their operation, i.e., where not all sensing, computation, and memory is integrated into a standalone system. This survey is organized around four potential benefits of the Cloud: 1) Big Data: access to libraries of images, maps, trajectories, and descriptive data; 2) Cloud Computing: access to parallel grid computing on demand for statistical analysis, learning, and motion planning; 3) Collective Robot Learning: robots sharing trajectories, control policies, and outcomes; and 4) Human Computation: use of crowdsourcing to tap human skills for analyzing images and video, classification, learning, and error recovery. The Cloud can also improve robots and automation systems by providing access to: a) datasets, publications, models, benchmarks, and simulation tools; b) open competitions for designs and systems; and c) open-source software. This survey includes over 150 references on results and open challenges. A website with new developments and updates is available at: http://goldberg.berkeley.edu/cloud-robotics/. Ben Kehoe, Sachin Patil, Pieter Abbeel, Kenneth Y. Goldberg	IEEE Trans Autom. Sci. Eng.	3
2015	Planning Curvature and Torsion Constrained Ribbons in 3D With Application to Intracavitary Brachytherapy abstract We present an approach for planning ensembles of channels, ribbons, within 3D printed implants for facilitating radiation therapy treatment of cancer. The ribbons are traced out by sweeping a constant width rigid body (cuboid) along spatial curves. We propose a method for planning multiple disjoint and mutually collision-free ribbons of finite thickness along curvature and torsion constrained curves in 3D space. This is equivalent to planning motions for the cross section of the ribbon along a spatial curve such that the cross section is oriented along the unit binormal to the curve defined according to the Frenet-Serret frame. We propose a two-stage planning approach. In the first stage, a customized sampling-based planner uses rapidly exploring random trees (RRTs) to generate feasible curvature and torsion constrained ribbons. In the second stage, the curvature and torsion along each ribbon is locally optimized using sequential quadratic programming (SQP). We use this approach to design curved radiation delivery channels inside custom 3D printed implants that allow temporary insertion of a high-dose radioactive source that is threaded through the channels using a wire and allowed to dwell for specified times to expose cancerous tumors for intracavitary brachytherapy treatment. Constraints on the curvature and torsion are required for 3D printing (to allow flushing of sacrificial material) and for smooth insertion of radioactive sources. In simulation experiments, this approach achieves an improvement of 46% in tumor coverage compared with a greedy approach that generates channels sequentially. Sachin Patil, Jia Pan 0001, Pieter Abbeel, Kenneth Y. Goldberg	IEEE Trans Autom. Sci. Eng.	3
2014	Optimization-Based Artifact Correction for Electron Microscopy Image Stacks Samaneh Azadi, Jeremy Maitin-Shepard, Pieter Abbeel	ECCV (2)	3
2014	Planning locally optimal, curvature-constrained trajectories in 3D using sequential convex optimization abstract 3D curvature-constrained motion planning finds applications in a wide variety of domains, including motion planning for flexible, bevel-tip medical needles, planning curvature-constrained channels in 3D printed implants for targeted brachytherapy dose delivery or channels for cooling turbine blades, and path planning for unmanned aerial vehicles (UAVs). In this work, we present a motion planning technique using sequential convex optimization for computing locally optimal, curvature-constrained trajectories to desired targets while avoiding obstacles in 3D environments. We report two main contributions in this work: (i) curvature-constrained trajectory optimization in 6D pose (position and orientation) space, and (ii) planning multiple trajectories that are mutually collision-free. We demonstrate the performance of our approach on two clinically motivated applications. Our experiments indicate that our approach can compute high-quality plans for medical needle steering in 1.6 seconds on a commodity PC, enabling re-planning during execution to correct for perturbations. Our approach can also be used for designing optimized channel layouts within 3D printed implants for intracavitary brachytherapy. Yan Duan, Sachin Patil, John Schulman, Kenneth Y. Goldberg, Pieter Abbeel	ICRA	5
2014	Autonomous multilateral debridement with the Raven surgical robot abstract Autonomous robot execution of surgical sub-tasks has the potential to reduce surgeon fatigue and facilitate supervised tele-surgery. This paper considers the sub-task of surgical debridement: removing dead or damaged tissue fragments to allow the remaining healthy tissue to heal. We present an autonomous multilateral surgical debridement system using the Raven, an open-architecture surgical robot with two cable-driven 7 DOF arms. Our system combines stereo vision for 3D perception with trajopt, an optimization-based motion planner, and model predictive control (MPC). Laboratory experiments involving sensing, grasping, and removal of 120 fragments suggest that an autonomous surgical robot can achieve robustness comparable to human performance. Our robot system demonstrated the advantage of multilateral systems, as the autonomous execution was 1.5× faster with two arms than with one; however, it was two to three times slower than a human. Execution speed could be improved with better state estimation that would allow more travel between MPC steps and fewer MPC replanning cycles. The three primary contributions of this paper are: (1) introducing debridement as a sub-task of interest for surgical robotics, (2) demonstrating the first reliable autonomous robot performance of a surgical sub-task using the Raven, and (3) reporting experiments that highlight the importance of accurate state estimation for future research. Further information including code, photos, and video is available at: http://rll.berkeley.edu/raven. Ben Kehoe, Gregory Kahn, Jeffrey Mahler, Jonathan Kim, Alex X. Lee, Anna Lee, Keisuke Nakagawa, Sachin Patil, W. Douglas Boyd, Pieter Abbeel, Kenneth Y. Goldberg	ICRA	10
2014	Predicting initialization effectiveness for trajectory optimization abstract Trajectory optimization is a method for solving motion planning problems by formulating them as non-convex constrained optimization problems. The optimization process, however, can get stuck in local optima that are in collision. As a consequence, these methods typically require multiple initializations. This poses the problem of deciding which initializations to use when given a limited computational budget. In this paper we propose a machine learning approach to predict whether a collision-free solution will be found from a given initialization. We present a set of trajectory features that encode the obstacle distribution locally around a robot. These features are designed for generalization across different tasks. Our experiments on various planning benchmarks demonstrate the performance of our approach. Jia Pan 0001, Pieter Abbeel	ICRA	3
2014	Gaussian belief space planning with discontinuities in sensing domains abstract Discontinuities in sensing domains are common when planning for many robotic navigation and manipulation tasks. For cameras and 3D sensors, discontinuities may be inherent in sensor field of view or may change over time due to occlusions that are created by moving obstructions and movements of the sensor. The associated gaps in sensor information due to missing measurements pose a challenge for belief space and related optimization-based planning methods since there is no gradient information when the system state is outside the sensing domain. We address this in a belief space context by considering the signed distance to the sensing region. We smooth out sensing discontinuities by assuming that measurements can be obtained outside the sensing region with noise levels depending on a sigmoid function of the signed distance. We sequentially improve the continuous approximation by increasing the sigmoid slope over an outer loop to find plans that cope with sensor discontinuities. We also incorporate the information contained in not obtaining a measurement about the state during execution by appropriately truncating the Gaussian belief state. We present results in simulation for tasks with uncertainty involving navigation of mobile robots and reaching tasks with planar robot arms. Experiments suggest that the approach can be used to cope with discontinuities in sensing domains by effectively re-planning during execution. Sachin Patil, Yan Duan, John Schulman, Kenneth Y. Goldberg, Pieter Abbeel	ICRA	5
2014	BigBIRD: A large-scale 3D database of object instances abstract The state of the art in computer vision has rapidly advanced over the past decade largely aided by shared image datasets. However, most of these datasets tend to consist of assorted collections of images from the web that do not include 3D information or pose information. Furthermore, they target the problem of object category recognition - whereas solving the problem of object instance recognition might be sufficient for many robotic tasks. To address these issues, we present a high-quality, large-scale dataset of 3D object instances, with accurate calibration information for every image. We anticipate that “solving” this dataset will effectively remove many perception-related problems for mobile, sensing-based robots. The contributions of this work consist of: (1) BigBIRD, a dataset of 100 objects (and growing), composed of, for each object, 600 3D point clouds and 600 high-resolution (12 MP) images spanning all views, (2) a method for jointly calibrating a multi-camera system, (3) details of our data collection system, which collects all required data for a single object in under 6 minutes with minimal human effort, and (4) multiple software components (made available in open source), used to automate multi-sensor calibration and the data collection process. All code and data are available at http://rll.eecs.berkeley.edu/bigbird. James Sha, Karthik S. Narayan, Tudor Achim, Pieter Abbeel	ICRA	5
2014	Combined task and motion planning through an extensible planner-independent interface layer abstract The need for combined task and motion planning in robotics is well understood. Solutions to this problem have typically relied on special purpose, integrated implementations of task planning and motion planning algorithms. We propose a new approach that uses off-the-shelf task planners and motion planners and makes no assumptions about their implementation. Doing so enables our approach to directly build on, and benefit from, the vast literature and latest advances in task planning and motion planning. It uses a novel representational abstraction and requires only that failures in computing a motion plan for a high-level action be identifiable and expressible in the form of logical predicates at the task level. We evaluate the approach and illustrate its robustness through a number of experiments using a state-of-the-art robotics simulator and a PR2 robot. These experiments show the system accomplishing a diverse set of challenging tasks such as taking advantage of a tray when laying out a table for dinner and picking objects from cluttered environments where other objects need to be re-arranged before the target object can be reached. Siddharth Srivastava 0001, Eugene Fang, Lorenzo Riano, Rohan Chitnis, Stuart Russell 0001, Pieter Abbeel	ICRA	6
2014	Unifying scene registration and trajectory optimization for learning from demonstrations with application to manipulation of deformable objects abstract Recent work [1], [2] has shown promising results in enabling robotic manipulation of deformable objects through learning from demonstrations. Their method computes a registration from training scene to test scene, and then applies an extrapolation of this registration to the training scene gripper motion to obtain the gripper motion for the test scene. The warping cost of scene-to-scene registrations is used to determine the nearest neighbor from a set of training demonstrations. Then once the gripper motion has been generalized to the test situation, they apply trajectory optimization [3] to plan for the robot motions that will track the predicted gripper motions. In many situations, however, the predicted gripper motions cannot be followed perfectly due to, for example, joint limits or obstacles. In this case the past work finds a path that minimizes deviation from the predicted gripper trajectory as measured by its Euclidean distance for position and angular distance for orientation. Measuring the error this way during the motion planning phase, however, ignores the underlying structure of the problem-namely the idea that rigid registrations are preferred to generalize from training scene to test scene. Deviating from the gripper trajectory predicted by the extrapolated registration effectively changes the warp induced by the registration in the part of the space where the gripper trajectories are. The main contribution of this paper is an algorithm that considers this effective final warp as the criterion to optimize for in a unified optimization that simultaneously considers the scene-to-scene warping and the robot trajectory (which were separated into two sequential steps by the past work). This results in an approach that adjusts to infeasibility in a way that adapts directly to the geometry of the scene and minimizes the introduction of additional warping cost. In addition, this paper proposes to learn the motion of the gripper pads, whereas past work considered the motion of a coordinate frame attached to the gripper as a whole. This enables learning more precise grasping motions. Our experiments, which consider the task of knot tying, show that both unified optimization and explicit consideration of gripper pad motion result in improved performance. Alex X. Lee, Sandy H. Huang, Dylan Hadfield-Menell, Eric Tzeng, Pieter Abbeel	IROS	5
2014	Learning Neural Network Policies with Guided Policy Search under Unknown Dynamics Sergey Levine, Pieter Abbeel	NIPS	2
2014	Scaling up Gaussian Belief Space Planning Through Covariance-Free Trajectory Optimization and Automatic Differentiation Sachin Patil, Gregory Kahn, Michael Laskey, John Schulman, Kenneth Y. Goldberg, Pieter Abbeel	WAFR	6
2014	Planning Curvature and Torsion Constrained Ribbons in 3D with Application to Intracavitary Brachytherapy Sachin Patil, Jia Pan 0001, Pieter Abbeel, Kenneth Y. Goldberg	WAFR	3
2014	The Path Inference Filter: Model-Based Low-Latency Map Matching of Probe Vehicle Data abstract We consider the problem of reconstructing vehicle trajectories from sparse sequences of GPS points, for which the sampling interval is between 1 s and 2 min. We introduce a new class of algorithms, which are altogether called the path inference filter (PIF), that maps GPS data in real time, for a variety of tradeoffs and scenarios and with a high throughput. Numerous prior approaches in map matching can be shown to be special cases of the PIF presented in this paper. We present an efficient procedure for automatically training the filter on new data, with or without ground-truth observations. The framework is evaluated on a large San Francisco taxi data set and is shown to improve upon the current state of the art. This filter also provides insights about driving patterns of drivers. The PIF has been deployed at an industrial scale inside the Mobile Millennium traffic information system, and is used to map fleets of data in San Francisco and Sacramento, CA, USA; Stockholm, Sweden; and Porto, Portugal. Timothy Hunter, Pieter Abbeel, Alexandre M. Bayen	IEEE Trans. Intell. Transp. Syst.	2
2013	Tracking deformable objects with point clouds abstract We introduce an algorithm for tracking deformable objects from a sequence of point clouds. The proposed tracking algorithm is based on a probabilistic generative model that incorporates observations of the point cloud and the physical properties of the tracked object and its environment. We propose a modified expectation maximization algorithm to perform maximum a posteriori estimation to update the state estimate at each time step. Our modification makes it practical to perform the inference through calls to a physics simulation engine. This is significant because (i) it allows for the use of highly optimized physics simulation engines for the core computations of our tracking algorithm, and (ii) it makes it possible to naturally, and efficiently, account for physical constraints imposed by collisions, grasping actions, and material properties in the observation updates. Even in the presence of the relatively large occlusions that occur during manipulation tasks, our algorithm is able to robustly track a variety of types of deformable objects, including ones that are one-dimensional, such as ropes; two-dimensional, such as cloth; and three-dimensional, such as sponges. Our implementation can track these objects in real time. John Schulman, Alex X. Lee, Jonathan Ho, Pieter Abbeel	ICRA	4
2013	Grounding spatial relations for human-robot interaction abstract We propose a system for human-robot interaction that learns both models for spatial prepositions and for object recognition. Our system grounds the meaning of an input sentence in terms of visual percepts coming from the robot's sensors in order to send an appropriate command to the PR2 or respond to spatial queries. To perform this grounding, the system recognizes the objects in the scene, determines which spatial relations hold between those objects, and semantically parses the input sentence. The proposed system uses the visual and spatial information in conjunction with the semantic parse to interpret statements that refer to objects (nouns), their spatial relationships (prepositions), and to execute commands (actions). The semantic parse is inherently compositional, allowing the robot to understand complex commands that refer to multiple objects and relations such as: “Move the cup close to the robot to the area in front of the plate and behind the tea box”. Our system correctly parses 94% of the 210 online test sentences, correctly interprets 91% of the correctly parsed sentences, and correctly executes 89% of the correctly interpreted sentences. Sergio Guadarrama, Lorenzo Riano, Dave Golland, Daniel Göhring, Yangqing Jia, Daniel Klein 0001, Pieter Abbeel, Trevor Darrell	IROS	7
2013	Sigma hulls for Gaussian belief space planning for imprecise articulated robots amid obstacles abstract In many home and service applications, an emerging class of articulated robots such as the Raven and Baxter trade off precision in actuation and sensing to reduce costs and to reduce the potential for injury to humans in their workspaces. For planning and control of such robots, planning in belief ssigma hullpace, i.e., modeling such problems as POMDPs, has shown great promise but existing belief space planning methods have primarily been applied to cases where robots can be approximated as points or spheres. In this paper, we extend the belief space framework to treat articulated robots where the linkage can be decomposed into convex components. To allow planning and collision avoidance in Gaussian belief spaces, we introduce the concept of sigma hulls: convex hulls of robot links transformed according to the sigma standard deviation boundary points generated by the Unscented Kalman filter (UKF). We characterize the signed distances between sigma hulls and obstacles in the workspace to formulate efficient collision avoidance constraints compatible with the Gilbert-Johnson-Keerthi (GKJ) and Expanding Polytope Algorithms (EPA) within an optimization-based planning framework. We report results in simulation for planning motions for a 4-DOF planar robot and a 7-DOF articulated robot with imprecise actuation and inaccurate sensors. These experiments suggest that the sigma hull framework can significantly reduce the probability of collision and is computationally efficient enough to permit iterative re-planning for model predictive control. Alex X. Lee, Yan Duan, Sachin Patil, John Schulman, Zoe McCarthy, Jur P. van den Berg, Kenneth Y. Goldberg, Pieter Abbeel	IROS	8
2013	A case study of trajectory transfer through non-rigid registration for a simplified suturing scenario abstract Suturing is an important yet time-consuming part of surgery. A fast and robust autonomous procedure could reduce surgeon fatigue, and shorten operation times. It could also be of particular importance for suturing in remote tele-surgery settings where latency can complicate the master-slave mode control that is the current practice for robotic surgery with systems like the da Vinci®. We study the applicability of the trajectory transfer algorithm proposed in [12] to the automation of suturing. The core idea of this procedure is to first use non-rigid registration to find a 3D warping function which maps the demonstration scene onto the test scene, then use this warping function to transform the robot end-effector trajectory. Finally a robot joint trajectory is generated by solving a trajectory optimization problem that attempts to find the closest feasible trajectory, accounting for external constraints, such as joint limits and obstacles. Our experiments investigate generalization from a single demonstration to differing initial conditions. A first set of experiments considers the problem of having a simulated Raven II system [5] suture two flaps of tissue together. A second set of experiments considers a PR2 robot performing sutures in a scaled-up experimental setup. The simulation experiments were fully autonomous. For the real-world experiments we provided human input to assist with the detection of landmarks to be fed into the registration algorithm. The success rate for learning from a single demonstration is high for moderate perturbations from the demonstration's initial conditions, and it gradually decreases for larger perturbations. John Schulman, Sibi Venkatesan, Mallory Tayson-Frederick, Pieter Abbeel	IROS	5
2013	Multimodal blending for high-accuracy instance recognition abstract Despite the rich information provided by sensors such as the Microsoft Kinect in the robotic perception setting, the problem of detecting object instances remains unsolved, even in the tabletop setting, where segmentation is greatly simplified. Existing object detection systems often focus on textured objects, for which local feature descriptors can be used to reliably obtain correspondences between different views of the same object. We examine the benefits of dense feature extraction and multimodal features for improving the accuracy and robustness of an instance recognition system. By combining multiple modalities and blending their scores through an ensemble-based method in order to generate our final object hypotheses, we obtain significant improvements over previously published results on two RGB-D datasets. On the Challenge dataset, our method results in only one missed detection (achieving 100% precision and 99.77% recall). On the Willow dataset, we also make significant gains on the prior state of the art (achieving 98.28% precision and 87.78% recall), resulting in an increase in F-score from 0.8092 to 0.9273. Ziang Xie, Justin Uang, Karthik S. Narayan, Pieter Abbeel	IROS	5
2013	Learning from Demonstrations Through the Use of Non-rigid Registration John Schulman, Jonathan Ho, Cameron Lee, Pieter Abbeel	ISRR	4
2013	Large-Scale Estimation in Cyberphysical Systems Using Streaming Data: A Case Study With Arterial Traffic Estimation abstract Controlling and analyzing cyberphysical and robotics systems is increasingly becoming a Big Data challenge. We study the case of predicting drivers' travel times in a large urban area from sparse GPS traces. We present a framework that can accommodate a wide variety of traffic distributions and spread all the computations on a cluster to achieve small latencies. Our framework is built on Discretized Streams, a recently proposed approach to stream processing at scale. We demonstrate the usefulness of Discretized Streams with a novel algorithm to estimate vehicular traffic in urban networks. Our online EM algorithm can estimate traffic on a very large city network (the San Francisco Bay Area) by processing tens of thousands of observations per second, with a latency of a few seconds. Timothy Hunter, Tathagata Das, Matei Zaharia, Pieter Abbeel, Alexandre M. Bayen	IEEE Trans Autom. Sci. Eng.	4
2012	Safe Exploration in Markov Decision Processes Teodor Mihai Moldovan, Pieter Abbeel	ICML	2
2012	A robot path planning framework that learns from experience abstract We propose a framework, called Lightning, for planning paths in high-dimensional spaces that is able to learn from experience, with the aim of reducing computation time. This framework is intended for manipulation tasks that arise in applications ranging from domestic assistance to robot-assisted surgery. Our framework consists of two main modules, which run in parallel: a planning-from-scratch module, and a module that retrieves and repairs paths stored in a path library. After a path is generated for a new query, a library manager decides whether to store the path based on computation time and the generated path's similarity to the retrieved path. To retrieve an appropriate path from the library we use two heuristics that exploit two key aspects of the problem: (i) A correlation between the amount a path violates constraints and the amount of time needed to repair that path, and (ii) the implicit division of constraints into those that vary across environments in which the robot operates and those that do not. We evaluated an implementation of the framework on several tasks for the PR2 mobile manipulator and a minimally-invasive surgery robot in simulation. We found that the retrieve-and-repair module produced paths faster than planning-from-scratch in over 90% of test cases for the PR2 and in 58% of test cases for the minimally-invasive surgery robot. Dmitry Berenson, Pieter Abbeel, Kenneth Y. Goldberg	ICRA	2
2012	A textured object recognition pipeline for color and depth image data abstract We present an object recognition system which leverages the additional sensing and calibration information available in a robotics setting together with large amounts of training data to build high fidelity object models for a dataset of textured household objects. We then demonstrate how these models can be used for highly accurate detection and pose estimation in an end-to-end robotic perception system incorporating simultaneous segmentation, object classification, and pose fitting. The system can handle occlusions, illumination changes, multiple objects, and multiple instances of the same object. The system placed first in the ICRA 2011 Solutions in Perception instance recognition challenge. We believe the presented paradigm of building rich 3D models at training time and including depth information at test time is a promising direction for practical robotic perception systems. Pieter Abbeel	ICRA	4
2012	Performance analysis and terrain classification for a legged robot over rough terrain abstract Minimally actuated millirobotic crawlers navigate unreliably over uneven terrain-even when designed with inherent stability-mostly because of manufacturing variabilities and a lack of good models for ground interaction. In this paper, we investigate the performance of a legged robot as it traverses three distinct rough terrains: tile, carpet, and gravel. Furthermore, we present an accurate, robust, low-lag, and efficient algorithm for terrain classification that uses vibration data from the on-board inertial measurement unit and motor control data from back-EMF sensing and magnetic encoders. Fernando Garcia Bermudez, Ryan C. Julian, Duncan W. Haldane, Pieter Abbeel, Ronald S. Fearing	IROS	4
2012	Risk Aversion in Markov Decision Processes via Near Optimal Chernoff Bounds abstract The expected return is a widely used objective in decision making under uncer- tainty. Many algorithms, such as value iteration, have been proposed to optimize it. In risk-aware settings, however, the expected return is often not an appropriate objective to optimize. We propose a new optimization objective for risk-aware planning and show that it has desirable theoretical properties. We also draw con- nections to previously proposed objectives for risk-aware planing: minmax, ex- ponential utility, percentile and mean minus variance. Our method applies to an extended class of Markov decision processes: we allow costs to be stochastic as long as they are bounded. Additionally, we present an efﬁcient algorithm for op- timizing the proposed objective. Synthetic and real-world experiments illustrate the effectiveness of our method, at scale. Teodor Mihai Moldovan, Pieter Abbeel	NIPS	2
2012	Machine Learning for Robotics Pieter Abbeel	ECML/PKDD (1)	1
2012	The Path Inference Filter: Model-Based Low-Latency Map Matching of Probe Vehicle Data Timothy Hunter, Pieter Abbeel, Alexandre M. Bayen	WAFR	2
2012	Learning the Dynamics of Arterial Traffic From Probe Data Using a Dynamic Bayesian Network abstract Estimating and predicting traffic conditions in arterial networks using probe data has proven to be a substantial challenge. Sparse probe data represent the vast majority of the data available on arterial roads. This paper proposes a probabilistic modeling framework for estimating and predicting arterial travel-time distributions using sparsely observed probe vehicles. We introduce a model based on hydrodynamic traffic theory to learn the density of vehicles on arterial road segments, illustrating the distribution of delay within a road segment. The characterization of this distribution is essentially to use probe vehicles for traffic estimation: Probe vehicles report their location at random locations, and the travel times between location reports must be properly scaled to match the map discretization. A dynamic Bayesian network represents the spatiotemporal dependence on the network and provides a flexible framework to learn traffic dynamics from historical data and to perform real-time estimation with streaming data. The model is evaluated using data from a fleet of 500 probe vehicles in San Francisco, CA, which send Global Positioning System (GPS) data to our server every minute. The numerical experiments analyze the learning and estimation capabilities on a subnetwork with more than 800 links. The sampling rate of the probe vehicles does not provide detailed information about the location where vehicles encountered delay or the reason for any delay (i.e., signal delay, congestion delay, etc.). The model provides an increase in estimation accuracy of 35% when compared with a baseline approach to process probe-vehicle data. Aude Hofleitner, Ryan Herring, Pieter Abbeel, Alexandre M. Bayen	IEEE Trans. Intell. Transp. Syst.	3
2011	Scaling the mobile millennium system in the cloud abstract We report on our experience scaling up the Mobile Millennium traffic information system using cloud computing and the Spark cluster computing framework. Mobile Millennium uses machine learning to infer traffic conditions for large metropolitan areas from crowdsourced data, and Spark was specifically designed to support such applications. Many studies of cloud computing frameworks have demonstrated scalability and performance improvements for simple machine learning algorithms. Our experience implementing a real-world machine learning-based application corroborates such benefits, but we also encountered several challenges that have not been widely reported. These include: managing large parameter vectors, using memory efficiently, and integrating with the application's existing storage infrastructure. This paper describes these challenges and the changes they required in both the Spark framework and the Mobile Millennium software. While we focus on a system for traffic estimation, we believe that the lessons learned are applicable to other machine learning-based applications. Timothy Hunter, Teodor Mihai Moldovan, Matei Zaharia, Samy Merzgui, Justin Ma, Michael J. Franklin, Pieter Abbeel, Alexandre M. Bayen	SoCC	7
2011	Bringing clothing into desired configurations with limited perception abstract We consider the problem of autonomously bringing an article of clothing into a desired configuration using a general-purpose two-armed robot. We propose a hidden Markov model (HMM) for estimating the identity of the article and tracking the article's configuration throughout a specific sequence of manipulations and observations. At the end of this sequence, the article's configuration is known, though not necessarily desired. The estimated identity and configuration of the article are then used to plan a second sequence of manipulations that brings the article into the desired configuration. We propose a relaxation of a strain limiting finite element model for cloth simulation that can be solved via convex optimization; this serves as the basis of the transition and observation models of the HMM. The observation model uses simple perceptual cues consisting of the height of the article when held by a single gripper and the silhouette of the article when held by two grippers. The model accurately estimates the identity and configuration of clothing articles, enabling our procedure to autonomously bring a variety of articles into desired configurations that are useful for other tasks, such as folding. Marco F. Cusumano-Towner, James F. O'Brien, Pieter Abbeel	ICRA	5
2011	Modeling and perception of deformable one-dimensional objects abstract Recent advances in the modeling of deformable one-dimensional objects (DOOs) such as surgical suture, rope, and hair show significant promise for improving the simulation, perception, and manipulation of such objects. An important application of these tasks lies in the area of medical robotics, where robotic surgical assistants have the potential to greatly reduce surgeon fatigue and human error by improving the accuracy, speed, and robustness of surgical tasks such as suturing. However, different types of DOOs exhibit a variety of bending and twisting behaviors that are highly dependent on material properties. This paper proposes an approach for fitting simulation models of DOOs to observed data. Our approach learns an energy function such that observed DOO configurations lie in local energy minima. Our experiments on a variety of DOOs show that models fitted to different types of DOOs using our approach enable accurate prediction of future configurations. Additionally, we explore the application of our learned model to the perception of DOOs. Shervin Javdani, Sameep Tandon, James F. O'Brien, Pieter Abbeel	ICRA	5
2011	Parametrized shape models for clothing abstract We consider the problem of recognizing the configuration of clothing articles when crudely spread out on a flat surface, prior to and during folding. At the core of our approach are parametrized shape models for clothing articles. Each clothing category has its own shape model, and the variety in shapes for a given category is achieved through variation of the parameters. We present an efficient algorithm to find the parameters that provide the best fit when given an image of a clothing article. The models are such that, once the parameters have been fit, they provide a basic parse of the clothing article, allowing it to be followed by autonomous folding from category level specifications of fold sequences. Our approach is also able to recover the configuration of a clothing article when folds are being introduced-an important feature towards closing the perception-action loop. Additionally, our approach provides a reliable method of shape-based classification, simply by examining which model yields the best fit. Our experiments illustrate the effectiveness of our approach on a large set of clothing articles. Furthermore, we present an end-to-end system, which starts from an unknown spread-out clothing article, performs a parametrized model fit, then follows a category-level (rather than article specific) set of folding instructions, closing the loop with perceptual feedback by re-fitting between folds. Mario Fritz, Trevor Darrell, Pieter Abbeel	ICRA	4
2011	EG-RRT: Environment-guided random trees for kinodynamic motion planning with uncertainty and obstacles abstract Existing sampling-based robot motion planning methods are often inefficient at finding trajectories for kinodynamic systems, especially in the presence of narrow passages between obstacles and uncertainty in control and sensing. To address this, we propose EG-RRT, an Environment-Guided variant of RRT designed for kinodynamic robot systems that combines elements from several prior approaches and may incorporate a cost model based on the LQG-MP framework to estimate the probability of collision under uncertainty in control and sensing. We compare the performance of EG-RRT with several prior approaches on challenging sample problems. Results suggest that EG-RRT offers significant improvements in performance. Léonard Jaillet, Judy Hoffman, Jur P. van den Berg, Pieter Abbeel, Josep M. Porta, Kenneth Y. Goldberg	IROS	4
2011	Perception for the manipulation of socks abstract We consider the perceptual challenges inherent in the robotic manipulation of previously unseen socks, with the end goal of manipulation by a household robot for laundry. The task poses challenging problems in modeling the appearance, shape and configuration of these textile items that tend to exhibit high variability in texture, design, and style while being highly articulated objects. Ping Chuan Wang, Mario Fritz, Trevor Darrell, Pieter Abbeel	IROS	5
2011	Grasping and Fixturing as Submodular Coverage Problems John Schulman, Kenneth Y. Goldberg, Pieter Abbeel	ISRR	3
2010	Superhuman performance of surgical tasks by robots using iterative learning from human-guided demonstrations abstract In the future, robotic surgical assistants may assist surgeons by performing specific subtasks such as retraction and suturing to reduce surgeon tedium and reduce the duration of some operations. We propose an apprenticeship learning approach that has potential to allow robotic surgical assistants to autonomously execute specific trajectories with superhuman performance in terms of speed and smoothness. In the first step, we record a set of trajectories using human-guided backdriven motions of the robot. These are then analyzed to extract a smooth reference trajectory, which we execute at gradually increasing speeds using a variant of iterative learning control. We evaluate this approach on two representative tasks using the Berkeley Surgical Robots: a figure eight trajectory and a two handed knot-tie, a tedious suturing sub-task required in many surgical procedures. Results suggest that the approach enables (i) rapid learning of trajectories, (ii) smoother trajectories than the human-guided trajectories, and (iii) trajectories that are 7 to 10 times faster than the best human-guided trajectories. Jur P. van den Berg, Daniel Duckworth, Humphrey Hu, Andrew Wan, Xiao-Yu Fu, Kenneth Y. Goldberg, Pieter Abbeel	ICRA	8
2010	On the influence of ship motion prediction accuracy on motion planning and control of robotic manipulators on seaborne platforms abstract Robotic manipulators on non-inertial platforms, such as ships, have to endure large inertial forces due to the non-inertial motion of the platform. When the non-inertial platform's motion is known, motion planning and control algorithms can eliminate these perturbations-in fact, in some situations the motion planning algorithms can even leverage the inertial forces to more cheaply move to a target point. However, for many non-inertial platforms, the motion is unknown. In this paper we investigate how prediction errors and the choice of the prediction horizon affect the motion planning and control of robots mounted on a non-inertial base with a particular focus on seaborne platforms. We study the following three aspects: (i) We study prediction of ship motion and how prediction errors affect the motion planning and control of the manipulator. (ii) We evaluate the relationship between prediction accuracy and control. In particular, we study what prediction horizon length is useful for motion planning and control. We also consider how uncertainties in the ship motion predictions map to uncertainties in the future state of the robot and how to include the variance in the cost function to increase the optimal horizon length. (iii) Finally, we study a receding horizon approach, which re-solves the optimal control problem on-line over a horizon as determined to be meaningful from (ii). Several simulations are presented and, to our knowledge, for the first time experiments of ship-manipulator systems based on real ship motion data are presented. Pål Johan From, Jan Tommy Gravdahl, Pieter Abbeel	ICRA	3
2010	Cloth grasp point detection based on multiple-view geometric cues with application to robotic towel folding abstract We present a novel vision-based grasp point detection algorithm that can reliably detect the corners of a piece of cloth, using only geometric cues that are robust to variation in texture. Furthermore, we demonstrate the effectiveness of our algorithm in the context of folding a towel using a general-purpose two-armed mobile robotic platform without the use of specialized end-effectors or tools. The robot begins by picking up a randomly dropped towel from a table, goes through a sequence of vision-based re-grasps and manipulations-partially in the air, partially on the table-and finally stacks the folded towel in a target location. The reliability and robustness of our algorithm enables for the first time a robot with general purpose manipulators to reliably and fully-autonomously fold previously unseen towels, demonstrating success on all 50 out of 50 single-towel trials as well as on a pile of 5 towels. Jeremy Maitin-Shepard, Marco F. Cusumano-Towner, Jinna Lei, Pieter Abbeel	ICRA	4
2010	Parameterized maneuver learning for autonomous helicopter flight abstract Many robotic control tasks involve complex dynamics that are hard to model. Hand-specifying trajectories that satisfy a system's dynamics can be very time-consuming and often exceedingly difficult. We present an algorithm for automatically generating large classes of trajectories for difficult control tasks by learning parameterized versions of desired maneuvers from multiple expert demonstrations. Our algorithm has enabled the successful execution of several parameterized aerobatic maneuvers by our autonomous helicopter. Nimbus Goehausen, Pieter Abbeel	ICRA	4
2010	On a Connection between Importance Sampling and the Likelihood Ratio Policy Gradient abstract Likelihood ratio policy gradient methods have been some of the most successful reinforcement learning algorithms, especially for learning on physical systems. We describe how the likelihood ratio policy gradient can be derived from an importance sampling perspective. This derivation highlights how likelihood ratio methods under-use past experience by (a) using the past experience to estimate {\em only} the gradient of the expected return $U(\theta)$ at the current policy parameterization $\theta$, rather than to obtain a more complete estimate of $U(\theta)$, and (b) using past experience under the current policy {\em only} rather than using all past experience to improve the estimates. We present a new policy search method, which leverages both of these observations as well as generalized baselines---a new technique which generalizes commonly used baseline techniques for policy gradient methods. Our algorithm outperforms standard likelihood ratio policy gradient algorithms on several testbeds. Pieter Abbeel	NIPS	2
2010	Gravity-Based Robotic Cloth Folding Jur P. van den Berg, Kenneth Y. Goldberg, Pieter Abbeel	WAFR	4
2010	LQG-Based Planning, Sensing, and Control of Steerable Needles Jur P. van den Berg, Sachin Patil, Ron Alterovitz, Pieter Abbeel, Kenneth Y. Goldberg	WAFR	4
2008	Learning for control from multiple demonstrations abstract We consider the problem of learning to follow a desired trajectory when given a small number of demonstrations from a sub-optimal expert. We present an algorithm that (i) extracts the---initially unknown---desired trajectory from the sub-optimal expert's demonstrations and (ii) learns a local model suitable for control along the learned trajectory. We apply our algorithm to the problem of autonomous helicopter flight. In all cases, the autonomous helicopter's performance exceeds that of our expert helicopter pilot's demonstrations. Even stronger, our results significantly extend the state-of-the-art in autonomous helicopter aerobatics. In particular, our results include the first autonomous tic-tocs, loops and hurricane, vastly superior performance on previously performed aerobatic maneuvers (such as in-place flips and rolls), and a complete airshow, which requires autonomous transitions between these and various other maneuvers. Adam Coates 0002, Pieter Abbeel, Andrew Y. Ng	ICML	2
2008	Apprenticeship learning for motion planning with application to parking lot navigation abstract Motion and path-planning algorithms often use complex cost functions for both global navigation and local smoothing of trajectories. Obtaining good results typically requires carefully hand-engineering the trade-offs between different terms in the cost function. In practice, it is often much easier to demonstrate a few good trajectories. In this paper, we describe an efficient algorithm which - when given access to a few trajectory demonstrations - can automatically infer good trade-offs between the different costs. In our experiments, we apply our algorithm to the problem of navigating a robotic car in a parking lot. Pieter Abbeel, Dmitri Dolgov, Andrew Y. Ng, Sebastian Thrun	IROS	1
2008	Max-margin Classification of Data with Absent Features Gal Chechik, Geremy Heitz, Gal Elidan, Pieter Abbeel, Daphne Koller	J. Mach. Learn. Res.	4
2007	Hierarchical Apprenticeship Learning with Application to Quadruped Locomotion abstract We consider apprenticeship learning—learning from expert demonstrations—in the setting of large, complex domains. Past work in apprenticeship learning requires that the expert demonstrate complete trajectories through the domain. However, in many problems even an expert has difﬁculty controlling the system, which makes this approach infeasible. For example, consider the task of teach- ing a quadruped robot to navigate over extreme terrain; demonstrating an optimal policy (i.e., an optimal set of foot locations over the entire terrain) is a highly non-trivial task, even for an expert. In this paper we propose a method for hier- archical apprenticeship learning, which allows the algorithm to accept isolated advice at different hierarchical levels of the control task. This type of advice is often feasible for experts to give, even if the expert is unable to demonstrate com- plete trajectories. This allows us to extend the apprenticeship learning paradigm to much larger, more challenging domains. In particular, in this paper we apply the hierarchical apprenticeship learning algorithm to the task of quadruped loco- motion over extreme terrain, and achieve, to the best of our knowledge, results superior to any previously published work. J. Zico Kolter, Pieter Abbeel, Andrew Y. Ng	NIPS	2
2006	Efficient L1 Regularized Logistic Regression Su-In Lee, Honglak Lee, Pieter Abbeel, Andrew Y. Ng	AAAI	3
2006	Using inaccurate models in reinforcement learning abstract In the model-based policy search approach to reinforcement learning (RL), policies are found using a model (or "simulator") of the Markov decision process. However, for high-dimensional continuous-state tasks, it can be extremely difficult to build an accurate model, and thus often the algorithm returns a policy that works in simulation but not in real-life. The other extreme, model-free RL, tends to require infeasibly large numbers of real-life trials. In this paper, we present a hybrid algorithm that requires only an approximate model, and only a small number of real-life trials. The key idea is to successively "ground" the policy evaluations using real-life trials, but to rely on the approximate model to suggest local changes. Our theoretical results show that this algorithm achieves near-optimal performance in the real system, even when the model is only approximate. Empirical results also demonstrate that---when given only a crude model and a small number of real-life trials---our algorithm can obtain near-optimal performance in the real system. Pieter Abbeel, Morgan Quigley, Andrew Y. Ng	ICML	1
2006	An Application of Reinforcement Learning to Aerobatic Helicopter Flight abstract Autonomous helicopter flight is widely regarded to be a highly challenging control problem. This paper presents the first successful autonomous completion on a real RC helicopter of the following four aerobatic maneuvers: forward flip and sideways roll at low speed, tail-in funnel, and nose-in funnel. Our experimental results significantly extend the state of the art in autonomous helicopter flight. We used the following approach: First we had a pilot fly the helicopter to help us find a helicopter dynamics model and a reward (cost) function. Then we used a reinforcement learning (optimal control) algorithm to find a controller that is optimized for the resulting model and reward function. More specifically, we used differential dynamic programming (DDP), an extension of the linear quadratic regulator (LQR). Pieter Abbeel, Adam Coates 0002, Morgan Quigley, Andrew Y. Ng	NIPS	1
2006	Max-margin classification of incomplete data abstract We consider the problem of learning classifiers for structurally incomplete data, where some ob jects have a subset of features inherently absent due to complex relationships between the features. The common approach for handling missing features is to begin with a preprocessing phase that completes the missing features, and then use a standard classification procedure. In this paper we show how incomplete data can be classified directly without any completion of the missing features using a max-margin learning framework. We formulate this task using a geometrically-inspired ob jective function, and discuss two optimization approaches: The linearly separable case is written as a set of convex feasibility problems, and the non-separable case has a non-convex ob jective that we optimize iteratively. By avoiding the pre-processing phase in which the data is completed, these approaches offer considerable computational savings. More importantly, we show that by elegantly handling complex patterns of missing values, our approach is both competitive with other methods when the values are missing at random and outperforms them when the missing values have non-trivial structure. We demonstrate our results on two real-world problems: edge prediction in metabolic pathways, and automobile detection in natural images. Gal Chechik, Geremy Heitz, Gal Elidan, Pieter Abbeel, Daphne Koller	NIPS	4
2006	Learning Factor Graphs in Polynomial Time and Sample Complexity abstract We study the computational and sample complexity of parameter and structure learning in graphical models. Our main result shows that the class of factor graphs with bounded degree can be learned in polynomial time and from a polynomial number of training examples, assuming that the data is generated by a network in this class. This result covers both parameter estimation for a known network structure and structure learning. It implies as a corollary that we can learn factor graphs for both Bayesian networks and Markov networks of bounded degree, in polynomial time and sample complexity. Importantly, unlike standard maximum likelihood estimation algorithms, our method does not require inference in the underlying network, and so applies to networks where inference is intractable. We also show that the error of our learned model degrades gracefully when the generating distribution is not a member of the target class of networks. In addition to our main result, we show that the sample complexity of parameter learning in graphical models has an O(1) dependence on the number of variables in the model when using the KL-divergence normalized by the number of variables as the performance criterion. Pieter Abbeel, Daphne Koller, Andrew Y. Ng	J. Mach. Learn. Res.	1
2005	Exploration and apprenticeship learning in reinforcement learning abstract We consider reinforcement learning in systems with unknown dynamics. Algorithms such as E3 (Kearns and Singh, 2002) learn near-optimal policies by using "exploration policies" to drive the system towards poorly modeled states, so as to encourage exploration. But this makes these algorithms impractical for many systems; for example, on an autonomous helicopter, overly aggressive exploration may well result in a crash. In this paper, we consider the apprenticeship learning setting in which a teacher demonstration of the task is available. We show that, given the initial demonstration, no explicit exploration is necessary, and we can attain near-optimal performance (compared to the teacher) simply by repeatedly executing "exploitation policies" that try to maximize rewards. In finite-state MDPs, our algorithm scales polynomially in the number of states; in continuous-state linear dynamical systems, it scales polynomially in the dimension of the state. These results are proved using a martingale construction over relative losses. Pieter Abbeel, Andrew Y. Ng	ICML	1
2005	Learning vehicular dynamics, with application to modeling helicopters abstract We consider the problem of modeling a helicopter's dynamics based on state-action trajectories collected from it. The contribution of this paper is two-fold. First, we consider the linear models such as learned by C I F E R (the industry standard in helicopter identification), and show that the linear parameterization makes certain properties of dynamical systems, such as inertia, fundamentally difficult to capture. We propose an alternative, acceleration based, parameterization that does not suffer from this deficiency, and that can be learned as efficiently from data. Second, a Markov decision process model of a helicopter's dynamics would explicitly model only the one-step transitions, but we are often interested in a model's predictive performance over longer timescales. In this paper, we present an efficient algorithm for (approximately) minimizing the prediction error over long time scales. We present empirical results on two different helicopters. Although this work was motivated by the problem of modeling helicopters, the ideas presented here are general, and can be applied to modeling large classes of vehicular dynamics. Pieter Abbeel, Varun Ganapathi, Andrew Y. Ng	NIPS	1
2005	Learning Factor Graphs in Polynomial Time & Sample Complexity Pieter Abbeel, Daphne Koller, Andrew Y. Ng	UAI	1
2004	Apprenticeship learning via inverse reinforcement learning abstract We consider learning in a Markov decision process where we are not explicitly given a reward function, but where instead we can observe an expert demonstrating the task that we want to learn to perform. This setting is useful in applications (such as the task of driving) where it may be difficult to write down an explicit reward function specifying exactly how different desiderata should be traded off. We think of the expert as trying to maximize a reward function that is expressible as a linear combination of known features, and give an algorithm for learning the task demonstrated by the expert. Our algorithm is based on using "inverse reinforcement learning" to try to recover the unknown reward function. We show that our algorithm terminates in a small number of iterations, and that even though we may never recover the expert's reward function, the policy output by the algorithm will attain performance close to that of the expert, where here performance is measured with respect to the expert's unknown reward function. Pieter Abbeel, Andrew Y. Ng	ICML	1
2004	Learning first-order Markov models for control abstract First-order Markov models have been successfully applied to many prob- lems, for example in modeling sequential data using Markov chains, and modeling control problems using the Markov decision processes (MDP) formalism. If a ﬁrst-order Markov model’s parameters are estimated from data, the standard maximum likelihood estimator considers only the ﬁrst-order (single-step) transitions. But for many problems, the ﬁrst- order conditional independence assumptions are not satisﬁed, and as a re- sult the higher order transition probabilities may be poorly approximated. Motivated by the problem of learning an MDP’s parameters for control, we propose an algorithm for learning a ﬁrst-order Markov model that ex- plicitly takes into account higher order interactions during training. Our algorithm uses an optimization criterion different from maximum likeli- hood, and allows us to learn models that capture longer range effects, but without giving up the beneﬁts of using ﬁrst-order Markov models. Our experimental results also show the new algorithm outperforming conven- tional maximum likelihood estimation in a number of control problems where the MDP’s parameters are estimated from data. Pieter Abbeel, Andrew Y. Ng	NIPS	1
2003	Link Prediction in Relational Data abstract Many real-world domains are relational in nature, consisting of a set of objects related to each other in complex ways. This paper focuses on predicting the existence and the type of links between entities in such domains. We apply the relational Markov network framework of Taskar et al. to deﬁne a joint probabilis- tic model over the entire link graph — entity attributes and links. The application of the RMN algorithm to this task requires the deﬁnition of probabilistic patterns over subgraph structures. We apply this method to two new relational datasets, one involving university webpages, and the other a social network. We show that the collective classiﬁcation approach of RMNs, and the introduction of subgraph patterns over link labels, provide signiﬁcant improvements in accuracy over ﬂat classiﬁcation, which attempts to predict each link in isolation. Ben Taskar, Ming Fai Wong, Pieter Abbeel, Daphne Koller	NIPS	3
2002	Discriminative Probabilistic Models for Relational Data Ben Taskar, Pieter Abbeel, Daphne Koller	UAI	2