Learning policies from high-dimensional visual inputs, such as pixels and point clouds, is crucial in various applications. Visual reinforcement learning is a promising approach that directly trains policies from visual observations, although it faces challenges in sample efficiency and computational costs. This study conducts an empirical comparison of State-to-Visual DAgger — a two-stage framework that initially trains a state policy before adopting online imitation to learn a visual policy — and Visual RL across a diverse set of tasks. We evaluate both methods across 16 tasks from three benchmarks, focusing on their asymptotic performance, sample efficiency, and computational costs. Surprisingly, our findings reveal that State-to-Visual DAgger does not universally outperform Visual RL but shows significant advantages in challenging tasks, offering more consistent performance. In contrast, its benefits in sample efficiency are less pronounced, although it often reduces the overall wall-clock time required for training. Based on our findings, we provide recommendations for practitioners and hope that our results contribute valuable perspectives for future research in visual policy learning.

Recent advancements in robot learning have used imitation learning with large models and extensive demonstrations to develop effective policies. However, these models are often limited by the quantity quality, and diversity of demonstrations. This paper explores improving offline-trained imitation learning models through online interactions with the environment. We introduce Policy Decorator, which uses a model-agnostic residual policy to refine large imitation learning models during online interactions. By implementing controlled exploration strategies, Policy Decorator enables stable, sample-efficient online learning. Our evaluation spans eight tasks across two benchmarks—ManiSkill and Adroit—and involves two state-of-the-art imitation learning models (Behavior Transformer and Diffusion Policy). The results show Policy Decorator effectively improves the offline-trained policies and preserves the smooth motion of imitation learning models, avoiding the erratic behaviors of pure RL policies. See our [project page](https://policydecorator.github.io/) for videos.

Long-horizon tasks in robotic manipulation present significant challenges in reinforcement learning (RL) due to the difficulty of designing dense reward functions and effectively exploring the expansive state-action space. However, despite a lack of dense rewards, these tasks often have a multi-stage structure, which can be leveraged to decompose the overall objective into manageable sub-goals. In this work, we propose DEMO$^3$, a framework that exploits this structure for efficient learning from visual inputs. Specifically, our approach incorporates multi-stage dense reward learning, a bi-phasic training scheme, and world model learning into a carefully designed demonstration-augmented RL framework that strongly mitigates the challenge of exploration in long-horizon tasks. Our evaluations demonstrate that our method improves data-efficiency by an average of 40% and by 70% on particularly difficult tasks compared to state-of-the-art approaches. We validate this across 16 sparse-reward tasks spanning four domains, including challenging humanoid visual control tasks using as few as five demonstrations.

The success of many RL techniques heavily relies on human-engineered dense rewards, which typically demands substantial domain expertise and extensive trial and error. In our work, we propose **DrS** (**D**ense **r**eward learning from **S**tages), a novel approach for learning *reusable* dense rewards for multi-stage tasks in a data-driven manner. By leveraging the stage structures of the task, DrS learns a high-quality dense reward from sparse rewards and demonstrations if given. The learned rewards can be *reused* in unseen tasks, thus reducing the human effort for reward engineering. Extensive experiments on three physical robot manipulation task families with 1000+ task variants demonstrate that our learned rewards can be reused in unseen tasks, resulting in improved performance and sample efficiency of RL algorithms. The learned rewards even achieve comparable performance to human-engineered rewards on some tasks. See our [project page](https://sites.google.com/view/iclr24drs) for more details.

In this paper, we examine the effectiveness of pre-training for visuo-motor control tasks. We revisit a simple Learning-from-Scratch (LfS) baseline that incorporates data augmentation and a shallow ConvNet, and find that this baseline is surprisingly competitive with recent approaches (PVR, MVP, R3M) that leverage frozen visual representations trained on large-scale vision datasets – across a variety of algorithms, task domains, and metrics in simulation and on a real robot. Our results demonstrate that these methods are hindered by a significant domain gap between the pre-training datasets and current benchmarks for visuo-motor control, which is alleviated by finetuning. Based on our findings, we provide recommendations for future research in pre-training for control and hope that our simple yet strong baseline will aid in accurately benchmarking progress in this area. Code: https://github.com/gemcollector/learning-from-scratch.

Training long-horizon robotic policies in complex physical environments is essential for many applications, such as robotic manipulation. However, learning a policy that can generalize to unseen tasks is challenging. In this work, we propose to achieve one-shot task generalization by decoupling plan generation and plan execution. Specifically, our method solves complex long-horizon tasks in three steps: build a paired abstract environment by simplifying geometry and physics, generate abstract trajectories, and solve the original task by an abstract-to-executable trajectory translator. In the abstract environment, complex dynamics such as physical manipulation are removed, making abstract trajectories easier to generate. However, this introduces a large domain gap between abstract trajectories and the actual executed trajectories as abstract trajectories lack low-level details and are not aligned frame-to-frame with the executed trajectory. In a manner reminiscent of language translation, our approach leverages a seq-to-seq model to overcome the large domain gap between the abstract and executable trajectories, enabling the low-level policy to follow the abstract trajectory. Experimental results on various unseen long-horizon tasks with different robot embodiments demonstrate the practicability of our methods to achieve one-shot task generalization.

In this article, we focus on the simulation of active stereovision depth sensors, which are popular in both academic and industry communities. Inspired by the underlying mechanism of the sensors, we designed a fully physics-grounded simulation pipeline that includes material acquisition, ray-tracing-based infrared (IR) image rendering, IR noise simulation, and depth estimation. The pipeline is able to generate depth maps with material-dependent error patterns similar to a real depth sensor in real time. We conduct real experiments to show that perception algorithms and reinforcement learning policies trained in our simulation platform could transfer well to the real-world test cases without any fine-tuning. Furthermore, due to the high degree of realism of this simulation, our depth sensor simulator can be used as a convenient testbed to evaluate the algorithm performance in the real world, which will largely reduce the human effort in developing robotic algorithms. The entire pipeline has been integrated into the SAPIEN simulator and is open-sourced to promote the research of vision and robotics communities.

We study how to learn a policy with compositional generalizability. We propose a two-stage framework, which refactorizes a high-reward teacher policy into a generalizable student policy with strong inductive bias. Particularly, we implement an object-centric GNN-based student policy, whose input objects are learned from images through self-supervised learning. Empirically, we evaluate our approach on four difficult tasks that require compositional generalizability, and achieve superior performance compared to baselines.

Nowadays, algorithms with fast convergence, small memory footprints, and low per-iteration complexity are particularly favorable for artificial intelligence applications. In this paper, we propose a doubly stochastic algorithm with a novel accelerating multi-momentum technique to solve large scale empirical risk minimization problem for learning tasks. While enjoying a provably superior convergence rate, in each iteration, such algorithm only accesses a mini batch of samples and meanwhile updates a small block of variable coordinates, which substantially reduces the amount of memory reference when both the massive sample size and ultra-high dimensionality are involved. Specifically, to obtain an ε-accurate solution, our algorithm requires only O(log(1/ε)/sqrt(ε)) overall computation for the general convex case and O((n+sqrt{nκ})log(1/ε)) for the strongly convex case. Empirical studies on huge scale datasets are conducted to illustrate the efficiency of our method in practice.