Reinforcement Learning (RL) has the potential to enable robots to learn from their own actions in the real world. Unfortunately, RL can be prohibitively expensive, in terms of on-robot runtime, due to inefficient exploration when learning from a sparse reward signal. Designing dense reward functions is labour-intensive and requires domain expertise. In our work, we propose Goal-Contrastive Rewards (GCR), a dense reward function learning method that can be trained on passive video demonstrations. By using videos without actions, our method is easier to scale, as we can use arbitrary videos. GCR combines two loss functions, an implicit value loss function that models how the reward increases when traversing a successful trajectory, and a goal-contrastive loss that discriminates between successful and failed trajectories. We perform experiments in simulated manipulation environments across RoboMimic and MimicGen tasks, as well as in the real world using a Franka arm and a Spot quadruped. We find that GCR leads to a more-sample efficient RL, enabling model-free RL to solve about twice as many tasks as our baseline reward learning methods. We also demonstrate positive cross-embodiment transfer from videos of people and of other robots performing a task. Website: https://gcr-robot.github.io/.

In reinforcement learning for legged robot locomotion, crafting effective reward strategies is crucial. Predefined gait patterns and complex reward systems are widely used to stabilize policy training. Drawing from the natural locomotion behaviors of humans and animals, which adapt their gaits to minimize energy consumption, we investigate the impact of incorporating an energy-efficient reward term that prioritizes distance-averaged energy consumption into the reinforcement learning framework. Our findings demonstrate that this simple addition enables quadruped robots to autonomously select appropriate gaits-such as four-beat walking at lower speeds and trotting at higher speeds-without the need for explicit gait regularizations. Furthermore, we provide a guideline for tuning the weight of this energy-efficient reward, facilitating its application in real-world scenarios. The effectiveness of our approach is validated through simulations and on a real Unitree Gol robot. This research highlights the potential of energy-centric reward functions to simplify and enhance the learning of adaptive and efficient locomotion in quadruped robots. Videos and more details are at https://sites.google.com/berkeley.edu/efficient-locomotion

Designing robotic agents to perform open vocabulary tasks has been the long-standing goal in robotics and AI. Recently, Large Language Models (LLMs) have achieved impressive results in creating robotic agents for performing open vocabulary tasks. However, planning for these tasks in the presence of uncertainties is challenging as it requires "chain-of-thought" reasoning, aggregating information from the environment, updating state estimates, and generating actions based on the updated state estimates. In this paper, we present an interactive planning technique for partially observable tasks using LLMs. In the proposed method, an LLM is used to collect missing information from the environment using a robot, and infer the state of the underlying problem from collected observations while guiding the robot to perform the required actions. We also use a fine-tuned Llama 2 model via self-instruct and compare its performance against a pre-trained LLM like GPT-4. Results are demonstrated on several tasks in simulation as well as real-world environments.

Automating the assembly of objects from their parts is a complex problem with innumerable applications in manufacturing, maintenance, and recycling. Unlike existing research, which is limited to target segmentation, pose regression, or using fixed target blueprints, our work presents a holistic multi-level framework for part assembly planning consisting of part assembly sequence inference, part motion planning, and robot contact optimization. We present the Part Assembly Sequence Transformer (PAST) – a sequence-to-sequence neural network – to infer assembly sequences recursively from a target blueprint. We then use a motion planner and optimization to generate part movements and contacts. To train PAST, we introduce D4PAS: a large-scale Dataset for Part Assembly Sequences consisting of physically valid sequences for industrial objects. Experimental results show that our approach generalizes better than prior methods while needing significantly less computational time for inference. Further details on our experiments and results are available in the video.

Most research on deformable linear object (DLO) manipulation assumes rigid grasping. However, beyond rigid grasping and re-grasping, in-hand following is also an essential skill that humans use to dexterously manipulate DLOs, which requires continuously changing the grasp point by in-hand sliding while holding the DLO to prevent it from falling. Achieving such a skill is very challenging for robots without using specially designed but not versatile end-effectors. Previous works have attempted using generic parallel grippers, but their robustness is unsatisfactory owing to the conflict between following and holding, which is hard to balance with a one-degree-of-freedom gripper. In this work, inspired by how humans use fingers to follow DLOs, we explore the usage of a generic dexterous hand with tactile sensing to imitate human skills and achieve robust in-hand DLO following. To enable the hardware system to function in the real world, we develop a framework that includes Cartesian-space arm-hand control, tactile-based in-hand 3-D DLO pose estimation, and task-specific motion design. Experimental results demonstrate the significant superiority of our method over using parallel grippers, as well as its great robustness, generalizability, and efficiency.

Animals are able to imitate each others' behavior, despite their difference in biomechanics. In contrast, imitating other similar robots is a much more challenging task in robotics. This problem is called cross domain imitation learning (CDIL). In this paper, we consider CDIL on a class of similar robots. We tackle this problem by introducing an imitation learning algorithm based on invariant representation. We propose to learn invariant state and action representations, which align the behavior of multiple robots so that CDIL becomes possible. Compared with previous invariant representation learning methods for similar purposes, our method does not require human-labeled pairwise data for training. Instead, we use cycle-consistency and domain confusion to align the representation and increase its robustness. We test the algorithm on multiple robots in the simulator and show that unseen new robot instances can be trained with existing expert demonstrations successfully. Qualitative results also demonstrate that the proposed method is able to learn similar representations for different robots with similar behaviors, which is essential for successful CDIL.

Robotic grasping for a diverse set of objects is essential in many robot manipulation tasks. One promising approach is to learn deep grasping models from large training datasets of object images and grasp labels. However, empirical grasping datasets are typically sparsely labeled (i.e., a small number of successful grasp labels**Labels refer to marking the image to indicate a successful robotic grasp. in each image). The data sparsity issue can lead to insufficient supervision and false-negative labels, and thus results in poor learning results. This paper proposes a Maximum Likelihood Grasp Sampling Loss (MLGSL) to tackle the data sparsity issue. The proposed method supposes that successful grasps are stochastically sampled from the predicted grasp distribution and maximizes the observing likelihood. MLGSL is utilized for training a fully convolutional network that generates thousands of grasps simultaneously. Training results suggest that models based on MLGSL can learn to grasp with datasets composing of 2 labels per image. Compared to previous works, which require training datasets of 16 labels per image, MLGSL is 8× more data-efficient. Meanwhile, physical robot experiments demonstrate an equivalent performance at a 90.7% grasp success rate on household objects. Codes and videos are available at [1].

Motion forecasting in highly interactive scenarios is a challenging problem in autonomous driving. In such scenarios, we need to accurately predict the joint behavior of interacting agents to ensure the safe and efficient navigation of autonomous vehicles. Recently, goal-conditioned methods have gained increasing attention due to their advantage in performance and their ability to capture the multimodality in trajec-tory distribution. In this work, we study the joint trajectory prediction problem with the goal-conditioned framework. In particular, we introduce a conditional-variational-autoencoder-based (CVAE) model to explicitly encode different interaction modes into the latent space. However, we discover that the vanilla model suffers from posterior collapse and cannot induce an informative latent space as desired. To address these issues, we propose a novel approach to avoid KL vanishing and induce an interpretable interactive latent space with pseudo labels. The proposed pseudo labels allow us to incorporate domain knowledge on interaction in a flexible manner. We motivate the proposed method using an illustrative toy example. In addition, we validate our framework on the Waymo Open Motion Dataset with both quantitative and qualitative evaluations.

The purpose of multi-task reinforcement learning (MTRL) is to train a single policy that can be applied to a set of different tasks. Sharing parameters allows us to take advantage of the similarities among tasks. However, the gaps between contents and difficulties of different tasks bring us challenges on both which tasks should share the parameters and what parameters should be shared, as well as the optimization challenges due to parameter sharing. In this work, we introduce a parameter-compositional approach (PaCo) as an attempt to address these challenges. In this framework, a policy subspace represented by a set of parameters is learned. Policies for all the single tasks lie in this subspace and can be composed by interpolating with the learned set. It allows not only flexible parameter sharing, but also a natural way to improve training.We demonstrate the state-of-the-art performance on Meta-World benchmarks, verifying the effectiveness of the proposed approach.

Proposing grasp poses for novel objects is an essential component for any robot manipulation task. Planning six degrees of freedom (DoF) grasps with a single camera, however, is challenging due to the complex object shape, incomplete object information, and sensor noise. In this paper, we present a 6-DoF contrastive grasp proposal network (CGPN) to infer 6-DoF grasps from a single-view depth image. First, an image encoder is used to extract the feature map from the input depth image, after which 3-DoF grasp regions are proposed from the feature map with a rotated region proposal network. Feature vectors that within the proposed grasp regions are then extracted and refined to 6-DoF grasps. The proposed model is trained offline with synthetic grasp data. To improve the robustness in reality and bridge the simulation-to-real gap, we further introduce a contrastive learning module and variant image processing techniques during the training. CGPN can locate collision-free grasps of an object using a single-view depth image within 0.5 second. Experiments on a physical robot further demonstrate the effectiveness of the algorithm. The experimental videos are available at [1].

Generating diverse and comprehensive interacting agents to evaluate the decision-making modules is essential for the safe and robust planning of autonomous vehicles (AV). Due to efficiency and safety concerns, most researchers choose to train interactive adversary (competitive or weakly competitive) agents in simulators and generate test cases to interact with evaluated AVs. However, most existing methods fail to provide both natural and critical interaction behaviors in various traffic scenarios. To tackle this problem, we propose a styled generative model RouteGAN that generates diverse interactions by controlling the vehicles separately with desired styles. By altering its style coefficients, the model can generate trajectories with different safety levels serve as an online planner. Experiments show that our model can generate diverse interactions in various scenarios. We evaluate different planners with our model by testing their collision rate in interaction with RouteGAN planners of multiple critical levels.

The Global Navigation Satellite System Occultation Sounder (GNOS) has been planned for the five Feng-Yun 3 series (FY3) weather satellites since 2013, the first of which, the FY3C satellite, was launched successfully at 03:07 UTC on September 23, 2013 from the Taiyuan Satellite Base, Shanxi province, China, into the orbit of 836-km altitude and 98.75° inclination. In addition to the Global Positioning System (GPS), the FY3C/GNOS is capable of tracking the occultation signal of the BeiDou Navigation Satellite System (BDS) (also called COMPASS) from space for the first time. The quality of BDS radio occultation (RO) has been verified in terms of signal-to-noise ratio. In this paper, the electron density profiles (EDPs) observed by FY3C/GNOS from both GPS RO and BDS RO, which were processed and archived in the National Satellite Meteorological Center of China Meteorological Administration, are compared with 32 globally distributed ionosonde observations, and then, we compare GPS RO EDPs with ionosonde observations at Mohe (52.0° N, 122.5° E), Beijing (40.3° N, 116.2° E), Wuhan (31.0° N, 114.5° E), and Sanya (18.3° N, 109.6° E). FY3C/GNOS EDPs show good agreement with ionosonde measurements, with larger discrepancies near the equatorial ionization anomaly region at Wuhan and Sanya. The ionospheric peak density (NmF2) and peak height (hmF2) derived from FY3C/GNOS EDPs are also compared with those obtained from the globally distributed ionosondes for the day of year 274-365 in 2013. In general, NmF2 and hmF2 have a higher correlation coefficient in the middle-high latitude than in the lower latitude region, due to the difference of ionospheric horizontal inhomogeneity. We also compared the NmF2 and hmF2 maps between FY3C/GNOS and the International Reference Ionosphere 2012 (IRI-2012) model. However, the wavenumber-4 structure, which can be indicated clearly from FY3C/GNOS observations, could not be reproduced well by IRI-2012. Further investigations show that the nighttime EDPs have obvious ionization enhancement around the ionospheric E layer over the Aurora and the South Atlantic Anomaly regions due to the energetic particle precipitation indicated by the Space Environment Monitor observations onboard FY3C.