Abstract:This paper presents a novel learning-based control framework that uses keyframing to incorporate high-level objectives in natural locomotion for legged robots. These high-level objectives are specified as a variable number of partial or complete pose targets that are spaced arbitrarily in time. Our proposed framework utilizes a multi-critic reinforcement learning algorithm to effectively handle the mixture of dense and sparse rewards. Additionally, it employs a transformer-based encoder to accommodate a variable number of input targets, each associated with specific time-to-arrivals. Throughout simulation and hardware experiments, we demonstrate that our framework can effectively satisfy the target keyframe sequence at the required times. In the experiments, the multi-critic method significantly reduces the effort of hyperparameter tuning compared to the standard single-critic alternative. Moreover, the proposed transformer-based architecture enables robots to anticipate future goals, which results in quantitative improvements in their ability to reach their targets.
Abstract:We study the problem of nonepisodic reinforcement learning (RL) for nonlinear dynamical systems, where the system dynamics are unknown and the RL agent has to learn from a single trajectory, i.e., without resets. We propose Nonepisodic Optimistic RL (NeoRL), an approach based on the principle of optimism in the face of uncertainty. NeoRL uses well-calibrated probabilistic models and plans optimistically w.r.t. the epistemic uncertainty about the unknown dynamics. Under continuity and bounded energy assumptions on the system, we provide a first-of-its-kind regret bound of $\setO(\beta_T \sqrt{T \Gamma_T})$ for general nonlinear systems with Gaussian process dynamics. We compare NeoRL to other baselines on several deep RL environments and empirically demonstrate that NeoRL achieves the optimal average cost while incurring the least regret.
Abstract:Legged locomotion has recently achieved remarkable success with the progress of machine learning techniques, especially deep reinforcement learning (RL). Controllers employing neural networks have demonstrated empirical and qualitative robustness against real-world uncertainties, including sensor noise and external perturbations. However, formally investigating the vulnerabilities of these locomotion controllers remains a challenge. This difficulty arises from the requirement to pinpoint vulnerabilities across a long-tailed distribution within a high-dimensional, temporally sequential space. As a first step towards quantitative verification, we propose a computational method that leverages sequential adversarial attacks to identify weaknesses in learned locomotion controllers. Our research demonstrates that, even state-of-the-art robust controllers can fail significantly under well-designed, low-magnitude adversarial sequence. Through experiments in simulation and on the real robot, we validate our approach's effectiveness, and we illustrate how the results it generates can be used to robustify the original policy and offer valuable insights into the safety of these black-box policies.
Abstract:We propose a self-supervised approach for learning physics-based subspaces for real-time simulation. Existing learning-based methods construct subspaces by approximating pre-defined simulation data in a purely geometric way. However, this approach tends to produce high-energy configurations, leads to entangled latent space dimensions, and generalizes poorly beyond the training set. To overcome these limitations, we propose a self-supervised approach that directly minimizes the system's mechanical energy during training. We show that our method leads to learned subspaces that reflect physical equilibrium constraints, resolve overfitting issues of previous methods, and offer interpretable latent space parameters.
Abstract:This work introduces a motion retargeting approach for legged robots, which aims to create motion controllers that imitate the fine behavior of animals. Our approach, namely spatio-temporal motion retargeting (STMR), guides imitation learning procedures by transferring motion from source to target, effectively bridging the morphological disparities by ensuring the feasibility of imitation on the target system. Our STMR method comprises two components: spatial motion retargeting (SMR) and temporal motion retargeting (TMR). On the one hand, SMR tackles motion retargeting at the kinematic level by generating kinematically feasible whole-body motions from keypoint trajectories. On the other hand, TMR aims to retarget motion at the dynamic level by optimizing motion in the temporal domain. We showcase the effectiveness of our method in facilitating Imitation Learning (IL) for complex animal movements through a series of simulation and hardware experiments. In these experiments, our STMR method successfully tailored complex animal motions from various media, including video captured by a hand-held camera, to fit the morphology and physical properties of the target robots. This enabled RL policy training for precise motion tracking, while baseline methods struggled with highly dynamic motion involving flying phases. Moreover, we validated that the control policy can successfully imitate six different motions in two quadruped robots with different dimensions and physical properties in real-world settings.
Abstract:We present SIM-FSVGD for learning robot dynamics from data. As opposed to traditional methods, SIM-FSVGD leverages low-fidelity physical priors, e.g., in the form of simulators, to regularize the training of neural network models. While learning accurate dynamics already in the low data regime, SIM-FSVGD scales and excels also when more data is available. We empirically show that learning with implicit physical priors results in accurate mean model estimation as well as precise uncertainty quantification. We demonstrate the effectiveness of SIM-FSVGD in bridging the sim-to-real gap on a high-performance RC racecar system. Using model-based RL, we demonstrate a highly dynamic parking maneuver with drifting, using less than half the data compared to the state of the art.
Abstract:We introduce PACOH-RL, a novel model-based Meta-Reinforcement Learning (Meta-RL) algorithm designed to efficiently adapt control policies to changing dynamics. PACOH-RL meta-learns priors for the dynamics model, allowing swift adaptation to new dynamics with minimal interaction data. Existing Meta-RL methods require abundant meta-learning data, limiting their applicability in settings such as robotics, where data is costly to obtain. To address this, PACOH-RL incorporates regularization and epistemic uncertainty quantification in both the meta-learning and task adaptation stages. When facing new dynamics, we use these uncertainty estimates to effectively guide exploration and data collection. Overall, this enables positive transfer, even when access to data from prior tasks or dynamic settings is severely limited. Our experiment results demonstrate that PACOH-RL outperforms model-based RL and model-based Meta-RL baselines in adapting to new dynamic conditions. Finally, on a real robotic car, we showcase the potential for efficient RL policy adaptation in diverse, data-scarce conditions.
Abstract:Collecting manipulation demonstrations with robotic hardware is tedious - and thus difficult to scale. Recording data on robot hardware ensures that it is in the appropriate format for Learning from Demonstrations (LfD) methods. By contrast, humans are proficient manipulators, and recording their actions would be easy to scale, but it is challenging to use that data format with LfD methods. The question we explore is whether there is a method to collect data in a format that can be used with LfD while retaining some of the attractive features of recording human manipulation. We propose equipping humans with hand-held, hand-actuated parallel grippers and a head-mounted camera to record demonstrations of manipulation tasks. Using customised and reproducible grippers, we collect an initial dataset of common manipulation tasks. We show that there are tasks that, against our initial intuition, can be performed using parallel grippers. Qualitative insights are obtained regarding the impact of the difference in morphology on LfD by comparing the strategies used to complete tasks with human hands and grippers. Our data collection method bridges the gap between robot- and human-native manipulation demonstration. By making the design of our gripper prototype available, we hope to reduce other researchers effort to collect manipulation data.
Abstract:Nonlinear metamaterials with tailored mechanical properties have applications in engineering, medicine, robotics, and beyond. While modeling their macromechanical behavior is challenging in itself, finding structure parameters that lead to ideal approximation of high-level performance goals is a challenging task. In this work, we propose Neural Metamaterial Networks (NMN) -- smooth neural representations that encode the nonlinear mechanics of entire metamaterial families. Given structure parameters as input, NMN return continuously differentiable strain energy density functions, thus guaranteeing conservative forces by construction. Though trained on simulation data, NMN do not inherit the discontinuities resulting from topological changes in finite element meshes. They instead provide a smooth map from parameter to performance space that is fully differentiable and thus well-suited for gradient-based optimization. On this basis, we formulate inverse material design as a nonlinear programming problem that leverages neural networks for both objective functions and constraints. We use this approach to automatically design materials with desired strain-stress curves, prescribed directional stiffness and Poisson ratio profiles. We furthermore conduct ablation studies on network nonlinearities and show the advantages of our approach compared to native-scale optimization.
Abstract:Precisely reconstructing and manipulating crumpled cloths is challenging due to the high dimensionality of the cloth model, as well as the limited observation at self-occluded regions. We leverage the recent progress in the field of single-view human body reconstruction to template-based reconstruct the crumpled cloths from their top-view depth observations only, with our proposed sim-real registration protocols. In contrast to previous implicit cloth representations, our reconstruction mesh explicitly indicates the positions and visibilities of the entire cloth mesh vertices, enabling more efficient dual-arm and single-arm target-oriented manipulations. Experiments demonstrate that our template-based reconstruction and target-oriented manipulation (TRTM) system can be applied to daily cloths with similar topologies as our template mesh, but have different shapes, sizes, patterns, and physical properties. Videos, datasets, pre-trained models, and code can be downloaded from our project website: https://wenbwa.github.io/TRTM/.