The fundamental limitation of the behavioral cloning (BC) approach to imitation learning is that it only teaches an agent what the expert did at states the expert visited. This means that when a BC agent makes a mistake which takes them out of the support of the demonstrations, they often don't know how to recover from it. In this sense, BC is akin to *giving the agent the fish* -- giving them dense supervision across a narrow set of states -- rather than teaching them *to fish*: to be able to reason independently about achieving the expert's outcome even when faced with unseen situations at test-time. In response, we explore *learning to search* (L2S) from expert demonstrations, i.e. learning the components required to, at test time, plan to match expert outcomes, even after making a mistake. These include *(1)* a world model and *(2)* a reward model. We carefully ablate the set of algorithmic and design decisions required to combine these and other components for stable and sample/interaction-efficient learning of recovery behavior without additional human corrections. Across a dozen visual manipulation tasks from three benchmarks, our approach SAILOR consistently out-performs state-of-the-art Diffusion Policies trained via BC on the same data. Furthermore, scaling up the amount of demonstrations used for BC by 5-10x still leaves a performance gap. We find that SAILOR can identify nuanced failures and is robust to reward hacking. Our code is available at [https://github.com/arnavkj1995/SAILOR](https://github.com/arnavkj1995/SAILOR).

We present Jade, a differentiable physics engine for articulated rigid bodies. Jade models contacts as the Linear Complementarity Problem (LCP). Compared to existing differentiable simulations, Jade offers features including intersection-free collision simulation and stable LCP solutions for multiple frictional contacts. We use continuous collision detection to detect the time of impact and adopt the backtracking strategy to prevent intersection between bodies with complex geometry shapes. We derive the gradient calculation to ensure the whole simulation process is differentiable under the backtracking mechanism. We modify the popular Dantzig’s algorithm to get valid solutions under multiple frictional contacts. We conduct extensive experiments to demonstrate the effectiveness of our differentiable physics simulation over a variety of contact-rich tasks. Supplemental materials and videos are available on our project webpage at https://sites.google.com/view/diffsim

Robot in-hand manipulation is an important skill for robots to carry out sophisticated tasks that require moving the grasped object within hand. In this work, we present the Roller Grasper V3, a nonanthropomorphic robot grasper with a steerable roller on each of its four fingertips, and a manipulation architecture that enables the Roller Grasper V3 to achieve full 6-DoF manipulation of the grasped object in$SE(3)$. The manipulation architecture consists of a high-level planner that searches for a feasible path with waypoints for the object to be manipulated, and a low-level control policy that is used to navigate the object in between the adjacent waypoints. The method was experimentally validated on the Roller Grasper V3 to manipulate multiple objects with different geometries and topologies.

Visual pre-training with large-scale real-world data has made great progress in recent years, showing great potential in robot learning with pixel observations. However, the recipes of visual pre-training for robot manipulation tasks are yet to be built. In this paper, we thoroughly investigate the effects of visual pre-training strategies on robot manipulation tasks from three fundamental perspectives: pre-training datasets, model architectures and training methods. Several significant experimental findings are provided that are beneficial for robot learning. Further, we propose a visual pre-training scheme for robot manipulation termed Vi-PRoM, which combines self-supervised learning and supervised learning. Concretely, the former employs contrastive learning to acquire underlying patterns from large-scale unlabeled data, while the latter aims learning visual semantics and temporal dynamics. Extensive experiments on robot manipulations in various simulation environments and the real robot demonstrate the superiority of the proposed scheme. Videos and more details can be found on https://explore-pretrain-robot.github.io.