Diffusion models have recently been successfully applied to a wide range of robotics applications for learning complex multi-modal behaviors from data. However, prior works have mostly been confined to single-robot and small-scale environments due to the high sample complexity of learning multi-robot diffusion models. In this paper, we propose a method for generating collision-free multi-robot trajectories that conform to underlying data distributions while using only single-robot data. Our algorithm, Multi-robot Multi-model planning Diffusion (MMD), does so by combining learned diffusion models with classical search-based techniques---generating data-driven motions under collision constraints. Scaling further, we show how to compose multiple diffusion models to plan in large environments where a single diffusion model fails to generalize well. We demonstrate the effectiveness of our approach in planning for dozens of robots in a variety of simulated scenarios motivated by logistics environments.

Model-based planners and controllers are commonly used to solve complex manipulation problems as they can efficiently optimize diverse objectives and generalize to long horizon tasks. However, they often fail during deployment due to noisy actuation, partial observability and imperfect models. To enable a robot to recover from such failures, we propose to use hierarchical reinforcement learning to learn a recovery policy. The recovery policy is triggered when a failure is detected based on sensory observations and seeks to take the robot to a state from which it can complete the task using the nominal model-based controllers. Our approach, called RecoveryChaining, uses a hybrid action space, where the model-based controllers are provided as additional nominal options which allows the recovery policy to decide how to recover, when to switch to a nominal controller and which controller to switch to even with sparse rewards. We evaluate our approach in three multi-step manipulation tasks with sparse rewards, where it learns significantly more robust recovery policies than those learned by baselines. We successfully transfer recovery policies learned in simulation to a physical robot to demonstrate the feasibility of sim-to-real transfer with our method.

Operating under real world conditions is challenging due to the possibility of a wide range of failures induced by execution errors and state uncertainty. In relatively benign settings, such failures can be overcome by retrying or executing one of a small number of hand-engineered recovery strategies. By contrast, contact-rich sequential manipulation tasks, like opening doors and assembling furniture, are not amenable to exhaustive hand-engineering. To address this issue, we present a general approach for robustifying manipulation strategies in a sample-efficient manner. Our approach incrementally improves robustness by first discovering the failure modes of the current strategy via exploration in simulation and then learning additional recovery skills to handle these failures. To ensure efficient learning, we propose an online algorithm called Meta-Reasoning for Skill Learning (MetaReSkill) that monitors the progress of all recovery policies during training and allocates training resources to recoveries that are likely to improve the task performance the most. We use our approach to learn recovery skills for door-opening and evaluate them both in simulation and on a real robot with little fine-tuning. Compared to open-loop execution, our experiments show that even a limited amount of recovery learning improves task success substantially from 71% to 92.4% in simulation and from 75% to 90% on a real robot.

Robots deployed in many real-world settings need to be able to acquire new skills and solve new tasks over time. Prior works on planning with skills often make assumptions on the structure of skills and tasks, such as subgoal skills, shared skill implementations, or task-specific plan skeletons, which limit adaptation to new skills and tasks. By contrast, we propose doing task planning by jointly searching in the space of parameterized skills using high-level skill effect models learned in simulation. We use an iterative training procedure to efficiently generate relevant data to train such models. Our approach allows flexible skill parameterizations and task specifications to facilitate lifelong learning in general-purpose domains. Experiments demonstrate the ability of our planner to integrate new skills in a lifelong manner, finding new task strategies with lower costs in both train and test tasks. We additionally show that our method can transfer to the real world without further fine-tuning.

We consider the problem of completing a set of$n$tasks with a human-robot team using minimum effort. In many domains, teaching a robot to be fully autonomous can be counterproductive if there are finitely many tasks to be done. Rather, the optimal strategy is to weigh the cost of teaching a robot and its benefit- how many new tasks it allows the robot to solve autonomously. We formulate this as a planning problem where the goal is to decide what tasks the robot should do autonomously (act), what tasks should be delegated to a human (delegate) and what tasks the robot should be taught (learn) so as to complete all the given tasks with minimum effort. This planning problem results in a search tree that grows expo-nentially with$n$- making standard graph search algorithms intractable. We address this by converting the problem into a mixed integer program that can be solved efficiently using off-the-shelf solvers with bounds on solution quality. To predict the benefit of learning, we use an approximate simulation model of the tasks to train a precondition model that is parameterized by the training task. Finally, we evaluate our approach on peg insertion and Lego stacking tasks- both in simulation and real-world, showing substantial savings in human effort.

Heuristic-based graph search algorithms like A* are frequently used to solve motion planning problems in many domains. For most practical applications, it is infeasible and unnecessary to pre-compute the graph representing the whole search space. Instead, these algorithms generate the graph incrementally by applying a fixed set of actions (frequently called motion primitives) to find the successors of every node that they need to evaluate. In many domains, it is possible to define actions (called adaptive motion primitives) that are not pre-computed but generated on the fly. The generation and validation of these adaptive motion primitives is usually quite expensive compared to pre-computed motion primitives. However, they have been shown to drastically speed up search if used judiciously. In prior work, ad hoc techniques like fixed thresholds have been used to limit unsuccessful evaluations of these actions. In this paper, we propose a learning-based approach to make more intelligent decisions about when to evaluate them. We do a thorough empirical evaluation of our model on a 3 degree-of-freedom (dof) motion planning problem for navigation using the Reeds-Shepp path as an adaptive motion primitive. Our experiments show that using our approach in conjunction with search algorithms leads to over 2x speedup in planning time.