ICRA 2023 Workshop on Multi-Robot Learning

ICC Capital Suite 11

ICRA London 2023

Although learning-based methods have become common-place for solving single-robot problems, they have only recently gained traction within the multi-robot domain. Practical multi-robot solutions lean on the progress of methods that provide scalability, that accommodate partial observability, and that can deal with imperfect information exchange. However, solving these problems is not only computationally hard, but also often involves hand-designing at least a part of the solution. By following a data-driven paradigm, learning-based methods allow us to offload the online computational burden to an offline learning procedure, thus not only alleviating the design task, but also promising to find solutions that balance optimality and real-world efficiency.

Work on multi-robot learning is nascent. In this workshop, we aim to bring together multi-robot, multi-agent, and machine learning researchers with varying theoretical foci (e.g., motion planning, game theory, POMDPs) who are applying a broad range of learning paradigms (e.g., reinforcement learning, imitation learning, federated learning, graph neural networks). The aim is to foster interaction and facilitate the communication of new insights. The program will consist of invited talks, spotlight presentations, a poster session, and a round-table discussion, allowing ample time for discussion.