Our mission is to develop solutions for collective intelligence in multi-robot and multi-agent systems. This research brings in methods from machine learning, planning and control, and has numerous applications, including automated transport, logistics, environmental monitoring, and search & rescue.
Research
All intelligence is collective intelligence… Our research builds on this premise, and focuses on synthesizing decentralized agent controllers and interaction policies that lead to cooperative and collaborative systems. Many of our methods build on Graph Neural Network (GNN) architectures, which provide a well-suited inductive bias for networked multi-agent teams and swarms. We are also interested in developing data-driven or hybrid approaches that find near-optimal solutions to hard coordination problems involving resource allocation, path finding, or graph optimization. Our solutions enable fast on-line decision-making, as typically required in robotics applications. As part of our work, we develop sim-to-real methods that facilitate the transfer of policies learned in simulation to real world systems.
Paper Highlights
M. Bettini, A. Shankar, A. Prorok, Heterogeneous Multi-Robot Reinforcement Learning, AAMAS 2023
Z. Gao, A. Prorok, Environment Optimization for Multi-Agent Navigation, IEEE ICRA 2023
J. Blumenkamp, Q. Li, B. Wang, Z. Liu, A. Prorok, See What the Robot Can’t See: Learning Cooperative Perception for Visual Navigation, IEEE/RSJ IROS 2023
A. Prorok, J. Blumenkamp, Q. Li, R. Korvelesy, Z. Liu, E. Stump, The Holy Grail of Multi-Robot Planning: Learning to Generate Online-Scalable Solutions from Offline-Optimal Experts, AAMAS 2022, May 9–13, 2022
University of Cambridge
William Gates Building
15 JJ Thompson Avenue
Cambridge, CB3 0FD
UK
asp45@cam.ac.uk