Abstract

My research group’s work aims to mitigate the curse of dimensionality in high degree-of-freedom (DOF) multi-agent robotic systems: as the number of agents (robots or DOFs) in the system grows, so does the combinatorial complexity of coordinating them. There are many solutions to managing this complexity growth, and our work favors decentralized approaches, whether it be for a team of mobile robots or a single articulated robot. Specifically, we have embraced advances in distributed reinforcement learning (dRL) to let multiple agents learn a common decentralized policy in a time-efficient manner, with or without explicit communications among agents. This has produced collaborative policies that naturally scale to large teams of agents while remaining near-optimal. In this talk, I will present dRL based approaches to 1) one-shot and lifelong multi-agent path finding (e.g., for warehouse automation), 2) the multi-agent traveling salesman problem (mTSP), and 3) communication learning for team-level cooperation in various tasks. I will present experiments on autonomous ground vehicles as well as in simulated environments that help validate our learned policies, and finally briefly go over some of my lab’s ongoing projects.

Speaker bio: Guillaume Sartoretti joined the Mechanical Engineering department at the National University of Singapore (NUS) as an Assistant Professor in 2019, where he founded and is directing the Multi-Agent Robotic Motion (MARMot) lab. Before that, he was a Postdoctoral Fellow in the Robotics Institute at Carnegie Mellon University (USA), where he worked with Prof. Howie Choset. He received his Ph.D. in robotics from EPFL (Switzerland) in 2016 for his dissertation on “Control of Agent Swarms in Random Environments,” under the supervision of Prof. Max-Olivier Hongler. He also holds a B.S. and an M.S. degree in Mathematics and Computer Science from the University of Geneva (Switzerland). His research focuses on the distributed/decentralized coordination of numerous agents, at the interface between stochastic modelling, conventional control, and artificial intelligence. Applications range from multi-robot systems, where independent robots and systems need to coordinate their actions to achieve a common goal, to high-DoF articulated robots, where joints need to be carefully coupled during locomotion in rough terrain. Guillaume was a Manufacturing Futures Initiative (MFI) postdoctoral fellow at CMU in 2018-2019, and was awarded an Outstanding Early Career Award from NUS ’ College of Design and Engineering in 2023.