Moderation Team
About the Talk: Contingency planning, wherein an agent generates a set of possible plans conditioned on the outcome of an uncertain event, is an increasingly popular way for robots to act under uncertainty. In this talk, it will be presented a game-theoretic perspective on contingency planning which is tailored to multi-agent scenarios in which a robot’s actions impact the decisions of other agents and vice versa. The resulting contingency game allows the robot to efficiently coordinate with other agents by generating strategic motion plans conditioned on multiple possible intents of other actors in the scene. Beyond this new problem formulation, it will be (i) discussed an efficient method for solving $N$-player contingency games with nonlinear dynamics and non-convex costs and constraints, (ii) shown that this framework recovers existing variants of game-theoretic planning under uncertainty as special case, and (iii) demonstrated quantitative performance gains over game-theoretic motion plans that do not account for future uncertainty reduction.
About the Speaker: Lasse Peters is a PhD candidate at the Department of Cognitive Robotics at Delft University of Technology, advised by Javier Alonso-Mora and Laura Ferranti. His research interests lie in combining methods from optimal control, game theory, and reinforcement learning to design safe and efficient strategies for multi-agent interaction. Before coming to Delft, he was full-time research scholar in the Photogrammetry & Robotics Lab of Cyrill Stachniss and a visiting research scholar at UC Berkeley the Hybrid Systems Lab of Claire Tomlin. He received a MSc in Mechatronics and a BSc in Mechanical Engineering from TU Hamburg, Germany, in 2020 and 2017, respectively.
GIPHY App Key not set. Please check settings