Robot Behavior with Deep RL and Friends
David Meger
Nov. 15, 2024, 2:30 p.m. - Nov. 15, 2024, 2:30 p.m.
MD 280
Agility and adaptability are critical factors for the success of biological and cyberphysical systems. Our group has been pursuing models and algorithms to enable efficient and robust behavior learning and adaptation for over a decade. Our methods for learning in continuous state and action systems are used ubiquiitously and are packaged in nearly every impactful RL libray. Our analyses have identified biases in optimization, improved data efficiency and representations along with successful deployments of learning robots in the Caribbean Ocean, and in Quebec's forests and lakes. This talk will summarize recent key results and reflect on achieving robot learning's ChatGPT moment.
David Meger is an Associate Professor at McGill SOCS, co-Director of the Mobile Robtics lab, member of the Centre for Intelligent Machines and Associate Member at Mila the Quebec AI Institute. His research spans robot perception, interaction and control including advances in 3D machine vision, deep reinforcement learning and field robotics.
