Next Thursday, we will have our next monthly meeting of the Boston Robot Hackers. Mark it on your calendar now, and please register for the event—select October 15, 7:00 to 9:00.
Reinforcement Learning for Multimodal Locomotion
Arjun’s MS thesis at Northeastern’s Silicon Synapse Lab tackled a robot that could both walk and fly: the Husky-Beta legged-aerial platform. Crossing rough terrain meant more than learning to walk — the robot also had to learn which mode of locomotion to use at all, switching between legged and aerial movement depending on the terrain ahead. He’ll walk through the hierarchical reinforcement learning approach that made this possible, and how it generalizes to multimodal robot behavior more broadly.
Speaker: Arjun Viswanathan
Arjun works on reinforcement learning for robot behavior at Boston Dynamics. He did his MS in Robotics at Northeastern’s Silicon Synapse Lab, teaching the Husky-Beta legged-aerial robot to cross rough terrain with hierarchical RL — learning not just how to walk, but which mode of locomotion to use at all.