Multi-Task Robotic Reinforcement Learning at Scale

Monday, July 12, 2021 - 11:30am to 12:30pm

Event Calendar Category

Uncategorized

Speaker Name

Karol Hausman

Affiliation

Google Brain & Stanford

Zoom meeting id

557641

Join Zoom meeting

https://mit.zoom.us/j/92561317547

Abstract

In this talk, I'll present two new advances for robotic RL at scale, MT-Opt, a new multi-task RL system for automated data collection and multi-task RL training, and Actionable Models, which leverages the acquired data for goal-conditioned RL. MT-Opt introduces a scalable data-collection mechanism that is used to collect over 800,000 episodes of various tasks on real robots and demonstrates a successful application of multi-task RL that yields ~3x average improvement over baselines. Additionally, it enables robots to master new tasks quickly through use of its extensive multi-task dataset (new task fine-tuning in <1 day of data collection). Actionable Models enables learning in the absence of specific tasks and rewards by training an implicit model of the world that is also an actionable robotic policy. This drastically increases the number of tasks the robot can perform (via visual goal specification) and enables more efficient learning of downstream tasks.

Biography

Karol Hausman is a Senior Research Scientist at Google Brain and an Adjunct Professor at Stanford working on robotics and machine learning. He is interested in enabling robots to autonomously acquire general-purpose skills with minimal supervision in the real world. He received his PhD in CS from the University of Southern California and Masters from the Technical University Munich. When he is not debugging robots at Google, he co-teaches Deep Multi-Task and Meta-Learning class at Stanford.