Real-World Reinforcement Learning of Active Perception Behaviors

Edward S. Hu, Jie Wang, Xingfang Yuan, Fiona Luo, Muyao Li, Gaspard Lambrechts, Oleh Rybkin, Dinesh Jayaraman

Keywords: robotics, reinforcement learning, imitation learning, active perception, privileged information

TL;DR: We propose a simple robot learning recipe leveraging privileged information to train active perception policies on real robots.

Abstract:

A robot's instantaneous sensory observations do not always reveal task-relevant state information. Under such partial observability, optimal behavior typically involves explicitly acting to gain the missing information.
Today's standard robot learning techniques struggle to produce such active perception behaviors.
We propose a simple real-world robot learning recipe to efficiently train active perception policies. Our approach, asymmetric advantage weighted regression (AAWR), exploits access to "privileged" extra sensors at training time. The privileged sensors enable training high-quality privileged value functions that aid in estimating the advantage of the target policy. Bootstrapping from a small number of potentially suboptimal demonstrations and an easy-to-obtain coarse policy initialization, AAWR quickly acquires active perception behaviors and boosts task performance. In evaluations on 8 manipulation tasks on 3 robots spanning varying degrees of partial observability, AAWR synthesizes reliable active perception behaviors that outperform all prior approaches. When initialized with a "generalist" robot policy that struggles with active perception tasks, AAWR efficiently generates information-gathering behaviors that allow it to operate under severe partial observability for manipulation tasks. Website:
https://penn-pal-lab.github.io/aawr/

Primary Area: Reinforcement learning (e.g., decision and control, planning, hierarchical RL, robotics)

Submission Number: 22397