Reactive Stepping for Humanoid Robots using Reinforcement Learning: Application to Standing Push Recovery on the Exoskeleton Atalante

TL;DR

This paper introduces a reinforcement learning framework that enables humanoid robots, specifically a medical exoskeleton, to learn robust standing push recovery that smoothly transfers from simulation to real hardware using only proprioceptive data.

Contribution

The authors develop a novel RL approach with termination conditions and policy smoothing that ensures stable, safe sim-to-real transfer for push recovery in humanoid robots.

Findings

Successful real-world push recovery on the Atalante exoskeleton

Stable transfer of learned policies from simulation to hardware

Insights into balance maintenance through reward engineering

Abstract

State-of-the-art reinforcement learning is now able to learn versatile locomotion, balancing and push-recovery capabilities for bipedal robots in simulation. Yet, the reality gap has mostly been overlooked and the simulated results hardly transfer to real hardware. Either it is unsuccessful in practice because the physics is over-simplified and hardware limitations are ignored, or regularity is not guaranteed, and unexpected hazardous motions can occur. This paper presents a reinforcement learning framework capable of learning robust standing push recovery for bipedal robots that smoothly transfer to reality, providing only instantaneous proprioceptive observations. By combining original termination conditions and policy smoothness conditioning, we achieve stable learning, sim-to-real transfer and safety using a policy without memory nor explicit history. Reward engineering is then used to give insights into how to keep balance. We demonstrate its performance in reality on the lower-limb medical exoskeleton Atalante.