End-to-End Reinforcement Learning for Torque Based Variable Height Hopping

TL;DR

This paper introduces an end-to-end reinforcement learning torque controller for legged robot hopping that automatically detects jump phases and successfully transfers from simulation to real robot without manual tuning.

Contribution

It presents a novel RL-based torque control method for hopping that eliminates manual phase detection and demonstrates effective simulation-to-reality transfer.

Findings

Successful real-world hopping with the learned controller

Automatic phase detection through RL eliminates heuristic requirements

Effective simulation-to-reality transfer for dynamic tasks

Abstract

Legged locomotion is arguably the most suited and versatile mode to deal with natural or unstructured terrains. Intensive research into dynamic walking and running controllers has recently yielded great advances, both in the optimal control and reinforcement learning (RL) literature. Hopping is a challenging dynamic task involving a flight phase and has the potential to increase the traversability of legged robots. Model based control for hopping typically relies on accurate detection of different jump phases, such as lift-off or touch down, and using different controllers for each phase. In this paper, we present a end-to-end RL based torque controller that learns to implicitly detect the relevant jump phases, removing the need to provide manual heuristics for state detection. We also extend a method for simulation to reality transfer of the learned controller to contact rich dynamic tasks, resulting in successful deployment on the robot after training without parameter tuning.