Learning Coordinated Terrain-Adaptive Locomotion by Imitating a Centroidal Dynamics Planner

TL;DR

This paper presents a method for training quadruped locomotion controllers that imitate trajectories generated by a non-linear solver on procedural terrains, enabling adaptive and precise movement over challenging terrains.

Contribution

It combines trajectory optimization and imitation learning to create terrain-adaptive controllers that transfer to unseen terrains and improve over standard RL methods.

Findings

Policies transfer successfully to unseen terrains.

Fine-tuning enhances traversal of challenging terrains.

Imitation of optimized trajectories improves locomotion precision.

Abstract

Dynamic quadruped locomotion over challenging terrains with precise foot placements is a hard problem for both optimal control methods and Reinforcement Learning (RL). Non-linear solvers can produce coordinated constraint satisfying motions, but often take too long to converge for online application. RL methods can learn dynamic reactive controllers but require carefully tuned shaping rewards to produce good gaits and can have trouble discovering precise coordinated movements. Imitation learning circumvents this problem and has been used with motion capture data to extract quadruped gaits for flat terrains. However, it would be costly to acquire motion capture data for a very large variety of terrains with height differences. In this work, we combine the advantages of trajectory optimization and learning methods and show that terrain adaptive controllers can be obtained by training policies to imitate trajectories that have been planned over procedural terrains by a non-linear solver. We show that the learned policies transfer to unseen terrains and can be fine-tuned to dynamically traverse challenging terrains that require precise foot placements and are very hard to solve with standard RL.