Efficient Learning of Locomotion Skills through the Discovery of Diverse Environmental Trajectory Generator Priors

TL;DR

This paper introduces EETG, a method that learns diverse environmental trajectory generator priors to improve quadruped locomotion across various terrains, significantly enhancing efficiency over single prior approaches.

Contribution

EETG is the first method to learn a diverse set of specialized trajectory generator priors using Quality-Diversity algorithms within a unified policy architecture.

Findings

EETG enables quadruped robots to traverse diverse terrains successfully.

Learning multiple specialized priors is 5 times more efficient than a single prior.

The approach improves robustness and adaptability in complex environments.

Abstract

Data-driven learning based methods have recently been particularly successful at learning robust locomotion controllers for a variety of unstructured terrains. Prior work has shown that incorporating good locomotion priors in the form of trajectory generators (TGs) is effective at efficiently learning complex locomotion skills. However, defining a good, single TG as tasks/environments become increasingly more complex remains a challenging problem as it requires extensive tuning and risks reducing the effectiveness of the prior. In this paper, we present Evolved Environmental Trajectory Generators (EETG), a method that learns a diverse set of specialised locomotion priors using Quality-Diversity algorithms while maintaining a single policy within the Policies Modulating TG (PMTG) architecture. The results demonstrate that EETG enables a quadruped robot to successfully traverse a wide range of environments, such as slopes, stairs, rough terrain, and balance beams. Our experiments show that learning a diverse set of specialized TG priors is significantly (5 times) more efficient than using a single, fixed prior when dealing with a wide range of environments.