Rapidly Adaptable Legged Robots via Evolutionary Meta-Learning

Learning adaptable policies is crucial for robots to operate autonomously in our complex and quickly changing world. In this work, we present a new meta-learning method that allows robots to quickly adapt to changes in dynamics. In contrast to gradient-based meta-learning algorithms that rely on sec...

Full description

Saved in:
Bibliographic Details
Published inProceedings of the ... IEEE/RSJ International Conference on Intelligent Robots and Systems pp. 3769 - 3776
Main Authors Song, Xingyou, Yang, Yuxiang, Choromanski, Krzysztof, Caluwaerts, Ken, Gao, Wenbo, Finn, Chelsea, Tan, Jie
Format Conference Proceeding
LanguageEnglish
Published IEEE 24.10.2020
Subjects
Heuristic algorithms
Intelligent robots
Learning systems
Legged locomotion
Noise measurement
Perturbation methods
Robots
Online AccessGet full text

Cover

More Information
Summary:Learning adaptable policies is crucial for robots to operate autonomously in our complex and quickly changing world. In this work, we present a new meta-learning method that allows robots to quickly adapt to changes in dynamics. In contrast to gradient-based meta-learning algorithms that rely on second-order gradient estimation, we introduce a more noise-tolerant Batch Hill-Climbing adaptation operator and combine it with meta-learning based on evolutionary strategies. Our method significantly improves adaptation to changes in dynamics in high noise settings, which are common in robotics applications. We validate our approach on a quadruped robot that learns to walk while subject to changes in dynamics. We observe that our method significantly outperforms prior gradient-based approaches, enabling the robot to adapt its policy to changes based on less than 3 minutes of real data.
ISSN:2153-0866
DOI:10.1109/IROS45743.2020.9341571