SEIL: Simulation-augmented Equivariant Imitation Learning

In robotic manipulation, acquiring samples is extremely expensive because it often requires interacting with the real world. Traditional image-level data augmentation has shown the potential to improve sample efficiency in various machine learning tasks. However, image-level data augmentation is ins...

Full description

Saved in:
Bibliographic Details
Published in2023 IEEE International Conference on Robotics and Automation (ICRA) pp. 1845 - 1851
Main Authors Jia, Mingxi, Wang, Dian, Su, Guanang, Klee, David, Zhu, Xupeng, Walters, Robin, Platt, Robert
Format Conference Proceeding
LanguageEnglish
Published IEEE 29.05.2023
Subjects
Automation
Data augmentation
Data models
Machine learning
Robots
Task analysis
Trajectory
Online AccessGet full text

Cover

More Information
Summary:In robotic manipulation, acquiring samples is extremely expensive because it often requires interacting with the real world. Traditional image-level data augmentation has shown the potential to improve sample efficiency in various machine learning tasks. However, image-level data augmentation is insufficient for an imitation learning agent to learn good manipulation policies in a reasonable amount of demonstrations. We propose Simulation-augmented Equivariant Imitation Learning (SEIL), a method that combines a novel data augmentation strategy of supplementing expert trajectories with simulated transitions and an equivariant model that exploits the O(2) symmetry in robotic manipulation. Experimental evaluations demonstrate that our method can learn non-trivial manipulation tasks within ten demonstrations and outperform the baselines by a significant margin.
DOI:10.1109/ICRA48891.2023.10161252