MILES: Making Imitation Learning Easy with Self-Supervision

Georgios Papagiannis, Edward Johns
Proceedings of The 8th Conference on Robot Learning, PMLR 270:810-829, 2025.

Abstract

Data collection in imitation learning often requires significant, laborious human supervision, such as numerous demonstrations, and/or frequent environment resets for methods that incorporate reinforcement learning. In this work, we propose an alternative approach, MILES: a fully autonomous, self-supervised data collection paradigm, and we show that this enables efficient policy learning from just a single demonstration and a single environment reset. MILES autonomously learns a policy for returning to and then following the single demonstration, whilst being self-guided during data collection, eliminating the need for additional human interventions. We evaluated MILES across several realworld tasks, including tasks that require precise contact-rich manipulation such as locking a lock with a key. We found that, under the constraints of a single demonstration and no repeated environment resetting, MILES significantly outperforms state-of-the-art alternatives like imitation learning methods that leverage reinforcement learning. Videos of our experiments and code can be found on our webpage: www.robot-learning.uk/miles.

Cite this Paper

BibTeX
@InProceedings{pmlr-v270-papagiannis25a, title = {MILES: Making Imitation Learning Easy with Self-Supervision}, author = {Papagiannis, Georgios and Johns, Edward}, booktitle = {Proceedings of The 8th Conference on Robot Learning}, pages = {810--829}, year = {2025}, editor = {Agrawal, Pulkit and Kroemer, Oliver and Burgard, Wolfram}, volume = {270}, series = {Proceedings of Machine Learning Research}, month = {06--09 Nov}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v270/main/assets/papagiannis25a/papagiannis25a.pdf}, url = {https://proceedings.mlr.press/v270/papagiannis25a.html}, abstract = {Data collection in imitation learning often requires significant, laborious human supervision, such as numerous demonstrations, and/or frequent environment resets for methods that incorporate reinforcement learning. In this work, we propose an alternative approach, MILES: a fully autonomous, self-supervised data collection paradigm, and we show that this enables efficient policy learning from just a single demonstration and a single environment reset. MILES autonomously learns a policy for returning to and then following the single demonstration, whilst being self-guided during data collection, eliminating the need for additional human interventions. We evaluated MILES across several realworld tasks, including tasks that require precise contact-rich manipulation such as locking a lock with a key. We found that, under the constraints of a single demonstration and no repeated environment resetting, MILES significantly outperforms state-of-the-art alternatives like imitation learning methods that leverage reinforcement learning. Videos of our experiments and code can be found on our webpage: www.robot-learning.uk/miles.} }
Endnote
%0 Conference Paper %T MILES: Making Imitation Learning Easy with Self-Supervision %A Georgios Papagiannis %A Edward Johns %B Proceedings of The 8th Conference on Robot Learning %C Proceedings of Machine Learning Research %D 2025 %E Pulkit Agrawal %E Oliver Kroemer %E Wolfram Burgard %F pmlr-v270-papagiannis25a %I PMLR %P 810--829 %U https://proceedings.mlr.press/v270/papagiannis25a.html %V 270 %X Data collection in imitation learning often requires significant, laborious human supervision, such as numerous demonstrations, and/or frequent environment resets for methods that incorporate reinforcement learning. In this work, we propose an alternative approach, MILES: a fully autonomous, self-supervised data collection paradigm, and we show that this enables efficient policy learning from just a single demonstration and a single environment reset. MILES autonomously learns a policy for returning to and then following the single demonstration, whilst being self-guided during data collection, eliminating the need for additional human interventions. We evaluated MILES across several realworld tasks, including tasks that require precise contact-rich manipulation such as locking a lock with a key. We found that, under the constraints of a single demonstration and no repeated environment resetting, MILES significantly outperforms state-of-the-art alternatives like imitation learning methods that leverage reinforcement learning. Videos of our experiments and code can be found on our webpage: www.robot-learning.uk/miles.
APA
Papagiannis, G. & Johns, E.. (2025). MILES: Making Imitation Learning Easy with Self-Supervision. Proceedings of The 8th Conference on Robot Learning, in Proceedings of Machine Learning Research 270:810-829 Available from https://proceedings.mlr.press/v270/papagiannis25a.html.

Related Material