Inferring Versatile Behavior from Demonstrations by Matching Geometric Descriptors

Niklas Freymuth, Nicolas Schreiber, Aleksandar Taranovic, Philipp Becker, Gerhard Neumann
Proceedings of The 6th Conference on Robot Learning, PMLR 205:1379-1389, 2023.

Abstract

Humans intuitively solve tasks in versatile ways, varying their behavior in terms of trajectory-based planning and for individual steps. Thus, they can easily generalize and adapt to new and changing environments. Current Imitation Learning algorithms often only consider unimodal expert demonstrations and act in a state-action-based setting, making it difficult for them to imitate human behavior in case of versatile demonstrations. Instead, we combine a mixture of movement primitives with a distribution matching objective to learn versatile behaviors that match the expert’s behavior and versatility. To facilitate generalization to novel task configurations, we do not directly match the agent’s and expert’s trajectory distributions but rather work with concise geometric descriptors which generalize well to unseen task configurations. We empirically validate our method on various robot tasks using versatile human demonstrations and compare to imitation learning algorithms in a state-action setting as well as a trajectory-based setting. We find that the geometric descriptors greatly help in generalizing to new task configurations and that combining them with our distribution-matching objective is crucial for representing and reproducing versatile behavior.

Cite this Paper

BibTeX
@InProceedings{pmlr-v205-freymuth23a, title = {Inferring Versatile Behavior from Demonstrations by Matching Geometric Descriptors}, author = {Freymuth, Niklas and Schreiber, Nicolas and Taranovic, Aleksandar and Becker, Philipp and Neumann, Gerhard}, booktitle = {Proceedings of The 6th Conference on Robot Learning}, pages = {1379--1389}, year = {2023}, editor = {Liu, Karen and Kulic, Dana and Ichnowski, Jeff}, volume = {205}, series = {Proceedings of Machine Learning Research}, month = {14--18 Dec}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v205/freymuth23a/freymuth23a.pdf}, url = {https://proceedings.mlr.press/v205/freymuth23a.html}, abstract = {Humans intuitively solve tasks in versatile ways, varying their behavior in terms of trajectory-based planning and for individual steps. Thus, they can easily generalize and adapt to new and changing environments. Current Imitation Learning algorithms often only consider unimodal expert demonstrations and act in a state-action-based setting, making it difficult for them to imitate human behavior in case of versatile demonstrations. Instead, we combine a mixture of movement primitives with a distribution matching objective to learn versatile behaviors that match the expert’s behavior and versatility. To facilitate generalization to novel task configurations, we do not directly match the agent’s and expert’s trajectory distributions but rather work with concise geometric descriptors which generalize well to unseen task configurations. We empirically validate our method on various robot tasks using versatile human demonstrations and compare to imitation learning algorithms in a state-action setting as well as a trajectory-based setting. We find that the geometric descriptors greatly help in generalizing to new task configurations and that combining them with our distribution-matching objective is crucial for representing and reproducing versatile behavior.} }
Endnote
%0 Conference Paper %T Inferring Versatile Behavior from Demonstrations by Matching Geometric Descriptors %A Niklas Freymuth %A Nicolas Schreiber %A Aleksandar Taranovic %A Philipp Becker %A Gerhard Neumann %B Proceedings of The 6th Conference on Robot Learning %C Proceedings of Machine Learning Research %D 2023 %E Karen Liu %E Dana Kulic %E Jeff Ichnowski %F pmlr-v205-freymuth23a %I PMLR %P 1379--1389 %U https://proceedings.mlr.press/v205/freymuth23a.html %V 205 %X Humans intuitively solve tasks in versatile ways, varying their behavior in terms of trajectory-based planning and for individual steps. Thus, they can easily generalize and adapt to new and changing environments. Current Imitation Learning algorithms often only consider unimodal expert demonstrations and act in a state-action-based setting, making it difficult for them to imitate human behavior in case of versatile demonstrations. Instead, we combine a mixture of movement primitives with a distribution matching objective to learn versatile behaviors that match the expert’s behavior and versatility. To facilitate generalization to novel task configurations, we do not directly match the agent’s and expert’s trajectory distributions but rather work with concise geometric descriptors which generalize well to unseen task configurations. We empirically validate our method on various robot tasks using versatile human demonstrations and compare to imitation learning algorithms in a state-action setting as well as a trajectory-based setting. We find that the geometric descriptors greatly help in generalizing to new task configurations and that combining them with our distribution-matching objective is crucial for representing and reproducing versatile behavior.
APA
Freymuth, N., Schreiber, N., Taranovic, A., Becker, P. & Neumann, G.. (2023). Inferring Versatile Behavior from Demonstrations by Matching Geometric Descriptors. Proceedings of The 6th Conference on Robot Learning, in Proceedings of Machine Learning Research 205:1379-1389 Available from https://proceedings.mlr.press/v205/freymuth23a.html.

Related Material