In-Distribution Barrier Functions: Self-Supervised Policy Filters that Avoid Out-of-Distribution States

Fernando Castañeda, Haruki Nishimura, Rowan Thomas McAllister, Koushil Sreenath, Adrien Gaidon
Proceedings of The 5th Annual Learning for Dynamics and Control Conference, PMLR 211:286-299, 2023.

Abstract

Learning-based control approaches have shown great promise in performing complex tasks directly from high-dimensional perception data for real robotic systems. Nonetheless, the learned controllers can behave unexpectedly if the trajectories of the system divert from the training data distribution, which can compromise safety. In this work, we propose a control filter that wraps any reference policy and effectively encourages the system to stay in-distribution with respect to offline-collected safe demonstrations. Our methodology is inspired by Control Barrier Functions (CBFs), which are model-based tools from the nonlinear control literature that can be used to construct minimally invasive safe policy filters. While existing methods based on CBFs require a known low-dimensional state representation, our proposed approach is directly applicable to systems that rely solely on high-dimensional visual observations by learning in a latent state-space. We demonstrate that our method is effective for two different visuomotor control tasks in simulation environments, including both top-down and egocentric view settings.

Cite this Paper

BibTeX
@InProceedings{pmlr-v211-castaneda23a, title = {In-Distribution Barrier Functions: Self-Supervised Policy Filters that Avoid Out-of-Distribution States}, author = {Casta\~neda, Fernando and Nishimura, Haruki and McAllister, Rowan Thomas and Sreenath, Koushil and Gaidon, Adrien}, booktitle = {Proceedings of The 5th Annual Learning for Dynamics and Control Conference}, pages = {286--299}, year = {2023}, editor = {Matni, Nikolai and Morari, Manfred and Pappas, George J.}, volume = {211}, series = {Proceedings of Machine Learning Research}, month = {15--16 Jun}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v211/castaneda23a/castaneda23a.pdf}, url = {https://proceedings.mlr.press/v211/castaneda23a.html}, abstract = {Learning-based control approaches have shown great promise in performing complex tasks directly from high-dimensional perception data for real robotic systems. Nonetheless, the learned controllers can behave unexpectedly if the trajectories of the system divert from the training data distribution, which can compromise safety. In this work, we propose a control filter that wraps any reference policy and effectively encourages the system to stay in-distribution with respect to offline-collected safe demonstrations. Our methodology is inspired by Control Barrier Functions (CBFs), which are model-based tools from the nonlinear control literature that can be used to construct minimally invasive safe policy filters. While existing methods based on CBFs require a known low-dimensional state representation, our proposed approach is directly applicable to systems that rely solely on high-dimensional visual observations by learning in a latent state-space. We demonstrate that our method is effective for two different visuomotor control tasks in simulation environments, including both top-down and egocentric view settings.} }
Endnote
%0 Conference Paper %T In-Distribution Barrier Functions: Self-Supervised Policy Filters that Avoid Out-of-Distribution States %A Fernando Castañeda %A Haruki Nishimura %A Rowan Thomas McAllister %A Koushil Sreenath %A Adrien Gaidon %B Proceedings of The 5th Annual Learning for Dynamics and Control Conference %C Proceedings of Machine Learning Research %D 2023 %E Nikolai Matni %E Manfred Morari %E George J. Pappas %F pmlr-v211-castaneda23a %I PMLR %P 286--299 %U https://proceedings.mlr.press/v211/castaneda23a.html %V 211 %X Learning-based control approaches have shown great promise in performing complex tasks directly from high-dimensional perception data for real robotic systems. Nonetheless, the learned controllers can behave unexpectedly if the trajectories of the system divert from the training data distribution, which can compromise safety. In this work, we propose a control filter that wraps any reference policy and effectively encourages the system to stay in-distribution with respect to offline-collected safe demonstrations. Our methodology is inspired by Control Barrier Functions (CBFs), which are model-based tools from the nonlinear control literature that can be used to construct minimally invasive safe policy filters. While existing methods based on CBFs require a known low-dimensional state representation, our proposed approach is directly applicable to systems that rely solely on high-dimensional visual observations by learning in a latent state-space. We demonstrate that our method is effective for two different visuomotor control tasks in simulation environments, including both top-down and egocentric view settings.
APA
Castañeda, F., Nishimura, H., McAllister, R.T., Sreenath, K. & Gaidon, A.. (2023). In-Distribution Barrier Functions: Self-Supervised Policy Filters that Avoid Out-of-Distribution States. Proceedings of The 5th Annual Learning for Dynamics and Control Conference, in Proceedings of Machine Learning Research 211:286-299 Available from https://proceedings.mlr.press/v211/castaneda23a.html.

Related Material