Learning Disturbances Online for Risk-Aware Control: Risk-Aware Flight with Less Than One Minute of Data

Prithvi Akella, Skylar X Wei, Joel W. Burdick, Aaron Ames
Proceedings of The 5th Annual Learning for Dynamics and Control Conference, PMLR 211:665-678, 2023.

Abstract

Recent advances in safety-critical risk-aware control are predicated on apriori knowledge of the disturbances a system might face. This paper proposes a method to efficiently learn these disturbances online, in a risk-aware context. First, we introduce the concept of a Surface-at-Risk, a risk measure for stochastic processes that extends Value-at-Risk — a commonly utilized risk measure in the risk-aware controls community. Second, we model the norm of the state discrepancy between the model and the true system evolution as a scalar-valued stochastic process and determine an upper bound to its Surface-at-Risk via Gaussian Process Regression. Third, we provide theoretical results on the accuracy of our fitted surface subject to mild assumptions that are verifiable with respect to the data sets collected during system operation. Finally, we experimentally verify our procedure by augmenting a drone’s controller and highlight performance increases achieved via our risk-aware approach after collecting less than a minute of operating data.

Cite this Paper

BibTeX
@InProceedings{pmlr-v211-akella23a, title = {Learning Disturbances Online for Risk-Aware Control: Risk-Aware Flight with Less Than One Minute of Data}, author = {Akella, Prithvi and Wei, Skylar X and Burdick, Joel W. and Ames, Aaron}, booktitle = {Proceedings of The 5th Annual Learning for Dynamics and Control Conference}, pages = {665--678}, 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/akella23a/akella23a.pdf}, url = {https://proceedings.mlr.press/v211/akella23a.html}, abstract = {Recent advances in safety-critical risk-aware control are predicated on apriori knowledge of the disturbances a system might face. This paper proposes a method to efficiently learn these disturbances online, in a risk-aware context. First, we introduce the concept of a Surface-at-Risk, a risk measure for stochastic processes that extends Value-at-Risk — a commonly utilized risk measure in the risk-aware controls community. Second, we model the norm of the state discrepancy between the model and the true system evolution as a scalar-valued stochastic process and determine an upper bound to its Surface-at-Risk via Gaussian Process Regression. Third, we provide theoretical results on the accuracy of our fitted surface subject to mild assumptions that are verifiable with respect to the data sets collected during system operation. Finally, we experimentally verify our procedure by augmenting a drone’s controller and highlight performance increases achieved via our risk-aware approach after collecting less than a minute of operating data.} }
Endnote
%0 Conference Paper %T Learning Disturbances Online for Risk-Aware Control: Risk-Aware Flight with Less Than One Minute of Data %A Prithvi Akella %A Skylar X Wei %A Joel W. Burdick %A Aaron Ames %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-akella23a %I PMLR %P 665--678 %U https://proceedings.mlr.press/v211/akella23a.html %V 211 %X Recent advances in safety-critical risk-aware control are predicated on apriori knowledge of the disturbances a system might face. This paper proposes a method to efficiently learn these disturbances online, in a risk-aware context. First, we introduce the concept of a Surface-at-Risk, a risk measure for stochastic processes that extends Value-at-Risk — a commonly utilized risk measure in the risk-aware controls community. Second, we model the norm of the state discrepancy between the model and the true system evolution as a scalar-valued stochastic process and determine an upper bound to its Surface-at-Risk via Gaussian Process Regression. Third, we provide theoretical results on the accuracy of our fitted surface subject to mild assumptions that are verifiable with respect to the data sets collected during system operation. Finally, we experimentally verify our procedure by augmenting a drone’s controller and highlight performance increases achieved via our risk-aware approach after collecting less than a minute of operating data.
APA
Akella, P., Wei, S.X., Burdick, J.W. & Ames, A.. (2023). Learning Disturbances Online for Risk-Aware Control: Risk-Aware Flight with Less Than One Minute of Data. Proceedings of The 5th Annual Learning for Dynamics and Control Conference, in Proceedings of Machine Learning Research 211:665-678 Available from https://proceedings.mlr.press/v211/akella23a.html.

Related Material