On exploration requirements for learning safety constraints

Pierre-François Massiani, Steve Heim, Sebastian Trimpe
Proceedings of the 3rd Conference on Learning for Dynamics and Control, PMLR 144:905-916, 2021.

Abstract

Enforcing safety for dynamical systems is challenging, since it requires constraint satisfaction along trajectory predictions. Equivalent control constraints can be computed in the form of sets that enforce positive invariance, and can thus guarantee safety in feedback controllers without predictions. However, these constraints are cumbersome to compute from models, and it is not yet well established how to infer constraints from data. In this paper, we shed light on the key objects involved in learning control constraints from data in a model-free setting. In particular, we discuss the family of constraints that enforce safety in the context of a nominal control policy, and expose that these constraints do not need to be accurate everywhere. They only need to correctly exclude a subset of the state-actions that would cause failure, which we call the critical set.

Cite this Paper

BibTeX
@InProceedings{pmlr-v144-massiani21a, title = {On exploration requirements for learning safety constraints}, author = {Massiani, Pierre-Fran\c{c}ois and Heim, Steve and Trimpe, Sebastian}, booktitle = {Proceedings of the 3rd Conference on Learning for Dynamics and Control}, pages = {905--916}, year = {2021}, editor = {Jadbabaie, Ali and Lygeros, John and Pappas, George J. and A. Parrilo, Pablo and Recht, Benjamin and Tomlin, Claire J. and Zeilinger, Melanie N.}, volume = {144}, series = {Proceedings of Machine Learning Research}, month = {07 -- 08 June}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v144/massiani21a/massiani21a.pdf}, url = {https://proceedings.mlr.press/v144/massiani21a.html}, abstract = {Enforcing safety for dynamical systems is challenging, since it requires constraint satisfaction along trajectory predictions. Equivalent control constraints can be computed in the form of sets that enforce positive invariance, and can thus guarantee safety in feedback controllers without predictions. However, these constraints are cumbersome to compute from models, and it is not yet well established how to infer constraints from data. In this paper, we shed light on the key objects involved in learning control constraints from data in a model-free setting. In particular, we discuss the family of constraints that enforce safety in the context of a nominal control policy, and expose that these constraints do not need to be accurate everywhere. They only need to correctly exclude a subset of the state-actions that would cause failure, which we call the critical set.} }
Endnote
%0 Conference Paper %T On exploration requirements for learning safety constraints %A Pierre-François Massiani %A Steve Heim %A Sebastian Trimpe %B Proceedings of the 3rd Conference on Learning for Dynamics and Control %C Proceedings of Machine Learning Research %D 2021 %E Ali Jadbabaie %E John Lygeros %E George J. Pappas %E Pablo A. Parrilo %E Benjamin Recht %E Claire J. Tomlin %E Melanie N. Zeilinger %F pmlr-v144-massiani21a %I PMLR %P 905--916 %U https://proceedings.mlr.press/v144/massiani21a.html %V 144 %X Enforcing safety for dynamical systems is challenging, since it requires constraint satisfaction along trajectory predictions. Equivalent control constraints can be computed in the form of sets that enforce positive invariance, and can thus guarantee safety in feedback controllers without predictions. However, these constraints are cumbersome to compute from models, and it is not yet well established how to infer constraints from data. In this paper, we shed light on the key objects involved in learning control constraints from data in a model-free setting. In particular, we discuss the family of constraints that enforce safety in the context of a nominal control policy, and expose that these constraints do not need to be accurate everywhere. They only need to correctly exclude a subset of the state-actions that would cause failure, which we call the critical set.
APA
Massiani, P., Heim, S. & Trimpe, S.. (2021). On exploration requirements for learning safety constraints. Proceedings of the 3rd Conference on Learning for Dynamics and Control, in Proceedings of Machine Learning Research 144:905-916 Available from https://proceedings.mlr.press/v144/massiani21a.html.

Related Material