Data-driven memory-dependent abstractions of dynamical systems

Adrien Banse, Licio Romao, Alessandro Abate, Raphael Jungers
Proceedings of The 5th Annual Learning for Dynamics and Control Conference, PMLR 211:891-902, 2023.

Abstract

We propose a sample-based, sequential method to abstract a (potentially black-box) dynamical system with a sequence of memory-dependent Markov chains of increasing size. We show that this approximation alleviates a correlation bias that has been observed in sample-based abstractions. We further propose a methodology to detect on the fly the memory length resulting in an abstraction with sufficient accuracy. We prove that, under reasonable assumptions, the method converges to a sound abstraction in some precise sense, and we showcase it on two case studies.

Cite this Paper

BibTeX
@InProceedings{pmlr-v211-banse23a, title = {Data-driven memory-dependent abstractions of dynamical systems}, author = {Banse, Adrien and Romao, Licio and Abate, Alessandro and Jungers, Raphael}, booktitle = {Proceedings of The 5th Annual Learning for Dynamics and Control Conference}, pages = {891--902}, 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/banse23a/banse23a.pdf}, url = {https://proceedings.mlr.press/v211/banse23a.html}, abstract = {We propose a sample-based, sequential method to abstract a (potentially black-box) dynamical system with a sequence of memory-dependent Markov chains of increasing size. We show that this approximation alleviates a correlation bias that has been observed in sample-based abstractions. We further propose a methodology to detect on the fly the memory length resulting in an abstraction with sufficient accuracy. We prove that, under reasonable assumptions, the method converges to a sound abstraction in some precise sense, and we showcase it on two case studies.} }
Endnote
%0 Conference Paper %T Data-driven memory-dependent abstractions of dynamical systems %A Adrien Banse %A Licio Romao %A Alessandro Abate %A Raphael Jungers %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-banse23a %I PMLR %P 891--902 %U https://proceedings.mlr.press/v211/banse23a.html %V 211 %X We propose a sample-based, sequential method to abstract a (potentially black-box) dynamical system with a sequence of memory-dependent Markov chains of increasing size. We show that this approximation alleviates a correlation bias that has been observed in sample-based abstractions. We further propose a methodology to detect on the fly the memory length resulting in an abstraction with sufficient accuracy. We prove that, under reasonable assumptions, the method converges to a sound abstraction in some precise sense, and we showcase it on two case studies.
APA
Banse, A., Romao, L., Abate, A. & Jungers, R.. (2023). Data-driven memory-dependent abstractions of dynamical systems. Proceedings of The 5th Annual Learning for Dynamics and Control Conference, in Proceedings of Machine Learning Research 211:891-902 Available from https://proceedings.mlr.press/v211/banse23a.html.

Related Material