Black-Box Control for Linear Dynamical Systems

Xinyi Chen, Elad Hazan
Proceedings of Thirty Fourth Conference on Learning Theory, PMLR 134:1114-1143, 2021.

Abstract

We consider the problem of black-box control: the task of controlling an unknown linear time-invariant dynamical system from a single trajectory without a stabilizing controller. Under the assumption that the system is controllable, we give the first {\it efficient} algorithm that is capable of attaining sublinear regret under the setting of online nonstochastic control. This resolves an open problem since the work of Abbasi-Yadkori and Szepesvari(2011) on the stochastic LQR problem, and in a more challenging setting that allows for adversarial perturbations and adversarially chosen changing convex loss functions. We give finite-time regret bounds for our algorithm on the order of $2^{poly(d)} + \tilde{O}(poly(d) T^{2/3})$ for general nonstochastic control, and $2^{poly(d)} + \tilde{O}(poly(d) \sqrt{T})$ for black-box LQR. To complete the picture, we investigate the complexity of the online black-box control problem and give a matching regret lower bound of $2^{\Omega(d)}$, showing that the exponential cost is inevitable. This lower bound holds even in the noiseless setting, and applies to any, randomized or deterministic, black-box control method.

Cite this Paper


BibTeX
@InProceedings{pmlr-v134-chen21c, title = {Black-Box Control for Linear Dynamical Systems}, author = {Chen, Xinyi and Hazan, Elad}, booktitle = {Proceedings of Thirty Fourth Conference on Learning Theory}, pages = {1114--1143}, year = {2021}, editor = {Belkin, Mikhail and Kpotufe, Samory}, volume = {134}, series = {Proceedings of Machine Learning Research}, month = {15--19 Aug}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v134/chen21c/chen21c.pdf}, url = {https://proceedings.mlr.press/v134/chen21c.html}, abstract = {We consider the problem of black-box control: the task of controlling an unknown linear time-invariant dynamical system from a single trajectory without a stabilizing controller. Under the assumption that the system is controllable, we give the first {\it efficient} algorithm that is capable of attaining sublinear regret under the setting of online nonstochastic control. This resolves an open problem since the work of Abbasi-Yadkori and Szepesvari(2011) on the stochastic LQR problem, and in a more challenging setting that allows for adversarial perturbations and adversarially chosen changing convex loss functions. We give finite-time regret bounds for our algorithm on the order of $2^{poly(d)} + \tilde{O}(poly(d) T^{2/3})$ for general nonstochastic control, and $2^{poly(d)} + \tilde{O}(poly(d) \sqrt{T})$ for black-box LQR. To complete the picture, we investigate the complexity of the online black-box control problem and give a matching regret lower bound of $2^{\Omega(d)}$, showing that the exponential cost is inevitable. This lower bound holds even in the noiseless setting, and applies to any, randomized or deterministic, black-box control method.} }
Endnote
%0 Conference Paper %T Black-Box Control for Linear Dynamical Systems %A Xinyi Chen %A Elad Hazan %B Proceedings of Thirty Fourth Conference on Learning Theory %C Proceedings of Machine Learning Research %D 2021 %E Mikhail Belkin %E Samory Kpotufe %F pmlr-v134-chen21c %I PMLR %P 1114--1143 %U https://proceedings.mlr.press/v134/chen21c.html %V 134 %X We consider the problem of black-box control: the task of controlling an unknown linear time-invariant dynamical system from a single trajectory without a stabilizing controller. Under the assumption that the system is controllable, we give the first {\it efficient} algorithm that is capable of attaining sublinear regret under the setting of online nonstochastic control. This resolves an open problem since the work of Abbasi-Yadkori and Szepesvari(2011) on the stochastic LQR problem, and in a more challenging setting that allows for adversarial perturbations and adversarially chosen changing convex loss functions. We give finite-time regret bounds for our algorithm on the order of $2^{poly(d)} + \tilde{O}(poly(d) T^{2/3})$ for general nonstochastic control, and $2^{poly(d)} + \tilde{O}(poly(d) \sqrt{T})$ for black-box LQR. To complete the picture, we investigate the complexity of the online black-box control problem and give a matching regret lower bound of $2^{\Omega(d)}$, showing that the exponential cost is inevitable. This lower bound holds even in the noiseless setting, and applies to any, randomized or deterministic, black-box control method.
APA
Chen, X. & Hazan, E.. (2021). Black-Box Control for Linear Dynamical Systems. Proceedings of Thirty Fourth Conference on Learning Theory, in Proceedings of Machine Learning Research 134:1114-1143 Available from https://proceedings.mlr.press/v134/chen21c.html.

Related Material