Geometric Exploration for Online Control
Abstract
We study the control of an \emph{unknown} linear dynamical system under general convex costs. The objective is minimizing regret vs the class of strongly-stable linear policies. In this work, we first consider the case of known cost functions, for which we design the first polynomial-time algorithm with $n^3\sqrt{T}$-regret, where $n$ is the dimension of the state plus the dimension of control input. The $\sqrt{T}$-horizon dependence is optimal, and improves upon the previous best known bound of $T^{2/3}$. The main component of our algorithm is a novel geometric exploration strategy: we adaptively construct a sequence of barycentric spanners in an over-parameterized policy space. Second, we consider the case of bandit feedback, for which we give the first polynomial-time algorithm with $poly(n)\sqrt{T}$-regret, building on Stochastic Bandit Convex Optimization.
BibTeX
@inproceedings{NEURIPS2020_565e8a41,
author = {Plevrakis, Orestis and Hazan, Elad},
booktitle = {Advances in Neural Information Processing Systems},
editor = {H. Larochelle and M. Ranzato and R. Hadsell and M.F. Balcan and H. Lin},
pages = {7637--7647},
publisher = {Curran Associates, Inc.},
title = {Geometric Exploration for Online Control},
url = {https://proceedings.neurips.cc/paper_files/paper/2020/file/565e8a413d0562de9ee4378402d2b481-Paper.pdf},
volume = {33},
year = {2020}
}