The Nonstochastic Control Problem
arXiv:1911.12178
Abstract
We consider the problem of controlling an unknown linear dynamical system in the presence of (nonstochastic) adversarial perturbations and adversarial convex loss functions. In contrast to classical control, the a priori determination of an optimal controller here is hindered by the latter's dependence on the yet unknown perturbations and costs. Instead, we measure regret against an optimal linear policy in hindsight, and give the first efficient algorithm that guarantees a sublinear regret bound, scaling as T^{2/3}, in this setting.
To appear at Algorithmic Learning Theory (ALT) 2020; small revisions from the last ver
References in corpus (7)
- Logarithmic Regret for Online Control
- Model-Free Linear Quadratic Control via Reduction to Expert Prediction
- Learning Without Mixing: Towards A Sharp Analysis of Linear System Identification
- Learning Linear Dynamical Systems with Semi-Parametric Least Squares
- Learning Linear-Quadratic Regulators Efficiently with only Regret
- Online Linear Quadratic Control
- Semiparametric Contextual Bandits
Cited by in corpus (7)
- Logarithmic Regret Bound in Partially Observable Linear Dynamical Systems
- Information Theoretic Regret Bounds for Online Nonlinear Control
- Non-Stochastic Control with Bandit Feedback
- Online Agnostic Boosting via Regret Minimization
- Streaming Linear System Identification with Reverse Experience Replay
- Generating Adversarial Disturbances for Controller Verification
- Geometric Exploration for Online Control