Online greedy identification of linear dynamical systems

TL;DR

This paper introduces an online greedy control policy for linear dynamical systems that maximizes information gain per step, offering a low-complexity alternative with competitive performance in limited-trial settings.

Contribution

It proposes a novel online greedy approach for exploration in linear dynamical systems, emphasizing low complexity and effectiveness with few experiments.

Findings

Low computational complexity compared to gradient-based methods

Experimentally competitive performance in limited trials

Effective information maximization per control step

Abstract

This work addresses the problem of exploration in an unknown environment. For linear dynamical systems, we use an experimental design framework and introduce an online greedy policy where the control maximizes the information of the next step. In a setting with a limited number of experimental trials, our algorithm has low complexity and shows experimentally competitive performances compared to more elaborate gradient-based methods.