Seminar: Jack Umenberger, Sept 19th 2019

When: Thurs 19th Sept, 1pm.

Where: Rose St Building seminar area

Title: Learning robust LQ-controllers using application oriented exploration

Abstract: This talk concerns the problem of learning robust linear quadratic (LQ) controllers, when the dynamics of the linear system are unknown. First, we propose a robust control synthesis method to minimize the worst-case LQ cost, with high probability, given empirical observations of the system. Next, we propose an approximate dual controller that simultaneously regulates the system and reduces model uncertainty. The objective of the dual controller is to minimize the worst-case cost attained by a new robust controller, synthesized with the reduced model uncertainty. The dual controller is subject to an exploration budget; i.e., a limit on the allowable worst-case cost incurred during exploration, given our current understanding of the system uncertainty. Numerical experiments demonstrate superior performance of the proposed robust LQ-controller over existing methods. Moreover, the dual control strategy is observed to significantly outperform common epsilon-greedy random exploration strategies.

Contacts

Sydney Institute for Robotics and Intelligent Systems
info@acfr.usyd.edu.au