Data Driven & Learning Based Control System

Controlling systems with unknown dynamics is challenging in general. Usually, the data or the distribution governing the system uncertainty is known. My research interest lies in studying algorithms that address uncertainties which might be adversarial (as in the H infinity setting) or stochastic (as in the H2 setting). Specifically, I am interested in studying the regret associated with any learning based control algorithms which might operate to maximise their performance compared to an optimal controller or a sequence of best control actions available in hindsight.