Data-driven Learning and Control - DDLC

Model-based control has played a vital role in many branches of engineering and sciences. The purpose of this talk is to present a different paradigm for control systems design. Instead of designing controllers from model, we learn desirable controllers directly from data, a new direction in control theory that arises from emerging applications in artificial intelligence and autonomous systems. Learning-based control is a direct control method aimed at developing computationally simple, analytically tractable (reinforcement) learning algorithms with guaranteed stability, robustness and optimality for the closed-loop system. In this talk, I will first review early developments in learning-based control for continuous-time linear and nonlinear systems with unknown dynamics. Then, I will present recent results in robustness of learning-based controllers. Finally, we illustrate the effectiveness of learning-based control via its applications to autonomous vehicles and biological motor control.