Grant: Discovery Grants Program - Individual

Machine learning is ubiquitous in most applications of the modern world. However, the use of machine learning for safety-critical control systems has still not found widespread use. This is due to the lack of formal results capable of proving safety, stability, performance, and robustness of the closed-loop control system when using machine learning algorithms. Based on this shortcoming, the proposed research will focus on controller and estimator design for nonlinear dynamical systems based on machine learning algorithms. In particular novel neural networks and reinforcement learning algorithms will be developed to design controllers that can stabilize a nonlinear dynamical system and offer performance  and robustness guarantees, as well as to design state estimators with error convergence guarantees. The focus of the research will be on the formulation of control and estimation problems as convex optimization programs, which can be solved efficiently both from a theoretical and practical point of view. The areas of advanced air mobility and energy systems will be the main applications of this research.