MACHINE LEARNING-BASED METHODS FOR DYNAMICS AND CONTROL
This course will discuss and study the utility of learning algorithms in dynamic systems and control applications. Approximately 1/6 of the course will study iterative learning control (useful in manufacturing, robotics, etc.), and the remainder split between shallow and deep neural networks for approximating and designing controllers and prediction algorithms (widely applicable in time series prediction), and a final portion spent on reinforcement learning and basic dynamic programming. It is meant to complement other data-driven courses in the department and add a dynamic systems and control flavor and perspective, and provide some survey of the state of the art via the projects. This course will mostly focus on linear and nonlinear ODE (difference, and differential equation) based systems as opposed to PDE-based ones.