(150b) The Next Epoch of Model Predictive Control: Exploiting Machine Learning Methods for Approximation and Design

Model predictive control (MPC) is an optimization-based predictive control strategy that has become very popular in the last few decades due to its broad capabilities that include accounting for systems with multivariable and interacting dynamics, nonlinearities, and constraints. Furthermore, its versatility in terms of its ability to provide robustness guarantees (by design) as well as consideration of economically oriented control objectives, MPC is increasingly being applied in many emerging systems including chemical process systems, energy systems, robotics, unmanned vehicles, and biomedical systems, to name a few. Yet, many MPC applications face important challenges related to the difficulty of modeling complex systems and the need for more advanced MPC strategies to provide provably safe and robust performance with low online computational and memory requirements. The proliferation of machine learning (ML) and artificial intelligence (AI), combined with the availability of increased computational and sensing capabilities in modern control systems, has led to a rapidly growing interest in the use of learning-based methods in MPC. While most research in this area has focused on the construction of the internal prediction model, the opportunities for ML/AI extend far beyond model improvement and adaptation. In this tutorial session, we will provide an overview of recent efforts for fusing ML/AI and MPC as well as discuss potential future research opportunities in this rapidly evolving field.