Title:
Adaptive Model Predictive Control: Robustness, Performance Enhancement and Parameter Estimation
Abstract
Control algorithms that combine online model identification with optimization of predicted performance have been a focus of research since the origins of Model Predictive Control (MPC) 40 years ago. However few control strategies based on MPC with online model identification provide guarantees of robust performance and constraint satisfaction. Recent developments in robust predictive control, set-based identification and convex optimization have led to a resurgence of interest in this direction.
This talk will outline recent work on computationally tractable robust adaptive MPC formulations for systems with uncertain models, additive disturbances, and state and control constraints. The approach has the potential to overcome a fundamental limitation of robust MPC, namely that the amount of uncertainty in system models and unknown disturbances restricts the achievable performance, even though model uncertainty can often be reduced by making use of information acquired during the closed loop operation of the controller.
We will explore conditions for parameter convergence and connections with the well-known dual control problem. Implications for safety and robustness in machine learning will be discussed and a framework for general model classes using imitation learning techniques will be described. The talk will also discuss how to balance the often conflicting requirements for achieving good tracking performance and improving parameter estimates by introducing convex constraints that ensure persistency of excitation.