On the design of robust model predictive controllers for constrained processes

he development of a framework for robust stability and performance analysis of Constrained Model Predictive Control (CMPC) for linear time invariant and open-loop stable processes, leading to a robust design methodology for CMPC is the main purpose of this dissertation. CMPC is a process model based control law that utilizes on-line optimization subject to possible constraints on process variables. For a computer controlled system based on the CMPC algorithm, the feedback controller is a piecewise linear operator because of the constraints. This piecewise linear operator can be thought of as an array of linear feedback controllers in parallel, handling different types of predicted active constraint situations. The corresponding control law is constructed to include "uncertainty" blocks which are obtained by decomposition of the terms corresponding to active constraint situations, to allow one to address robust analysis and design.;All practically relevant active constraint situations in a CMPC closed-loop system can be found by search algorithms developed in this dissertation. Also, an algorithm is constructed for identifying and separating the corresponding controllers for these active constraint situations to groups that are independent or dependent on the tuning parameters of the on-line optimization.;A necessary condition for contraction is established as an effective indicator for the CMPC closed-loop robust stability. The "uncertainty" blocks of the control law are one-side real-valued uncertainties in the CMPC closed-loop system. The computational technique of structured singular value with phase information is used to analyze the necessary contraction stability condition of the CMPC closed-loop system after incorporation of model-plant mismatch. In addition, a computer-aided off-line tuning method is developed to automatically satisfy the necessary contraction condition by solving a minimum maximum problem.;In this dissertation, some additional properties of CMPC are also revealed with implications for the dynamic resiliency of constrained processes. Methods are also developed for studying the quadratic stability and strongly...