Explicit Dead-Time Compensation in Linear Parameter Varying Model Predictive Control

Abstract

Model Predictive Control (MPC) is able to directly deal with dead-time (DT) phenomena. Yet, implicit delay compensation heavily affects computational aspects of these algorithms, and stability and feasibility analyses become numerically tougher. The Linear Parameter Varying (LPV) framework has been shown suitable to model complex, nonlinear dynamics, with corresponding MPC algorithms being developed over the last few years. Thus, we propose a novel MPC method for DT LPV systems, using a DT-free model. The scheme explicitly accounts for the DT via a compensation loop, thus avoiding augmented state-space models. The algorithms is able to ensure input-to-state stability, recursive feasibility, and robust constraint satisfaction w.r.t. model-mismatches and delay estimation uncertainties. A solar collector benchmark example is used to illustrate the advantages of the method, which is compared against a regular LPV MPC algorithm (with standard implicit DT compensation).