Feasibility issues in linear model predictive control

Abstract The problem of infeasibility caused by inconsistent hard state constraints in model predictive control is considered. The difficulties that may arise are analyzed, and the inherent trade‐off a controller must address at times of infeasibility is shown. From a multiobjective viewpoint, it is demonstrated that the two approaches for handling infeasibility documented in the literature have significant limitations. Two new techniques for handling infeasibility that overcome the difficulties encountered with current approaches are presented.