ABSTRACT
This chapter looks at how a Model Predictive Control algorithm can be modified in the event that not all the desired constraints can be satisfied. It discusses how infeasiblility arises and various mechanisms for avoiding it or tackling it. Feasibility is usually a term applied to optimisation problems and describes whether a solution exists. Hard constraints are constraints which must be satisfied. For instance, they may be limits on actuators or on valves open. Soft constraints are those which should be satisfied if possible. Clearly terminal constraints are a form of soft constraint in that they are applied on far future predictions and arise from the desire for a guarantee of stability. For the certain case recursive feasibility is relatively straightforward to establish, as it can be shown that the maximal admissible set is a possible choice for the terminal region.
