Use of barrier functions for handling operating constraints in predictive control

This paper discusses the problem of handling output operating constraints in model-based predictive control. Possible absence of compatibility of output constraints requires special ways of treating them. The use of internal penalty (or barrier) functions is suggested. Unlike external penalty functions, they ensure that the constraints arc· satisfied although do not fully prevent infeasibility. Therefore, an approach based on the combined penalty functions as a way to handle infeasibility is also suggested. These techniques are used in combination with generalised predictive control (GPC) and with stabilising constrained receding-horizon predictive control (CRHPC).