A systematic selection of an alternative parameterisation for predictive control

Alternative parameterisations have been shown to improve the feasible region for a predictive control law when the number of degrees of freedom is limited. One question yet to be resolved is: which alternative parameterisation is best for a particular problem and what choice of parameter(s) within each parameterisation will lead to an improved feasible region with good performance? This paper tackles this question and demonstrates two systematic approaches to select the best alternative parameterisations. These approaches are based on multiobjective optimisation and a pragmatic selection. Numerical examples demonstrate the efficacy of both methods.