Comparisons and combinations of interpolation methods with conventional predictive control

Many papers have demonstrated that interpolation methods can often enlarge feasible regions significantly while also reducing computational complexity in MPC algorithms (Bacic et al, 2003; Rossiter et al, 2004). However, the simplest interpolations give non-convex feasible regions and thus may have limited benefit in some state directions. This paper presents a new interpolation method which combines the advantages of simple interpolations with conventional predictive control methods to enable a better compromise balance between the volume and shape of the feasible region and the number of degrees of freedom required.