Optimal prioritized infeasibility handling in model predictive control: Parametric preemptive multiobjective linear programming approach (Q5949891)

A strategy which relaxes optimally an infeasible model-based predictive control (MPC) problem with prioritized constraints into a feasible one by solving an additional single-objective linear program (LP) is proposed. It is shown that in the case when all constraints have different priorities, the sequence of LP problems can be reduced to a single LP by proper selection of weights. Moreover, it is possible to design the weights such that the optimum becomes equal to the lexicographic minimum of the set of feasible constraint violations for all possible right-hand sides of the constraints. These weights are obtained off-line in two stages. First, a finite set of preemptive optimal bases (such that all possible right-hand sides of the constraints are associated with one of these bases) is computed. In a second stage, weights (which satisfy optimally the given prioritization) are calculated from this set of bases. The existence of optimal weights is proved under nonrestrictive assumptions. As an illustration, the optimal weight design problem is solved for an idealized mass-spring system.