Efficient Zero-Order NMPC with Feasibility and Stability Guarantees

Abstract

This paper discusses systems theoretic and computational aspects of a feasible, but suboptimal, nonlinear modelpredictive control scheme based on fixed sensitivities of thefunctions representing the constraints and cost of the underlying nonlinear programs. In particular, it will be shown how, byfreezing the sensitivities computed at the desired steady state ofthe system, an efficient, structure-exploiting scheme is obtainedthat can considerably speed up the computations required forboth construction and solution of the quadratic subproblems.Moreover, the local stability properties of the converged solutionare analysed using results on pseudoexpansions of generalizedequations present in the literature. The effectiveness of theproposed scheme is demonstrated on a non-trivial benchmarkwhere large speedups can be achieved.