N-GKLS: An NMPC problem generator and a test platform for NMPC solvers

Abstract

Nonlinear Model Predictive Control (NMPC) solves a series of optimization problems online during the entire operational time of the controlled plant, and the performance of NMPC depends on the superiority of the solver embedded in NMPC. Traditionally, gradient-based deterministic methods are used as NMPC solvers, but metaheuristics are now becoming more and more popular for its ability to find the global optimal solution. In order to verify the superiorities of different algorithms to be NMPC solvers, a novel NMPC problem generator named NMPC-based GKLS generator (N-GKLS) is proposed. With the help of Optimal-Replace Method (ORM), a reliable comparison of the closed-loop performances result from receding horizon strategy in NMPC is guaranteed. In addition, comparison techniques named accumulated operational characteristic and accumulated operational zone are proposed to directly compare different optimization algorithms for solving various NMPC problems generated by N-GKLS. An auxiliary comparison table is also proposed to report numerical comparative results. In this way, N-GKLS is not only an NMPC problem generator, but also a test platform for verifying the superiorities of different algorithms as NMPC solvers. Finally, simulation results based on 8 metaheuristics and 3 deterministic methods are used to illustrate the effectiveness of the proposed N-GKLS. A corresponding MATLAB APP demo is designed and is available online: https://github.com/JiahongXu123/N-GKLS.