Section: New Results

Particle Swarm Optimization based Approach for Model Predictive Control Tuning

Participants : Mohamed Lotfi Derouiche [CentraleSupelec, L2S, Ecole Nationale d'Ingénieur de Tunis] , Guillaume Sandou [correspondent] , Soufiene Bouallegue [Ecole Nationale d'Ingénieur de Tunis] , Joseph Haggège [Ecole Nationale d'Ingénieur de Tunis] .

In this work, a new Model Predictive Controller (MPC) parameters tuning strategy is proposed using a perturbed Particle Swarm Optimization (pPSO) approach. This original LabVIEW implementation of this metaheuristic algorithm is firstly validated on some test functions in order to show its efficiency and validity. The optimization results are compared with the standard PSO as well as a LabVIEW implemented Genetic Algorithm (GA) approaches. The parameters tuning problem, i.e. the weighting factors on the output error and input increments of the MPC algorithm, is after that formulated and systematically resolved, using the proposed LabVIEW pPSO algorithm. The case of a Magnetic Levitation (MAGLEV) system is investigated to illustrate the robustness and superiority of the proposed pPSO-based tuning MPC approach. All obtained simulation results, as well as the statistical analysis tests, are compared and discussed in order to improve the effectiveness of the proposed pPSO-based MPC tuning methodology.