T-adaptive an Online Tuning Technique Coupled to MOEA/D Algorithm: A Comparative Analysis with Offline Parameter Tuning Techniques

Appropriate design decisions and finding effective parameter values are well-known obstacles to improving the performance of metaheuristics. Therefore, the adjustment of parameters to obtain better-quality solutions is a relevant task. In this chapter, two different parameter adjustment methods, well known in the literature, are compared against our proposal, an adaptive method called T-adaptive, which is coupled to the MOEA/D multiobjective algorithm to solve the Electric Fleet Size and Mix Vehicle Routing Problem with Time Windows and Recharging Stations (E-FSMFTW). T-adaptive and DOE Taguchi provide the best results in the analyzed test cases. One advantage of both techniques, compared to EVOCA, is that they do not increase the problem complexity in the tuning process, as EVOCA does, since no new parameters are required.