DMulti-MADS: mesh adaptive direct multisearch for bound-constrained blackbox multiobjective optimization

The context of this research is multiobjective optimization where conflicting objectives are present. In work, these only available as the outputs a blackbox for which no derivative information available. This work proposes new extension mesh adaptive direct search (MADS) algorithm to derivative-free with bound constraints. method does not aggregate and keeps list non dominated points converges (local) Pareto set long unfolds. As in single-objective MADS algorithm, built around step poll step. Under classical assumptions, it proved that so-called DMulti-MADS generates multiple subsequences iterates converge local stationary points. Finally, computational experiments suggest approach competitive compared state-of-the-art algorithms optimization.