Multi-objective optimal computing budget allocation for multi-objective particle swarm optimisation with particle-dependent weights

In this paper, we develop a multi-objective optimal computing budget allocation method with multiple weights (MOCBAmw) assigned to each particle in multi-objective particle swarm optimisation based on weighted scalarising functions (MPSOws) algorithm under the stochastic environment. By intelligently allocating computing budget among all particles instead of simple equal allocation (EA), we are able to improve the probability of correctly selecting the global best designs under limited computing budget. Improvement of correct leading particles identification in each generation of the MPSOws procedure helps to facilitate the convergence of the swarm to the Pareto front under the stochastic environment. Test results from bi-objective ZDT problems and tri-objective DTLZ problems have shown that MOCBAmw achieves a better convergence rate and a higher hypervolume than EA under the same noise setting.