A novel coevolutionary multi-objective particle swarm optimization based on decomposition

To improve the performance of particle swarm optimization (PSO) and balance convergence and diversity, we propose a coevolutionary multi-objective particle swarm optimization based on decomposition (CMOPSO). CMOPSO includes 3 strategies, coevolutionary mechanism, best-effort strategy and weighted maximum approach. Coevolutionary mechanism is used to maintain convergence, while PSO operator focuses on diversity. Best-effort strategy allows operators that perform well enough to execute again, which improves the utilization of computing resources. Weighted maximum approach is an environmental selection strategy based on decomposition, which selects by comparing the maximum of weighted objective values of the subproblem. Each new individual will be compared with the best individual of all subproblems, not only in the sub domain, which helps the PSO operator to maintain the diversity in search process. The CMOPSOD is tested against 6 other algorithms on the ZDT and UF test problems, the results show that the proposed CMOPSOD has significant advantages in terms of convergence and diversity.