An effective solution to finding global best guides in particle swarm for typical MOPs

It is of critical importance for convergence and diversity of final solutions that finding out a feasible global best guide for each particle of the current swarm in multi-objective particle swarm optimization (MOPSO). An improved approach for determining the best local guide in MOPSO is proposed, where the Pareto archive with size limit is used to store the non-dominated solutions. While selecting the local best particle, a random number is used to judge whether the crowding distance is taken into account or not. A new solution is referred to overcome the problem that it is much harder to generate a new particle dominating the original one in MOPs than in single-objective optimal problems. In addition, to improve the efficiency of search and avoid precocity, the inertial weight changes in the iteration process. The proposed approach is applied to some typical testing functions, and the experimental results of Pareto fronts for these functions are satisfied.