HANDLING DYNAMIC MULTIOBJECTIVE PROBLEMS WITH PARTICLE SWARM OPTIMIZATION

In this paper the hyperplane distribution and Pareto dominance were incorporated into a particle swarm optimization algorithm in order to allow it to handle dynamic multiobjective problems. When a change in a dynamic multiobjectve function is detected, the proposed algorithm reinitializes (in different ways) the PSO's velocity parameter and the archive where the non-dominated solutions are beeing stored such that the algorithm can follow the dynamic Pareto front. The proposed approach is validated using two dynamic multiobjective test functions and an standard metric taken from the specialized literature. Results indicate that the proposed approach is highly competitive which can be considered as a viable alternative in order to solve dynamic multiobjective optimization problems.