Non-dominated Sorting Based Fireworks Algorithm for Multi-objective Optimization

Multi-objective optimization is one of the most important problem in the mathematical optimization. Some researchers have already proposed several multi-objective fireworks algorithms, of which S-metric based multi-objective fireworks algorithm (S-MOFWA) is the most representative work. S-MOFWA takes the hypervolume as the evaluation criterion of external archive updating, which is easy to implement but ignores the landscape information of the population. In this paper, a novel multi-objective fireworks algorithm named non-dominated sorting based fireworks algorithm (NSFWA) is proposed. The proposed algorithm updates the external archive with the selection operator based on the fast non-dominated sorting approach, which is specially designed for the spark generation characteristic of FWA to improve the diversity. A multi-objective guided mutation operator is also designed to enhance the efficiency of population information utilization and improve the search capability of the algorithm. Experimental results on the benchmarks demonstrate that NSFWA outperforms other multi-objective swarm intelligence algorithms of S-MOFWA, NSGA-II and SPEA2.