A Novel Epsilon-Dominance Multi-objective Evolutionary Algorithms for Solving DRS Multi-objective Optimization Problems
Liu Liu, Minqiang Li, Dan Lin · 2007
A new kind of multiobjective optimization model is constructed in this paper, which contains various solutions apart from the true Pareto-optimums but hardly dominated. These solutions are defined as dominance resistant solutions (DRSs). It is proved that the evolutionary algorithms based on Pareto- dominance relationship fail to find the true Pareto fronts for the DRS MOP. Hence a new algorithm based on epsiv-dominance relationship, called epsiv-dominance MOEA (EDMOEA), is proposed to improve the DRSs in population effectively. Finally, experiments on a set of DRS MOOPs and other regular test functions are conducted, the EDMOEA outperforms the NSGA-II, and can be applied easily to complex multiobjective optimization problems.