-domination based Random Elitist and Non- uniform Domination Strategy

MOEA/D multi-objective optimization algorithm features the shortcoming of losing partial of the excellent individuals when it updates sub-problems and its inefficiency of convergence speed. In order to overcome such shortcoming, we propose a multi-objective evolutionary optimization algorithm based on adaptive epsilon-domination and random elitist strategy in this paper. This algorithm uses archive population that is updated by adaptive epsilon-domination to achieve the optimization goals. The algorithm is able to keep non-inferior solutions, reduce the losing of excellent individuals in the evolution process and hold archive population to maintain a certain size, and ensure the convergence speed and the uniformity of the distribution of noninferior solutions; in addition, the algorithm uses the archive population to update each subproblem of the evolution population at certain probability and the number of domination so as to speed up the convergence speed. The experiment results show that the new algorithm proposed by us is more effective than MOEA/D and NSGA-II in ensuring the uniformity of the distribution of non-inferior solution and the convergence speed for Multi-objective optimization.