An External Archive Guided Adaptive Multiobjective Evolutionary Algorithm

Multi-objective optimization problems(MOPs) are very pervasive in the real world applications. MOPs aim to optimize multiple conflicting objectives simultaneously. Different from single objective optimization which has only one optimal solution, a set of trade-off solutions exist in a multi-objective optimization problem. Evolutionary algorithms is a kind of intelligent optimization algorithm which simulates the process of biological evolution. Compared with traditional deterministic algorithms, evolutionary algorithms are more suitable for solving complex NP-hard MOPs. This thesis proposes an external archive guided adaptive multi-objective Evolutionary Algorithm for MOPs. Specifically, this paper includes the following aspects.First, the advantages and disadvantages of two classes of multiobjective evolutionary algorithms(MOEAs), decomposition and domination based ones, are introduced and analyzed.Second, a framework that adopts both decomposition and domination based approaches is proposed. This framework uses two population: one is the working population, and another is the external archive. Decomposition and domination based MOEAs are applied to two populations, respectively. Comparative experiments show that the hybrid algorithm has the better performance than each of any single algorithms.Third, this paper further uses the external archive to guided the search process of the working population. This new adaptive MOEA, named external archive guided multiobjective evolutionary algorithm based on decomposition(EAG-MOEA/D), use the information obtained from the external archive to guide the search direction for the working population: the external archive allocates the computational resources to each subproblem in the working population in an adaptive manner, based on its previous performance in the external archive. The experiments show that the performance of EAG-MOEA/D is better than other compared algorithms. In addition, the effects of the external archive are analyzed in details.