Space Congestion Control Based Multi-objective Evolutionary Optimization

Evolutioary algorithm achieves excellent performance in handling multi-objective optimization problem, which makes it become one hotspot in the filed of multi-objective optimization. In order to improve the capacity of solving complexity multi-objective optimization problem, and based on the theory of multi-objective optimization and the method of evolutionary computation,this paper payed key research on how to improve the effect of the congestion control in the objective-space, so that the capacity of diversity-keeping can be improved during the environmental selection, uniformity and spread of the solutions as well. The main contents are listed below:Firstly, this paper introduces the basic concept of multi-objective optimization problem and its research significance. The relevant theories of evolutionary multi-objective optimization algorithm were given in chapter 2, and the design goals and the key points of the former algorithms were analyzed in detail. In addition, the benchmark problems and performance metric of the multi-objective optimization algorithms followed.Secondly, this paper proposed a cooperative multi-objective optimization algorithm with polymorphous populations. The new approach designs a novel coevolutionary frame for polymorphous populations. In addition, by introducing the minimum vectorial angle, which is capable of measuring the similarity between different Pareto-ranked solutions, a novel selection strategy of suboptimum non-dominated solutions was proposed to enhance the diversity of populations.Furthermore, the new approach puts forward a new population deletion tactic, which is based on an ordered link-list. The tactic further improves the uniformity and spread of the solutions.Thirdly, for many-objective optimization problem, the proportion of the non-dominated individuals increases dramatically with the increase of target dimension, which may seriously reduce the population evolutionary pressures. In order to efficiently control the congestion among the very large numbers of non-dominated solutions and improve its diversity, and on the basis of previously mentioned minimum vectorial angle, this paper firstly defined the concept of open angle, based on which a novel space congestion control strategy proposed. It’s called CCSOA here, along with the adaptive adjustment strategy for the open angle threshold and the introduction of buffer-pool strategy. The new approaches improved the capability of evolutionary algorithm for solving many-objective optimization problem.Finally, the thesis research work is summarized, some views on its research prospects are further suggested.