Evolutionary Algorithms For Multi-Objective Optimization Problems

There are many multi-objective optimization problems which objectives often conflict with each other in the engineering problems. No single solution can be found to optimize all the objectives simultaneously. This situation is different from the one in single objective optimization problems. Thus, for multi-objective optimization problems, the key issue is to find a set of well distributed and close-to-Pareto-optimal solutions. According to their advantages, evolutionary algorithms for multi-objective optimization problems have become a hot research topic.The main works in this thesis are as follows:First, note that the distribution of the initial population directly affects the search results of the evolutionary algorithms. In this thesis, uniform design and chaos mapping are jointly employed to generate the initial population. The individuals in the generated population are scattered uniformly over the feasible decision space, and also have better diversity which is good for an algorithm to explore the decision space. Moreover, a new crossover operator is designed, which can improve the search ability of the proposed algorithms by using property of chaos. Based on these two schemes, we propose two multi-objective optimization evolutionary algorithms named UCMOEA with Logistical mapping and USMOEA with Sinusoidal mapping. The simulation results indicate that the non-dominated solutions obtained have better quality and good distribution.Second, a new multi-objective optimization evolutionary algorithm with local search named LSMOEA is proposed. This algorithm can find the sparse area of the non-dominated front. Concretely, the search was especially focused on the sparse area using uniform design method, and finally the non-dominated solutions which have been found can scatter uniformly along the non-dominated front. The simulation results indicate the effectiveness of the proposed algorithm.