Enhancing MOEA with component-emphasizing mechanism for multi-objective optimization

Multi-objective optimization is an important and challenging topic in the field of industrial design and scientific research because real-world problems usually involve several conflicting objectives. Since a multi-objective evolutionary algorithms (MOEA) is able to obtain an approximation to the Pareto optimal set and provide substantial information of the tradeoff between objectives, it is becoming one of the most successful methods for multi-objective optimization. Usually, an MOEA generates new trial solutions (offspring) with some candidate decision vectors (parents) to search for the promising areas and make the population evolve towards the Pareto optimal set. Moreover, in the reproduction procedure, most of MOEAs view the decision vector as a whole, and do not recognize the effects of a single component on the new trial solutions. In this paper, we propose the component-emphasizing mechanism for enhancing the search ability of MOEAs. In this mechanism, each component of a decision vector is viewed as an independent factor affecting the quality of the solution. Based on the mechanism, a new MOEA is presented. Finally, the performance of this new algorithm is compared with two other promising MOEAs, namely, NSGA-II and GDE3, on a set of test instances. The experimental results have shown that the proposed algorithm outperforms the others in solution quality and time cost.