摘要

In the real-life multiobjective optimization, it is often required to generate a well-distributed Pareto set. Only a few methods are capable of tackling such a problem in a quite general formulation. The Directed Search Domain method (DSD) proved to be efficient and, therefore, attracted much attention. In this paper, two modifications to the rotation strategy of DSD algorithm are proposed. The first modification is meant to improve its computational efficiency. The second modification is to enhance the evenness of the Pareto set with a number of additional Pareto points. These points are obtained according to some specific rotation angles calculated in a particular way. The modified approach is verified on several test cases with three objectives, including an engineering case. The proposed algorithm is compared with both the original DSD and DSD-II algorithms. It is shown that the new approach can maintain the distribution of the Pareto set at a high level with a relatively low computational cost.

  • 出版日期2018-2
  • 单位中国空间技术研究院