基于类圆映射的多目标粒子群优化算法

Multi-objective particle swarm algorithm based on quasi-circular mapping

针对多目标优化求解的问题,提出一种基于类圆映射的多目标粒子群优化算法(qMOPSO)。首先,利用类圆映射技术将高维空间的解集映射到二维坐标平面,监控粒子种群的进化状态。其次,为了兼顾种群的收敛性和分布性,采用类圆占优和类圆扇块距离的概念设计了新的档案集管理策略。另外,根据种群分布熵变化情况,选择全局最优粒子,指导种群进化方向。最后,基于换维思想和淘汰机制,采用一种新的综合管理策略,提高种群寻优性能。所提算法采用三类测试函数和五种对比算法进行了对比实验。仿真实验证明,该方法是正确、有效的。

For the problem of multi-objective optimization,this paper proposed a multi-objective particle swarm optimization algorithm based on quasi-circular mapping(qMOPSO).Firstly,qMOPSO projected the solutions in high-dimensional objective space into the divided 2-dimensional quasi-circular space in order to monitor evolutionary status of the particle population.Then,the algorithm adopted quasi-circular dominant and individual distance based on sector blocks to design an archive maintaining algorithm which could balance diversity and convergence of solutions in the archive.In addition,it introduced an adaptive global best selection mechanism based on the solution distribution entropy to analyze the evolutionary tendency.Finally,the algorithm used a general management strategy based on changing dimension and elimination mechanism to enhance population management.qMOPSO was compared with five multi-objective optimization algorithms on three kinds of test suites.The simulation results show the correctness and effectiveness of the proposed approach.