嵌入Circle映射的混合策略多目标粒子群算法

A Hybrid Strategy Multi-Objective Particle Swarm Optimization Embedded with Circle Mapping

多目标粒子群算法具有收敛速度快、原理简单和易于实现等优点,被广泛应用于解决多目标优化问题。然而,它存在容易过早收敛、陷入局部最优等缺点。针对上述问题,提出了一种嵌入Circle映射的混合策略多目标粒子群算法(CMEMOPSO)。当粒子当前位置与其个体历史最优位置互不支配时,在一定概率下,利用Circle映射调整粒子位置,使其找到更优的解。同时,若外部存档达到预定阈值,则使用结合个体密度和拐点距离的混合评价指标评估非劣解的综合性能,去除较差的非劣解以实现对外部存档的更新,提高算法的综合性能。最后,利用所提出的算法在12个典型测试函数上进行实验。实验结果表明CMEMOPSO具有良好的收敛性和更快的收敛速度,它在大多数测试函数上优于其他比较算法。

Multi-objective particle swarm optimization has the advantages of fast convergence speed,simple principle and easy implementation,so it is widely used to solve multi-objective optimization problems.However,it is easy to converge prematurely and fall into local optimum.To solve these problems,a hybrid strategy multi-objective particle swarm optimization embedded with Circle mapping(CMEMOPSO)was proposed.When the current position of a particle and its individual historical optimal position do not dominate each other,circle mapping is used to adjust the position of the particle to find a better solution under a certain probability.At the same time,if the external archive reaches the predetermined threshold,a blend evaluation index combining individual density and inflection point distance is used to evaluate the comprehensive performance of non-dominated solutions,and poor non-dominated solutions are removed to realize the update of the external archive and effectively improve the comprehensive performance of the algorithm.Finally,the proposed algorithmis used to performexperiments on 12 typical test functions.Experimental results showthat CMEMOPSOhas good convergence and faster convergence rate,and it is superior to other comparison algorithms on most test functions.