基于参考点的高维多目标粒子群算法

高维多目标优化问题一般指目标个数为4个 或以上时的多目标优化问题.由于种群中非支配解数量随着目标数量的增加而急剧增多,导致进化算法的进化压力严重降低,求解效率低.针对该问题,提出一种基于粒子群的高维多目标问题求解方法,在目标空间中引入一系列的参考点,根据参考点筛选出能兼顾多样性和收敛性的非支配解作为粒子的全局最优,以增大选择压力.同时,提出了基于参考点的外部档案维护策略,以保持最后所得解集的多样性.在标准测试函数DTLZ2上的仿真结果表明,所提方法在求解高维多目标问题时能够得到收敛性和分布性都较好的解集.

Reference-point-based particle swarm optimization algorithm for many-objective optimization

Many-objective optimization problems generally refer to problems with four or more objectives. Because the number of nondominated solutions increases quickly with the increase of the number of objectives, the evolutionary pressure and efficiency of multi-objective evolutionary algorithms decrease heavily. To solve this problem, this paper proposes a reference-point-based particle swarm optimization algorithm for many-objective optimization. A structured set of reference points is generated in the objective space, and then the solutions which have both good convergence and diversity are selected as the global best to increase the selection pressure. Meanwhile, a truncation method for the external archive based on the reference points is proposed to maintain the diversity of the solution set. The results of simulations on the DTLZ2 test problem show that the proposed algorithm can obtain a solution set which has both good convergence and diversity for many-objective optimization problems.