摘要
针对粒子群算法求解多目标问题极易收敛到伪Pareto前沿(等价于单目标优化问题中的局部最优解),并且收敛速度较慢的问题,提出一种ε占优的自适应多目标粒子群算法(εDMOPSO).在εDMOPSO算法中,每个粒子的邻居根据粒子的运行动态地组建,且粒子的速度不由其邻居中运行最好的粒子来调整,而是由其所有邻居共同调整.同时,采用外部存档保存非劣解,并利用ε占优更新非劣解.模拟结果表明了εDMOPSO算法的有效性.
Multi-objective particle swarm optimizers（MOPSOs） easily converge to a false Pareto front （the equivalent of a local optimum in single objective optimization）, and converge slowly when applied to solve multi-objective optimization problems（MOPs）. Therefore, this paper presents a self-adaptive multiobjective particle swarm optimizer based on e- domination（eDMOPSO） to handle MOPs. In the eDMOPSO algorithm, the neighborhood of each particle is dynamically changed in terms of the performances of the particles, and the velocity of each particle is not adjusted by the best performing particle in its neighborhood, but by all particles in its neighborhood including itself. Finally, external archive is employed to store the nondominated solutions and e-dominance is applied to update non-dominated solutions in external archive. Simulation results show the effectiveness of the proposed eDMOPSO algorithm.
出处
《控制与决策》
EI
CSCD
北大核心
2011年第1期89-95,共7页
Control and Decision
基金
广东省自然科学基金项目(9451806001002294)
深港创新圈基金项目(200810220137A)
贵州省教育厅社科基金项目(2007018)
关键词
多目标优化
粒子群算法
ε占优
动态邻居
multi-objective optimization： particle swarm optimizer： e-domination： dynamic neighbor topology
