混合分解的多目标粒子群优化算法

Hybrid Particle Swarm Algorithm for Multi-Objective Optimization Based on Decomposition

针对多目标粒子群算法全局最优值的选取缺陷以及多样性保留缺陷,提出了一种基于分解和拥挤距离的多目标粒子群优化算法(Smoeadpso).算法采用切比雪夫分解机制,将邻居向量对应的子问题的中的最优解来作为某个粒子全局最优值的候选解了更有效限制粒子飞行速度以避免粒子飞行超出解空间界限,引入了新的速度限制因子维持了种群多样性.本文算法与经典的多目标进化算法在10个测试函数上的对比结果表明,Smoeadpso求得的Pareto解集与真实Pareto解集的逼近程度有明显提升并且对于3目标问题求解的均匀性也比同类粒子群算法优秀.

To deal with the problems of the way for selecting the global best position and reserve the diversity, a multi-objective particle swarm optimization algorithm based on decomposition and crowding distance was proposed. We introduced the Tchebycheff decompostion mechnisam and choose the best solution which comes form the neighbour weight vectors to be this particle＇s global best solution. To confine the flying of the particle ,this paper introduced a new speed restriction factor. Comparing with three state-of-the-art multi-objective optimizers on ten test Problems, Smoeadpso outperforms the other algorithms as regards the coverage and approximation to the real pareto front.Meanwhile, the uniformity of the solution set to the 3 objective problems performs better than other particle algorithms.