基于分解的改进自适应多目标粒子群优化算法

An Improved Self-Adaptive Multi-Objective Particle Swarm Optimization Based on Decomposition

为提高粒子群算法的搜索效率,克服分解方法处理复杂多目标问题的不足,通过考虑父代解的选择和种群的更新对算法收敛性及解的分布均匀性的重要影响,提出了一种基于分解的改进自适应多目标粒子群优化算法。首先,为提高算法收敛速度,在分解方法确保进化种群多样性的前提下,设计了新的适应度评价方法以评价个体的优劣,并将在竞争中获胜的优质后代解添加到父代候选解中;其次,为避免算法陷入局部最优,在更新粒子时,从当前粒子的邻居或邻居外随机选择个体最优和全局最优位置;最后,引入外部文档,将其作为候选的输出种群,并采用拥挤距离维持多样性,增强了算法处理复杂问题的能力。用12个测试函数的数值实验,并与5种多目标优化算法的比较,表明了所提算法的优越性。

In order to improve the search efficiency of particle swarm optimization and overcome the weakness of the decomposition method to deal with complex multi-objective problems,an improved self-adaptive multi-objective particle swarm optimization based on decomposition is pro-posed by considering the important influence of parent solutions selection and population updating on the convergence of algorithm and the distribution uniformity of solutions.To improve the convergence speed,a new fitness evaluation method is first designed to estimate solutions'quality and the quality offspring solution won in the competition is added to the parent candidate solutions under the premise of ensuring diversity of evolutionary population by decomposition method.Next,to avoid the algorithm falling into local optimum,the personal optimal and global optimal positions are randomly selected from current particles'neighbors or outside of neighbors when updating the particles.Last,to enhance the ability of algorithm to deal with complex problems,external archive is introduced as a candidate output population and crowding distance is used to maintain its diversity.The numerical experiments are carried out on twelve test functions and compared with five multi-objective optimization algorithms that can show the superiority of proposed algorithm.