基于旋转基技术的多目标粒子群优化算法

为了解决多目标优化求解的问题，提出一种基于旋转基技术的多目标粒子群优化算法（rtMOPSO）。改进了旋转基可视化技术，并将Pareto前沿映射到改进的旋转基扇形平面上，采用差熵指标监测种群进化状态。为平衡归档集的收敛性和多样性，提出了角度支配和角度支配力度两种新的概念，并设计归档集新的排序方法。在融合了旋转基角度和距离的概念的基础上，提出了一种改进的全局引导粒子的选择策略。改进算法采用两个类别的测试函数，与五种多目标优化算法进行了对比实验。实验结果表明，改进算法在收敛性和多样性方面优势明显。

Multi-objective particle swarm algorithm based on rotation basis

This paper proposed a multi-objective particle swarm algorithm based on rotation basis(rtMOPSO) in order to solve problems about multi-objective optimization. Firstly, it detected Pareto front maps to the improved sectorial plane of rotation basis so that evolutionary status of the population by the entropy and its difference entropy of the population. Then, it proposed angle dominance and strength of angle dominance to design an archive maintaining strategy which could balance diversity and convergence. Finally, it proposed selecting the global best solution based on rotation angle and distance of the rotation basis. It compared the improved algorithm with 5 multi-objective optimization algorithms on two kinds of test suites. The Simulation results show that improved algorithm has big advantages over the other competitors in terms of diversity and convergence.