引入欧椋鸟群飞行机制的改进粒子群算法

传统粒子群算法存在早熟、精度低等不足,许多改进算法尽管性能略有提高,但依然存在原有弊端。生物学家对欧椋鸟群的最新研究发现:鸟群飞行机制中个体间存在拓扑相互作用,与距离远近无关。受这一研究成果启发,提出一种引入欧椋鸟飞行机制的改进粒子群算法。该算法在进化策略上引入拓扑作用和猎食动物的惊扰机制,在参数选择上提出粒子群动能的概念,在线性递减权重框架下通过粒子动能自适应更新惯性权重,拓扑作用集合采用最近邻粒子构成,将拓扑因子与惯性权重相联系,达到进化状态的平衡和自适应性。仿真实验表明,改进算法在精度、成功率和效率上具有一定的优势,尤其是对多模态优化问题。

Improved particle swarm algorithm introducing flight mechanism offlock of European starlings

The main defects of conventional particle swarm algorithms are premature and low precision. Many improved algorithms promoted the performance of algorithm slightly while the defects aforementioned remained in a sense. In recent research on the flock of European starlings, biologists reveal that the flying mechanism of the flock lies on a topological interaction among members, not the distance between them. Enlightened by this new discovery, this paper proposed an improved algorithm by introducing two strategies of the flock of starlings flight mechanism: the topological interaction and the disturbance of predator. With respect to the selection of parameters, it put forward a new concept of particle's kinetic energy to adjust the inertia factor adaptively in the frame of linear decreased weights. The topological interaction set consists of the near neighbors in the flock. It set the factors of inertia weight and topological interactions to affect each other. This made the evolution state more harmonic and adaptive. The numerical experiments indicate that the improved PSO algorithm surpasses the conventional ones on aspects of the optimization accuracy, success rates and overcoming the premature problem in a sense, especially in the case of multi-modal scenarios.