基于多阶段调度框架的麻雀搜索优化算法

Sparrow Search Optimization Algorithm Based on Multi-Stage Scheduling Framework

本文提出一种多阶段调度框架,实现对麻雀种群的初始位置、觅食、侦查与反捕食不同阶段的多策略调度.利用Halton序列与Tent映射提升种群个体质量与初始位置的分布均匀性.在觅食阶段,针对发现者与加入者因位置争夺导致种群质量劣化,设计最佳适配比调控二者数量关系,对超出适配比的加入者采用碰撞反弹算子改变其优化轨迹.满足适配比后则通过侦查判断是否存在天敌,若有则进入反捕食阶段,并利用Levy飞行并结合指数分布设计随机迁移机制,生成潜在的全局最优解区域;当连续多次没有发现天敌时为避免种群陷入局部极值,建立模拟预警机制并采用蝗虫算法进行多路径开发,避免寻优方向单一化.不同策略与机制的交替运行、协同调度,平衡了算法的多样性与收敛性.实验结果表明,与最近麻雀变体算法和元启发改进算法相比,该算法在寻优效率与收敛精度上显著优于对比方法.

This paper proposes a multi-stage scheduling framework to realize multi-strategy scheduling of sparrow populations in different stages of initial location,foraging,detection,and anti-predation.Halton sequence and Tent mapping are used to improve the quality of the population individuals and the distribution uniformity of initial position.In the forag⁃ing stage,aiming at the deterioration of the population quality caused by the position competition between the finder and the joiner,the best fit ratio is designed to control the quantitative relationship between the two,and the collision rebound opera⁃tor is used to change the optimal trajectory of the joiner beyond the fit ratio.After the adaptation ratio is met,judge whether there is a natural enemy through investigation,and if there is,enter the anti-predation stage,and use Levy flight and com⁃bine exponential distribution to design a random migration mechanism to generate a potential global optimal solution area;when no natural enemy is found for many times in a row in order to prevent the population from falling into local extre⁃mum,an early warning mechanism is established and the locust algorithm is used for multi-path development to avoid a sin⁃gle optimization direction.The alternate operation and coordinated scheduling of different strategies and mechanisms bal⁃ance the diversity and convergence of the algorithm.Experimental results show that,compared with the latest sparrow vari⁃ant algorithm and meta-heuristic improved algorithm,the algorithm is significantly better than the comparison methods in terms of optimization efficiency and convergence accuracy.