樽海鞘群算法基于动力学模型的改进

Improvements of slap swarm algorithm based on dynamic model

针对樽海鞘群算法(salp swarm algorithm,SSA)中参数含义不明确、收敛性不确定的问题,构建了SSA的差分动力学模型,定义了领导者选择机制、领导者游走机制、跟随者偏序学习机制,重点针对跟随者偏序学习机制分析了系统的收敛性,面向领导者游走机制与跟随者偏序学习机制提出了算法全局最优收敛的充分条件。基于动力学分析结果提出了樽海鞘群异质定常跟随率与偏序多驱动机制的改进方法,仅对算法结构与参数进行了调整,在未增加计算量的情况下提高了算法的性能,通过仿真分析验证了改进的有效性。

Focusing on the problems of unclear parameter meaning and uncertain convergence in salp swarm algorithm(SSA),the differential dynamics model of SSA is constructed,including leader selection mechanism,leader walking mechanism,and follower partial order learning mechanism.The convergence of the system is analyzed with emphasis on the follower partial order learning mechanism.The sufficient conditions for the global optimal convergence of the algorithm are proposed for the leader-walking mechanism and the follower partial order learning mechanism.Based on the analytical result of dynamic behavior,an improved method of heterogeneous steady following rate and partial order multi-drive mechanism of salp swarm is proposed.Only the structure and parameters of the algorithm are adjusted,and the performance of the algorithm is improved without increasing the amount of calculation.The effectiveness of the proposed algorithm is verified by simulation analysis.