摘要
针对标准引力搜索算法存在收敛速度过快,容易陷入局部最小值等问题,提出一种改进的基于莱维飞行的引力搜索算法,在引力搜索算法框架下引入莱维飞行产生随机步长,进一步更新种群位置。莱维步长缩放因子动态调整,随迭代次数增加,莱维更新逐步发挥作用,使算法继续保持较好的全局搜索性能。莱维更新只选择适应度无退化粒子参与下次计算,且飞行步长更新以历史最优粒子的位置为指导,增强算法记忆性,促使粒子向更优适应度方向进化。对多个标准测试函数进行仿真,结果表明改进算法较标准引力搜索算法和布谷鸟算法具有更好的全局搜索能力和寻优精度。
Aiming at the problem that the standard gravitational search algorithm converges too fast and easily falls into the local minimum, an improved gravitational search algorithm based on Levi’s flight is proposed. This algorithm introduces Levy flight to generate random steps and further updates the population position within the framework of the gravity search algorithm. Levi’s step size scaling factor is dynamically adjusted. As the number of iterations increases, Levy update gradually takes effect, so that the algorithm continues to maintain a good global exploration performance. Levi’s update only selects non-degrading particles for the next calculation, and the flight step update is guided by the historical optimal particle position to enhance the memory of the algorithm and promote the evolution of particles towards a better fitness direction. Simulation experiments on multiple standard test functions show that the algorithm has better global search ability and optimization accuracy than the standard gravitational search algorithm and cuckoo algorithm.
作者
刘紫阳
庞志华
陶佩
郑韩飞
LIU Zi-yang;PANG Zhi-hua;TAO Pei;ZHEGN Han-fei(College of Computer Science,North China Institute of Aerospace Engineering,Langfang Hebei 065000,China;Institute of Mechanical Science,Beijing 100044,China)
出处
《计算机仿真》
北大核心
2022年第1期312-317,共6页
Computer Simulation
基金
国家自然科学基金(51875018)
国家重点研发计划(2018YFB1004100)
廊坊市科技支撑计划项目(2019011020)
北华航天工业学院青年基金(KY-2018-40)。
关键词
引力搜索
莱维飞行
记忆增强
群体智能优化
Gravitational search algorithm
Levy flight
Memory enhancement
Swarm intelligence optimization
