面向路径规划问题的虚拟多任务共生生物搜索算法  

Virtual multi-tasking symbiotic organisms searchalgorithm for path planning problem

作　　者：徐江 程美英 Xu Jiang;Cheng Meiying(School of Information Engineering,Huzhou University,Huzhou Zhejiang 313000,China;School of Economics&Management,Huzhou University,Huzhou Zhejiang 313000,China)

机构地区：[1]湖州师范学院信息工程学院,浙江湖州313000 [2]湖州师范学院经济管理学院,浙江湖州313000

出　　处：《计算机应用研究》2023年第12期3599-3605,3613,共8页Application Research of Computers

基　　金：国家自然科学基金青年基金资助项目(62102148)。

摘　　要：针对现有共生生物搜索(SOS)算法在求解路径规划等离散型优化问题时存在性能较差、收敛速度慢等问题,提出虚拟多任务共生生物搜索(VMTSOS)算法。首先根据双向映射解码策略,实现个体连续空间位置和离散城市序列转换;然后引入多任务优化思想构建虚拟多任务环境,设计多种群同时优化同一任务,并通过停滞阈值控制种群间信息迁移频率,当主种群达到停滞阈值时,将辅助种群中部分优秀个体替换为主种群劣质个体;最后对VMTSOS算法时间和空间复杂度进行分析。仿真实验表明,VMTSOS算法在求解多数TSP时均能快速收敛至各测试实例目前的最优解,而在求解冷链物流配送问题时,具有多种群辅助机制的VMTSOS算法能较大程度地降低最优总成本。Aiming at the problems of poor performance and slow convergence of existing symbiotic organisms search(SOS)algorithm in solving discrete optimization problems such as path planning,this paper proposed the virtual multi-task SOS(VMTSOS).Firstly,according to the bidirectional mapping decoding strategy,it realized the transformation between individual continuous spatial position and discrete city sequence.Secondly,it introduced the idea of multi-task optimization to construct a virtual multi-task environment,designed multiple populations to optimize the same task simultaneously,and the stagnation threshold controlled the frequency of information transfer between populations.Once the main population reached the stagnation threshold,it replaced some excellent individuals in the auxiliary population with inferior individuals in the main population.Finally,it analyzed the time and space complexity of VMTSOS.Experiment results show that VMTSOS converges rapidly to the current optimal solution of each test case when solving most TSP problems,while VMTSOS with multiple groups assistance mechanisms can greatly reduce the optimal total cost when solving cold chain logistics distribution problems.

关 键 词：共生生物搜索算法 多任务优化 信息迁移 路径规划 冷链物流配送 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]