基于Pareto支配的改进人工大猩猩部队多目标优化  

Multi⁃Objective Optimization of Improved Artificial Gorilla Troops Based on Pareto Domination

作　　者：杨模 刘紫燕 梁静[1] 东文 吴颖 YANG Mo;LIU Ziyan;LIANG Jing;DONG Wen;WU Ying(College of Big Data and Information Engineering,Guizhou University,Guiyang Guizhou 550025,China;State Key Laboratory of Public Big Data,Guizhou University,Guiyang Guizhou 550025,China)

机构地区：[1]贵州大学大数据与信息工程学院,贵州贵阳550025 [2]贵州大学公共大数据国家重点实验室,贵州贵阳550025

出　　处：《传感技术学报》2023年第4期590-601,共12页Chinese Journal of Sensors and Actuators

基　　金：贵州省科学技术基金资助项目(黔科合基础[2016]1054);贵州省联合资金资助项目(黔科合LH字[2017]7226号);贵州大学2017年度学术新苗培养及创新探索专项资助项目(黔科合平台人才[2017]5788);贵州省科技计划(黔科合SY字[2011]3111)。

摘　　要：针对二维和三维多目标优化的解易陷入次优和分布不规则问题,引入Pareto支配方法并提出了结合改进策略的人工大猩猩部队优化算法(MOGTO)以改善多目标优化问题。首先,利用外部档案集存储互不支配解集以防止种群经过一次寻优迭代后,删除劣解的同时错误去除非支配解。其次,加入蒙特卡洛树搜索对种群探索阶段的三种机制进行优化,增加算法全局搜索能力。同时,结合天牛须算法的左右须寻优原理及黄金正弦寻优策略对开发阶段的两种机制进行优化,生成左右须解和促进个体位置更新,引导个体扩大搜索范围以防止最优解陷入局部最优。最后,通过12个基准测试函数对所提算法进行验证并将所提算法与其他6种常见算法进行对比,实验结果表明所提算法在多目标问题的寻优能力较对比算法有较大的提升。通过曲柄摇杆机构优化设计案例测试分析,验证了所提算法在实际工程应用中的可行性和实用性。Aiming at the fact that the solutions of 2D and 3D multi⁃objective optimization are easy to trap into suboptimal and irregular distributions,after the Pareto dominance method is presented,an artificial gorilla troops optimization algorithm(MOGTO)combined with an improved strategy is proposed to improve the multi⁃objective optimization problem.Firstly,an external archive set is used to store mu⁃tually exclusive solutions to prevent the population from removing non⁃dominant solutions by mistake when deleting inferior solutions af⁃ter one optimization iteration.Secondly,Monte Carlo tree search is added to optimize the three mechanisms in the population exploration stage to increase the global⁃search ability of the algorithm.At the same time,the two mechanisms in the development stage are optimized by combining the left and right whisker optimization principle of the beetle algorithm and the golden sine optimization strategy to gener⁃ate left and right whisker solutions and promote individual position update,and guide individuals to expand the search range to prevent the optimal solution from falling into local optimum.Finally,the proposed algorithm is verified with 12 benchmark functions and com⁃pared with other 6 common algorithms,the experimental results demonstrate that the proposed algorithm achieves better performance and greater improvement,and it possesses higher optimization ability in multi⁃objective optimization compared with the current algorithms.The feasibility and practicability of the proposed algorithm in practical engineering applications are verified through the test and analysis of the optimal design case of the crank⁃rocker mechanism.

关 键 词：机械设计 多目标优化 人工大猩猩部队算法 PARETO支配 黄金正弦策略 蒙特卡洛树搜索 

分 类 号：TP301[自动化与计算机技术—计算机系统结构] TP302[自动化与计算机技术—计算机科学与技术]