Abstract:

To overcome the shortcomings of slow convergence， low accuracy and being easy to fall into local optimum in Dung Beetle Optimizer （DBO） algorithm， a Dung Beetle Optimizer algorithm with restricted reverse learning and Cauchy-Gauss variation （SI-DBO） was proposed. Firstly， Circle mapping was used to initialize the population to make distribution of the population more uniform and diverse， which improved the convergence speed and optimization accuracy of the algorithm. Secondly， restricted reverse learning was used to update the locations of dung beetles， so as to improve the search ability of dung beetles. Finally， Cauchy-Gauss variation strategy was used to help the population escape from the local optimal location and find the global optimal location. To verify the performance of SI-DBO， simulation experiments were carried out on the test functions and Wilcoxon rank-sum test was performed on the experimental results， and the algorithm was used to solve robot gripper problem. Experimental results show that SI-DBO achieves higher optimization accuracy and convergence speed than Black Widow-Dung Beetle Optimization （BWDBO） algorithm and Sparrow Search Algorithm （SSA） on the test functions. Meanwhile， SI-DBO performs better than Particle Swarm Optimization （PSO） algorithm for solving robot gripper problem， indicating better optimization performance and engineering practicability of SI-DBO.

Key words: Dung Beetle Optimizer (DBO) algorithm, restricted reverse learning, Cauchy-Gauss variation, Wilcoxon rank-sum test, robot gripper problem

摘要：

针对蜣螂优化（DBO）算法中存在的收敛速度慢、精度不高以及容易陷入局部最优的问题，提出一种融入限制反向学习与柯西-高斯变异的蜣螂优化算法（SI-DBO）。首先，用Circle映射初始化种群，以使种群的分布更加均匀和具有多样性，从而提升算法的收敛速度和寻优精度；其次，使用限制反向学习对蜣螂的位置进行更新，以提升蜣螂的搜索能力；最后，使用柯西-高斯变异策略帮助种群逃逸出局部最佳位置并寻找全局最佳位置。为了验证SI-DBO的性能，在测试函数上进行仿真实验并对实验结果进行Wilcoxon秩和检验，而且将该算法用于求解机器人夹持器问题。实验结果表明，与黑寡妇-蜣螂优化算法（BWDBO）和麻雀搜索算法（SSA）相比，SI-DBO在测试函数上均获得了较高的寻优精度和收敛速度，同时，SI-DBO在求解机器人夹持器问题时的效果优于粒子群优化（PSO）算法，验证了SI-DBO具有更好的寻优性能和工程实用性。

关键词: 蜣螂优化算法, 限制反向学习, 柯西-高斯变异, Wilcoxon秩和检验, 机器人夹持器问题