Abstract: In order to overcome the limitations lying in slow convergence, poor accuracy and the propensity of getting into a local best answer in Artificial Fish-Swarm Algorithm（AFSA）, improved measures, which adaptively tuning algorithm parameter of artificial fish-swarm, are presented. This measure accords the comparison between the adaptive value of individual fish and the average adaptive value of the fish-swarm, dividing the whole fish into three groups which including better fish, general fish and poor fish, adaptively turning the step length and view field of fish in each group, in order to optimize and improve the basic algorithm. It analyzes the adapting fish-swarm step length and view field, optimization accuracy, convergence speed and stability of algorithm through four typical test function simulation. It can get the conclusion that adapting fish-swarm’s field of view and steps can make the algorithm converge quickly to the global better solution, and has good optimization performance.

Key words: artificial fish-swarm, swarm intelligence, adaptive adjustment, function optimization

摘要： 为了克服基本人工鱼群算法收敛速度慢、求解精度不高和易陷入局部最优的不足，提出了自适应调整人工鱼群算法参数的方法，该方法采用个体鱼适应值与整个鱼群的平均适应值作比较，将整个鱼群分为三组，再采用自适应调整每组鱼群的视野范围和步长的方法，对基本鱼群算法进行了优化和改进。应用四个典型的测试函数进行仿真实验，分析算法的寻优精度、收敛速度及稳定性。实验结果表明改进后的算法能够较快地收敛至全局较优解，并具有较好的寻优性能。

关键词: 人工鱼群, 群体智能, 自适应调整, 函数优化