基于克隆选择原理,引入混沌机制和小生境技术,提出了一种新的人工免疫算法——改进型克隆选择算法(ICSA)。该算法设计了一种自适应混沌变异算子,有效地避免了搜索的盲目性,提高了算法的收敛速度。利用随机过程鞅理论,分析了算法所形成抗体种群的平均适应度鞅的性质,并且当种群为有限状态时,证明了该算法能以概率1确保在有限步内收敛到全局最优解。对多模态函数优化的仿真实验表明,该算法能有效地抑制早熟,具有更好的全局收敛性。 Based on the theory of clonal selection, chaos mechanism and niche technology are introduced, and a new artificial immune algorithm, called Modified Clonal Selection Algorithm (ICSA), is proposed. The algorithm designed an adaptive chaotic mutation operator, which effectively avoided the blindness of search and improved the convergence speed of the algorithm. Using the stochastic process martingale theory, the properties of the average fitness martingale of the antibody population formed by the algorithm are analyzed. When the population is finite, it is proved that the algorithm can guarantee the convergence to the global optimal solution in a finite step with probability 1. Simulation experiments on multi-modal functions show that this algorithm can effectively restrain prematurity and have better global convergence.