基于特异性免疫策略的遗传算法及应用

Genetic Algorithm Based on Specific Immunity Strategy and Application

针对标准遗传算法在进化后期收敛速度慢,易陷入未成熟收敛的问题,借鉴免疫应答机理,提出一种基于特异性免疫策略的遗传算法.算法的核心在于保持种群的多样性和执行特异性免疫策略,即引入小生境技术维持种群的多样性,对遗传参数自适应调节以适应种群的实际变化;利用高亲和度抗体搜寻更优秀的抗体,并发掘低亲和度抗体寻优的潜力;通过优良记忆库实现精英保留策略,保证算法搜索的快速性及有效性.理论上证明了算法的收敛性.仿真结果表明,算法能有效地改善种群多样性,具有较强的全局收敛能力.以二级倒立摆为被控对象,将该算法应用于Takagi-Sugeno模糊神经网络控制器的优化,实物控制结果表明该方法具有良好的动稳态性能和抗干扰能力.

Standard genetic algorithm has a slow convergence velocity in late evolution and gets premature convergence easily. To solve these problems, a new genetic algorithm based on specific immunity strategy was proposed in view of mechanism of immune response. The key to this algorithm lies in maintaining the diversity of population and executing the strategy of specific immunity. The niche mechanism was introduced to maintain diversity of population. To adapt the actual change of population, genetic parameters were adjusted adaptively. The antibody population with high affinity was used to seek more excellent antibodies, and the antibody population with low affinity was inspired to seek optimum. Under the action of excellent memory cell to actualize elitist strategy, the search of algorithm is rapid and effective. It is proven that the algorithm can guarantee the convergence towards the global optimum. Simulation results show the algorithm can improve population diversity effectively and global convergence ability. Applying the algorithm to the optimal design of T-S fuzzy neural network controller, the controller can control a double inverted pendulum system well. Experiment results demonstrate the method has ideal dynamic, steady performance and anti-disturbance.