基于改进的遗传算法神经网络控制器

Neural Network Controller Based On Genetic Algorithm

基于遗传算法的神经网络控制器,对神经网络中连接权值采用实值编码,通过随机选择,可变变异率和Elitism方法形成改进遗传算法。以倒立摆系统神经网络控制器参数寻优为例,对二进制编码和实值编码的遗传算法进化过程进行仿真,结果表明本算法寻优效果优,在多参数、多目标、多约束条件下,寻优效果更佳。

Based on genetic algorithm, NN controller uses real coding, stochastic sampling, invariable mutation rate and elitism method to improve genetic algorithm. Take inverse pendulum system as an example, it uses binary coding genetic algorithm and real coding genetic algorithm to simulate. The results show that the method has good optimization effect, and under multi-parameter, multi-object and multi-constraint conditions, it is more feasible.