Wavelet network controller based on improved genetic algorithm

Wavelet neural network is a kind of feed forward neural network introducing wavelet analysis theory. The combination between genetic algorithm and wavelet neural network can get the way of learning and training including the good global optimization search and the local time-frequency characteristic. A wavelet neural network controller based on improved genetic algorithm is proposed in this paper. This method can overcome a series of problems of the basic genetic algorithm such as slow convergence speed, early mature convergence and poor calculation stability etc,by increasing the probability of crossover, mutation reduces the probability that converge population, increasing the convergence speed.Hence, the performance of the wavelet neural network controller is further improved. Finally, the controller´s effectiveness is demonstrated through the simulation and the real control of the double inverted pendulum.