Design of Wavelet Neural Network Controller Based on PSO Algorithm
The Particle Swarm Optimization (PSO)is a kind of bionic evolutional algorithm inspired by the swarms of birds in nature. The error back-propagation algorithm of wavelet neural network is easily trapped into the local minimum point and has high requirement to the initial values of parameters. For the PSO algorithm is not easy to fall into the local minima and has a fast convergence speed, this paper presents a genetic variation factor based on an improved particle swarm neural network learning algorithm with the inertia weight factor added, which uses the linear regression strategy to increase the convergence precision. And the genetic algorithm crossover and mutation factors are combined to optimize the connection weights of hidden layer and output layer, and in each step of the genetic algorithm, using the elite preservation strategy, a very good optimization result is obtained. The PSO and generic algorithms are combined to control an inverted pendulum, and the simulation lab shows that the improved PSO optimizes the design of inverted pendulum controller and improves the performance of control. Finally, the effectiveness of the algorithm is verified by real control of a double inverted pendulum.
Wavelet Neural Network Particle Swarm Optimization Genetic Algorithm Double Inverted Pendulum
Qingkun Song Jiankai Gao Teng Zhou Lina Liu
Automation College Harbin University of Science and Technology Harbin, P.R.China
国际会议
哈尔滨
英文
321-328
2010-08-01（万方平台首次上网日期，不代表论文的发表时间）