Double Inverted Pendulum Based On Particle Swarm Algorithm Of Neural Network Control

Wavelet neural network use the wavelet analysis theory. It’s a kind offeedforward neural network which put wavelet function into artificial neuralnetwork, and a kind of optimized neural network. In the analysis process, putwavelet basis function which based on fourier analysis, high frequencycharacteristics, great multi-scale resolution into artificial neural network withself-organizing, self-learning and self-adaption. So it replace traditional Sigmoidfunction and get better approximation function, faster rate of convergence, moreflexible fault-tolerant ability, and better adaptivity by scaling factor and shiftfactor. So the wavelet neural network has been widely used in pattern recognition,signal analysis, cybernation and many other fields of science.The particle swarm optimization that often used in the wavelet neuralnetwork is a kind of bionics algorithm. It is a kind of global optimizationintelligence algorithm that appeared by imitate the birds action of predation. Andit’s also a kind of group searching algorithm that regeneration population byrandom optimization. But the basic particle swarm optimization has slowerconvergence speed, poor generalization performance, low arithmetic stability,long training time, gets into the local convergence and series of problems. So tosolve these problems, this paper optimized the particle swarm optimization. Inthe process of optimized, this paper introduce the inertia weight factor andmake it gradual step-down, and also replacing personal best particle with theaverage of personal best particles in swarm. By this way, it can enhance thesolution space, individual experience and convergence efficiency. Then a moreeffective learning and training way can be get that makes the control performanceof wavelet neural network controller can be better.In this paper, set up the double inverted pendulum system model, and use the wavelet neural network controller that has been designed before to performthe double inverted pendulum system simulation experiment. In the simulationexperiment, the wavelet neural network controller can stable control the waveletneural network and has great capacity of resisting disturbance and faster restorebalance speed. So, it can explain that the wavelet neural network controller bythis approach has good results.