Research On Control System Based Hopfield Neural Network

Hopfield network as a nonlinear dynamic system, posed by the nonlinear elements fully connected network for dynamic feedback, the feedback from the output return to its input of Hopfield. So,The motivation of the input will lead to contibued change of the entire network. When there is input, we can have the output from the Hopfield net, the output feedback to the input to generate a new output, This feedback process has been going.If the Hopfield network is a converged and steady network, the network feedback and iterative calculation of the changes produced by the process of getting smaller and smaller, once in a stable equilibrium, then the Hopfield network will output a stable constant. If the steady state as a memory of the sample,then it is to find the optimal sample that from the beginning the process towards stability and convergence process. Initial state given a sample of some of the information, the evolution of neural networks can find all the information from the initial part of the information, in order to achieve associative memory. If the steady state corresponds to a calculation of the objective function optimization, as the minimal point Neural network energy function,that the Optimization process is find the minima energy function minima From initial state.This paper use the feature of Hopfield network to design neural network controller, The Hopfield network stable state corresponds to the objective function, the Operation of the network is to track the course of the process of objective function. First we describes the working principle of Hopfield network.Then we discuss the stability of the Hopfield network, once the network running will move towards to energy reduction automatically, the final output of the network is the equilibrium point of the network and it is also the minima point of energy. Finally, we design the controller of Hopfield network, comparise the Hopfield network controller and PID controller, the results of Simulation show that Hopfield network controller has good rapidity, stability and robustness.