Research On The Application Of Neuron Network In Course-control Autopilot

First, the development of autopilot and Neutral Network are discussed, furthermore, the Matlab6.1 is used to construct a athletic of submarine's mathematical model, this mathematical model construct a development environment of controller of submarine. Sencond, in this development environment research into the Neural Network Algorithm and Genetic Algorithm for the control of the autopilot.Though conformation of traditional PID Algorithm is very simple, it is extremely depend upon the accurate model of the controller. To involve this problem, Neural Network Algorithm combined the PID Algorithm format a Neuron PID Algorithm, and a PSD Algorithm of adaptive Control of autopilot of submarine is applicator to control the submarine's direction. The Genetic Algorithm was used to search for the optimize parameter of the PID Algorithm.In order to optimize the BP online algorithm, this text according to the law of evolution of the Neural Networks' authority and base on the Terminal Attractors, author propose a method of Globally Convert Adaptive Quick Back Propagation and measures of improving the algorithm's performance are discussed.