Improved Evolutionary Programming and Its Application in Navigation System

In designing BP neural network, it is difficult to determine the network parameters and weights, or to achieve the best effect under random perturbation, and impossible to satisfy the real-time requirement of system. To solve these problems, this paper introduces an evolutionary neural network and puts forward some improvements on conventional evolutionary programming. This algorithm is then used in the integrated navigation system. Experiment results indicate that BP neural network based on evolutionary programming can not only overcome the shortcomings of BP artificial neural network, but also avoid problems of genetic algorithms caused by binary-coded and cross operation. Also the navigation system based on this algorithm can get rid of the problems of kalman filter when the outside observation data are unreliable. Finally, the normal operation of kalman filter is ensured with a higher accuracy.