Game Design of Self-Automation Based on Artificial Neural Nets and Genetic Algorithms

This paper put forward the realization of the self-automation role, which has leaning ability and dynamical acclimatization. First of all, BP algorithm of artificial neural net (ANN) is improved, the self-adjusted algorithm of all parameters has been proposed for the back-propagation learning, which can make the selection of hidden layer units and rate of studying easily in the course of training, reduce artificial influence and improve the adaptive ability of rate of studying and neural net. Secondly, genetic algorithms (GA) has been optimized from primitive colony, selective manipulation, intercross manipulation. At the same time, methodology of ANN was integrated with GA and self-learning models of NPC were created to control their behaviors. At last, the experimental results have shown that self-learning system of NPC provides artificial behaviors with more automation and intelligence.