Based On Neural Network And Genetic Algorithm Research And Application Of Artificial Intelligence Game

Artificial Intelligence is the research and development for simulation, extension and expansion of human intelligence, which consists of theory, methods, techniques and applications to new technological sciences. AI is not supposed to develop games at early stage. However, facing the pressure and competition that increasing games spring up in market, introducing artificial intelligence into the game has gradually become an effective and essential means to attract players, for it improves the plays’ sense of achievement, enables the virtual world to appear more realistic and assists other features."Learning", with some difficulty to perform, obviously belongs to the scope of a non-qualitative AI. It is also the goal that major game companies attempt to achieve. To be specific, the methods of non-qualitative AL have been adopted by several mainstream games such as "Creatures" and "Black&White". Undoubtedly, the success of these games extremely stirs people’s interest of "learning", which can be exemplified by the numerous studies on the decision tree, neural networks, genetic algorithms (GA) and probabilistic methods.This paper describes the genetic algorithm and artificial neural network concepts and status, genetic algorithms and artificial neural network application in the game are described, based on genetic algorithms and artificial neural networks, independent of the NPC with the design, construction of the new algorithm NPG algorithm.The paper brings in the situation and determines the optimal decision of the evaluation method by designing process which can play Tetris automatically. The method we employ has three most authoritative valuation functions at present. First, we use the method to enumerates the overall possible variant situations of current and upcoming boxes in their falling period as well as all probable results they generates. Moreover, we count the box height, the number of rows which may be eliminated and the possible number of holes in terms of the attacking and defending capabilities thereby making a comprehensive assessment.Eventually, we obtain a weighted average to select the maximum weight as the decisive parameters of optimal strategy. What is worth noticing is that the decision-making process will go throughout the game. By experiments we find that the approach is very effective and efficient.