Research On Role Control Of Intelligent Game Based On Evolutionary Behavior Trees

In today's society,with the continuous development of computer software and hardware level,the game industry has achieved rapid development in the past ten years.For a long time due to the dual constraints of artificial intelligence theory research and computer real-time performance of development,development speed caused by the development of game artificial intelligence technology is relatively backward in graphics and real-time interaction.As gamers pursue deeper levels of gameplay,the gameplay needs to be taken into account in game design and development,and the performance of the game's artificial intelligence highlights this.At the same time in different games for the game of artificial intelligence demand is different,the implementation process and the means of game artificial intelligence is not the same,it can integrate and improve the function of the game artificial intelligence game engine will not appear.This thesis introduces and analyzes the realization technology of game artificial intelligence,which begins with the realization of the technology state machine of the game artificial intelligence,and extends to the behavior tree.Through the realization mechanism of the state machine and the behavior tree and the complexity of comparing the two,the application scope of the state machine and the behavior tree is analyzed.By analyzing the theoretical basis of genetic algorithm and BP neural network,the neural network based on genetic algorithm is used to evolve the behavior tree nodes,so that the behavior tree itself is constantly evolving.In this thesis,the effectiveness of behavior tree evolution is validated by designing the implementation on the Unity 3D game engine.Due to the need to simulate the real scene and interactive behavior of game player and game character of non game player to control the game characters(NPC)the number may be very large,in order to ensure the operation of the game in a refresh rate within the allowable range,so to achieve the game artificial intelligence behavior tree model complexity set limit requirements.The preform design required to complete the game project game scene and game player and game characters,and builds a model according to the requirements of different behavior and role behavior game player game tree,after the implementation of neural network based on genetic algorithm on the basis of the evolution of the behavior of the tree model,finally using evolutionary behavior to construct tree to control the behavior of NPC.In the test,the genetic behavior tree model constructed by genetic algorithm based neural network can achieve the purpose of different NPC attributes and further expand to apply to the complex NPC game AI architecture.