Research And Implementation Of Intelligent Agent Behavior Control System In Games

With computer calculation speed improved substantially and the improvement of gamers' requirements in gameplay and experience,games have higher requirements for artificial intelligence.In the paper,starting from the demand of game players,it researchs on path planning algorithm of intelligent Agent,cooperation and confrontation method of Multi-Agent System,and puts forward a new solution for the problems of the path planning and behavioral control of NPC group in the game,specific work is as follows:First of all,it contrasts and analysis several rasterize map path planning algorithm commonly used in the game and points out the deficiency of the A* algorithm.Relative to the path search algorithms of edge restricted,the path generated by any-angle path-finding algorithm is shorter and more realistic.Theta* is a typical representative of any-angle path-finding algorithm,however it doesn't achieve the optimal balance between run time and path length,its running time is short,but the finding path is not optimal.The paper presents improved Theta* which uses Goal-directed heuristic function to re-expand the vertexes generated by Theta* algorithm.Compared with other path-finding algorithms,improved Theta* algorithm has a better balance between the path planning runtime and the length of the resulting path,to a certain extent,it improves the execution efficiency of the algorithm and capacity of obstacle avoidance.Secondly,as the next generation of game AI mainstream,behavior tree has been extensively studied in recent years.However,it still need further research in support of group intelligence.Taking advantage of the hierarchical characteristics of behavior tree itself,combined with independent decision-making and group confrontation method of Multi-Agent system,the paper puts forward a Multi-Agent behavior control system based on behavior tree,designs and completes behavior tree of intelligent Agent and the console structure of hierarchical decision-making based on behavior tree.Combining with Unreal Engine 4 game engine,a tank war game has developed in order to simulate path navigation of intelligent Agent,the group behavior and AI control of Multi-Agent,and it achieves the ideal experimental results.