Research On Key Techniques Of Game Al

With the development of modern game industry, techniques upon computer game including graphics rendering, realistic interaction, game artificial intelligence(game AI) have been stepped up. Among these techniques, game AI falls behind others due to the restriction of real-time system and theoretical AI. Besides, game AI could not be dependent on game engine and there is no game engine which contains complete game AI module, because different games require different game AI.Within the area of computer games, leisure games generally have less requirement on game AI, and chess&card games have fixed AI solutions. However, real-time Strategy and First-Person Shooter games ask for a high standard of game AI, due to subjects of these games are mainly about a realistic warfare. AI solutions for these games own features as follow:(1) In RTS and FPS games, there are a huge number of Non-Player-Controlled Characters (NPC), usually dozens or even hundreds. Such many NPC's intelligence computing is completely executed by game AI system, which forces a restriction on complexity of AI mathematical model.(2) In RTS games, the virtual camera in the scene overlooks in a high position and has a wide visual field on the battle ground, which makes player be able to observe the group behaviour of NPC. In FPS games, the final image of rendering process is a simulation of human vision. More strict conditions such as smooth movement and animation playback are required, in order to give a better user experience.In this thesis, an integrated battle game AI system is designed in terms of RTS and FPS games' AI requirements, which contains not only some low level technologies such as movement, sense, animation control, but also a tripple levels architecture to control battle behaviour of multiple NPC groups. And this system is realized on the game engine Unity3D. The main contributions of the thesis are as follow:(1)A movement and animation control sub-system is designed. This sub-system modifies the Flock algorithm and uses A* path finding to manage NPC's movement.NPC can follow a best path and move to any position in the map. With the animation controller integrated, correct animations are played during movement.(2)A sense and memory sub-system is designed to provide input information for NPC. The model uses the integrated physics engine PhysX in Unity3D, and a message route mechanism makes all data s are shared by each NPC, which is a simulation of communication of troops.(3)A map analysis sub-system is presented. This system generates data for movement decisions, NPC could using his memory and vision to judge the situation of each area of the map.(4) A hierarchical control architecture is devised. Including commander, captain, squad member, information report and orders assignment are carried out by message system. We add order machines to each level of the architecture, and designed specific orders for it. After decision-making module push orders to the queue of the order machines, these orders can be executed in sequence and displays by NPC movement and animation playback.