Representation of models for solving real-world physics problems
Proceedings of the sixth conference on Artificial intelligence applications
AfriGraph '01 1st International Conference on Virtual Reality, Computer Graphics and Visualization in Southern Africa ( formerly known as SAGA 2001 )
Minimal hierarchical collision detection
VRST '02 Proceedings of the ACM symposium on Virtual reality software and technology
A Tutorial on Support Vector Machines for Pattern Recognition
Data Mining and Knowledge Discovery
IEEE Spectrum
Intelligent Self -learning Characters for Computer Games
EGUK '02 Proceedings of the 20th UK conference on Eurographics
| Hi-index | 0.00 |
This paper presents the intelligent game agent that gives effective intelligence to NPCs (Non Player Characters) for which intelligence did not exist. Generally, non-player characters (NPCs), or agents, such as monsters, enemy guards, or friendly wingmen, can be controlled by a finite-state machine. To overcome the shortcoming of NPC’s restricted action, we applied a LSVM (Linear Support Vector Machine) as pattern recognition for the intelligent game agent, and processed all data in XML format to handle the data efficiently. The intelligent agent is executed in the base of the game physics engine. A lot of NPCs that act in a game learn physics values that are produced in the game, and change NPS’s action intelligently. We applied two pattern recognition algorithms to estimate the algorithm’s performance through comparison. As indicated by experiments, when the M-BP has a fixed number of input layers (number of physical parameters) and output layers (impact value), it shows the best performance when the number of hidden layers is 3 and the learning count number is 30,000. The pattern recognizer that applied LSVM shows the best performance when the learning count number is 25000, and the LSVM shows better performance than the M-BP in the intelligent game agent.