GENERAL ARTIFICIAL INTELLIGENCE MODEL FOR DEVELOPING ADAPTIVE NON PLAYABLE CHARACTERS IN COMPUTER GAMES

Abstract

In 2005, through inauguration of the annual AAAI General Game Playing Competition the artificial intelligence (AI) community was introduced to AI in computer games. Since then, the efforts have been made to make AI systems understand the rules of arbitrary computer games without any human help. This paper will introduce a new model based on neural networks and genetic algorithm concepts which has the potential to replace behavior tree model currently used by game developers for making AI systems. This model introduces the possibility for AI systems to be trained in the game instead of creating complex nested if-else tree to determine the next move. This adaptive learning approach for AI systems could potentially lead to development of more powerful and engaging AI systems known as Non playable characters in gaming industry. The model demonstrated in this paper is developed using Python 3 language and the game environment in which it is tested are developed in Python 3 environment as well.