| This project attempts to show the practicality and effectiveness of machine learning in video game Artificial Intelligence (AI) by developing an adaptive AI system for a fighting game called Alpha Fighter. The use of sequential prediction, n-grams, hidden Markov models, and dynamic Bayesian networks are discussed in the context of the game and considered for their suitability for the AI system. The design and implementation of the AI system as well as more notable aspects of the game itself are also discussed. Results on a survey of the effectiveness of the adaptive and non-adaptive versions of the AI are given, showing that the adaptive AI was perceived to be more intelligent and challenging. The adaptive AI was also perceived to provide slightly better game-play. |
