Implementation of dynamic artificial intelligence in game development

Abstract

The aim of this paper is to create an adaptive Artificial Intelligence implemented within our own video game which learns from players and uniquely adapts its playstyle in order to counter the play style of the current player in real time. This is done by creating a basic AI using the AI packages that come with Unreal Engine, which is what is being used to implement this technique. This AI is then trained with many different possible moves in the game represented within a tree. The end goal is to have the AI learn the moves that the user is using most frequently or most effectively and begin countering it more and more effectively as the level rises. This is done by training the AI to various movesets and giving it a sizeable sample space in order to understand and predict as required.