SmarTee: a neural network based navigational bot

Abstract

Most modern computer controlled players in computer games are controlled by static rule-based systems. Rule-based computer controlled players often do not act in a natural human way. To overcome this problem I propose using an artificial neural network, trained on human player input data, instead. This bachelor thesis assesses the performance of an artificial neural network controlled player and its generalization capabilities by performing real-time simulations and comparing the results to human data.