Machine Learning techniques applied to the development of a strategy battle video game

Treball Final de Grau en Disseny i Desenvolupament de Videojocs. Codi: VJ1241. Curs acadèmic: 2016-2017In the last years, the research of Artificial Intelligence techniques in the field of robotics has made a huge advance that, joined to the evolution of the processing capacity of CPUs, allows other fields to take advantage of this progress as the video game development. The use of these new Artificial Intelligence discoveries has led video game designers to create autonomous characters that are able to evolve and improve themselves, adding a whole new dimension in the design of the behaviour of non playable characters improving the interaction with the game world. In this document, the video game Assault Intelligence is proposed. It consist of the development of a strategy battle video game focused in the research and application of Machine Learning techniques. This video game will aim to create a challenging and smart opponent who really makes the player feels the learning and evolution of an autonomous character. This adversary will adapt its behaviour to the user way of playing and will learn new strategies to become each match a new challenge. To achieve this goal, two well known Reinforcement learning techniques will be implemented and tested: the Monte Carlo approach and the Q-Learning one. An evaluation and comparative of the results of each technique will be presented