Please use this identifier to cite or link to this item: http://hdl.handle.net/10609/119086
Title: Desarrollo de un agente mediante Deep Q-Learning en un entorno de juegos de plataformas
Author: Buedo Risueño, Álvaro
Director: Ventura, Carles  
Tutor: Kanaan-Izquierdo, Samir  
Abstract: Nowadays, the interaction between humans and computer systems is turning into a greater and greater tendency. Thus, the significance of machine learning techniques in our lives is constantly increasing, since they could be applied in almost every sphere. In the present work we employ Deep Reinforcement Learning technique in the platform game Super Mario Bros. In particular, the research consists in implementing the Q-Learning algorithm, one of the most relevant Deep Reinforcement Learning technique algorithms. Furthermore, due to the problem complexity, convolutional neural networks have been introduced, what leads to Deep Q-Learning (DQN). By using this algorithm we try to achieve that the videogame character is able to pass the levels which are set out to him by means of the experiences he has acquired throughout his exploration of those levels. In order to apply this algorithm, Python programming language has been used, which is the most powerful one to develop artificial intelligence systems, with its libraries TensorFlow and Keras for the neural networks implementation. Along this work, we expound the analysis, the design and the system implementation, in addition to the results and their interpretation. Finally, a series of conclusions are described as well as a future research proposal with the aim of increasing and improving the present one.
Keywords: reinforcement learning
neural networks
Document type: info:eu-repo/semantics/masterThesis
Issue Date: Jun-2020
Publication license: http://creativecommons.org/licenses/by-nc-nd/3.0/es/  
Appears in Collections:Bachelor thesis, research projects, etc.

Files in This Item:
File Description SizeFormat 
abuedorTFM0620memoria.pdfMemoria del TFM1,56 MBAdobe PDFThumbnail
View/Open