Decomposing Temporal Equilibrium Strategy for Coordinated Distributed Multi-Agent Reinforcement Learning

A major portion of the research presented in this document was conducted within the scope of the project "Learning Algorithms for Cooperative Multi-Agent Systems" which was funded by the German Research Foundation (DFG). I therefore would first like to thank the DFG for their financial support. Many people have influenced my work during the previous years and it is nearly impossible to list all of them. However, I would like to mention a few persons to whom I am particularly indebted. First of all, I would like to thank my supervisor, Prof. Dr. Martin Riedmiller, for his commitment to serve as my doctoral advisor and for his continued support in scientific and occupational questions. Moreover, he has nurtured the development of inspiring and comfortable working conditions in his research group that, in my opinion, were beyond compare. Furthermore, I would like to thank Prof. Dr. Héctor Muñoz-Avila for his great job as a co-referee of this thesis, as well as Prof. Dr. Volker Sperschneider, Prof. Dr. Werner Brockmann, and Dr. Andreas Nüchter for serving on the board of disputation examiners. I would also like to thank all my former students as well as my colleagues, Roland Hafner, Sascha Lange, Martin Lauer, and Stephan Timmer, for a harmonious working atmosphere, for interesting and encouraging discussions, as well as exciting conference travels. I am also greatly indebted to Thorsten Kundoch for his technical and IT support as well as to Anna Rushing-Jungeilges for her administrative support and her proofreading of nearly all of my publications and parts of this thesis. I owe special thanks to Susanne Linster for proofreading the final version of this thesis. Last but not least, I owe many thanks to my parents and family for supporting me in countless ways throughout the years.