Master Thesis
Agent Performance Comparison using Reinforcement Learning in an Interactive Setting
This paper presents the comparison of six different Reinforcement Learning based agents using the
SARSA algorithm. A short explanation and presentation of the most recent Reinforcement Learning
research prepares for the methodology behind the showcased algorithms. Two single-agent algorithms,
two decentralised and two centralised multi-agent algorithms are created and tested, using differing
state-actions pairs in an adaptable and expandable test setting. Results are evaluated and compared
to another, and future outlooks are shared with a short discussion of the outcome of the experiment.
For now, this webpage only presents a summarised overview of my work, but the entire thesis can be read
here.