Project 2: CS8803-O03 Reinforcement Learning

The purpose of this project report is to experimentally replicate Multi-agent Correlated Q-Learning put forward by Amy Greenwald and Keith Hall in their ’Correlated Q-Learning’ paper published in 2003. The focus of this report will be to apply Correlated Q, Friend & Foe Q and classic Q-Learning to the ”Soccer Game” example discussed by them in the second half of the paper. This piece of research attempts to learn equilibrium policies for Markov games as does classical Q-learning for regular MDPs. Following parts of this section cover important concepts pertaining to the report before we delve into the experiments.