Andrew Ferris Undergraduate Dissertation 2014/15

Multi-Agent Reinforcement Learning

Supervised by E.Vasilaki

Abstract

Reinforcement learning is an area of machine learning in which the machine learns through interaction with its environment rather than through direct instruction. This paper aims to compare and contrast a few similar implementations of a hunter-prey reinforcement learning task with multiple agents to see how well agents perform when their state-space is increased to include each others' position compared to when they are only aware of themselves and their goal.