Learning Communication in Multi-Agent Reinforcement Learning

In the context of a course on multi agent systems this fall I have been looking into multi agent reinforcement learning. In particular I have explored the problem of learning communication between agents. This resulted in a short report that gives a concise introduction and overview of the research area.

The report is written as a survey paper, covering the motivation for learning communication, some underlying theory of reinforcement learning and an overview of existing approaches to the problem. It also includes a discussion part with some thoughts on practical applications and future research.

The report is avaliable here and some slides with descriptive illustrations can be found here.