Learning Emergent Discrete Message Communication for Cooperative Reinforcement Learning
This work addresses the need for more interpretable communication protocols in cooperative AI systems, though it is incremental as it builds on existing MARL frameworks.
The paper tackles the problem of interpretability in multi-agent reinforcement learning by proposing a method for agents to learn emergent discrete message communication, achieving performance comparable to continuous communication with a much smaller vocabulary size.
Communication is a important factor that enables agents work cooperatively in multi-agent reinforcement learning (MARL). Most previous work uses continuous message communication whose high representational capacity comes at the expense of interpretability. Allowing agents to learn their own discrete message communication protocol emerged from a variety of domains can increase the interpretability for human designers and other agents.This paper proposes a method to generate discrete messages analogous to human languages, and achieve communication by a broadcast-and-listen mechanism based on self-attention. We show that discrete message communication has performance comparable to continuous message communication but with much a much smaller vocabulary size.Furthermore, we propose an approach that allows humans to interactively send discrete messages to agents.