In this project, we implement Communication-aware Collective Intelligence(CoCI) in two different environment(e.g., highway-env and multiagent-particle-env). By comparing our model with baseline models such as MADDPG, ACML, ATOC, etc, we prove that CoCI has a better performance in uncertain communication channel(e.g.,network delay, packet loss). Code is available in the highway and mpe folders.

Binary-env is a small environment in which we test some conjectures about the efficiency of communication in MARL. Code is available in the binary folder.

Moreover, there is a ReadMe.md under the folder for each environment, which contains the detail of parameters and code structure. You can refer to these documents and carry out your experiment.