A Multi-Agent Reinforcement Learning Framework with Recurrent Communication Module for Traffic Light Control | IEEE Conference Publication | IEEE Xplore