An Improved Traffic Signal Control Method Based on Multi-agent Reinforcement Learning | IEEE Conference Publication | IEEE Xplore