Collaborative Traffic Signal Automation using Deep Q-Learning

Every metropolitan trip is punctuated by traffic signals, which have an immediate effect on drivers, the environment, and the economy whether the route is crowded or not. Traffic signal automation to reduce traffic delay is a major issue all over the world. Nevertheless, the current solutions to reduce exponentially rising traffic issues are not completely dealing with the problem. Companies, traffic engineers and researchers have suggested several Traffic Signal control systems. The main function of the traffic signal management system is to coordinate individual traffic signals to accomplish operational goals for the entire network. The single junction-based systems are unable to reduce the waiting time of exponentially increasing traffic load on the roads. To deal with this, we propose collaborative signal automation on a traffic simulator based on reinforcement learning techniques. The model utilized a q-learning technique that depicts composing units of addressed issues: agents, surrounding and response. The collaborative network takes advantage of traffic flow prediction with signal automation. Multi-junction road environments and vehicles are fed to the network as input. The proposed system suggests optimal signal automation to alleviate delay time and sequence length of traffic. Q-learning-based model decreases the wait time and leads to a steady flow of vehicles with several significances in composite traffic areas.