Deep Recurrent Q Networks for Urban Traffic Signal Control

Traditional traffic signal control methods have not played an effective role in alleviating traffic congestion. In order to make better signal control decisions, this paper combines deep recurrent Q network (DRQN) to carry out in-depth research on the optimization of signal controls. The core idea of this approach is to set up a deep neural network to learn the Q-function of reinforcement learning from state inputs and performance output. Specific state, action and reward functions were defined in reinforcement learning, a long short-term memory (LSTM) network was used to fit the state, then the appropriate signal control strategy was proposed. SUMO software was used for simulation experiments to create a single intersection traffic signal control.