Signal Synchronization of Traffic Lights Using Reinforcement Learning

Today, the traffic problem is a serious problem, especially in big cities. The increasing number of cars with the increasing population further increases the traffic problem. This traffic problem increases travel times, increases fuel consumption, causes many accidents, and negatively affects human psychology. One of the reasons that increase traffic on the roads the most is traffic lights. Since traffic lights are used at most intersections, large traffic occurs at intersections. The signal periods of most traffic lights are predetermined using data from the intersection. However, these traffic lights are not adaptive to different situations that may occur on the road. In the study, we tried to make non-adaptive traffic lights adaptive using deep reinforcement learning. In the study, the signal periods of traffic lights were managed by using a reinforcement learning agent trained on simulation. When the training performances of the Reinforcement learning agent suggested in the study were examined, it was seen that the training was successful. It has been observed that there is a decrease in queue length and average delays experienced by vehicles at the intersection. It has been observed that the reinforcement learning algorithm can work well at intersections.