PROJECT TITLE :

Traffic Signal Control Using End-to-End Off-Policy Deep Reinforcement Learning

ABSTRACT:

However, road intersections have historically been among the most significant traffic bottlenecks that have contributed to traffic congestion. An effective transportation system can provide significant benefits to our society. It's possible that better timing of traffic signals that are adapted to real-time traffic could help relieve some of this congestion. However, the majority of the existing methods for controlling traffic signals require a massive amount of information about the road, such as the positions of the vehicles. In this article, we zero in on a specific road intersection and work toward reducing the typical amount of time spent waiting there. A traffic signal control (TSC) system that is based on an end-to-end off-policy Deep Reinforcement Learning (deep RL) agent and background removal residual networks is something that we propose. The input for the agent are real-time images taken at the intersection of the roads. After receiving adequate training, the agent will be able to perform (nearly) optimal traffic signaling based on the current conditions of the traffic. Experiments are run on a variety of intersection scenarios, and different TSC methods are compared to one another. The results of the experiments show that our end-to-end deep RL approach is superior to other TSC methods in terms of its performance and its ability to adapt to the dynamic traffic based on the traffic images.