Deep neural network dynamic traffic routing system for vehicles

Traffic grids have always suffered from a lack of dynamic routing and path planning algorithms and relied only on static characteristics of the roads like the number of lanes, distance and speed limits to avoid and resolve traffic congestion, by routing traffic to a lighter traffic path. However, with the increased number of vehicles in urban areas these algorithms may have reached their limitation due to the huge increase in the state space in a limited computing power and memory environment. In this research we will introduce a dynamic routing system for traffic in intersections based on real-time traffic conditions such as individual vehicle speed, destination and traffic light status to provide the fasted path between a source and a target point. This system will exploit the recent advancements in the field of machine learning by leveraging the power of deep learning especially deep convolutional neural networks. Simulation shows that the proposed model results in a path that are generally fast and avoids frequent red light stops.