Deep Learning Enabled Fine-Grained Path Planning for Connected Vehicular Networks

In this paper, to alleviate the ever-increasing traffic congestion in urban areas by accommodating higher road traffic, we develop traffic prediction framework together with path planning method for connected vehicular networks. First, through the employment of convolutional neural network (CNN) and residual unit (RN), deep learning (DL) based fine-grained traffic prediction algorithm is designed to obtain the spatial-temporal characteristics of vehicular traffic. The regionally fine-grained traffic prediction framework can realize real-time traffic prediction of future changing trends at each road with a high accuracy and reliability. Second, we propose a gridded path planning method by making use of the traffic prediction information. The accuracy of selected path, complexity of path calculation, and adaptive path adjustment are jointly taken into consideration by achieving the refined traffic regulation in different gridded section. Finally, we utilize the actual vehicle data from the city of Beijing and digital map on OpenStreetMap to validate the effectiveness and reliability of the proposed traffic prediction framework and path planning method. Simulation results demonstrate that the proposed approach is capable of relieving urban traffic congestion based on the existing roadway systems, which can provide methodological guidance for data-intensive traffic management.

Automated summary

In this paper, a traffic prediction framework and path planning method for connected vehicular networks are developed to alleviate urban traffic congestion. By using deep learning algorithm, the spatial-temporal characteristics of vehicular traffic are obtained, and the path planning is refined based on the traffic prediction information. The proposed approach has been validated using actual vehicle data and digital map, showing its effectiveness in relieving urban traffic congestion and providing guidance for data-intensive traffic management.