Large-Scale Full-Coverage Traffic Speed Estimation under Extreme Traffic Conditions Using a Big Data and Deep Learning Approach: Case Study in China

Abstract

For both travelers and traffic operation centers, especially under extremely large traffic volumes, full-coverage traffic state monitoring of a major corridor is urgently needed. In the present paper, a traffic speed estimation method is proposed using a big data and deep learning approach under extreme traffic conditions. Particularly, a geospatial mapping method is proposed in this paper. This method ensures the scalability and easy-deployment, extracts phone speed (PSP) and phone count (PC) from raw cellular data, and estimates the traffic speed using a deep long short-term memory (DLSTM) neural network. The proposed method is used to estimate traffic speed for a major expressway in China that is installed with limited roadside equipment. The field test, which gives a promising performance, was performed during the Golden Week, the Chinese national holiday in 2014 (00:00 October 1 to 23:59 October 7) on the nearly 250-km-long busy freeway, G42, for both directions. The results suggest that the proposed cellular-based system can be an alternative and supplement solution for monitoring various practical traffic states, especially when only limited conventional roadside equipment is installed.