Deep Learning-Based Probability Model for Traffic Information Estimation

This study proposes a novel traffic information estimation method based on deep learning and cellular floating vehicle data (CFVD). In this paper, a probabilistic analysis model based on deep learning is proposed to consider the relationship between vehicle speed and communication behaviors. Then, a vehicle speed estimation method based on the proposed probabilistic analytical model is proposed to estimate vehicle speed. For traffic flow estimation, normal location update is adopted to estimate traffic flow. The estimated vehicle speed and the estimated traffic flow can be gathered to estimate traffic density. The proposed method is verified by the simulation tool and the experiment results revealed that the accuracies of estimated vehicle speed, estimated traffic flow, and estimated traffic density are 96.36%, 99.80%, and 96.45%, respectively. Thus, this method estimates traffic information accurately and helps improve the performance of the intelligent transportation system (ITS). © 2022, Taiwan Ubiquitous Information CO LTD. All rights reserved.