Traffic monitoring is still a major challenge for traffic management systems. Traffic monitoring services are nowadays performed with traditionally wired systems that have high installation costs. Therefore, they are not adequate for dense deployment. In this paper, we propose a traffic monitoring approach that is passive and exploits wireless signals sent in a vehicular ad hoc network. We utilize machine learning techniques to make inferences on traffic conditions, directly based on raw samples observed at the receiver. To verify the feasibility of this approach, we first created simulation data, using a ray-tracing simulator and a traffic simulator. The data contain wireless channel realizations under realistic traffic conditions created by the traffic simulator. Next, we collected data in the form of dedicated short-range communications (DSRC) signals transmitted from a roadside unit (RSU) deployed at an intersection and replicated the same experiments with real data. The results show that our approach successfully distinguishes different traffic intensities with an accuracy of 91.6%, 86.4% on simulation and real-world data, respectively. It also tracks the number of vehicles on the road with a mean absolute error (MAE) of 1.3 on both datasets. The proposed approach can be employed alongside the current monitoring systems to increase the situational awareness in a vehicular network without requiring additional investment in infrastructure.
Increasing Situational Awareness in Vehicular Networks: Passive Traffic Sensing Based on Machine Learning
2020-05-01
1425022 byte
Conference paper
Electronic Resource
English
Acoustic Vector Sensors Increasing UAV's Situational Awareness
SAE Technical Papers | 2009
|Situational Awareness in Air Traffic Control
British Library Conference Proceedings | 1994
|Event-based Sensing for Space Situational Awareness
British Library Conference Proceedings | 2017
|Event-based Sensing for Space Situational Awareness
Online Contents | 2019
|Event-based Sensing for Space Situational Awareness
Springer Verlag | 2019
|