Reliable Traffic Density Estimation in Vehicular Network

Traffic density estimation relying on vehicular ad hoc networks can facilitate many applications, including traffic management, infrastructure planning, pollutant measurement, etc. Traditional density estimation is achieved by counting the number of vehicles located in a certain area with inductive loop detectors and cameras, which suffers from limited coverage and high cost. In this paper, we propose to fuse vehicle spacing information collected through the vehicular network and compute average spacing in a specific place within a short period. With received information on spacing, a Data Center (DC) can estimate average spacing with a maximum likelihood estimator. More importantly, modern vehicles can be attacked through their open interfaces, sending modified information to the DC. We analyze the estimations under different kinds of Byzantine attacks and propose corresponding estimation methods. The estimation mechanism proposed in this paper can exploit the historical prior probability to obtain unbiased spacing estimation without the necessity of identifying whether a specific vehicle is attacked or not. The theoretical analyses in normal states and under Byzantine attacks are presented, respectively. Finally, we carry out experiments with the US Highway 101 data and show that the average spacing estimation is consistent with the real mean value.