Mobile Sampling for Traffic Estimation
Mark Christopher Johnson and Kannan Ramchandran
Fannie and John Hertz Foundation, National Science Foundation 0729237 and National Science Foundation 0635114
We study the problem of estimating the traffic conditions on a highway from the perspective of sampling theory. The standard method of collecting traffic data has been through loop detectors embedded in the road surface, which count the number of cars passing a fixed point. In our work, we examine the alternative concept of using "probe vehicles" which report their position and velocity, either once or periodically, to a traffic management system. Possible methods of collecting such data include triangulating the position of cellular phones in the vehicles, or through GPS receivers. The velocity samples measured by the probe vehicles are then fused with a model of traffic dynamics to reliably estimate the velocity field on the road. Our preliminary results show that by incorporating probe vehicles into existing systems, we can achieve the same estimation error performance with a smaller number of loop detectors. The fundamental challenges involve dealing with the uncertainty in the position associated with each velocity sample, and localizing each vehicle to the correct road.
More information: http://www.eecs.berkeley.edu/~mjohnson