Filtering techniques for urban traffic data

There is evidence that the algorithms for estimating traffic flows from sensor data need to be improved before computer controlled traffic responsive urban traffic control systems can reach their full potential effectiveness. A large part of the problem appears to be that the data from traffic sensors is, in the statistical jargon, a marked point process. It is only very recently that the theoretical techniques for estimation based on point process data have reached the sophistication needed for traffic problems. In this paper, these techtechniques are used to derive several recursive algorithms for filtering traffic sensor data and predicting urban traffic flows. These filters and predictors are then evaluated and compared using simulated traffic data.