Operational filtering of traffic data

An operational filter of traffic state variables is presented for use in designing computer-aided traffic surveillance and control systems. A total of 166 data sets from three traffic surveillance systems were used in the filter development. All the data sets were best represented by an ARIMA (0,1,3) filter. This filter has the following advantages: (1) it yields minimum mean-square-error forecasts if stationarity of the observations can be obtained; (2) it provides much better results than the existing ad hoc filters; (3) it is computationally tractable; and (4) it requires modest computer storage of data. Suggestions and implications for the use of this filter are given.