The primary function of an automatic surveillance system is to gather information that describes the behaviour of the traffic stream. This is accomplished by the detection, or measurement, of traffic variables and then by data processing the detector' outputs. The measurable traffic variables are stochastic processes because the behaviour of the individual vehicles in the traffic stream is random. The stochastic nature of the available information allows a data-handling system to be designed such that it implements the best linear estimator of the desired information. Since the desired information has statistics that are nonstationary in both time and space, the optimum location of detector sites is a function of the time and the set of points along the freeway at which the information is being estimated. A solution of the problem of optimum detection sites and the corresponding data-handling system is presented. The solution of the estimation problem is an optimisation algorithm that may be used during the preliminary design stage of a surveillance and control project, or as an analysis technique for evaluating the effectiveness of existing surveillance systems. An example is included to illustrate the design procedure presented.
Abstract