In the Netherlands there is a lot of traffic data available. Most of these data is measured using dual loop detectors, which are placed in the asphalt. These detectors measure the speed and length of a vehicle and combine that with the passage time of a vehicle. This information is used to calculate other values, such as a time-headway or a flow. Usually this data is saved in a database, aggregated over a one-minute interval. Only in special cases the individual vehicle data is saved, mainly because the database would become too big to handle. The individual vehicle data is however used for certain applications, for example the Automatic Incident Detection, which uses the speed of successive vehicles to determine a maximum speed. This maximum speed is then displayed on matrix-signs above the highway, which most will have seen in the western part of The Netherlands.

The detectors are usually placed on a distance of 500 meters between each pair. Thus only each 500 meter (or more, depending on the actual distance between two successive detectors) the state of traffic is known in the form of speeds, flows, densities, etc. For example for incident detection, this distance is still quite big, because depending on the actual location of the incident, it could take up to 3 minutes before it is detected. If information about the traffic state between the detectors was available, the detection of an incident may be done much earlier. Another possible application of more information about the traffic between detectors is a ramp-metering algorithm that uses the time headway between successive vehicles upstream of an on-ramp, instead of values such as flow and occupancy.

These applications indicate that there is a need for information about the traffic stream, especially between the detectors. In this project an investigation is performed, considering a reconstruction of a spatial temporal traffic stream on Dutch highways, using the information already available by the current detectors. This is a method to gather information about the traffic state between detectors.

On a macroscopic level, such a method is already available. This method is called the adaptive smoothing method, developed by Treiber and Helbing. Because such a method is already available, the method developed in this research is at a microscopic level.

Three different options for the method were developed, out of which one was selected for further elaboration. These three methods are:


Simulation using global statistics;

This method is based on the use of a microscopic simulation; the vehicles are simulated based on the global statistics of the measured traffic stream such as an average flow and its deviation. These global statistics are based on an interval, which can vary from twenty seconds to a full day. In contrast to the next method using simulation, this will not provide an exact reconstruction of the traffic stream.


Simulation using vehicle matching;

This method is also based on the use of a microscopic simulation in which the simulated vehicles are altered (speed, headway, etc.) to match the measured vehicles. To make sure this alteration is possible, the measured vehicles are matched on each detector. This means that a vehicle can be traced on a highway section.


Interpolation of a traffic flow signal;

This method is based on signal theory. A number of measured vehicle speeds can be seen as a signal. Several signal theories can be applied to interpolate the traffic flow signal between detectors to create the information between these detectors.

The method simulation using global statistics was selected for further elaboration, because this method seems the most promising of all three. Simulation using vehicle matching is heavily dependent on the actual matching, which has never been tested in Europe (with its different vehicle and fleet charachteristics from the USA). Interpolation of a traffic flow signal has got underlying assumptions, which have not been tested, causing it to be not preferred above the method involving simulation using global statistics.

The microscopic simulation that is used is Fosim (Freeway Operations SIMulation). The most common use of Fosim in the Netherlands is the estimation of capacities for highway sections, including merging areas and on- and off-ramps. All the information needed to run Fosim is described in detail and the underlying car following and lane changing models are presented in this report.

The method is tested on a highway section of the A2 between Maarssen and Abcoude where individual vehicle data were available. Of all the available days one day was selected, based on criteria such as the existence of congestion and the sort of day (bank holiday, workday, etc). This selected day is the 26th of June 1998 and of this day only the morning peak was chosen, between 6:00 and 10:00. Fosim was set up with the data that was gathered from the available data from this day and then calibrated for this data.

The first calibration aimed at reconstructing the traffic on a global level first and then trying to improve the local detail. However the global traffic state that was found in the measurements could not be reconstructed in the simulation without great alterations in the input data. This meant that this calibration did not work. Another calibration set was done, aiming at reconstructing the traffic flow on only the first four detectors between which are no external influences such as on- and off-ramps. Again the measurements of these four detectors could not be reconstructed. Yet another try with two successive detectors with only 500 meter distance between them, gave the same results; the reconstruction of the spatial temporal traffic flow is not possible using Fosim. Several reasons for this inability to reconstruct the traffic flow have been found in the research. However, the fact that Fosim is not able to resemble the measurements does not necessarily mean that each microscopic simulation is unable to resemble these measurements.

Since the microscopic approach to the reconstruction did not work, the macroscopic approach was used to reconstruct the traffic flow. The adaptive smoothing method was programmed and applied to the same dataset as was used with the microscopic approach. This method gave much more promising results. It is able to reconstruct the spatial temporal traffic flow on a Dutch highway. However, it has to be mentioned that the results of the method are not as robust as the developers of the method have shown. Several reasons for this are presented.

Final conclusion is that it is possible to reconstruct a spatial temporal traffic stream on a Dutch highway using dual loop detectors on a macroscopic level. To do the same at a microscopic level more research is necessary.