Information Technology Reference
In-Depth Information
third step, a thorough analysis of the underlying causes is needed. The fourth
step is the model adjustment. This can either be an adjustment of the pa-
rameters (parameter variation), or a change in the underlying mathematical
functions relating the parameters to each other (structural variation). After
having adjusted the model we have to start with step one again and perform
a new simulation run.
In the case the deviation during calibration is below the threshold, the val-
idation of microscopic model parameters immediately follows the calibration.
The calibrated model is transferred into a new, but comparable situation. For
this new situtation a simulation run is performed. The microscopic simula-
tion results are to be compared to an additional set of measured (microscopic)
data. In case the previously defined threshold is exceeded a re-calibration is
needed, thus the causal analysis is the subsequent step. The model gives valid
results on the considered microscopic level when the previously determined
error measurement bound is not exceeded.
As soon as microscopic parameters have been successfully validated a val-
idation of macroscopic variables (e.g. trac flow, trac density, mean speed)
is possible. First a microscopic simulation run is performed. In a second step
the macroscopic simulation results are compared against empirical data. In
case the results stay within the permissible range the calibration and valida-
tion on microscopic and macroscopic level has been successful. Otherwise a
re-calibration on microscopic level becomes necessary. Now it becomes obvi-
ous that calibration is only possible on the microscopic level (as microscopic
variables are independent variables whereas macroscopic variables are depen-
dent variables).
6
Data acquisition concept
To carry out the two-level calibration and validation both macroscopic and
microscopic real-trac variables need to be gathered in such a manner that
they share the same time and geographical reference. Additionally, the equiv-
alent microscopic and macroscopic simulation results are required.
As input for simulation and for validation on macroscopic level an
empir-
ical data acquisition on macroscopic level
is necessary. The knowledge about
the trac flow within the considered section of the highway is essential. Along
the track this data is available at trac control centers. However, the avail-
able data is insucient and needs to be amended with empirical data from
on- and off-ramp trac flow. For this reason a use of additional (temporary)
sensors is proposed. For an installation nearby the track the radar sensor as
well as the passive infrared sensor is suitable [9].
To describe the individual driver behavior,
empirical data acquisition on
the micro-scopic level
is necessary. The current focus of investigation of indi-
vidual driver behavior is the car-following behavior. In the selected approach
an experimental vehicle is equipped with sensor systems based on radar and
lidar technology, which measure distance and relative speed to the preced-
