Information Technology Reference
In-Depth Information
In order to analyze
inter-driver-variability
of parameters used to describe
car-following behavior in trac simulations a total of 85 car-following tra-
jectories from the vehicles, following the equipped vehicle were used for cali-
bration. In this case the identified values for parameter b indicate that those
drivers are inclined to exert a more severe braking (apply higher rates of de-
celeration) compared to the test driver of the equipped vehicle. Smaller values
for parameter b are due to the smaller distance the following vehicles kept
to the equipped vehicle. Because the driver of the equipped vehicle knows
that he takes part in an investigation, it can be assumed that the test driver
was driving more carefully than the following vehicles. This reactivity is also
known as the so called “Hawthorne effect”. The identified parameter values are
distributed from
m/sec
2
indicating the
inter-driver-variability
.
In a second step a comparison of aggregated empirical distance behavior
(expressed in terms of the time headway in seconds) of the vehicles following
the equipped vehicle, the simulation results using the standard parameter
set of AIMSUN and the simulation results using the identified parameter
set for car-following behavior identified for
inter-driver-variability
has been
performed. An adjustment of the identified parameters was necessary as AIM-
SUN assumes that parameters are distributed normally, but some identified
parameter distributions were skewed to the right. For all other parameters
(not considering car-following) the default parameters of AIMSUN were used.
In the case using the standard parameter set of AIMSUN it becomes ob-
vious that the mean value differs from empirical data and also the standard
deviation is lower than the one derived from empirical data. This compari-
son demonstrates the need to calibrate and validate the simulation model. In
the case using the identified parameters for the car-following model the time
gaps still differ from the empirical data. In reality, shorter headways could
be observed. However, in comparison to the standard parameters a closer
approximation of simulative data with empirical results could be achieved.
Particularly the
inter-driver-variability
expressed by the larger standard de-
viation is represented much more precisely. The systematic weakness of the
simulation tool seems to be the main reason for the remaining deviation.
Thus, the adaption of the identified parameters for the car-following model
in AIMSUN mentioned above was necessary.
m/sec
2
to
2
8
8
Conclusion
It has become obvious that previous approaches towards a calibration and
validation of trac simulation models do not serve the needs for investigation
and optimization of TAS. Therefore a new two-level approach was presented,
which combines a calibration on the microscopic level with a validation on
the microscopic and macroscopic level. The paper presented a data acqui-
sition method, which allows the consideration of measurement data of the
microscopic and macroscopic level originating from the same time span and
the same road network. To obtain the empirical microscopic data a vehi-
