Information Technology Reference
In-Depth Information
ously discussed they can be further differentiated into
elementary properties
and
emergent properties
:
In the domain of trac modelling different model structures developed,
which may be classified with respect to different levels of emergence [3], [5],
[6].
Nanoscopic trac models
simulate the cognitive processes within the
driver or the mechanics of the vehicle.
Microscopic trac models
simulate
the behavior of single driver-vehicle units and describe their interactions by
rule-bases that specify acceleration or velocity. Generally the dynamics are
based on microscopic variables like distance or relative speed to the front
or rear vehicle.
Macroscopic tra
c models
neglect individual vehicles. They
are devoted to aggregate state variables like tra
c density and tra
c flow
for representing the collective behavior of vehicles. The equations are derived
from the laws of nature and are structurally often similar to fluid dynamics.
3
Requirements for investigation of TAS
In recently developed TASs the local (individual) vehicle behavior is influ-
enced in order to globally optimize trac flow [7]. In order to obtain valid
conclusion the following requirements for the investigation of TAS need to
taken into consideration [2]:
For quantifying the benefit of these systems in real tra
c a lot of test ve-
hicles would be necessary. To equip so many vehicles is almost not possible.
Therefore a
simulative approach
is necessary for the evaluation and optimiza-
tion of these systems. Simulations have advantages relating to expenditure
of time, costs and the possibiltiy to test different system alternatives.
The behavior of the driver-vehicle units is generally determined by the
models in use, the interactions of the sub-models and especially by the pa-
rameters for the models. In order to obtain reliable simulation results, it is
necessary to prove the validity of the selected parameters for each individual
application [12]. Therefore, it is necessary to perform a
calibration
, i.e. ad-
justing the model parameters so that the simulation is suciently accurate
when compared to actual behavior, as well as a
validation
, i.e. proving the
model with simulation results obtained with the identified parameters which
are compared to a second measurement data set. Quantitative statements
can only be made from a model validated suciently.
Since the TAS changes the following behavior of a vehicle to its ahead-
driving vehicle, it is absolutely necessary to use a trac model for testing
such a system which includes the headway behavior of the human driver
as well as the TAS. Therefore in principle only a nanoscopic or microscopic
simulation model can be used, which constitutes the
two-level approach
in-
troduced in this paper. For simulating a lot of vehicles in a reasonable time
span, nanoscopic simulation models are too complex. Therefore a simulation
at microscopic level is necessary. A sole consideration of the microscopic level
is not sucient. Any adjustment of local microscopic trac variables by the
TAS emerges a certain global macroscopic behavior. In order to obtain reli-
