Information Technology Reference
In-Depth Information
4.1 Guidance of the Vehicles
A real driver usually has the intention to reach some goal with his driving.
This makes it necessary to incorporate routes in the modeling. In principle,
there are two different strategies to solve this problem. One can assign an origin
and a destination to the road user and then guide him through the network
according to this route [2,4]. For our network origin-destination information
with a su
cient temporal and spatial resolution is not available. Therefore, the
vehicles are guided in the network according to the probabilities calculated on
the basis of the measured data. This means that a car is not guided through the
whole network, but every time it reaches a new link it will decide in accordance
with the measured probabilities how it leaves the link.
To implement this we use forced lane changes. Forced lane changes are nec-
essary so that the cars can drive from on-ramps on the autobahn, from the
autobahn on off-ramps, when the autobahn narrows, and when cars drive from
one particular section of the autobahn on another over an intersection. Forced
lane changes differ from free lane changes in a fundamental way. While free lane
changes give vehicles the opportunity to overtake vehicles driving slower and
thus reduce disturbances, forced lane changes stem from the need to reach a
node and are obviously an additional source for disturbances.
The simulator uses gradually increasing harsh measures to force lane changes.
At the beginning of an area where a car could change to the target lane, it does
so, if the gap is su
ciently large and no car is severely hindered. At the end
of the area it will bully into any gap regardless of velocity differences. Further,
a vehicle driving on its target lane should not leave the lane to overtake. An
e
cient implementation of this strategy is to store the lane change information
in the cells. This gives a fast access through the coordinates of a vehicle. Of
course this information depends on the node chosen and whether the vehicle is
a lorry or a passenger car. Because of this every link has several versions of the
lane change information.
4.2 Sensor Input
To incorporate the real world measurements from the loop detectors into the sim-
ulation vehicle-moving, inserting, and removing algorithms have to be applied.
This is done at the so-called checkpoints, which are located at those places in the
network where a complete cross-section is available, i.e., all lanes are covered by
a loop detector. Every time, when checkpoint-data is provided the simulator uses
the measured values to adjust the tra
c state in the simulation. The first step
is to try to move vehicles behind the checkpoint in front of it and vice versa.
If this is not enough to adjust the tra
c state, cars are inserted or removed.
This should be preferred to pure insert/removal strategies, because these can
completely fail due to positive feedback if a non-existing tra
c jam is produced
by the simulation. In this case the simulation measures a low flow in comparison
with the real data, so cars are added periodically to the ever growing tra
c jam
leading to a total breakdown.
For realistic results it is further important to minimize the perturbation of
the dynamics present in the network due to the data integration. Therefore,
