Information Technology Reference
In-Depth Information
Fig. 2.
Fixed graphs used for parameter optimization and evaluation. Left: Random generated,
used for parameter optimization (1200m x 1200m). Right: Graph generated based on an existing
road network of the city of Heidelberg, Germany used for evaluation (1200m x 1200m).
Ta b l e 2 .
Values of the various weights used during the evaluation. The three on the left side
denote the weights for the different heuristics while the two on the right were used for score
computation of Heuristic 2 (see Algorithm 3).
Heuristic Weight Heuristic 2 score
weight
value
score
value
α
0
.
5
ω
2
β
0
.
3
τ
1
γ
0
.
2
E
are consid-
ered as wireless connection between two vehicles. The abstract defined functions
f
,
w
and
g
of Section 3 are then implemented as follows: Given a time
t
,
f
assigns vehicles
to specific road segments. Weight is calculated out of the distance to the beginning of
the assigned road segment. In our application domain,
f
is responsible for controlling
and updating positions of vehicles therefore realizing vehicle movements over time.
Function
w
returns the distance between a vehicle in transmission range and a junction
of the city map. The vehicles further include the road segments which they have passed
as well as their most probable path in their position updates. This realizes function
g
.
V
, edges
e
vehicular ad-hoc network. Vehicles are represented by
v
∈
∈
4.2
Scenario
Evaluation was done in a scenario where every vehicle starts one shortest path com-
putation to a given car, e.g. the center of the underlying city map. The destination car
remained stationary while all others were driving a route on the map.
We optimized weights for heuristics 1 - 3 introduced in Section 3 on a randomly
generated map shown in Figure 2 (left). Weights and values for
ω
and
τ
,whichwere
found to be optimal for our algorithm, are given in Table 2. We optimized for high PDR.
After optimization, evaluation was done on a map generated out of an existing city
environment (Heidelberg, Germany, Figure 2, right). Evaluation runtime was 120 sec-
onds where we started shortest path computation after 20 seconds simulation time.
This ensured a fair distribution of vehicles on the map.
n
vehicles per second were
placed on the map by the simulator and removed after they completed their route where
