Information Technology Reference
In-Depth Information
Fig. 5.
Screenshot showing part of the motorway (blue), containing the VMS and the
junction (green). The arrows indicate the direction of travel.
dEI(20) <= 0.019705
|
meanV <= 101.907816
| flow <= 214: a0 (62.0/18.0)
| flow > 214: a1 (1724.0/391.0)
meanV > 101.907816
|
dEI(20) <= 0.01273: a0 (13773.0/4413.0)
.
flow <= 14
|
|
flow <= 2: a1 (5.0)
|
|
|
dEI(17,18) > 0.004072: a1 (1311.0/526.0)
|
|
flow > 2: a0 (30.0/2.0)
|
|
dEI(17,18) > 0.011574: a1 (3771.0/1061.0)
|
flow > 14: a0 (660.0/66.0)
|
dEI(20) > 0.02811: a1 (3818.0/433.0)
with
a0 = no speedlimit
a1 = speed limit 130km/h
with
a0 = turn off VMS
a1 = turn on VMS
dEI(x) =
density
(
EI
x
)
Fig. 6.
Classifier of static and dynamic scenario
described in the last subsection. After this time slice, the result is transmitted
to the server and situation after having applied the better action is used for the
subsequent simulation. The server lets the clients simulate until 10,000 data sets
are generated.
The classifier is tested on the same scenario and with a duration of 48h sim-
ulation time (172,800 iterations). The classifier determines the action to take
for a period of 60s. The trac volumes on the two starting points vary over
time. The simulation is performed 100 times and the mean overall velocities for
the different strategies are computed: “turn off VMS”: 125
.
12, “turn on VMS”:
125
.
44, and classifier: 126
.
41km/h. The corresponding boxplot is shown in Fig-
ure 4 on the right. Statistical tests with the same setting as before lead to the
same result (
p<
2
.
2
e
16 in both cases comparing the classifier to the fixed
settings) and thus, indicate a greater mean velocity of the classifier. The trained
classifiers and the confusion matrices of both scenarios can be seen in Figures 6
and 7.
The average travel times per car in the different settings are:
a
0
: 289
.
53,
a
1
: 282
.
15, and
classifier
: 279
.
89 seconds. Scaling these numbers to the
simulated time of 48 hours and the average number of cars in each scenario
(43148
.
09), using the classifier approx. 120h are saved in comparison to the
fixed
a
0
and 27h in comparison to the fixed
a
1
setting.
−
