Multilayer Modeling of Driver Behavior Based on Hierarchical Mode Segmentation - Digital Signal Processing for In-Vehicle Systems and Safety - page 59

Fig. 4.3 Dendrogram of clustering (Examinee A)
• The expressway is endless and has two lanes - the cruising lane and passing lane.
• There are ten cars on the cruising lane. Five of them are in front of the examinee's
car. The remaining five are behind the examinee's vehicle. Their velocities vary
from 70 to 85 km/h. Once the examinee's car overtakes the lead vehicle, then the
tail-end car is moved in front of the lead car. The examinee is not aware of the
switch.
• There are ten cars on the passing lane. Five of themare ahead of the examinee's car.
The remaining five cars are behind the examinee's vehicle. Their velocities vary
from 90 to 110 km/h. Once the lead car passes the examinee's car, then the tail end
is moved in front of the lead car. The examinee is not aware of this change.
• The range between cars is set at 50-300 m, and there is no collision between cars
except the examinee's car.
• The examinee's car change lanes while the other vehicles stay on their lanes.
Five examinees performed the test driving using the driving simulator. Note
that the examinees were provided with the instruction “Drive the car according to
your usual driving manner.” Since this instruction is “broad,” the examinees did not
concern themselves much with the environmental information. As a result, each
examinee drove his/her usual way.
4.3.2 Observed Behavioral Data and Clustering Results
The unsupervised clustering based on the feature vector shown in the previous
section has been applied to the observed driving behavioral data. The dendrogram
obtained from the proposed strategy is shown in Fig. 4.3 wherein the vertical axis
represents the dissimilarity between clusters.