The study was conducted in April and May 2010 using the dynamic driving simulator at BMW Group’s Research and Innovation Center in Munich. Details regarding test design, subject sample, findings as well as further implications are discussed in this chapter.
4.2 Test design and subject sample
Driving data as well as subjective ratings and comments were collected during the experi-ment.
The experiment lasted about 110 min. and was structured into the following segments (the situations are described in detail below):
• Welcome; collection of demographical data.
• “Acclimatization” to the driving simulator (≈12 min.):
– Start on a rural road, then transition to urban environment.
– Two normal situations: Parking bay. Pedestrian crossing.
• Experimental section (≈20 min.):
– Urban environment.
– Four normal situations: Parking bay (2x). Pedestrian crossing (2x).
– Two highly critical crossing situations with and without system. The sequence of those two situations is varied between subjects.
• Interview.
• Subjects are instructed regarding research questions and the specific functions of the system under investigation.
• Acceptance section (≈22 min.):
– Different false system actions are presented to the driver.
– The sequence is not varied between the subjects.
• Interview.
• “Clearance test”: Artificial situation for investigation of the normal passing clearance for a pedestrian (≈5 min.).
The subject sample was reduced to 20 persons for the acceptance section and the clearance section. The subject sample itself is introduced after a discussion of the specific situations.
Test design specifics: normal situation
The normal situations cannot be identified by the subjects as experimental situations as they are implemented as common uncritical interactions with pedestrians without any system actions involved. The data collected are used to gain knowledge about the normal interaction and draw conclusions on the perception of discomfort or hazard while interact-ing with pedestrians.
In thefirst normal situation, the driver has to pass a pedestrian walking beside the street along a parking bay (Fig. 4.1). The lateral position of the pedestrian is different in each of the three occurrences of the situation in the experiment:
P-1. Walking on the road marking.
P-2. Walking left of the road marking (0.2 m left compared to P-1.).
P-3. Walking right of the road marking (0.5 m right compared to P-1.).
Thesecond normal situation is a common crossing scenario where a pedestrian crosses the street from the right (see Fig. 4.2). A key parameter characterizing the mitigation of a potential conflict is the time-to-collision (TTC), which is defined here as the distance to the projected collision point divided by the current vehicle speed. The TTC when the pedestrian enters the street in this situation is varied during the experiment:
C-1. Low TTC: approx. 5.4 s.
C-2. Medium TTC: approx. 6.6 s.
C-3. High TTC: approx. 7.8 s.
The TTC values here represent the average for specific situations over all subjects. There is a small variation between the subjects due to a technical characteristic of the driving simulator: Pedestrians have a given (unalterable) motion characteristic where the starting point, the final speed, and the trajectory can be defined. That means that once the pedestrian is in motion, his speed cannot be adjusted depending on the current motion parameters of the vehicle. As a result, the actual TTC where the pedestrian steps on the street varies due to small deviations in vehicle speed between the subjects while approaching the pedestrian.
Test design specifics: clearance test
20 randomly chosen subjects participated in the clearance test. The task was to pass 10 pedestrians each. This task resembles an artificial (i.e., unrealistic) situation, where the pedestrians walk on an empty highway segment towards the vehicle (see Fig. 4.3).
Each pedestrian has a different lateral clearance to the road marking. The drivers are instructed to drive at 50 kph and pass the pedestrian to the left at a clearance that is still acceptable for them and to ignore all road markings.
Test design specifics: acceptance
The acceptance of false system actions was investigated in the second part of the ex-periment. All subjects were informed about the experiment and the system and were confronted with several situations that could trigger undesired system responses. Seven situations selected by an internal expert panel were presented to the subjects. The subjects had to give ratings
• regarding hazard of the undesired system action for the traffic situation as a whole and
• their individual acceptance of the false system action.
4.2 Test design and subject sample
Figure 4.1: Normal situation “passing” with variations P-1 to P-3 (from top to bottom).
Figure 4.2: Normal situation “crossing”. Figure 4.3: Clearance test on the highway.
The system response in those situations was triggered in order to obtain a reliable pre-sentation for every subject. In order to get a more realistic feeling in some situations, the TTC of the optical pre-warning was set earlier than described above. After each situation, the subjects had to stop the vehicle and had to go through an interview. The situations are described in detail in the following:
Situation 1 (Fig. 4.4): False system response due to a pedestrian on atraffic island: The driver approaches an urban environment at approx. 60 kph. The system reacts because of a pedestrian standing at the edge of a traffic island. The pedestrian is directly in front of the vehicle at the moment of the warning.
Situation 2 (Fig. 4.4): False system response while negotiating anevasive maneuver due to a pedestrian on the left side of the street: The driver has to make an evasive maneuver because of a parked car in his lane. The system reacts because of a pedestrian standing at the edge of the left sidewalk.
Situation 3 (Fig. 4.4): False system response due to a pedestrian on the left side of the street while negotiating a rightcurve: The driver follows the street in a curve to the right.
The system reacts because of a pedestrian walking on the left sidewalk.
Situation 4 (Fig. 4.5): False system response due to a pedestrian on the opposite side of an intersection while negotiating a right curve: The driver follows the street in a right curve while crossing an intersection.
Situation 5 (Fig. 4.5): False system response due to a pedestrian on the opposite side of aT-intersection: The driver approaches a T-intersection and the pedestrian is standing on the opposite sidewalk directly in front of the vehicle.
Situation 6 (Fig. 4.5): False system response due to a pedestrian walking at the side of the street in a parking bay: The driver passes the pedestrian and the system reacts.
4.2 Test design and subject sample
Figure 4.4: Acceptance situations: Traffic island (1); Evasive maneuver (2); Curve (3), (from top to bottom).
The situations described above represent undesired actions of the system. The next situation does not present an undesired system action. In this situation, the driver makes a right turn at an intersection while a pedestrian is running across the street from behind (Fig. 4.6). The system does not show any reaction; this is a situation with a false-negative response. The question here is whether the subjects expect the system to handle this situation or not.
Test design specifics: highly critical situation
The last situation included in the experiment was a highly critical situation. The subject drives for a long time through the city and has to fulfill a secondary task several times without anything happening. At the end of the first half of the experiment, the driver is confronted with a highly critical situation, which represents the most common real-world accident scenario (see [61] or Section 3.2, p. 44). A pedestrian is crossing the street from the right (Fig. 4.7). The presence of other pedestrians could have an influence on the situation, but is not recorded in accident data. At that time, the driver is working on the secondary task (see description below). The situation has to be managed twice by each driver (with and without system). The sequence of the highly critical situation was varied between the subjects (regarding the system), i.e., 20 persons had the first highly critical situation with the system, the other 20 without. The actual number of measurements per situation can vary as not every person went through the whole experiment.
Test design specifics: secondary task
The secondary task is a visual loading task, which does not force the driver to take his hands off the steering wheel. The loading task has been successfully used in previous internal studies. It is not interruptible and produces a constant visual distraction. The loading is constant and it is hardly possible to develop individual strategies for solving the task. The driver is confronted with single letters displayed in the central information display of the vehicle. Each letter is displayed for a very short time, which makes constant monitoring by the driver necessary. As soon as a number appears instead of a letter, the driver has to press a button on the steering wheel within 1 s. The beginning of the task is introduced by an acoustical signal. The task appears several times within the experiment and lasts about 1 min. each time. The task is used to produce visual distraction and
“prepare” the subject in a better way for the critical situation.
Subject sample characteristics
The subject sample consisted of 40 persons, aged 22 to 60 years (average 37.3 years, SD 10.9 years). 13 persons were female, 27 male. All subjects were BMW Group employees not working or acquainted with driver assistance or systems of active safety.
4.2 Test design and subject sample
Figure 4.5: Acceptance situations: Intersection curve (4); T-intersection (5); Parking bay (6), (from top to bottom).
Figure 4.6: Acceptance situation: pedestrian running at intersection.
Figure 4.7: Highly critical situation: pedestrian crossing from the right.