The Three Vehicle Platooning Task (3VPT) is a more realistic driving task and allows measuring many different dependent variables, but is still controlled. It demands the driver to control his lateral as well as longitudinal position and additionally to detect and respond to several events. NOTE: This driving task is only available in the professional version of OpenDS.
General idea and goals of 3VPT
The main idea behind the development of the 3VPT was to create a driving scenario that approximated real driving in a superior way than LCT and the ConTRe task. Therefore, the question was which aspects of real driving could have been underrepresented in those previous driving tasks and how an alternative scenario could look like. These thoughts included realism of the environment, occurring events and stimuli, unpredictability of these events on the one hand and the variation of the actual driving speed and keeping an appropriate safety distance on the other hand. Also, like for ConTRe, there should be the possibility to vary the scenario for different applications.
This setting was created on the basis of the Driver Workload Metrics Project (DWM Project), as published in Task 2 Final Report (LS Angell, Auflick, Austria, & Kochhar, 2006). Research scope of this study was the driver workload associated with use of several in-vehicle systems while driving. The primary goal was to ‘develop performance metrics and test procedures to assess the visual, manual, and cognitive aspects of driver workload’ (Angell et al., 2006). Therefore, the setting described above was realized both on a highway and a test track.
Description of 3VPT
The scenario was set on a straight highway-like road with two lanes per direction which heads to mountain scenery. The weather conditions were configured to sunny and dry. Besides the three vehicles and start and finish sings there were no other objects within this scenario. The driver’s car was steered from ego-perspective. The visual elements in the head-on-display were: speedometer, rev meter, gear indicator and rear mirror (cf. Figure 1). In addition, a distance indicator was presented during the trial run and briefly at the beginning of every test trial.
The driver’s task was to drive in the middle position of a car platoon consisting of three cars overall. The other two computer-controlled vehicles were programmed to react to the driver’s speed. In addition to take care of the instructed speed of 90 km/h (56 mph) and keeping an appropriate distance to the leading car of about 35 meters (115 feet), the driver had to react to three different events: brake light activation of the leading vehicle, deceleration of the leading vehicle, and turn signal activation of the following vehicle.
Figure 1. Screenshot of the 3VPT.
The nominal instructed speed for all test runs was 90 km/h (= 25 m/s, 56 mph), what bases on the scenario of the DWM Project. One run had duration of about three minutes over a distance of 4250 meters (start – finish, 2.64 miles). The initial position of the vehicles was 250 meters (820 feet) ahead of the start sign so that the driver was able to speed up to 90 km/h. In this speed-up phase, there was also a distance feedback until reaching the start sign (cf. Figure 1).
Another parameter that was taken from the DWM Project is the longitudinal distance between the vehicles. The driver is instructed to keep a safety distance to the leading vehicle of about 35 meters (115 feet) all the time. The leading vehicle is programmed to keep a distance of 35 to 45 meters in front of the driver’s car. Also, the leading car had a maximum speed of 90 km/h, so it did not keep the above mentioned programmed distances above this velocity. The driver’s car has a maximum speed of 120 km/h (75 mph) so it was, on principle, possible to get very close to the leading car and even overtake or crash into it. This behavior was eliminated through instruction.
The following vehicle was set to keep a distance of 10-15 meters to the driver’s car and had a maximum speed of 120 km/h to ensure that it consistently followed the driver’s car at every possible velocity.
Events and demanded reactions
There were three different events which are presented four times each, so an event frequency of twelve resulted. This approximated the event frequency of the Lane Change Task (18) during an equivalent trial duration of about three minutes (Stefan Mattes, 2003).
The events which are also related to the DWM-Project were:
Brake light activationof the leading car for 2.5 seconds or until the driver reacts through pressing a button on the steering wheel with the left thumb (cf. Figure 1).
Figure 2. 3VPT brake light activation event.
Turn signal activationof the following vehicle for 2.5 seconds or until the driver reacts through pressing the same button as mentioned above (cf. Figure 2).
Figure 3. 3VPT turn signal activation event.
Deceleration of the leading vehicleto 70 km/h (43 mph) for 100 meters (328 feet) (cf. Figure 3). The driver has to react via pressing the brake pedal and keeping an appropriate safety distance. After that, the leading vehicle accelerates again so the driver has to keep up and reach the initial velocity and safety distance.
Figure 4. 3VPT leading vehicle deceleration event.
The timeout for successful reactions was set to five seconds.
Conception of different tracks
To avoid effects of implicit learning when reacting to the events ten different track versions were created. These included differing sequences of events as well as unsteady time intervals between their presentations.
Within the logging-area on the track (start-finish, 4250 meters, about 170 seconds) there were set 16 possible event positions on each track. In consequence, a minimum distance gap of 250 meters resulted (about 10 seconds) so it was ensured that the events did not overlap. By using a random generator, the twelve events (4x brake light activation, 4x turn signal activation, 4x deceleration) as well as four “voids” (no event; to serve arbitrariness and unpredictability) were varied for each track version. In addition, an aberration of 20 meters (about 800 milliseconds) was added randomly to each event so that time-based implicit learning could be widely eliminated.
For every participant, the values were averaged in every trial to get one single score for each condition. Generally, there were three discrete types of dependent variables. Firstly, variables that describe the driver car position: lateral steering deviation, which represented the standard deviations of the ideal track (center line of the right lane); as well as longitudinal deviation, which is the standard deviation of the ideal (instructed) distance to the leading vehicle. Secondly, variables that describe the actual velocity. Therefore, average speed and maximum speed in each trial was observed. Thirdly, reactions to the three events. Miss rates as well as reaction times for correct reactions were calculated and analyzed.