IUCT GROUP (InKemia)
10. Ingresos y Gastos
Experiment 2 used a mixed design with three within-participants independent variables and one between-participants independent variable, each with two levels. To test hypotheses H I, H2, H5 and H6, the acceleration available from the simulated vehicle and its responsiveness were varied. The other within-participants variable was included to test hypotheses H3(a), H3(b), H7(a) and H7(b), that driver behaviour and perception of available vehicle performance would be affected by
differences in driving behaviour that would result in different opportunities to perceive those performance differences). It involved manipulating the activation level o f the goal to complete the journey as quickly as possible.
The between-participants variable involved comparing two groups of drivers, general drivers and performance-oriented drivers. It was intended that the latter group would tend to have more active symbolic and affective goals, that would tend to lead to more dynamic driving styles, so this variable provided a test of hypotheses H4(a), H4(b), H8(a) and H8(b).
It is not possible to manipulate personality traits experimentally. However hypotheses H4(a) and H8(a) were investigated further by measuring personality traits via a self-report questionnaire, and exploring associations between them and driving behaviour13.
A full factorial design was not practical, since it would have required that all participants experience each combination of within-participants conditions, making 2 x 2 x 2 = eight drives in all. Allowing time for briefing, familiarization, and questionnaire completion, participation would have required around 6 hours per participant, which experience at TRL with other simulation studies suggested would lead to substantial fatigue effects, difficulty with recruitment, and high rates of failure to complete participation. Accordingly a less ambitious design was adopted, which only required participants to complete four experimental drives each. In this design participants from each driver group were randomly allocated to three blocks. In each block, two o f the within- participants variables were varied, and one was fixed. This design, shown in Table 5-1, enabled within-participants comparisons to be made with 16 participants per block (in fact slightly fewer participants actually completed all the experimental drives). All main effects, two-way interactions between any pair of the three within-participants variables, and three-way interactions between any pair of them and driver group, could be assessed. In contrast to a full factorial design, however, it
13 Such analyses cannot test causal relationships: hence I refer to “investigation” rather than “testing” of H3(a), H3(b), H6(a) and H6(b).
was not possible to assess three-way interactions between the within-participants variables, or four way interactions between all variables.
Experiment
Block Within-participant variables Fixed condition
Between- participant
variable
1
Goal condition (2 levels: relaxed vs. time pressured
X
Vehicle acceleration (2 levels: low vs. high) High responsiveness 2 Driver groups: performance- oriented (N=8) and general (N=8) 2
Goal condition (2 levels: relaxed vs. time pressured)
X
Responsiveness (2 levels: low vs. high) High vehicle acceleration 2 Driver groups: performance- oriented (N=8) and general (N=8) 3
Vehicle acceleration (2 levels: low vs. high)
x
Responsiveness (2 levels: low vs. high) Time pressured 2 Driver groups: performance- oriented (N=8) and general (N=8)
Table 5-1: Experimental design
To keep the sample size manageable it was not practical to use full randomization o f the within- participants conditions. Instead counterbalancing was achieved using a Latin Squares scheme, in which each condition occurred once in each ordinal position, and once before and once after each other condition.
Method
The TRL DigiCar driving sim ulator facility
Experiment 2 was carried out in the DigiCar driving simulator at TRL. Figure 5-1 shows the general configuration of the simulator, and Figure 5-2 shows an example of the forward view (taken from the rear seat).
Figure 5-1. General view of TRL DigiCar driving sim ulator
The simulator vehicle was a Flonda Civic. Three projection screens provided a 210° forward field of view and another provided a 60° rear field of view. The resolution of the visual scene was 1280 x 1024 pixels per channel (one channel for each of the four screens), lower than a “natural” scene in real driving, but providing sufficient visual information to generate realistic driving behaviour from participants. The visuals were refreshed at 60Flz so that the driver perceived a seemingly continuous driving experience. Engine, road and traffic sounds were generated by a stereo sound system with speakers inside and outside the vehicle. The simulator used a motion system with three degrees of freedom (pitch; roll and heave) delivered by rams attached to axles under each wheel. These imparted a limited range of motion in the three axes, providing the driver with an impression of the acceleration forces and vibrations that would be experienced when driving a real vehicle. The motion system did not simulate motion in yaw, surge and sway, so the motion experience of
when cornering was somewhat limited. Compared with simulators with more elaborate motion systems, however, it had the advantage that participants walked up to and entered the car as they would a real vehicle, encouraging expectations of a close-to-normal driving experience.
Figure 5-2. Forw ard view from rear seat of DigiCar sim ulator vehicle (taken during development of the simulated route: road markings, etc. not fully implemented)
All control interfaces had a realistic feel and the manual gearbox could be used in the normal manner, enabling participants to drive the simulated vehicle through a simulated environment, which contained the road, roadside features, and other traffic. The vehicle dynamics were based on a Renault model and the simulation was implemented using OKTAL SCANeR II. The software interpreted the driver’s control inputs, related them to the current vehicle status and computed a prediction of how a real vehicle would behave in the given circumstances. The system then responded to present to the driver its optimal representation of how this behaviour would be perceived through the visual, sound, and motion sub-systems. Data was then recorded relating to all control inputs made by the driver, including steering, pedals, gear, indicators; vehicle parameters such as speed and engine speed; and parameters to assess behaviour in relation to other vehicles, such as distance and time headways. As there was no particular requirement for high temporal resolution, data was recorded at 20 Hz.
Other traffic was included in the simulation in the form of “autonomous traffic vehicles” selected from a pre-existing library of different vehicle types including cars, trucks, buses, emergency vehicles, bicycles, and pedestrians. Each autonomous vehicle obeyed specific driving rules, so as to behave in a normal manner with respect to other traffic vehicles. The autonomous vehicles also had dynamic properties of their own - they appeared to pitch realistically under acceleration and braking, and vehicle graphics included body tilt and roll under braking, acceleration and turning; speed dependent rotating wheels and fully working brake, indicator, fog, and head lights. These provided additional cues to the driver and enhanced the realism of a scene.
An in-car colour LCD display was used to provide task-related information, as described later. An adjacent interview room was used for questionnaire completion and debriefing, and a waiting room was available to participants between experimental drives.
Simulated route
The simulated route consisted o f 10.2 km o f generic rural single carriageway A-road, with road side features such as trees, bushes, fences, walls and buildings. It was designed to create repeated natural opportunities for accelerations. The route included appropriate signs positioned in accordance with the UK Department for Transport Traffic Signs Manual Chapter 4 (Warning Signs) 2004. Chevron signs indicated sharp turns and were presented at each o f the critical bends.
The analysed section of the route was preceded by a familiarisation section and followed by a 1 km run-out section. Table 5-2 summarises the route design.
Lead-in section
The lead-in section consisted o f 2.8km of unchallenging rural road with gentle bends and slight up and down gradients. It enabled participants to become familiar with the handling characteristics of the vehicle before tackling the analysed section of the route.
Analysed section
The analysed section was a winding rural road 10.2km in length. It contained 18 bends, six of which were sharp bends (Radius 45m, turning through 100-105°) and 12 were more gentle bends (Radius 90m, turning through 70°). All were flat (no gradient) and there were equal numbers of bends to the left and to the right. The route had several 500m sections that were straight with no gradient, giving adequate sight lines to facilitate the overtaking of slower moving vehicles, if necessary. The route also contained three 700m, straight, 1:8 (12.5%; 7.125°) uphill gradients that also represented good overtaking zones if required. The bends and other features were separated by short straight sections of between 100 and 300m. These sections had minor gradients and changes in direction.
Run-out section
The run-out section was 1.0km long, giving participants time in which to bring the vehicle to a halt, following verbal instruction over an intercom that they had completed the route.
Section Description Length Notes
1 Familiarisation 2.8km Undemanding road to allow familiarisation with
vehicle and environment
2 Analysed section 10.2km 6 sharp bends; 12 gentle bends; 3 steep hills; 3 long
straights
3 Run out
Total
1.0km 14.0km
Short straight section where the simulated drive is brought to an end
Within-participants independent variables
Vehicle performance: vehicle acceleration
Two vehicle acceleration conditions were used. In the first, low acceleration, the basic dynamics model was used without modification. In the second, high acceleration, a modified version of the dynamics model was used that gave higher acceleration across all speed ranges (Table 5-3).
Speed range (mph)
Difference in time taken to accelerate across speed
range (%) between low and high acceleration
conditions
16-30 18
30-50 9
50-70 8
Table 5-3. Percentage acceleration difference between conditions, in three speed ranges
These differences were chosen to be somewhat smaller than the typical differences between manufacturers’ claimed 0-60mph acceleration times for medium sized family hatchback cars with 1.6 litre or 1.8 litre engines. They were somewhat larger than those typically obtained with cars driven on premium versus standard grade market fuels (in the range 1-4%), and similar in magnitude to the differences in acceleration from diesel-engine vehicles with clean diesel injectors compared with those from vehicles with injectors heavily fouled with deposits.
Vehicle performance: Responsiveness
Responsiveness was varied by introducing a 1.5s delay between the accelerator pedal action and the vehicle response in a low responsiveness condition, compared with a high responsiveness condition with no delay. The pedal delay simulated poor responsiveness as a consequence o f fuel hold-up in the inlet port of a port-injected spark-ignition engine during acceleration transients (especially when the engine is cold), or “turbo lag” in a turbocharged diesel engine, but was larger than the delays typically experienced in real vehicles. The large difference was intended to be
perceivable to a majority of participants, and would therefore enable the effects of interactions with the goal condition and driver group variables to be studied.
Driving goal condition
There were two driving goal conditions:
Time pressured: Simulated drive in which participants were placed under explicit time pressure. They were briefed to imagine that they were late for an important meeting and should complete the route as fast as they could, while driving as they normally would in such a situation. A countdown timer and remaining distance indicator were visible on a display attached to the dashboard (Figure 5-3). Participants also encountered three overtaking challenges, intended to provide additional need to accelerate, and reinforce the overall sense of time pressure.
Relaxed: Simulated drive with light oncoming traffic, and a lead vehicle that adjusted its speed to remain in view but was not approached (so no overtaking was required). Participants were briefed to drive as they normally would.