Carta de presentación
10. Preguntas frecuentes
Method
Driving simulator and experimental route
Experiment 3 was carried out using the TRL DigiCar driving simulator, described in Chapter Five. The experimental route was a long and relatively straight section of simulated UK rural dual carriageway, which could be negotiated safely at speeds up to 70mph without deceleration at any point. A simulated overtaking target vehicle was present in the simulated environment, in a position ahead of the simulator vehicle. The speed o f the target vehicle was constant, at the same speed as the initial speed of the simulator vehicle at the start of the acceleration manoeuvre. In the majority of drives the participant caught up with and passed the target vehicle by the time the required target speed was reached. Figure 6-3 shows the forward view of the route, at the start o f the acceleration manoeuvre, with the target vehicle visible ahead in the distance.
The simulator vehicle began each drive in the right hand lane, so that the entire drive and acceleration manoeuvre could be completed with minimal steering. This prevented any disorientation or distraction of the participant that might result from the absence of lateral force simulation in the DigiCar simulator when carrying out lateral displacement manoeuvres such as lane changing.
Figure 6-3. Experimental route as seen at start of acceleration manoeuvre. Overtaking target vehicle visible in left lane ahead
Participants
50 participants completed Experiment 3: 25 were male and 25 female. All had a full UK driving licence, and a large majority (47) had more than 5 years driving experience. The mean age of the sample was 38.6 ± 9.6 years. The mean number of years of driving experience was 19.7 ± 9.3. All participants drove more than 8,000 miles per year. The mean annual mileage of the sample was 14,061 ± 12,261 miles. Potential participants who drove diesel vehicles were excluded, since the simulated vehicle had the characteristics of a car with a spark-ignition (gasoline fuelled) engine. Those who worked in advertising, marketing, market research, car or motoring industries, petroleum industry, public relations, journalism, TV or media, were also excluded, again for reasons of sponsor confidentiality.
Independent variable: sensory modality
The independent variable was the sensoiy modalities available to convey acceleration cues to the participant. Three modalities were represented in the DigiCar simulator: Visual, Sound and Motion. The experimental route provided two types o f visual acceleration cue. The first type o f cue was optic flow of roadside features such as trees and the roadside crash barrier, and the central road markings, which moved outwards from the centre o f the field o f view with angular velocities which gave cues to speed. The second type of cue was the target vehicle, which grew in angular size as it was approached. The visual environment also contained a horizon that offered a visual tilt cue. The sound system conveyed acceleration information in two ways: through the frequency of the engine sound, and through loudness. Increasing speed is associated with higher engine frequency and loudness. The motion system was able to convey acceleration information using a tilt cue, corresponding to the tilt backwards of a vehicle at the beginning o f an acceleration (and the equivalent tilt forwards on braking). There was no system of tilt co-ordination so it was not possible to provide a vestibular acceleration cue.
There were four different sensory modality conditions:
• Visual only
• Visual and Motion
• Visual and Sound
• Visual, Motion and Sound
Independent variable: Maximum Available Vehicle Acceleration
Maximum Available Vehicle Acceleration (VAmax) was modified by specifying different levels o f a variable, engine yield, in the simulator’s vehicle dynamics model. Adjusting the engine yield variable simulated different levels of engine output and caused the simulated vehicle to move through the simulated environment with higher apparent acceleration.
Two different VAmax conditions were used in Experiment 3. The first (VAO) was a baseline condition in which engine yield was maintained at the default setting o f the system, which
simulated the dynamics of a 1.61 Renault Megane. For the second, the value of engine yield was adjusted to deliver a maximum acceleration over the required speed range of 9% higher than baseline (VA9).
Experimental design
In Experiment 3 all drives involved mid-range acceleration, from 30mph to 50 mph. Each participant completed 16 drive pairs, 4 in each sensory modality condition. Within the group of 4, 2 drives had the order VAO first, then VA9 second, the other two had the opposite order. Constraints on the time taken to change between sensory modality conditions meant that all 4 drive pairs within the same sensory modality condition were carried out in sequence for each participant. However, the order of sensory modality conditions, and the sequence of drive pairs with a sensory modality condition, was counterbalanced across the participant pool.
Dependent variable: frequency o f correct pair-wise judgments
Participants completed pairs of drives in succession...After each pair, the participant made a pair wise judgment as to the drive in which the vehicle accelerated fastest. Individual judgments were categorised as correct or incorrect. The percentage o f correct responses for each VAmax pair was calculated by totalling responses between participants.
Simulator outputs
Three simulator outputs, logged at 20 Hz, were recorded:
• Gear: the gear selected
• Accelerator Pedal Depression: the position of the accelerator pedal, from 0.0 (no depression) to 1.0 (fully depressed)
• Speed: the longitudinal ground speed of the simulated vehicle relative to the simulated environment (in mph)
Gear and Accelerator Pedal Depression were logged to act as a validity filter, enabling checking of the data and rejection of individual drives if the experimental conditions had not been met (for
instance, if the participant delayed depressing the accelerator pedal significantly after receiving the “go” signal). Speed was used to calculate the actual acceleration rate produced in each drive.
Procedure
On arrival at TRL participants completed a consent form and short demographics questionnaire. Participants then completed one familiarisation drive, using the same familiarization route as Experiment 2, to enable them to quickly become comfortable with the simulated environment and car controls. Immediately afterwards, they were verbally briefed on how to carry out the experiment, and experienced six practice pairs of drives.
Achieve required
i n i t i a l s n e e d
Accelerate to target seconds
Engage correct gear
Figure 6-4. Procedure for individual drives
The procedure for individual drives is shown in Figure 6-4. Participants started each individual drive with the simulator vehicle stationary in the right hand lane. They accelerated to the required initial speed and engaged the required gear. The visual display indicated the required gear and speed, the current gear, and whether to speed up or slow down to attain the required initial speed (Figure 6-5).
Once the required initial speed had been reached the visual display indicated the message “Prepare to accelerate to (target final speed)”. Once the required initial speed had been maintained continuously for 2 seconds (within ± 5 kph (3.13 mph) the visual display indicated the instruction “GO”. At this point participants depressed the accelerator pedal fully. This process ensured that experimental drives started in the correct gear and from a steady-state initial speed. Once the target final speed had been reached, the visual display changed to indicate the message “Target Speed Reached”. At this point the participant slowed the simulated vehicle to a halt.
After each pair of drives, participants were asked over the intercom “on which drive did the vehicle accelerate the fastest, the first drive or the second?”
Time S tart of drive
Participant is yet to begin, therefore must ‘speed lip’ to reach the required initial speed as well as engage the correct gear.
Participant exceeds required speed and is in too high a gear
A c celerate to t a r g e t initial sp e e d a n d s e le c t t a r g e t g e a r Gear
1
I Speed up ■ Target gear: 3 Initial speed: 30 A c celerate to ta r g e t initial sp e e d a n d s e le c t t a r g e t g e a r G e a r 4 H Slow d o w n ■ Target gear: 3 1 Initial speed: 30 A c celerate to t a r g e t initial sp e e d a n d s e le c t ta r g e t g e a r Gear Target gear: 3 Initial speed: 30 Prepare to accelerate to 5 0 m p nFigure 6-5. Example of Visual Display sequence for drive in Experim ent 3 (3I(I gear, 30mph to 50 mph acceleration)
Results
Figure 6-6 shows the percentage of correct and incorrect responses for each of the four different sensory modality conditions. The frequency of correct responses was larger in the Visual only condition than any of the other three. This result contradicts the naive hypothesis that acceleration difference would be less readily perceived when less sensory modalities were available. The
differences between correct and incorrect responses were statistically significant for all four conditions (X2 test, p < 0.05 criterion; Table 6-1).
80
30
Visual only Visual & Motion Visual & Auditory Visual, Auditory & Motion Incorrect Correct
Sensory modality condition
Figure 6-6. Perception of acceleration difference using different combinations of sensory cues: C orrect and Incorrect response rates
Sensory modalities C orrect responses Incorrect responses X2(l) P DL75 relative to standard stimulus (%) (%) VA0 (% ) Visual only 67 33 11.56 0.001 13.9 Visual & Motion 62.5 37.5 6.19 0.013 19.1 Visual & Auditory 61 39 4.84 0.028 20.1 Visual & Auditory & 62 38 5.76 0.016 19.4 Motion
Table 6-1. Experim ent 3: Perception of acceleration difference using different com binations of sensory cues: X2 tests and difference thresholds
Table 6-1 also shows the difference thresholds (DL75) for each cue combination, calculated by assuming that the data lie on a cumulative normal distribution. The DL75 value for the Visual Only condition was slightly lower than that for the other three.
Discussion
The results of Experiment 3 do not support the naive hypothesis, that difference thresholds decrease as more sensory cues become available: the opposite was found to be the case. The results can, however, be explained using the Bayesian model of cue combination. The results for the presence/absence of motion cueing can be understood if we assume that the motion cue was assigned zero weight when it was absent, and some finite weight when present. The motion cue signal represented a cue for lower acceleration than the visual cue, because o f the limitations of the motion system described in the method section. The Bayesian cue combination model predicts that an inconsistent, lower acceleration estimate from the vestibular system would reduce the overall, multi-sensory perception of acceleration, as observed. A similar argument can be advanced for the auditory cues. While the auditory frequency cue (representing engine speed) was realistically simulated, the results suggest that the auditory loudness was an inconsistent cue, signalling lower acceleration than the visual cue. This is entirely plausible, and would result if the auditory signals were quieter in the simulator than those participants had learned to associate with acceleration in their usual vehicles.
The implication was that the auditory and motion cues for acceleration in the driving simulator were somewhat inconsistent with the visual cues. This suggested that the simulator did not have absolute validity for perception of acceleration and hence difference thresholds. Accordingly, in Exper 5, the absolute difference threshold for mid-range (30-50mph) acceleration was measured using a real vehicle on a test track. However it was not feasible to measure difference thresholds for 50mph-70mph accelerations in this way, because of safety considerations and a 60mph speed limit for drivers without specialist training on the TRL test track. Neither was it feasible to perform an accurate test track experiment for the 0-20mph speed range, because of the large contribution to variance that results from drivers having to control the vehicle clutch as well as the accelerator
relative validity for acceleration perception in different speed ranges (the inconsistencies being similar across these conditions), in Experiment 4 I measured the relative difference thresholds for accelerations in all three different speed ranges of interest. I could then use the ratios between them, and the absolute value for 30-50mph accelerations from Experiment 5, to estimate the absolute difference thresholds for 0-20mph and 50-70mph accelerations.