I have read before about car-to-car or vehicle-to-vehicle communication, and apparently the US Department of Transportation is way ahead. V2V “allow cars to share information about their position, speed and heading with each other and alert a driver when there is potential for danger. That could be a car speeding through an intersection ahead or a truck in your blind spot when trying to change lanes.”
These systems have been mentioned in here because this study tries to find the optimal time and place to show potential alerts. V2V may bring great benefits to road traffic, but before the technology perfects itself, we should have well established the ideal time to activate alert systems.
Wolffsohn, J.S., Edgar, G.K., & McBrien, N.A. (1998). The effect of viewing a car head up display on ocular accommodation and response times. In A.G. Gale, I.D. Brown, C.M. Haselgrave, & S.P. Taylor, Vision in Vehicles – VI (pp.143-151)
This paper was at a Vision in Vehicles VI Proceedings and it’s from 1998. It reminded me of this one in the sense that both present two overlayed stimuli to attend.
The authors manipulated the distance at which they collimated the HUD in order to verify differences in accommodation and in reaction time to detect changes in the environment. They were constantly measuring the accomodation of the eyes whereas the drivers were attenting the environment only or the HUD+ environment.
The question under this research is: Is there an optimal distance at which an HUD should be focused? A driver, unlike a pilot, focuses on stimuli at very different distances, varying from 1m to infinity.
Results showed that although there was a trend to higher accommodative levels when the HUD was focused closer to the subject, there was not a significant difference in accommodative level over the range of distances at which the virtual HUD image was focused.
The response times of subjects for detecting changes in the external environment (traffic lights) was consistently faster than for detecting changes in the HUD indicators under all conditions.
However, subject response times to detecting changes in the traffic lights and HUD indicators did not significantly vary with the distance at which the virtual HUD image was focused.
When performing the HUD assisted driving task, there was a significant increase in the percentage of traffic light changes missed and HUD indicator changes missed over all distances tested.
So these results are very important, as they suggest that, with a HUD, vital danger signs, such as warning boards, are more likely to be missed.
Porter, B.E. (Ed.) Handbook of Traffic Psychology. Elsevier Inc.
While reading the chapter on Ergonomics and Human Factors I found the statement I was looking for. It’s hard to find definite answers of course, but since the first time I heard about In-vehicle information systems (IVIS) I wondered if they really were effective.
Although the safety potential is huge, the ultimate effects are definitely smaller than expected. If these systems are included in the vehicle, they would work perfectly if everything remained the same. But when one thing inside the vehicle is changed, this will inherently change the behavior of the driver: everything is connected.
Some of the possible negative effects of the IVIS – the book mentions ITS, Intelligent Transport Systems – are:
1) Underload and diminished attention level;
2) Information overload (Google Glass and all AR things, please do watch out);
3) Incorrect interpretation of information;
4) Overreliance on the system;
5) Risk compensation.
This is definitely interesting, although quite intuitive for anyone with minimal knowledge on cognition.
Do engineers consider this when they conceive these systems? I know nowadays most of them have psychologists or ergonomists on their teams, but some of them still don’t.
And these systems worked rather well on aeronautics contexts, but it seems as if they were blindly applied to the automotive sector.
More on this on the future, definitely.
Werneke, J., & Vollrath, M. (2013). How to present collision warnings at intersections? – A comparison of different approaches. Accident Analysis and prevention 52, 91-99. http://dx.doi.org/10.1016/j.aap.2012.12.001
Keywords: Top-down warnings; Bottom-up warnings; Car-to-car communication; Peripheral warnings; Driving behavior; Visual scanning strategies; T-intersection
Specific subject: This paper aims to study the best approach to warnings while driving at T-intersections
I really enjoyed reading this article both for the subject and the methodology used. Werneke is on my check-out list.
It analyses the dynamic and complex situation of driving at intersections.
If we carefully think about all those little safety measures we need to attend to in a short ammount of time, we would list: looking for crossing vehicles, vehicles driving ahead, vehicles arriving, traffic signs, pedestrians, cyclists, etc. This implies a large ammount of cognitive resources in order to process visually and spacially all these stimuli.
Several studies analysing the origin of accidents at intersections revealed that, in most cases, these are due to recognition errors:
1) “looked-but-failed-to-see” accidents, where drivers often fail to yield the right of way, using improper scanning strategies;
2) “failed-to-look” accidents, where drivers allocate their attention to areas where they expect to see relevant objects.
So, according to the authors, we could invest in developping proper warning systems in order to support drivers:
One strategy could be to inform drivers, at an early stage, of a potentially dangerous event so they are ready to react when the situation occurs; Other strategy could be to warn the driver immediately before the event so they can react faster when seing the object.
Technically, the first strategy implies a predictive technology, still under development. So our current efforts should go to the second strategy.
Great, so we know when, now where?
Where to present these warnings? Should we present them in the driver’s periphery, leading to an attentional change in focus? Or should we present them in the center of the driver’s view?
The researchers investigated three warning signals, all presented on a simulated head-up display (HUD).
1) A warning signal was presented earlier, while approaching the critical intersection (early-middle)
2) A warning sign – a flashing orange circle -symbolized the road side at which the critical incident was presented (late-middle)
3) A warning sign that didn’t show the road, itwas presented in augmented reality at the spot where the incided occurred.
These last two signs were supposed to capture driver’s attention as they appeared quite suddenly (approximately 18.5 m away from the entering car when drivers passed the limit line). However, this warning strategy is only useful when the warnings are given near the critical event at a point in time when the critical incident can be perceived by drivers.
So 4 groups were used: control, early middle warning signal, group with late-middle and late-sideways.
A T-intersection situation with a yield sign was chosen where drivers had to turn right and had to yield to oncoming cars from the left-hand side of the intersection and driving straight (Fig. 2). A vehicle parked at the right-hand side of the road suddenly left the parking space and entered the road whilst drivers were actively executing the right turn. The driving scenario had a length of approximately 50 km and took 45 till 50 min.
The dependent variables were driving and gaze data in the two critical incident situations.
The analysis of the collision frequency in the first critical event showed no significant differences between the four groups. Looking at the pairwise comparisons, there were significantly fewer collisions in the group with the early middle warning signal, as compared to the group with the late-middle warning (p = .025). The difference of the group with the early-middle warning signal to the control group was not significant (p = .132).
Fewest collisions occurred in the group with the early-middle warning signal.
In the group in which the drivers were supported by the late-middle warning signal the collision frequency was quite high. Only half of the drivers (n = 6 out of 12) could avoid a collision.
Drivers with the early-middle warning signal stopped 6.5 m (SD = 2.8 m) before the entering vehicle. Drivers with the later warnings showed similar pattern of the minimal distance to the entering vehicle. Both groups had a mean minimal distance of approximately 3.0 m. However, differences in the standard deviation of the minimal distances were found.
The TTC (Time do Collision) of the group with the early-middle warning signal was by trend larger compared to the other three groups.
Drivers warned by the early-middle warning signal had a mean TTC of over 4 s (SD = 1.5 s). In the other three groups, the TTC was between 3.2 s and 3.5 s. On average, drivers reacted nearly 0.5 s slower to the entering vehicle (M = 2.4 s, SD = 0.4 s) compared to the other three groups.
Drivers waited longer at the intersection when they were already warned about an unexpected hazard (group with the early-middle warning signal)
This early warning was very effective to adapt drivers’ behavior toward safer driving. They waited longer and accelerated slower when turning right which enabled them to stop with a much larger minimum of distance toward the entering vehicle.
The two later warnings (late-middle and late sidewise warning signal) were not effective in improving drivers’ reaction in the critical event. The number of collisions did not differ from the control group.
It may also be that the late-sidewise warning did not provide any additional information to the entering vehicle. Similarly, in a study of Muhrer et al. (2011) a visual warning signal did not prove effective when a car in front was braking suddenly. In this unexpected situation, the brake lights of the front car provided sufficient warning for drivers so that an additional in-vehicle warning system was not beneficial for them.
This kind of early information may be provided by car-to-car technology (e.g. broadcasting the information that a car will enter the road in the near future) or car-to-infrastructure technology (e.g. detecting by a video monitoring of the intersection that a car is likely to enter the road and broadcasting this information).
None. Except it made me think on the utility of in-vehicle information systems on critical situations…At least they don’t harm.