Friday, March 27, 2009

Evaluation of a sleepiness warning system – a test track study

January 2009

Sleep related crashes have received increasing attention over the past decade. Driver support systems that are able to detect sleepiness and warn the driver could be a potential countermeasure to reduce sleepiness related crashes. The aim of this study was to evaluate a warning system in an experimental investigation performed in a real time car driving application. An additional aim was also to examine the suitability of using an experimental vehicle at test track for evaluation of a sleepiness warning system. The modalities used for warning was a combination of sound, vibrations in belt and spoken messages. In addition a hand worn confirmation button was used.

In total 40 participants drove at a closed test track during night (00h–05h). They were instructed to sleep between 01.00 and 07.00 for two nights prior to the night of the experiment. The design of the study was a between group, with 10 participants in each group. The four different groups experienced one of the following: A: No warning – baseline; B: SENSATION warning system with an early trigger; C: SENSATION warning system with a late trigger; D: Feedback (DMS – Siemens) and warning with a late trigger. The trigger of the warnings was based on observer registrations of the driver and driver behaviour.

The drivers were instructed to drive a distance of 110 kilometres. The car was a Volvo 850, equipped with sensors in order to measure driving behaviour (lateral position, speed, steering wheel angle). As sleepiness indicators blink duration was used (mean and sd), measured with EOG. After the drive the participants answered a questionnaire. The experiment focused on comprehension, usability, effectiveness, and acceptance.

There was no significant observable difference between the groups that received the warning at an early point, at a late point, or at a late point in combination with a feedback system. Regarding the self reported measures, it should be noticed that no participant considered the warning to be too early. The sound was seen as most disturbing and most effective. The belt vibration was seen as least disturbing, but also as least effective. These results seem to be highly related to the level of the sound and the amplitude of the vibrations. In total, 27 out of 30 drivers in group B-D reported they had experienced warnings. About 50 per cent felt that the warning had come at the correct time, and no driver found it to be too early. More than 70 per cent stated that the warning had influenced their driving and 85 per cent thought it had made them more awake; 55 per cent stated that the warning had helped them to stay awake. The warning was easy to understand, and the drivers felt that it was clear what was expected from them. No clear feedback was given when pressing the confirmation button, and so some users pressed the button several times. Confirmation feedback to the user is necessary. Almost all drivers thought that the warning system would contribute to increased traffic safety, and that it would contribute to the prevention of crashes caused by sleepiness.

Even if driving for real on a test track gives higher ecological validity than a driving simulator there is still problems to be solved. Drivers, prepared to be sleepy during night time will most truly accept and agree with a given warning. This will influence the drivers’ acceptance of the system. Additionally, sensors used do not have a high technical reliability for speeds under 50 km /h. However, the present study shows that a test during real driving can be done without any risky situations. The recommendation is to take the next step and move out on real roads with experimental cars equipped with double commands in order to have a high external validity, but still a high control, at least regarding the participants’ preparation and actions under the experiment.

Acceder al documento

No comments: