Advancing Situation Awareness estimation: a driving simulator experiment at TNO

In our news article of October 2023 we explained the why behind the research into driver Situation Awareness (SA): to make automated driving safer by tracking the driver’s SA and either urge the driver to regain SA, or take system safety measures. The latter can range from taking a different approach to Transition of Control to ultimately performing a safety manoeuvre. Recently a major step into the research was taken: performing a driving simulator experiment.

The main research question driving the experiment was to evaluate if there is a relation between driver gaze behaviour and the SA of the driver. The gaze behaviour was captured using an eye tracking device. Registering the SA of a driver was, however, more difficult. After many deliberations we opted for an approach in which the driving simulator would be paused in a specific driving situation, after which the driver was queried about this situation. To be more specific, we used the Situation Awareness Global Assessment Technique (SAGAT) from Endsley[1]. It considers three levels of situation awareness: level 1: perception, level 2: comprehension and level 3: projection to the near future. For now we focussed mainly on level 1, since this level has the most direct connection to gaze behaviour. We designed a questionnaire for the participants that they answered when the simulator paused.

The scenario was a two-lane highway setting with slower traffic on the right lane (just below the maximum speed) and faster traffic on the left (overtaking) lane. The driver was asked to drive in the right lane as much as possible, but to overtake slower traffic (i.e., to follow the maximum speed as much as possible). To manipulate different SA of the driver, some of the drivers had to perform a distracting, secondary task on a tablet at the centre console. Other variables in the experiment were the speed limit, density of traffic in the left lane and the moment at which the experiment was paused. The pause in the experiment was timed in relation to the vehicle that had to be overtaken. This way we tried to capture different required situation awarenesses of the driver (tracking the vehicle ahead, seeing if the left lane has a space to merge into, performing the lane change, planning for merging back into the right lane).

A total of 47 participants took part in the study and 38 ‘good’ datasets were acquired. Analysis of the data is currently ongoing. Based on the findings the Situation Awareness model will be further developed and tuned.

[1] Endsley, Mica R. "Direct measurement of situation awareness: Validity and use of SAGAT." Situational awareness. Routledge, 2017. 129-156.