Determining driver situational awareness in the context of partially autonomous vehicles

Determining driver situational awareness is one of the most difficult tasks in driver performance profiling. Endley's model of determining situational awareness defines three levels of engagement with the environment in hierarchical order, such that (i) Perception, (ii) Comprehension and (iii) Projection. Perception, being the first level of driver situational awareness, is perhaps the only objectively quantifiable factor for driver inattention detection. In the context of this thesis, driver perception is defined as the ability to comprehend the operating environment and peripheral contextual information. One of the most reliable means of determining driver perception is through eye tracking and gaze concentration. An advanced multi-camera remote eye-tracking solution specifically designed for monitoring driver behaviour in the challenging environment of a vehicle cockpit is used to monitor driver gaze behaviour in a complete naturalistic driving environment. A set of carefully chosen in-vehicle gaze targets are used to determine driver gaze concentration on and off the road. The time drivers spend looking at these predefined gaze targets has been used to characterise driver visual behaviour at roundabouts, signalised and un-signalised intersections with adjacent cycle lanes. This study shows that on the approach to roundabouts, driver gaze direction and gaze concentration were often diverted away from the direction of travel of the vehicle and for some drivers it took a significant amount of time to return their gaze towards the direction of travel of the vehicle. This lack of attention to the front and left poses risks to other road users. Observations indicate that side-window visibility becomes more critical as drivers enter the roundabout. Additionally, analysis of driver gaze behaviour during left-turn maneuvers at intersections with cycle lanes reveals that many drivers fail to check their left-view mirror, relying instead on memory, which increases collision risks with cyclists. Eye-tracking data shows that while drivers glance at mirrors, these glances are often too brief to ensure proper hazard detection, especially for fast-moving cyclists or e-scooters travelling along side on cycle lane. Additionally, drivers tend to focus on the side window rather than mirrors when making turns, creating blind-spots. The study highlights the need for an advanced algorithm to generate alerts for drivers who fail to monitor key visual targets, enhancing safety at intersections. A gaze-analysis algorithm was developed to identify unsafe behaviour, offering insights for enhancing Advanced Driver-Assistance Systems (ADAS) to improve road safety. The algorithm, adaptable to left turns, right turns, and roundabouts, can intervene when necessary to prevent collisions. Future improvements will integrate advanced sensors for higher accuracy.

Funder

Irish Research Council
Science Foundation Ireland

Publisher

University of Galway

Rights

CC BY-NC-ND

cbn

Collections

University of Galway Theses (PhD Theses)