Tactile Takeover Requests Based on Seat Belt Tension Patterns During Autonomous Driving

Background Level 3 autonomous vehicles must request drivers to take over control when a dangerous situation arises during autonomous driving. However, drivers are often engaged in non-driving tasks (e.g., using a smartphone or watching a movie) during these situations, making it less likely they will notice auditory or visual notifications. Therefore, it is crucial to design takeover request notifications that drivers can reliably perceive in any situation. This study investigates whether seat belt tension patterns could serve as effective takeover request notifications in autonomous driving situations. Methods A driving simulator capable of adjusting seat belt tension was developed for this study. Based on prior research, four distinct seat belt tension patterns with varying levels of urgency were designed. Afterwards, an experiment was conducted to determine whether takeover was effectively performed at different lead times (4, 6, and 8 seconds) when the seat belt tension patterns were used as takeover request notifications while drivers performed non-driving tasks during autonomous driving. Results The results showed that shorter lead times led to faster takeover responses, resulting in quicker reaction times and longer periods of manual driving. Additionally, when seat belt tension patterns (T) were used alongside visual and auditory notifications (VA), the takeover reaction time was significantly faster compared to using visual and auditory notifications alone. In particular, The fastest reaction time occurred when the high-urgency seat belt tension pattern (VA+T_HU1) was used. Manual driving times were shortest when the low-urgency seat belt tension pattern (VA+T_LU), and longest when the high-urgency pattern (VA+T_HU1), likely due to the faster driver response to the more urgent signal. There were no significant differences in lane departures, collisions, or non-driving task performance. However, subjective satisfaction with the seat belt tension patterns—regarding effectiveness, reliability, and overall satisfaction—was highly positive, averaging above 4.5 points. Conclusions This study demonstrates that seat belt tension patterns are effective as tactile takeover request notifications in Level 3 autonomous driving. Different tension patterns allow drivers to distinguish varying levels of urgency. These findings suggest that more diverse seat belt tension patterns can be designed and applied to takeover request notifications in Level 3 autonomous driving.