A Preliminary Evaluation of a Driver-Infrastructure Interface for a Left-Turn Warning

Abstract

To examine the viability of an infrastructure-based, left-turn warning, twenty test participants drove through a test-track intersection with oncoming traffic. Participants decided, both with and without the aide of an infrastructure-based real-time warning, whether or not there was enough time to complete the turn in front of the oncoming traffic. In order to allow for natural driving behaviors such as turning without stopping, a new metric, the predicted trailing buffer (spare time), was introduced to describe the synthesis of vehicle movements, and as a basis for a warning algorithm. This study also examined the effect of or the timeliness of three different warning onset points.

Gap (lag) acceptance increased as the predicted trailing buffer increased with almost all gaps being accepted with greater than one second of predicted spare time. Overall, the activation of an infrastructure-based warning reduced the percentage of turns made by about twenty percent. As the predicted spare time decreased, the activation of a warning was even more effective in persuading the drivers to both stop and to agree with the warning. Drivers were largely insensitive to the onset of the warning; however, a warning onset of 32 m or 4 s before reaching the stop bar was mostly considered acceptable.