Field assessment of driver decision making at intersections: A real-time wireless application to manipulate encroachment time

Abstract

Accidents at intersections are of the most common causes of fatalities on roads. Statistics from the EU and the USA show that fatalities at intersections represent more than 20% of all traffic fatalities. A left turn across the path of a vehicle coming from the opposite direction (LTAP/OD) is one of the riskiest scenarios at intersections. In this scenario, we can refer at the vehicle turning left as the Subject vehicle and to the one going straight as Confederate vehicle. In order to understand driver behavior in this scenario during field trials, the interaction between the two vehicles should be manipulable by the experimenters. In this study, a system able to manipulate LTAP/OD scenarios by controlling the difference in time-tointersection between the Subject and the Confederate vehicles has been developed. By means of a human machine interface installed in the Confederate vehicle, an experimenter driver was instructed about the velocity he/she should follow to control the difference in time-to-intersection with respect to the Subject vehicle. This velocity instruction was determined from 1) the position of the vehicles, 2) their kinematics and 3) historical intersection data. The system comprised of 1) a single board computer with one GPS device and one 3G modem for each vehicle, 2) a cloud application, 3) a computational server and 4) a computer to render the human machine interface. The algorithm providing velocity instruction addressed six experimental phases { waiting, be ready, start-up, regulation, stabilization and releasing. This algorithm relied on predictions: profiles extracted from historical intersection data were used to estimate the current time-to-intersection, which was the base for the whole evaluation. The novelty of this algorithm consisted in estimating in real-time the optimal velocity that Confederate vehicle needed to follow to control for a certain difference in time-to-intersection with the Subject vehicle. The velocity estimation was based on wireless communication of GPS position and velocity. The algorithm was validated in the real world by showing that the actual difference in time-to-intersection achieved by following the velocity instruction from the algorithm was very close to the one aimed for. Furthermore, the system used in this study is affordable and accessible to anyone; therefore this system can be easily reproduced and employed to understand driver behavior at intersection as well as for developing active safety system to support other road users at intersections.