Use of Field Observations in Developing Collision-Avoidance System for Arterial Red Light Running

Abstract

Red light running (RLR) behaviors at arterial intersections are different from RLR behaviors at high-speed intersections. The majority of RLRs at arterial intersections are not caused by drivers trapped in the dilemma zone; at high-speed intersections, drivers' becoming caught in the dilemma zone is an important contributory factor to RLR. Consequently, collision-avoidance systems (CAS) based on detection of actual RLR trajectories after yellow light onset have been conceived. To form a set of parameters and to detect or predict RLR and optimize the performance of CAS, data collection and analysis are needed. Research was done at a well-designed arterial intersection in California and data were collected with multiple discrete point sensors. The work focused on empirical observation of vehicle trajectories to arrive at statistical descriptions that could be used in RLR CAS. Empirical data showed that for more than 90% of RLRs, drivers were not trapped in the dilemma zone at yellow light onset. Instead, they could have stopped safely. Further headway analysis of the empirical data showed that most RLR vehicles belonged to a platoon, and that the average headways of RLR vehicles were 5% to 10% less than the headways of vehicles traveling in the yellow light phase. This suggests that those RLR drivers either were trying to beat the yellow light or were following a leading vehicle, and so drivers were inattentive to the traffic signal. An intervehicle analysis found that following a leading vehicle contributes significantly to stopping probabilities. It is concluded that parameters such as approaching speed, acceleration or deceleration rate, and intervehicle characteristics should be included in an effective RLR CAS.