Turning Left at Urban Intersections: Turning Patterns and Gap Acceptance

Kurzfassung

Turning left through oncoming traffic is one of the most safety-critical traffic manoeuvres. This work aims to describe and understand the interaction behaviour between left-turning and oncoming traffic by analysing video and trajectory data collected at the Application Platform for Intelligent Mobility (AIM) research intersection in Braunschweig, Germany. At this intersection, motorists turning left from Rebenring into Brucknerstraße must give way to oncoming traffic as well as crossing cyclists and pedestrians. In the following analyses, video and trajectory data of thirteen days (recorded in the period of 20th of May to 2nd of June 2019) were included. In a first step, potential interactions between left-turning and oncoming traffic were extracted based on post encroachment time (PET), which indicates by how many seconds two intersecting road users miss each other. In order to investigate both critical and less critical interactions between left-turning and oncoming traffic, left-turns with a PET ≤ 2 seconds between the left-turning vehicle and oncoming traffic were randomly selected for further analysis (80 left-turns per day). In two cases a left turn was detected by mistake, reducing the number of cases to n = 1038. The video data of these left-turns were then analysed to determine the turning pattern. Four turning patterns were identified: The left-turning motorist turned a) before all oncoming traffic, b) after all oncoming traffic, c) between oncoming traffic (i.e. both before and after oncoming traffic passed the intersection within a traffic light phase), and d) while the traffic light for oncoming traffic turned red (i.e. while oncoming traffic slowed down or stopped in front of the traffic light). Due to the selection by means of PET, left-turns without oncoming traffic were not included. The analysis revealed that left-turning motorists turned between oncoming traffic (pattern c) in about 50 % of cases, and were the first to turn left within one traffic light phase in 85 % of cases. In a second step, the gap acceptance was determined for a subset of cases (n = 191; all motorists who turned left first within one traffic light phase with pattern c that passed between 6 and 20 o’clock within seven days) by first calculating the gap size of all accepted and rejected gaps. The gap size was defined as the time between the point in time when the rear of the first vehicle crossed the future path of the left-turning vehicle and the point in time when the front of the second vehicle crossed this path. Including 191 accepted and 830 rejected gaps, gap acceptance was then estimated by means of logistic regression. The analysis showed that all gaps larger than 8.03 seconds were accepted and all gaps smaller than 2.74 seconds were rejected. Gap acceptance was 50 % for a gap size of 4.70 seconds. In a next step, cases with gap sizes around 4.70 seconds will be examined in more detail to determine factors influencing the decision to turn (besides gap size). Of special interest is the behaviour of oncoming traffic (e.g. speed and acceleration patterns of oncoming motorists). If, for example, specific driving patterns affect left-turn decisions, these patterns could be implemented in automated vehicles to facilitate safe turning behaviour. In general, these findings might support the selection of suitable infrastructure measures and the development of assistance systems and autonomous driving functions to promote safety.