Speed behaviour of right‐turn drivers at signalized intersections in China

Ensuring human-like driving performance is fundamental to the widespread adoption and acceptance of intelligent driving systems. The speed behaviour of right-turn drivers at signalized intersections was studied to improve the human-like driving capability. The study was conducted using natural driving data and 545 data for drivers turning right at signalized intersections were extracted. YOLOv4 was used to identify the different types of road users. A one-way ANOVA was used to study the influence of the scene complexity, number of lanes, and intersection shape on the speed selection behaviour. Linear regression and binary logistic regression analyses were used to study the influence of road users, number of lanes, intersection shape, and vehicle motion state on the speed control behaviour. The main results indicate the following: (1) Drivers drove slower when there were more traffic participants or fewer lanes. (2) The intersection shape did not have any effect on the speed behaviour of right-turn drivers, either speed selection behaviour or speed control behaviour. (3) The reaction distance was affected only by the approach speed in this study; the greater the speed, the longer the reaction distance. (4) In addition to the vehicle motion state, external environmental factors also play a critical role in the braking characteristics, that is, pedestrians, cyclists, lateral traffic, and number of lanes after the turn. (5) The driver's intention to brake after entering the intersection stemmed from road users who were in collision conflict with the host vehicle and fewer lanes after the turn, rather than the speed at the stop line. This research has a guiding significance for the design of human-like driving of right-turn driving assistance systems. The target classification method and the measurement method of the driving environment complexity also help improve the human-like perception of the right-turn driving assistance system.