Human-Like Automatic Braking Algorithm Based on Kinematic Cue: Yield Behavior Analysis and Fuzzy Controller Design

<div>Pedestrians are among the most vulnerable participants in traffic, particularly when crossing the road. Extensive research has been conducted globally on the yielding behavior analysis of vehicle–pedestrian interaction and the design of automatic vehicle braking systems to mitigate pedestrian casualties. However, few studies have comprehensively addressed lateral risks using implicit kinematic cues in pedestrian–vehicle interactions. Moreover, the design of collision avoidance systems has rarely taken into account driving behavior, along with the pedestrian’s kinematics and crossing behavior. This article presents a human-like automatic braking fuzzy control strategy for pedestrian–vehicle collision avoidance, combining the advantages of professional driver emergency braking behavior and kinematic interaction cues. First, a high-fidelity driving simulator is used to investigate the yielding behavior of pedestrian–vehicle interaction when pedestrians cross the road. Second, the intrusion position (<i>X<sub>P</sub></i>), as a new lateral risk index, is designed to overcome the limitation of lateral distance in complex pedestrian–vehicle interaction scenarios. Various metrics are considered to analyze driver emergency braking behavior using statistical methods from both lateral and longitudinal aspects. Subsequently, based on driver braking behavior, the human-like automatic braking fuzzy control strategy is proposed. Finally, simulation examples verify the reliability of the analysis results and the proposed controller’s effectiveness. Compared with a conventional automatic braking system, the timing of interventions of the proposed system is on average 2.9 s earlier, and the braking deceleration is reduced by 3.59 m/s<sup>2</sup>.</div>