Dynamic Cumulative Human-Like Brake Control Modeling for Autonomous Vehicle Collision Analysis

Rear-end collision where a human-driven vehicle (HV) hits an autonomous vehicle (AV) from behind (called HV-AV collision) represents a critical scenario in the context of mixed traffic including both AVs and HVs. In the present paper, the Human-like Brake Initiation and Modulation (HLBIM) model is first proposed to emulate the driver brake control, and further employed to investigate HV-AV collision risk in the Stop-in-Lane (SiL) scenario where an HV follows an AV. The HLBIM model combines human driving intention, vision-based expectancy and certain inherent characteristics of human driving based on a novel dynamic accumulation of braking-needed evidence. Especially, the HLBIM model incorporates a high-level expectancy derived from practical driving expertise to ensure precise timing for braking initiation. Furthermore, the combined distribution of Off-Road-Glance and Time-Headway is originally proposed and integrated into the HLBIM model, to emulate drivers' diverted attention during their driving activities. Our approach addresses a main issue, which is often neglected in the existing studies on HV-AV collision, which corresponds to the coupling effect of human driving intention, vision-based expectancy and inherent characteristic of human driving on HV-AV collision risk. Moreover, we analyze HV-AV collision risk in two specified SiL cases, which has rarely been investigated in existing related studies. The model performance comparison shows that our HLBIM model outperforms the referred models. Furthermore, the case study comprehensively reveals how the HV-AV collision probability changes with respect to various collision speeds, initial speeds, decelerations of cars under test, percentages of overlap detection of the AV in front, and lane change durations. The HLBIM model can be used to simulate driver-in-the-loop car-following experiment. Moreover, it can be fitted in automatic braking systems within AVs to achieve human-like braking.