Yang Zhen, Gong Zhe, Qin Yimei, Zheng Ruiping
The Key Laboratory of Road and Traffic Engineering, Ministry of Education, Tongji University, Shanghai, China.
Traffic Inj Prev. 2025;26(3):291-299. doi: 10.1080/15389588.2024.2405647. Epub 2024 Oct 17.
This study aims to develop a model for quantifying perceived risk in obstacle avoidance, emphasizing how drivers' perceived risk characteristics influence their driving decisions. The research addresses the lack of attention given to modeling risk from the perspective of drivers' risk perceptions.
Monte Carlo methods are employed to account for the uncertainties and complexities of driving behavior, restoring the probabilistic nature of risk. The proposed method quantifies perceived risk by incorporating drivers' fuzzy perceptions, enabling a quantitative evaluation during obstacle avoidance. A logit model is used to link perceived risk with driving decisions, identifying key factors influencing driver behavior in obstacle avoidance scenarios.
Experimental data revealed significant variations in vehicle trajectories and speed distributions due to differences in drivers' experience and proficiency. The perceived risk indicator (PRI) values for leftward bypasses were higher compared to rightward bypasses, and the receiver operating characteristic (ROC) curve confirmed the PRI's strong predictive ability with an area under the curve (AUC) of 0.820. The logit model showed that both PRI and speed significantly influenced the probability of choosing a rightward bypass, achieving 90% accuracy. Building on the model, the study predicted and visualized the probability of vehicles turning right to avoid obstacles at different positions and speeds within 200 m of the obstacle.
The research offers a framework for traffic professionals to understand driver-perceived risk and decision-making mechanisms. This understanding is beneficial for improving traffic safety and highlights the importance of considering drivers' risk perceptions in modeling driving behavior.
本研究旨在开发一种用于量化避障过程中感知风险的模型，重点关注驾驶员的感知风险特征如何影响其驾驶决策。该研究解决了从驾驶员风险感知角度对风险建模缺乏关注的问题。
采用蒙特卡罗方法来考虑驾驶行为的不确定性和复杂性，恢复风险的概率性质。所提出的方法通过纳入驾驶员的模糊感知来量化感知风险，从而在避障过程中实现定量评估。使用逻辑回归模型将感知风险与驾驶决策联系起来，确定在避障场景中影响驾驶员行为的关键因素。
实验数据显示，由于驾驶员经验和熟练程度的差异，车辆轨迹和速度分布存在显著变化。向左绕行的感知风险指标（PRI）值高于向右绕行，接收器操作特征（ROC）曲线证实了PRI具有较强的预测能力，曲线下面积（AUC）为0.820。逻辑回归模型表明，PRI和速度均显著影响选择向右绕行的概率，准确率达到90%。基于该模型，研究预测并可视化了车辆在距障碍物200米内不同位置和速度下右转避障的概率。
该研究为交通专业人员理解驾驶员感知风险和决策机制提供了一个框架。这种理解有助于提高交通安全，并突出了在驾驶行为建模中考虑驾驶员风险感知的重要性。
