[论文解读] Analyzing Highly Volatile Driving Trips Taken by Alternative Fuel Vehicles

Mohsen Kamrani, Ramin Arvin+1 more|arXiv (Cornell University)|Jul 10, 2018

Vehicle emissions and performance参考文献 21被引用 9

一句话总结

本研究利用2012–2013年加州家庭出行调查（N=62,839次出行）的速度数据，分析了替代燃料汽车的驾驶波动性。采用OLS和分位数回归识别高波动性出行（第90百分位数）的相关因素，发现轿车驾驶员、年轻驾驶员以及短途出行表现出更高的波动性，而皮卡、掀背车、敞篷车和多功能休旅车则波动性较低，为通过行为干预和设计改进提升交通安全性提供了可操作的见解。

ABSTRACT

Volatile driving, characterized by fluctuations in speed and accelerations and aggressive lane changing/merging, is known to contribute to transportation crashes. To fully understand driving volatility with the intention of reducing it, the objective of this study is to identify its key correlates, while focusing on highly volatile trips. First, a measure of driving volatility based on vehicle speed is applied to trip data collected in the California Household Travel Survey during 2012-2013. Specifically, the trips containing driving cycles (N=62839 trips) were analyzed to obtain driving volatility. Second, correlations of volatility with the trip, vehicle, and person level variables were quantified using Ordinary Least Squares and quantile regression models. The results of the 90th percentile regression (which distinguishes the 10% highly volatile trips from the rest) show that trips taken by pickup trucks, hatchbacks, convertibles, and minivans are less volatile when compared to the trips taken by sedans. Moreover, longer trips have less driving volatility. In addition, younger drivers are more volatile drivers than old ones. Overall, the results of this study are reasonable and beneficial in identifying correlates of driving volatility, especially in terms of understanding factors that differentiate highly volatile trips from other trips. Reductions in driving volatility have positive implications for transportation safety. From a methodological standpoint, this study is an example of how to extract useful (volatility) information from raw vehicle speed data and use it to calm down drivers and ultimately improve transportation safety.

研究动机与目标

• 识别替代燃料汽车中高波动性驾驶出行的关键相关因素。
• 量化出行层面、车辆层面及个人层面变量对驾驶波动性的影响。
• 通过统计建模将高波动性出行（第90百分位数）与波动性较低的出行区分开来。
• 通过揭示与激进驾驶相关的驾驶行为和车辆特定模式，为交通安全管理策略提供依据。

提出的方法

• 将基于速度的驾驶波动性度量方法应用于加州家庭出行调查（2012–2013年）的出行级车辆速度数据。
• 使用普通最小二乘法（OLS）回归模型分析整体波动性相关性。
• 在第90百分位数处采用分位数回归，以隔离并分析最波动的10%出行。
• 分析波动性与车辆类型、出行长度及驾驶员年龄等变量之间的关系。
• 比较不同车辆类别（如轿车与皮卡）及不同人口统计群体（如年轻驾驶员与年长驾驶员）的驾驶行为。
• 通过统计建模从原始速度数据中提取可操作的见解，用于交通安全管理应用。

实验结果

研究问题

• RQ1 哪些车辆类型与替代燃料汽车中更高的驾驶波动性相关？
• RQ2 出行长度与驾驶波动性之间存在何种关联，特别是在最波动的10%出行中？
• RQ3 驾驶员年龄和人口统计因素在多大程度上影响驾驶波动性？
• RQ4 不同车辆类型和出行特征下，速度波动和激进驾驶行为有何差异？
• RQ5 哪些统计方法能有效从原始速度数据中隔离并分析高波动性驾驶行为？

主要发现

• 与皮卡、掀背车、敞篷车或多功能休旅车相比，轿车出行的波动性显著更高。
• 年轻驾驶员的驾驶波动性高于年长驾驶员。
• 较长出行与较低的驾驶波动性相关，表明驾驶更趋平稳，速度波动更小。
• 第90百分位数分位数回归模型成功隔离了最波动的10%出行，凸显了极端驾驶行为。
• 驾驶波动性与车辆类型和驾驶员年龄密切相关，表明针对性的安全干预措施可降低事故风险。
• 本研究证明，原始速度数据可有效转化为可操作的波动性指标，用于交通安全管理应用。