基于车辆车载诊断数据的山地城市道路异常驾驶行为空间分布特征

为了探究山地城市异常驾驶行为的空间分布规律，本文基于车辆OBD（On Board Diagnostics）数据建立了异常驾驶行为空间分布规律模型。首先，以重庆市6条主干道130个路段为研究对象，定性分析道路坡度、弯度、公交站和开口与异常驾驶行为间的关联性；然后，分别构建Possion回归模型和零膨胀Possion回归模型（ZIP）、零膨胀负二项回归模型（ZINB），对急加速、急减速、急转弯和超速行为发生频率的空间分布特征进行了描述；最后，随机选取路段进行模型验证，结果表明：对于急加速率、急减速率空间分布特征，Possion回归模型拟合效果较优，绝对误差在10%~10%；急转弯率适宜采用ZINB回归模型，80.77%的绝对误差分布在（-0.005~0.005）；超速率适宜采用ZIP回归模型，71.15%的绝对误差基本分布在（-0.002~0.002）区间。

Spatial distribution characteristics of abnormal driving behavior on mountain roads based on vehicle OBD data

In order to explore the spatial distribution of abnormal driving behavior in mountain cities, a spatial distribution model of abnormal driving behavior was established based on vehicle OBD data. Firstly, 130 sections of 6 main roads in Chongqing were taken as research objects, and the correlation between road slope, curve, bus station and opening on abnormal driving behavior was qualitatically analyzed. Then, Possion regression model, zero expansion Possion regression model (ZIP) and zero expansion negative binomial regression model (ZINB) were constructed to describe the spatial distribution characteristics of the occurrence frequency of sharp acceleration, sharp deceleration, sharp turning and overspeed behaviors. Finally, random sections were selected for model verification. The results showed that Possion regression model had better fitting effect for spatial distribution characteristics of rapid acceleration rate and rapid deceleration rate, with absolute error of 10%-10%. ZINB regression model is suitable for sharp turn rate, and the absolute error distribution of 80.77% is (-0.005~0.005).ZIP regression model is suitable for super speed, and the absolute error of 71.15% is basically distributed in the range of (-0.002~0.002).