基于自然驾驶数据的高速公路出口换道决策模型

Modeling LaneChanging Behavior in Freeway OffRamp Areas Using Naturalistic Driving Data

摘要:

为研究高速公路出口换道行为特性，分析驾驶员换道决策机理，依托上海自然驾驶实验所采集的驾驶行为样本和车辆运行参数，采用Google Earth标定行驶路径及高速公路出口范围以筛选出口样本，并根据车道偏移参数和方向盘转角识别换道行为；以单向4车道高速公路为例，综合考虑行驶路径信息和交通流环境，基于随机效用理论，采用Binary Logit(BL)模型拟合构建换道决策模型，得到车道效用函数；基于效用函数作出高速公路出口范围内在自由流、稳定流和拥挤流水平下的分车道效用分布图，并进行同质性和异质性分析.结果表明,换道决策模型准确率达到86.21%，各变量影响均可得到合理解释；根据效用分析，出匝车辆的换道行为是出匝意愿与通行环境改善需求两方面平衡的结果，兼有强制性换道与自由性换道的行为特性，且随着交通流状态由自由流过渡到拥挤流，后者影响逐渐增强，表现为上游向左换道行为趋于活跃、下游向右换道位置接近出口.

Abstract:

The main focus of this paper is to study lanechanging decisionmaking characteristics in freeway offramp areas and analyze the mechanism of lanechanging decisionmaking. Using driving behavior samples and vehicle operating data collected from Shanghai Naturalistic Driving Study, the departure samples are selected according to the driving path and specific range of exit with Google Earth. Then lanechanging behaviors can be identified by lane offset position and steering wheel angle. Furthermore, illustrated by the example of the 8lane freeway (four in each direction), this paper comprehensively considers driving route information and traffic flow environment as factors. On this basis, relying on the random utility theory, the binary logic (BL) model is adopted for fitting lanechanging decisionmaking model so that the lane utility function is obtained. After graphing utility of different lanes in offramp areas respectively in free flow, steady flow and congested flow, the homogeneity and heterogeneity analysis are conducted. The result indicates that the accuracy of the model comes to 86.21% and the influence of each factor can be reasonably explained. Besides, according to utility analysis, the lanechanging behavior of leaving vehicles is a consequence of the balance between leaving willingness and requirement for improving driving environment. In other words, it combines the characteristics of mandatory and discretionary incentives. Moreover, in the process of traffic transforming from free flow to heavy flow, the influence of the latter increases: the position at which drivers start to perform right lanechanging gradually approaches the diverging area and left lanechanging behavior is more active in upstream section.