摘要
汇入行为是导致高速公路、快速路汇入区瓶颈失效的根本原因,也是引发车辆冲突、交通事故的重要诱因。基于上海虹许路汇入区高精度轨迹数据,在经典的3类汇入行为(自由、协作、压迫)分类基础上发现并定义了一类新的汇入行为:主动-回应汇入,即在目标车道插车间隙不足的条件下,汇入车辆通过横向偏移主动表达汇入意图,引发后车合作让行,从而实现汇入。由于传统的换道类型划分方法已不适用于描述瓶颈区汇入过程的复杂交互现象,因此根据汇入车辆的横向位置将汇入过程划分为4个时段,以目标车道后车让行的时段不同对4类汇入行为进行了重新划分。基于此对虹许路汇入区轨迹数据中的汇入行为进行分类,并比较主动-回应汇入行为与经典的3类汇入在汇入位置与汇入间隙变化分布的差异。进一步,采用风险空间理论判断目标车道间隙的可汇入程度,并结合混合高斯-隐马尔可夫模型实现对后车的让行意图识别,建立了主动-回应汇入模型。最后构建了瓶颈交通流仿真原型系统以验证模型有效性。结果表明:在宏观交通流层面,主动-回应汇入行为在汇入位置、速度和间隙的分布上与实证数据一致,均值没有显著差异;在个体行为层面,也能再现定义描述的“主动试探-两车博弈-回应让行”的过程。该研究成果对解析瓶颈交通流早发性失效机理、揭示汇入风险变化过程以及设计符合人类驾驶人交互特征的自动驾驶汇入策略均具有重要的指导意义。
Merging behavior is deemed as the primary trigger of on-ramp breakdown on freeways and expressways and an important cause of traffic conflicts and accidents.Based on the observations of high-resolution trajectory data from the Hongxu Road bottleneck in Shanghai,a new type of merging behavior that differs from the three conventional types(free/forced/cooperative),active-responsive merge,was found and defined.In the case of an unacceptable merge gap,the on-ramp vehicle actively expresses the intent of merging by laterally approaching before entering the target lane,forcing the lag vehicle on the target lane to yield as a response,thereby successfully merging.Given that the classification of different lane-change types in previous studies cannot accurately represent the complicated interactions occurring at high-density expressway on-ramp bottlenecks,a new method is proposed in which each lane-change process is divided into four phases,and the lane-change type is determined by the phase at which the lag vehicle yields its way.The characteristics of the active-responsive merge were analyzed and compared with those of the three conventional types.Furthermore,to establish an active-responsive merge model,the risk field theory was introduced to evaluate the merging risk,and the mixed Gaussian-hidden Markov model was introduced to recognize the yield intention of the lag vehicle.Finally,the model was validated using a high-density traffic-flow simulation prototype.The results show that the distributions of the merging position/speed/gap in the field data and simulation do not significantly differ.Meanwhile,from the perspective of individual vehicle behavior,the interactive process,which includes “active approach-interaction-responsive yield”,is well represented.This study is notably relevant to achieve a deeper understanding the mechanism of early onset bottleneck breakdown,revealing the changing process of the merging risk.It is also helpful for designing an autonomous driving merging strategy that conforms to the interaction characteristics of human drivers.
作者
叶颖俊
倪颖
孙剑
YE Ying-jun;NI Ying;SUN Jian(Key Laboratory of Road and Traffic Engineering of the Ministry of Education,Tongji University,Shanghai 201804,China)
出处
《中国公路学报》
EI
CAS
CSCD
北大核心
2022年第8期278-290,共13页
China Journal of Highway and Transport
基金
国家重点研发计划项目(2019YFB1600200)
国家自然科学基金项目(52125208)。
关键词
交通工程
主动-回应汇入
隐马尔可夫模型
汇入瓶颈
让行意图识别
traffic engineering
active-responsive merge
hidden Markov model
expressway on-ramp bottleneck
yield intention recognition
