【作者】 罗强；

【导师】 许伦辉；

【作者基本信息】 华南理工大学 ， 交通信息工程及控制， 2014， 博士

【摘要】 随着经济的快速发展，汽车保有量的迅猛增长，城市交通问题不断加剧，主要表现为道路交通事故数量增加、交通拥堵以及相关的环境污染严重等。尤其是随着高速公路建设水平的不断提高，汽车行驶速度的不断提高，交通事故数和伤亡人数逐年增长，交通安全越来越受到广泛关注，已经成为了全世界所面临的重大问题。因此，展开对辅助驾驶员安全行驶的相关技术的研究，对于改善道路交通安全状况具有十分重要的理论意义和实际应用价值。本文在对车辆安全辅助驾驶技术的发展及研究现状回顾与分析的基础上，选取对高速公路上车辆横向驾驶行为的安全性进行研究，主要研究内容包括车辆车道偏离预警系统和车辆安全换道模型两大部分，对其中关键理论与技术进行了相关的研究与探讨，完成的主要科研工作与取得的研究成果概括如下：1、建立了车辆车道偏离时间TLC的估算模型。以车辆行驶过程中方向盘转角是否零，将车辆的行驶轨迹划分成直线和曲线轨迹两种情况；同时，结合车辆运动学理论、几何知识和三角函数知识，对车辆分别行驶在直线和曲线路段上的实际轨迹进行了研究，得到了车道偏离时间TLC的计算方法，由此建立了车辆向左和向右两种偏离情况下偏离时间TLC的估算模型。分别就车辆横向距离、相对偏航角和曲线道路半径等参数对偏离时间TLC的影响进行了数值仿真分析，得到了相应的仿真结果；利用自主开发的车辆换道模拟系统进行仿真实验，得到了不同计算方法的TLC预测值及与真实TLC值的差值。数值仿真结果对于驾驶员的安全驾驶和道路线形的设计有一定的指导作用；实验结果表明本文建立的模型计算得到的TLC值比其他相似方法所得值更接近真实值，误差更小。2、提出了一种对驾驶员的横向驾驶行为进行分类的模糊决策系统。通过对驾驶员的历史驾驶数据的分析，结合车辆车道偏离的运动特性，文中选用横向速度和横向位置作为评价驾驶员横向驾驶风格的主要因素，分别求取它们的标准方差，并作为决策系统的输入变量。然后利用基于Mamdami inference推理的模糊决策方法对驾驶员驾驶风格类型进行了分类，从而为建立具有自适应特性的车道偏离预警系统打下基础。3、建立了基于驾驶员驾驶风格类型和实时驾驶情况的可变虚拟车道边界偏离预警模型。结合驾驶员驾驶风格的类型、驾驶过程中实时横向位移、横向速度三个参数，文中利用Mamdami inference推理的模糊决策方法给出可变虚拟车道边界的调整方案，能够适应不同的驾驶员类型和实时变化的驾驶情况。然后，综合分析车道偏离预警系统的特性和现有研究中采用的预警性能指标，本文选用了虚警率和预警时间作为可变虚拟车道边界偏离预警模型的评价指标。最后，由于实验设备和条件有限，本文设置了虚拟实验仿真，详细介绍了实验方法、评价方法以及跟其他方法的结果对比的表现形式。4、建立了基于最小安全距离的车辆换道模型。首先，从车辆换道行为特性和运动学特性的角度出发，详细描述车辆换道过程，给出了换道的三阶段和碰撞情况的分类，同时定义了车辆换道环境、运动学参数和时间参数等。然后，从车辆换道的运动学角度出发，分析了换道车辆M与其周围相邻车辆（车辆Mol、Mdl、Mdf）在换道过程中的运动位置关系，以换道过程中避免碰撞为目标，建立了换道车辆M与相关联车辆之间的最小初始安全换道距离模型。5、对换道车辆M以匀速和加速两种换道策略进行了仿真分析。以建立的最小安全距离换道模型为基础，分别对换道车辆M设置了两种不同的换道策略，分别为：一是车辆M以匀速状态进行换道；二是车辆M以加速状态进行换道。利用Matlab仿真软件对两种不同换道的情况进行了仿真分析，给出了车辆M与周围相邻车辆间的安全换道区域和非安全换道区域。对仿真结果进行了分析，得到了影响车辆安全换道的相关因素，同时，通过对比分析车辆M在匀速换道和加速换道情况下的安全区域范围可知，车辆M采取加速度换道策略，可以拓宽安全区域范围，换道安全更可靠和安全。更多还原

【Abstract】 With the rapid development of economy, vehicle registrations are growing fast. Thus theurban traffic problem are aggravating, mainly for growing number of road traffic accidents,more serious traffic congestion and associated with environmental pollution, etc. Especiallywith the constant improvement of highway construction level, vehicle speed are increasing,the number of traffic accidents and casualties are increasing year by year. Traffic safety hastaken more and more attention, become a major problem which the whole world faces.Therefore, research of vehicle safety assistant driving technology has important theoreticalsignificance and practical application value for realizing the auto safety driving andimproving road traffic safety.On the basis of review and analysis of development and research status of the vehiclesafety assistant driving technology, this dissertation selects horizontal driving behavior’ssecurity on the motorway to be the research object of which main research content includesthe vehicle lane departure warning systems and vehicle safety lane changing model, studiesand discusses the key theory and technology. The main research work and research results inthis dissertation are summarized as follows:1. The estimation model of vehicle Time to Lane Crossing (TLC) is established.According to whether the vehicle’s steering wheel angle is zero or not when moving, themovement track of vehicle is divided into two kinds: straight trajectory and curve trajectory.At the same time, on the basis of vehicle kinematics theory, geometry knowledge andtrigonometric function knowledge, the calculation method of TLC is studied when vehiclesare driving in a straight line and curve sections respectively, then the estimation models ofTLC are established in the cases of vehicle departure to left and right. Finally, the simulationanalysis with effect of vehicle lateral distance, relative yaw angle and curve radius of road onthe TLC is carried out, the corresponding simulation results are obtained, which have animportant guiding role for driver’s safe driving and road alignment design. The experimentresults show that TLC value that is obtained from established model of this dissertation ismore approximate to true value, and the error is less than that of other methods.2. The fuzzy decision system of driver classification based on the transverse drivingbehavior is proposed. Through the analysis of driver’s driving history data, this dissertationselects the lateral velocity and lateral position as the main factors of evaluating driving lateraldriving styles by combining with the characteristics of vehicle departure. Then, the driver’sdriving styles classified system is obtained based on the Mamdami inference reasoning method, which lays the foundation to establish adaptive characteristics of lane departurewarning model.3. The lane departure warning model of variable virtual lane boundary based on drivertype and real-time driving situation is established. Combining with driver’s driving style,real-time horizontal displacement and lateral velocity, this dissertation gives the adjustmentscheme of variable virtual lane boundary by using Mamdami inference reasoning method,which can adapt to different kinds of drivers and real-time change of driving conditions. Then,both false alarm rate and warning time are choose as the evaluation index of variable virtuallane boundary departure warning model, after comprehensive analysis of characteristics oflane departure warning system and warning performance index being used in existingresearches. In the end, this dissertation sets up a virtual experiment simulation due to thelimited experimental equipment and simulation conditions, and detailed introduce theexperiment setting, experiment evaluation method and comparison results manifestation withother methods.4. The vehicle lane changing models based on the minimum safe distance are established.First of all, this dissertation details the vehicle lane changing process, gives three stages oflane changing and collision classification from the vehicle lane-changing behaviorcharacteristic and the kinematics characteristics of perspective. At the same time, the vehiclelane changing environment, kinematics parameters and time parameters are defined. Then, onthe basis of the goal of avoiding collisions, the kinematics relation between lane-changingvehicle M and its surrounding adjacent vehicle (vehicle Mol, Mdl, Mdf) in the process of lanechanging is analyzed from the view of the vehicle lane changing kinematics, then theminimum initial safety lane changing distance model between lane-changing vehicle M andassociated adjacent vehicles is set up.5. The simulation analysis is performed about the vehicle M changes lane according tochanging strategies of constant speed and accelerating. On the basis of the minimum initialsafety lane changing distance model, this dissertation sets up two different lane changingstrategies of vehicle M respectively, as follows: one way is that the vehicle M changes lane ata constant speed, the other way is that vehicle M changes lane at the condition of acceleration.Then, the simulation analysis of two different lane changing situation is carried out by usingthe Matlab simulation software. The relational graphs of safety region and non-safety regionare given between vehicle M and adjacent vehicles (vehicle Mol, Mdl, Mdf) in the process oflane changing. After analysising the simulation results, this dissertation gets relevant factorsthat affect vehicles’ safety lane-changing. At the same time, after contrasting of safety region between vehicle M and adjacent vehicles in the cases of lane changing at the constant speedand acceleration changing strategies, the analysis results show that the security region will bebroaden when vehicle M taking acceleration lane changing strategy, and there will be morereliable safety and security for lane-changing.更多还原