大型营运客车在高速公路的换道安全性辨识

为了给大型营运客车换道预警系统设计提供参考,采用毫米波雷达、激光雷达、车道线识别传感器、GPS、视频监控系统以及控制器局域网（CAN）总线数据采集仪等设备,基于小型乘用车搭建浮动车采集平台。通过在试验线路上进行1.5×10⁴ km的驾驶试验,获取1 200余次营运客车的真实换道数据。以Jula提出的换道安全性模型为基础,结合营运客车的换道行为特征,通过分析换道进程结束后客车需要与周围车辆保持的安全距离,建立适合于营运客车的3类换道安全性识别模型（客车与自车道前方车辆、目标车道前方车辆、目标车道后方车辆）,并利用真实数据对3类模型进行验证。研究结果表明：客车换道持续时间均值为10.4 s,换道起始时刻与目标车道后方车辆的距离为10.0~40.0 m;所有换道样本中,73.3%的换道过程中客车速度要高于目标车道后方车辆,且超过90%的换道过程是由前方慢车引起;不同的速度区间下,车速和航向角联合变化情况下,驾驶人控制营运客车的横向偏移速度保持稳定,可认为客车驾驶人的心理预期换道进程存在固定经验模式,这与小型车换道的研究结论存在较大差异,传统的TTC预警算法识别率较低,在不同速度区间情况下,所提出的模型对客车与自车道前方车辆、目标车道前方车辆、目标车道后方车辆的换道安全识别评价准确率均超过了90%。     

Vehicle dynamics and road curvature estimation for lane departure warning system using robust fuzzy observers: experimental validation

This paper describes a new approach to estimate vehicle dynamics and the road curvature in order to detect vehicle lane departures. This method has been evaluated through an experimental set-up using a real test vehicle equipped with the RT2500 inertial measurement unit. Based on a robust unknown input fuzzy observer, the road curvature is estimated and compared to the vehicle trajectory curvature. The difference between the two curvatures is used by the proposed lane departure detection algorithm as the first driving risk indicator. To reduce false alarms and take into account driver corrections, a second driving risk indicator based on the steering dynamics is considered. The vehicle nonlinear model is deduced from the vehicle lateral dynamics and road geometry and then represented by an uncertain Takagi–Sugeno fuzzy model. Taking into account the unmeasured variables, an unknown input fuzzy observer is proposed. Synthesis conditions of the proposed fuzzy observer are formulated in terms of linear matrix inequalities using the Lyapunov method.     

Vision-based fusion of robust lane tracking and forward vehicle detection in a real driving environment

With the goal of developing an accurate and fast lane tracking system for the purpose of driver assistance, this paper proposes a vision-based fusion technique for lane tracking and forward vehicle detection to handle challenging conditions, i.e., lane occlusion by a forward vehicle, lane change, varying illumination, road traffic signs, and pitch motion, all of which often occur in real driving environments. First, our algorithm uses random sample consensus (RANSAC) and Kalman filtering to calculate the lane equation from the lane candidates found by template matching. Simple template matching and a combination of RANSAC and Kalman filtering makes calculating the lane equation as a hyperbola pair very quick and robust against varying illumination and discontinuities in the lane. Second, our algorithm uses a state transfer technique to maintain lane tracking continuously in spite of the lane changing situation. This reduces the computational time when dealing with the lane change because lane detection, which takes much more time than lane tracking, is not necessary with this algorithm. Third, false lane candidates from occlusions by frontal vehicles are eliminated using accurate regions of the forward vehicles from our improved forward vehicle detector. Fourth, our proposed method achieved robustness against road traffic signs and pitch motion using the adaptive region of interest and a constraint on the position of the vanishing point. Our algorithm was tested with image sequences from a real driving situation and demonstrated its robustness.     

高速公路相机自动标定及道路坐标系构建

为解决高速公路场景下利用视频监控系统正确描述车辆相对于道路的空间位置问题，通过引入Frenet坐标系概念，提出一种基于相机自动标定的道路坐标系模型。在相机自动标定阶段，利用线分段拟合方法从曲线车辆轨迹中提取平行于直线路段的轨迹点，并通过级联霍夫变换精确估计道路方向的消失点。然后,根据多车辆三维模型约束，对相机参数进行迭代优化。基于标定结果，将车辆轨迹映射到世界坐标系平面上，并用3次样条插值进行拟合。根据大量运动车辆在道路平面内形成的轨迹域分布特征，综合边界约束估计道路中心点。最后，结合道路中心线在各点处的法线向量与车道宽度信息确定平移量，并利用点平移运动拟合车道线，实现道路坐标系的自动建立。使用真实高速公路视频数据，在多种道路条件下进行试验。研究结果表明:在标定阶段，构建方法对不同高速公路场景的最大标定误差不超过11.55%;与最新的方法相比，直线道路平均标定误差分别降低6.68%和3.58%，弯曲道路平均标定误差分别降低7.43%和2.61%;在道路坐标系构建阶段，构建方法的平均投影距离为0.077 m，接近最新方法的0.069 5 m；而其平均精度为0.916，显著优于最新方法的0.663；所提道路坐标系能够自适应道路形态的变化，有效解决了从监控视频中描述车辆与道路之间相对位置关系的问题。     

基于扩展卡尔曼滤波的车道融合跟踪

车道检测算法的研究是智能车辆自动导航的首要环节。与目前基于视觉的车道检测与跟踪系统不同,本文提出一种基于扩展卡尔曼滤波的车道融合跟踪方法。该方法利用毫米波雷达探测到前方车辆的距离信息,并采用扩展卡尔曼滤波技术和图像处理技术,建立车道跟踪的动态视觉窗口,提取车道边界,并判断前方车辆相对于车道的位置。该方法大大缩减了处理时间,且增强了系统的鲁棒性。     

Dynamic Estimation of Road Bank Angle

Road bank angles have a direct influence on vehicle dynamics and lateral acceleration measurement. A vehicle stability control system that knows road bank angle has an advantageous capability in achieving desired control sensitivities for maneuvers on ice and snow, among all surfaces, while avoiding false/nuisance activation on a banked road. Since neither lateral velocity nor road bank angle are directly measurable in current vehicle systems due to economical reasons, the major challenge is to differentiate the bias induced by road bank disturbances from actual effect of vehicle lateral dynamics in current measurements. This paper proposes a method of road bank estimation and provides theoretical background for the decoupling effort of lateral dynamics and road disturbances involved in bank estimation.     

Dynamic Estimation of Road Bank Angle

Road bank angles have a direct influence on vehicle dynamics and lateral acceleration measurement. A vehicle stability control system that knows road bank angle has an advantageous capability in achieving desired control sensitivities for maneuvers on ice and snow, among all surfaces, while avoiding false/nuisance activation on a banked road. Since neither lateral velocity nor road bank angle are directly measurable in current vehicle systems due to economical reasons, the major challenge is to differentiate the bias induced by road bank disturbances from actual effect of vehicle lateral dynamics in current measurements. This paper proposes a method of road bank estimation and provides theoretical background for the decoupling effort of lateral dynamics and road disturbances involved in bank estimation.     

车队认知能力增强的通信区块间隙优化方法

为了在单车超越车队的过程中缩短超车车辆与车队间通信范围，减少车队通信压力，锁定影响车辆入队的关键车队区块，同时通过将待进入关键区块的车队进行间隙优化调整，为驾驶人提供定制化换道入队引导服务，提出了基于驾驶人超车风格特征参数的车队内信息传输关键区块锁定算法，通过分析影响驾驶人换道入队位置范围的关键因素，将驾驶人换道入队过程分为本车道速度调整过程与入队速度调整过程，利用非参数贝叶斯算法获取驾驶人超车换道特征数据并提出基于关键区块所在车队位置序列的车辆间隙优化调整策略。研究结果表明：超车车辆加速度、与前车预计碰撞时间、与车队相对速度是影响驾驶人换道入队范围的关键因素；通过非参数贝叶斯算法将超车车辆运行数据分类获取的驾驶人换道入队驾驶操作基元，可准确提供驾驶人行为特征关键参数；通过将驾驶人换道特征分为48个子类型，可锁定驾驶人换道入队范围且车队关键区块范围随着超车车辆与车队速度差值不同在各个特征类型上呈现不同变化趋势；针对驾驶人入队特征对待进入车队关键区块的车辆间隙进行优化调整，不仅可以为驾驶人提供可接受的驾驶辅助信息，同时减少了车队间隙产生过程中车辆加速度范围，提升了车队运行的舒适性。     

Vehicle velocity estimation using effective inertia for an in-wheel electric vehicle

In this study, a vehicle velocity estimation algorithm for an in-wheel electric vehicle is proposed. This algorithm estimates the vehicle velocity using the concept of effective inertia, which is based on the motor torque, the angular velocity of each wheel and vehicle acceleration. Effective inertia is a virtual mass that changes according to the state of a vehicle, such as acceleration, deceleration, turning or driving on a low friction road. The performance of the proposed vehicle velocity estimation algorithm was verified in various conditions that included straight driving, circle driving and low friction road driving using the in-wheel electric vehicle that was equipped with an in-wheel system in each of its rear wheels.     

In-depth understanding of lane changing interactions for in-vehicle driving assistance systems

Lane-changing events are often related with safety concern and traffic operational efficiency due to complex interactions with neighboring vehicles. In particular, lane changes in stop-and-go traffic conditions are of keen interest because these events lead to higher risk of crash occurrence caused by more frequent and abrupt vehicle acceleration and deceleration. From these perspectives, in-depth understanding of lane changes would be of keen interest in developing in-vehicle driving assistance systems. The purpose of this study is to analyze vehicle interactions using vehicle trajectories and to identify factors affecting lane changes with stop-and-go traffic conditions. This study used vehicle trajectory data obtained from a segment of the US-101 freeway in Southern California, as a part of the Next Generation Simulation (NGSIM) project. Vehicle trajectories were divided into two groups; with stop-and-go and without stop-and-go traffic conditions. Binary logistic regression (BLR), a well-known technique for dealing with the binary choice condition, was adopted to establish lane-changing decision models. Regarding lane changes without stop-and-go traffic conditions, it was identified based on the odd ratio investigation that he subject vehicle driver is more likely to pay attention to the movement of vehicles ahead, regardless of vehicle positions such as current and target lanes. On the other hand, the subject vehicle driver in stop-and-go traffic conditions is more likely to be affected by vehicles traveling on the target lane when deciding lane changes. The two BLR models are adequate for lane-changing decisions in normal and stop-and-go traffic conditions with about 80 % accuracy. A possible reason for this finding is that the subject vehicle driver has a tendency to pay greater attention to avoiding sideswipe or rear-end collision with vehicles on the target lane. These findings are expected to be used for better understanding of driver’s lane changing behavior associated with congested stop-and-go traffic conditions, and give valuable insights in developing algorithms to process sensor data in designing safer lateral maneuvering assistance systems, which include, for example, blind spot detection systems (BSDS) and lane keeping assistance systems (LKAS).     

Vehicle longitudinal velocity estimation during the braking process using unknown input Kalman filter

In this paper, vehicle longitudinal velocity during the braking process is estimated by measuring the wheels speed. Here, a new algorithm based on the unknown input Kalman filter is developed to estimate the vehicle longitudinal velocity with a minimum mean square error and without using the value of braking torque in the estimation procedure. The stability and convergence of the filter are analysed and proved. Effectiveness of the method is shown by designing a real experiment and comparing the estimation result with actual longitudinal velocity computing from a three-axis accelerometer output.     

考虑过街行人运动轨迹不确定性的人-车碰撞概率预测方法

为了准确预测人-车冲突中的碰撞风险,研究了利用碰撞概率评估人-车碰撞风险的预测方法。基于车辆运动特征建立车辆运动学模型,通过采集行人实际过街运动轨迹并提取不确定性特征,采用一阶马尔科夫模型和高斯白噪声建立行人随机运动模型,在此基础上构建人-车冲突距离模型;运用蒙特卡洛抽样,提取行人过街过程中的人-车最短距离和碰撞时间（time to collision,TTC）分布特征,通过拟合这些特征来估算最短距离和TTC的概率密度函数,建立人-车碰撞概率预测模型;结合2起人-车深度事故案例和3种不同制动特性的自动紧急制动（automatic emergency braking,AEB）系统,对比验证人-车碰撞概率预测模型的有效性。结果显示：建立的行人随机运动模型,其模拟的行人运动速度的均值和标准差与实际值的绝对误差在2%以内,模型精度较高;在事故案例仿真中,车辆与行人在发生碰撞时刻对应的碰撞概率为100%;在车辆加装AEB的仿真中,激进型AEB,法规型AEB以及保守型AEB在触发时刻对应的碰撞概率分别为超过了80%,在30%~40%之间,以及不足5%,这表明人-车碰撞概率预测模型可有效预测2起真实案例中行人和车辆在不同时刻的碰撞风险,且与使用固定触发阈值的AEB相比,建立的人-车碰撞概率预测模型能够更加准确直观地反应人-车碰撞风险。     

基于双视角学习原理的交通视频车辆事件鲁棒检测

提出了1种基于双视角学习原理的高速公路交通视频车辆事件鲁棒检测算法.针对道路交通结构化特点提出了分车道外极面图(Epipolar Plane Image,简称 EPI),以此反映交通断面车流整体特征.基于双视角学习原理,融合现有广泛应用的反映车辆独立行为的行驶轨迹特征,实现高速公路车辆事件鲁棒检测.针对多种典型车辆事件(包括交通拥堵,车辆逆行,车辆违规停车,交通事故等),本文算法总体检测率为94.09%,误检率为4.51%,漏检率为1.40%,其性能与传统单视角方法比较有较大的提高.     

基于分段学习模型的自动驾驶行为决策算法研究

在具有车道线的特定自动驾驶场景中,针对目前端到端的行为决策算法直接输入原始图像进行决策导致的网络模型迁移性差、预测精度欠佳、泛化能力不足等问题,提出一种基于分段学习模型的车辆自动驾驶行为决策算法。首先,基于GoogLeNet建立一种端到端的车道线检测网络模型,并引入车道中心线作为决策的重要线索提高算法的迁移能力,同时利用YOLOv3目标检测模型对本车道内前方最近障碍物进行位置检测;而后,经几何测量模型将两者检测结果转换成环境状态信息向量为决策做支撑;最后,构建基于长短期记忆(LSTM)网络的驾驶行为决策模型,根据编码的历史状态信息刻画出动态环境中车辆的运动模式,并结合当前时刻的状态推理得到驾驶行为参量。使用建立的真实驾驶场景数据集对模型分别进行训练、验证与测试,离线测试结果显示车道线检测模型的检测位置误差小于1.3%,车道内前方障碍物检测模型的检测精度达98%以上,驾驶行为决策网络模型表征预测优度的决定系数 大于0.7。为进一步验证算法的有效性,搭建了Simulink/PreScan联合仿真平台,多种工况下的仿真验证试验中多个评价指标均达到工程精度要求,实车测试的试验结果也表明该算法可实现复杂驾驶场景下平稳、准确无偏航的预测效果并满足实时性要求,且与传统端到端模式的算法相比,具有更好的迁移性和泛化能力。     

Vehicle detection system design based on stereo vision sensors

Vehicle detection is a crucial issue for driver assistance system as well as for autonomous vehicle guidance function and it has to be performed with high reliability to avoid any potential collision. The vision-based vehicle detection systems are regarded promising for this purpose because they require little infrastructure on a highway. However, the feasibility of these systems in passenger car requires accurate and robust sensing performance. In this paper, a vehicle detection system using stereo vision sensors is developed. This system utilizes feature extraction, epipoplar constraint and feature matching in order to robustly detect the initial corresponding pairs. The proposed system can detect a leading vehicle in front and can estimate its position parameters such as the distance and heading angle. After the initial detection, the system executes the tracking algorithm for the vehicles in the lane. The proposed vehicle detection system is implemented on a passenger car and its performances are verified experimentally.     

基于Matlab的汽车偏离车道报警系统

为提高汽车的主动安全性并克服现有的汽车偏离车道报警系统所存在的结构复杂和成本高等缺点,文章设计了一种基于Matlab的汽车偏离车道报警系统,利用摄像头获取车道图像并实时传给车内计算机系统,经过对图像一系列的处理分析判断汽车是否偏离车道而进行报警。经验证,该系统能够实时检测出左右车道标志线,可以根据车道夹角法判断汽车是否偏离车道,满足了汽车偏离车道报警的需要。该系统结构简单高效,能显著提高汽车行驶的主动安全性。     

Road curvature estimation for vehicle lane departure detection using a robust Takagi–Sugeno fuzzy observer

In this paper, a lane departure detection method is studied and evaluated via a professional vehicle dynamics software. Based on a robust fuzzy observer designed with unmeasurable premise variables with unknown inputs, the road curvature is estimated and compared with the vehicle trajectory curvature. The difference between the two curvatures is used by the proposed algorithm as the first driving risk indicator. To reduce false alarms and take into account the driver corrections, a second driving risk indicator is considered, which is based on the steering dynamics, and it gives the time to the lane keeping. The used nonlinear model deduced from the vehicle lateral dynamics and a vision system is represented by an uncertain Takagi–Sugeno fuzzy model. Taking into account the unmeasured variables, an unknown input fuzzy observer is then proposed. Synthesis conditions of the proposed fuzzy observer are formulated in terms of linear matrix inequalities using Lyapunov method. The proposed approach is evaluated under different driving scenarios using a software simulator. Simulation results show good efficiency of the proposed method.     

基于边界点分布特征的夜间道路检测算法研究

提出一种在结构化道路情况下的夜间道路标志线检测算法。选择邻域均值滤波、Sobel算子和最大熵分割算法对道路图像进行预处理。结合道路灰度图像与道路边界图像,分析夜间道路边界点的分布特征,将道路边界点划分为到4个点集。针对虚假道路边界点,应用多方向搜索方法予以消除。选用2D的直线模型,采用改进的Hough变换从预处理后的图像中获取道路几何特征参数,最终检测出道路边界。试验结果表明本算法可靠、有效,户外试验中具有很好的鲁棒性。     

基于无人机视频拍摄的高速公路小型车换道行为特性

为了研究高速公路小型车的换道行为特性,采用2台无人机同时在200 m的高空对交通流进行拍摄,获取交通流运行状态。构建拍摄路段的高精度地图,获取每一时刻车辆的精确运行状态数据,在此基础上对2个视频进行拼接,最终获得车道位置、速度、车辆编号等8项关键指标,共提取换道行为1 520条,筛选后得到完整的自由换道数据942条。采用车辆轨迹是否持续偏移作为判断换道行为起终点的依据,在此基础上分析换道的时间长度、空间长度、与周边车辆的相互状态以及换道行为的安全性等16个特征参数。得出平均换道时间长度为6.09 s,平均换道空间距离为148.08 m,换道时间与空间长度均符合对数正态分布。换道车辆与目标车道后方车辆的平均距离最小（34.29 m）,其相对距离在10 m以内的占28.24%,驾驶人为了加快行驶,在与目标车道后方车辆相对距离较小的情况下,依然采取换道措施。与正前方车辆的相对速度差最大,平均值为10.2 km·h^-1,并且在83%的情况下,本车的速度大于前车,说明车辆自由换道是由于前方车辆行驶速度较慢所引起。采用TTC,MTC分别对换道起始时刻的安全性进行分析,并将安全状态划分为4种类型：严重-紧急状态、严重-非紧急状态、非严重-紧急状态、非严重-非紧急状态。其中严重-非紧急,非严重-非紧急这2种状态占比最高。该研究成果对了解中国驾驶人在高速公路上的换道行为特性,以及对建立适用于中国实际交通环境特征的换道行为模型具有一定参考意义。     

Driving environment assessment using fusion of in- and out-of-vehicle vision systems

Because the overall driving environment consists of a complex combination of the traffic Environment, Vehicle, and Driver (EVD), Advanced Driver Assistance Systems (ADAS) must consider not only events from each component of the EVD but also the interactions between them. Although previous researchers focused on the fusion of the states from the EVD (EVD states), they estimated and fused the simple EVD states for a single function system such as the lane change intent analysis. To overcome the current limitations, first, this paper defines the EVD states as driver’s gazing region, time to lane crossing, and time to collision. These states are estimated by enhanced detection and tracking methods from in- and out-of-vehicle vision systems. Second, it proposes a long-term prediction method of the EVD states using a time delayed neural network to fuse these states and a fuzzy inference system to assess the driving situation. When tested with real driving data, our system reduced false environment assessments and provided accurate lane departure, vehicle collision, and visual inattention warning signals.