CN1775601A

CN1775601A - Vehicle Trajectory Prediction and Lane Departure Evaluation Method

Info

Publication number: CN1775601A
Application number: CN 200510017293
Authority: CN
Inventors: 管欣; 高振海; 张立存; 柯斌; 杨得军
Original assignee: Jilin University
Current assignee: Jilin University
Priority date: 2005-11-18
Filing date: 2005-11-18
Publication date: 2006-05-24

Abstract

The invention relates to a method for pre-estimating vehicle trace and evaluating lane deviation, for overcoming the defect of the existing technique not fully reflecting true behavioral characteristic of a driver, announcing a method for evaluating the danger of lane deviation based on behavioral simulation of a driver, using the image sensor-collected lane mark line information of front road and combining the vehicle sensor-provided information such as automobile speed, acceleration, and turning state, to simulate the foreseeing action of the driver (i.e. as driving, the driver controls the present driving of the automobile according to the front road information), forecasting an expected running trace of the automobile in a future period of time, and evaluating the danger of lane deviation with respect to the relative position of the expected running trace and the front road mark lines, thus judging if a lane deviation accident will happen. If the danger of the lane deviation exceeds a certain threshold value, it can give out lane deviation early warning information to the driver by sound and vehicle LCD.

Description

Vehicle Trajectory Prediction and Lane Departure Evaluation Method

所属技术领域Technical field

本发明涉及车辆行驶轨迹预估及车道偏离评价方法。The invention relates to a method for estimating vehicle driving trajectory and evaluating lane departure.

背景技术Background technique

交通安全历来是人们最为关心的问题之一，它直接关系到人民生命和财产的损失。在高速公路上，每年都会发生很多由于驾驶员注意力分散或是操作原因所引起的交通事故，造成严重的人员伤亡和财产损失。因此，开发智能辅助驾驶系统，利用传感器系统感知道路交通环境信息进行决策规划，给驾驶员提出驾驶建议或部分代替驾驶员进行车辆控制操作，具有十分重要的意义。Traffic safety has always been one of the most concerned issues, which is directly related to the loss of people's lives and property. On the expressway, there are many traffic accidents caused by drivers' distraction or operation reasons every year, causing serious casualties and property losses. Therefore, it is of great significance to develop an intelligent assisted driving system, use the sensor system to perceive road traffic environment information for decision-making planning, provide driving suggestions to the driver or partially replace the driver for vehicle control operations.

作为智能辅助驾驶系统的一个子系统，车道偏离预警系统主要功能是在高速或者类似的公路环境中，辅助过度疲惫或者长时间单调驾驶的驾驶员保持车辆在车道内行驶。当由于驾驶员疏忽、打瞌睡、精神不集中、疲劳、打电话、寻找物品、操作仪表盘、或与乘客交谈等原因可能造成汽车不按驾驶员意愿而偏离其行驶车道时，系统向驾驶员发出报警，提醒并鼓励驾驶员进行校正操作，从而防止汽车偏离行驶车道，提高汽车的主动安全性。As a subsystem of the intelligent assisted driving system, the main function of the lane departure warning system is to assist the driver who is tired or monotonously driving for a long time to keep the vehicle in the lane in the high-speed or similar highway environment. When the driver's negligence, dozing off, lack of concentration, fatigue, making a phone call, looking for items, operating the instrument panel, or talking to passengers may cause the car to deviate from its driving lane against the driver's will, the system will notify the driver Send an alarm to remind and encourage the driver to perform corrective operations, thereby preventing the car from deviating from the driving lane and improving the active safety of the car.

如何根据当前的汽车行驶状态及道路信息预测汽车未来的行驶轨迹，以及如何建立评价指标判断汽车是否发生车道偏离，是本发明的关键。目前车道偏离预警系统的评价方法主要有三种：基于TLC(Time to Lane Crossing)的评价方法，基于预测轨迹偏差的评价方法，以及基于EDF(edge distribution function)的评价方法。TLC是指从车辆当前位置开始到车辆与车道线开始接触止的运动时间，其基本思想是如果TLC小于设定的时间阈值，则认为车辆将发生车道偏离。基于预测轨迹偏差的评价方法，假设偏离过程中汽车的航向角始终保持不变，从而计算一定时间后汽车的行驶轨迹，并与目标行驶轨迹比较。如果车辆的预期行驶轨迹与目标行驶轨迹之间的偏差大于设定的阈值则系统报警。基于EDF的评价方法，对车道标志线作出一定假设，根据边缘分布函数(EDF)的两个重要形状特征——局部极大值和对称轴判断汽车是否发生偏离。上面所提及的车道偏离预警系统的评价方法，从图像处理的角度出发，利用车辆当前时刻的状态判断车辆是否发生车道偏离，并没有充分考虑驾驶员的驾驶行为特性，与真实驾驶员驾驶汽车时对车道偏离行驶行为的理解有一定差异。How to predict the future driving trajectory of the automobile according to the current automobile driving state and road information, and how to establish an evaluation index to judge whether the automobile lane deviation occurs are the key points of the present invention. At present, there are three main evaluation methods for lane departure warning systems: evaluation methods based on TLC (Time to Lane Crossing), evaluation methods based on predicted trajectory deviation, and evaluation methods based on EDF (edge distribution function). TLC refers to the movement time from the current position of the vehicle to the start of contact with the lane line. The basic idea is that if the TLC is less than the set time threshold, the vehicle is considered to be lane departure. Based on the evaluation method of predicted trajectory deviation, assuming that the heading angle of the vehicle remains unchanged during the deviation process, the trajectory of the vehicle after a certain period of time is calculated and compared with the target trajectory. If the deviation between the expected driving trajectory of the vehicle and the target driving trajectory is greater than the set threshold, the system will give an alarm. Based on the evaluation method of EDF, certain assumptions are made on the lane markings, and whether the vehicle deviates is judged according to two important shape characteristics of the edge distribution function (EDF): the local maximum value and the axis of symmetry. The evaluation method of the lane departure warning system mentioned above, from the perspective of image processing, uses the current state of the vehicle to judge whether the vehicle has lane departure, and does not fully consider the driving behavior characteristics of the driver, which is different from that of a real driver driving a car. There is a certain difference in the understanding of lane departure driving behavior.

发明内容Contents of the invention

本发明其目的在于克服现有技术没有充分反映真实驾驶员行为特性的缺陷，提出一种基于驾驶员行为模拟的车道偏离危险性评价方法。The purpose of the present invention is to overcome the defect that the prior art does not fully reflect the real driver's behavior characteristics, and propose a lane departure risk assessment method based on driver behavior simulation.

为实现上述目的，本发明主要包括以下步骤：To achieve the above object, the present invention mainly comprises the following steps:

信息感知步骤：通过车载传感器提供车辆当前运动状态及行车环境信息，其中行车环境信息指根据道路图像获得的前方道路的车道标志线信息，车辆运动状态信息包括车辆纵向速度

，横向速度

，方向盘转角δ_sw，油门开度或者制动踏板行程比α；Information perception step: provide the vehicle’s current motion state and driving environment information through the on-board sensor, where the driving environment information refers to the lane marking information of the road ahead obtained from the road image, and the vehicle motion state information includes the vehicle’s longitudinal speed

, lateral velocity

, steering wheel angle δ _sw , accelerator opening or brake pedal stroke ratio α;

轨迹预估步骤：通过模拟驾驶员的前视作用，利用车辆稳态响应特性对汽车在未来一段时间内的预期行驶轨迹进行预估；Trajectory estimation step: By simulating the forward-looking effect of the driver, the vehicle's steady-state response characteristics are used to estimate the expected driving trajectory of the car in a certain period of time in the future;

车道偏离危险性评价步骤：是根据汽车预期行驶轨迹以及前方道路的车道标志线，建立预期轨迹点处汽车到左侧道路和右侧道路的横向安全距离以及汽车到前方道路的纵向安全距离三个车道偏离危险性评价指标，通过这三个指标对预期轨迹点的偏离危险性进行分析。Lane departure risk assessment steps: According to the expected driving trajectory of the car and the lane markings of the road ahead, establish the lateral safety distance from the car to the left road and the right road at the expected trajectory point and the longitudinal safety distance from the car to the road ahead. Lane departure risk evaluation index, through these three indicators to analyze the departure risk of the expected track point.

本发明的有益效果是，从驾驶员驾驶行为特性建模的角度出发，将驾驶员操纵行为特性应用于汽车高速公路车道偏离预警系统，更为精确地预测汽车未来的预期行驶轨迹；建立了可充分反映驾驶员行为特性的车道偏离危险性评价指标，提高了相应评价方法对汽车复杂行驶环境的适应性，为车道偏离预警系统的研究开发提供了一种新的方法，可以有效防止高速行驶汽车偏离其行驶车道，提高了高速汽车的主动安全性。The beneficial effect of the present invention is that, from the perspective of modeling the driver's driving behavior characteristics, the driver's manipulation behavior characteristics are applied to the vehicle expressway lane departure warning system, so as to more accurately predict the future expected driving trajectory of the vehicle; The lane departure risk evaluation index that fully reflects the driver's behavior characteristics improves the adaptability of the corresponding evaluation method to the complex driving environment of the vehicle, and provides a new method for the research and development of the lane departure warning system, which can effectively prevent high-speed vehicles from Deviating from its driving lane improves the active safety of high-speed vehicles.

附图说明Description of drawings

下面结合附图和实施例对本发明进一步说明。The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

图l为应用于车道偏离预警系统的偏离危险性评价方法结构图；Fig. 1 is the structural diagram of the method for evaluating the risk of departure applied to the lane departure warning system;

图2为基于驾驶员行为模拟的汽车预期行驶轨迹的预估示意图；Fig. 2 is the estimated schematic diagram of the expected driving trajectory of the car based on the driver's behavior simulation;

图3为车道偏离危险性评价指标计算的流程图；Fig. 3 is the flowchart of calculating the lane departure risk evaluation index;

图4为车道偏离危险性评价指标的示意图；4 is a schematic diagram of a lane departure risk evaluation index;

图5为横向距离评价指标及其特征值；Figure 5 shows the lateral distance evaluation index and its eigenvalues;

图6为纵向距离评价指标及其特征值；Figure 6 is the longitudinal distance evaluation index and its eigenvalues;

图7为Sigmoid函数及其参数确定的示意图；Fig. 7 is the schematic diagram that Sigmoid function and parameter thereof determine;

图8为本发明的一个优选实施例的流程图。Fig. 8 is a flowchart of a preferred embodiment of the present invention.

具体实施方式Detailed ways

实施例1：Example 1:

首先，在步骤S00中，系统从车载传感器中获得汽车在当前时刻t的参数：汽车的纵向速度横向速度

方向盘转角δ_sw，油门开度或者制动踏板行程比α(当汽车处于驱动工况时，α为正值或等于0，表示油门开度，此时0％≤α≤100％；当汽车处于制动工况时，α为负值，表示制动踏板行程比，此时-100％≤α＜0％)，以及车道标志线信息，并读取人机界面中驾驶员对系统参数的设定。由于上述的

以及车道标志线信息是在车体坐标系(以汽车质心为原点，x轴沿着汽车纵轴线向前，y轴向左，z轴垂直向上，参考图2)下的获得的，因此在同一坐标系下汽车在t时刻的纵向位置x₀，横向位置y₀和横摆角₀均为零。First, in step S00, the system obtains the parameters of the car at the current moment t from the on-board sensors: the longitudinal speed of the car lateral velocity

Steering wheel angle δ _sw , accelerator opening or brake pedal travel ratio α (when the car is in driving condition, α is a positive value or equal to 0, indicating the accelerator opening, at this time 0%≤α≤100%; when the car is in In the braking condition, α is a negative value, indicating the brake pedal stroke ratio, at this time -100%≤α<0%), and the lane marking information, and read the driver’s setting of the system parameters in the man-machine interface Certainly. Due to the above

And the lane marking line information is obtained under the car body coordinate system (with the center of mass of the car as the origin, the x-axis moves forward along the longitudinal axis of the car, the y-axis goes to the left, and the z-axis goes up vertically, refer to Figure 2), so in the same In the coordinate system, the car's longitudinal position x ₀ , lateral position y ₀ and yaw angle  ₀ at time t are all zero.

其次，在步骤S11中，将根据方向盘转角δ_sw和油门开度或者制动踏板的行程比α获得汽车在当前时刻t的纵向加速度稳态值

和横向加速度稳态值

Secondly, in step S11, the steady-state value of the longitudinal acceleration of the car at the current moment t will be obtained according to the steering wheel angle δ _sw and the accelerator opening or the stroke ratio α of the brake pedal

and lateral acceleration steady-state value

由于前视时间T_p(通常为1～2s，本实施例取1.5s)远大于汽车的瞬态反应时间，因此汽车在t时刻到t+T_p时刻的这段时间内按照稳态特性运动。虽然汽车的纵向和横向加速度是随着时间不断变化的，但由于前视时间较短，在t时刻到t+T_p时刻的这段时间内，可以把汽车的速度控制特性和方向控制特性用一阶线性参考模型近似，从而对应于当前时刻t的方向盘转角和油门开度或者制动踏板行程比，汽车都有一个横向加速度和纵向加速度与之对应。Since the look-ahead time _Tp (usually 1-2s, 1.5s in this embodiment) is much longer than the transient response time of the car, the car moves according to the steady-state characteristic during the period from time t to time t+T _p . Although the longitudinal and lateral accelerations of the car are constantly changing with time, due to the short look-ahead time, during the period from time t to t+T _p , the speed control characteristics and direction control characteristics of the car can be used as The first-order linear reference model is approximated, so that corresponding to the steering wheel angle and accelerator opening or brake pedal travel ratio at the current moment t, the car has a corresponding lateral acceleration and longitudinal acceleration.

利用上述简化方法，根据最小二乘辨识原理，可以建立不同车速、不同油门开度或者制动踏板行程比所对应的纵向加速度稳态增益G_ax，以及不同车速、不同方向盘转角对应的汽车侧向加速度稳态增益G_ay两个二维数表。根据这两个二维数表，通过二维插值的方法，可以确定对应于当前时刻油门开度或者制动踏板行程比α和方向盘转角δ_sw的汽车纵向加速度稳态增益G_ax以及侧向加速度稳态增益G_ay，从而纵向加速度和横向加速度的稳态值计算如下：Using the above simplified method, according to the principle of least squares identification, the longitudinal acceleration steady-state gain G _ax corresponding to different vehicle speeds, different accelerator openings or brake pedal stroke ratios, and the vehicle lateral acceleration corresponding to different vehicle speeds and different steering wheel angles can be established. Acceleration steady-state gain G _ay two two-dimensional number tables. According to these two two-dimensional number tables, through the method of two-dimensional interpolation, the vehicle longitudinal acceleration steady-state gain G _ax and lateral acceleration corresponding to the current moment accelerator opening or brake pedal travel ratio α and steering wheel angle δ _sw can be determined The steady-state gain G _ay , and thus the steady-state values of longitudinal acceleration and lateral acceleration are calculated as follows:

$\overset{\cdot \cdot \cdot &Center Dot;}{x x} = = {G G}_{ax ax} \cdot \cdot α α$

$\overset{\cdot \cdot \cdot \cdot}{y the y} = = {G G}_{ay ay} \cdot &Center Dot; {δ δ}_{sw sw}$

在步骤S12中，根据步骤S11得到的汽车纵向和横向加速度稳态值

和，以及步骤S00值获得的汽车状态和位置参数，通过模拟驾驶员的前视作用，对汽车在未来一段时间T_p(前视时间)内的预期行驶轨迹进行预估。In step S12, according to the vehicle longitudinal and lateral acceleration steady-state values obtained in step S11

and , and the car state and position parameters obtained from the value of step S00, by simulating the driver's forward-looking effect, the expected driving trajectory of the car in a certain period of time T _p (forward-looking time) in the future is estimated.

根据实际驾驶经验，驾驶员在驾驶过程中，通常会向前观察汽车前方一定范围内的道路状况，并根据自己对汽车动力学和运动学特性的认识和掌握，利用汽车的当前状态以及汽车的稳态响应特性来估计汽车的预期行驶轨迹，从而控制汽车的当前运动，即所谓的驾驶员的前视作用。驾驶员确定汽车未来要到达的位置后，就会给方向盘一定的转角，并给油门或者制动踏板一定输入。从汽车动力学的角度讲，方向盘转角引起汽车横向加速度的变化，而油门或者制动踏板的输入会引起汽车纵向加速度的变化。从汽车运动学的角度讲，汽车在未来t+T_p时刻所能到达的位置可以认为是由汽车当前运动状态和汽车的纵向、横向加速度决定的。因此，可以通过模拟驾驶员的前视作用，根据当前时刻t时汽车的状态(

x₀，y₀，₀)预测未来时刻t+T_p汽车状态(

x_J，y_J，_J)，具体方法如下：According to the actual driving experience, the driver usually observes the road conditions in front of the car within a certain range during driving, and uses the current state of the car and the Steady-state response characteristics are used to estimate the expected driving trajectory of the car, so as to control the current movement of the car, which is the so-called forward-looking effect of the driver. After the driver determines the position the car will reach in the future, he will give the steering wheel a certain rotation angle and give a certain input to the accelerator or brake pedal. From the perspective of vehicle dynamics, the steering wheel angle causes changes in the vehicle's lateral acceleration, while the input of the accelerator or brake pedal causes changes in the vehicle's longitudinal acceleration. From the perspective of car kinematics, the position that the car can reach at time t+T _p in the future can be considered to be determined by the current state of motion of the car and the longitudinal and lateral acceleration of the car. Therefore, by simulating the forward-looking effect of the driver, according to the state of the car at the current time t (

x ₀ , y ₀ ,  ₀ ) to predict the future time t+T _p car state (

x _J , y _J ,  _J ), the specific method is as follows:

将前视时间T_p划分成均等的离散时间片，在每个时间片内，由于时间很短，使用刚体运动学原理计算车辆状态，然后逐步累加获得t+T_p时刻车辆的状态。由于车辆横摆运动的影响，相邻两个时刻之间车辆的航向角发生了变化，因此车体坐标系在不同时间片是不同的。但是汽车状态都是在前一时刻的车体坐标系下进行计算的，因此必须进行坐标变换，将其从不同车体坐标系转换到初始坐标系下才能累加，获得该时刻的车辆状态。在计算t+T_p时刻车辆状态的过程中始终以t时刻车辆坐标系为基准，参考图2，从坐标系(x_j，y_j)到坐标系(x₀，y₀)的转换矩阵为：The forward-looking time T _p is divided into equal discrete time slices. In each time slice, due to the short time, the vehicle state is calculated using the principle of rigid body kinematics, and then gradually accumulated to obtain the state of the vehicle at time t+T _p . Due to the influence of the yaw motion of the vehicle, the heading angle  of the vehicle changes between two adjacent moments, so the vehicle body coordinate system is different in different time slices. However, the car state is calculated in the car body coordinate system at the previous moment, so coordinate transformation must be carried out to convert it from different car body coordinate systems to the initial coordinate system to accumulate and obtain the vehicle state at that moment. In the process of calculating the vehicle state at time t+T _p , the vehicle coordinate system at time t is always taken as the reference. Referring to Figure 2, the transformation matrix from the coordinate system (x _j , y _j ) to the coordinate system (x ₀ , y ₀ ) is :

从而汽车在预期轨迹点P_J的状态由下列计算过程获得：Therefore, the state of the car at the expected trajectory point P _J is obtained by the following calculation process:

$[\begin{matrix} {\overset{\cdot &Center Dot; \cdot &Center Dot;}{x x}}_{j j} \\ {\overset{\cdot \cdot \cdot &Center Dot;}{y the y}}_{j j} \end{matrix}] = = [\begin{matrix} {\overset{\cdot &Center Dot; \cdot \cdot}{x x}}_{00} \\ {\overset{\cdot &Center Dot; \cdot \cdot}{y the y}}_{00} \end{matrix}]$

$[\begin{matrix} {\overset{\cdot \cdot}{x x}}_{j j} \\ {\overset{\cdot &Center Dot;}{y the y}}_{j j} \end{matrix}] = = [\begin{matrix} {\overset{\cdot \cdot}{x x}}_{j j - - 11} \\ {\overset{\cdot &Center Dot;}{y the y}}_{j j - - 11} \end{matrix}] + + [\begin{matrix} {\overset{\cdot \cdot \cdot \cdot}{x x}}_{j j} \\ {\overset{\cdot &Center Dot; \cdot &Center Dot;}{y the y}}_{j j} \end{matrix}] \times \times Δ Δ {t t}_{p p}$

$[\begin{matrix} {x x}_{j j} \\ {y the y}_{j j} \end{matrix}] = = [\begin{matrix} {x x}_{j j - - 11} \\ {y the y}_{j j - - 11} \end{matrix}] + + {A A}_{j j - - 1,0 1,0} (\begin{matrix} [\begin{matrix} {\overset{\cdot \cdot}{x x}}_{j j - - 11} \\ {\overset{\cdot \cdot}{y the y}}_{j j - - 11} \end{matrix}] \times \times Δ Δ {t t}_{p p} + + \frac{11}{22} \end{matrix} \times \times [\begin{matrix} {\overset{\cdot \cdot \cdot &Center Dot;}{x x}}_{j j} \\ {\overset{\cdot &Center Dot; \cdot \cdot}{y the y}}_{j j} \end{matrix}] \times \times Δ Δ {t t}_{}^{p p})$

式中j＝1，2，…，J，(J为等分时间片的个数)。其中x₀，y₀表示初始汽车位置。

表示初始纵向和横向速度。表示初始纵向和横向加速度的稳态值，₀是初始车辆航向角，Δt_p＝T_p/J，T_p为前视时间。In the formula, j=1, 2, ..., J, (J is the number of equally divided time slices). Where x ₀ , y ₀ represent the initial car position.

Indicates the initial longitudinal and lateral velocities. Indicates the steady-state value of the initial longitudinal and lateral acceleration,  ₀ is the initial vehicle heading angle, Δt _p =T _p /J, and T _p is the look-ahead time.

在步骤S13中，根据步骤S12得到的汽车在t+T_p时刻的质心位置(x_j，y_j)将汽车简化成具有四个角点的矩形，并确定汽车四个角点的位置坐标。参考图2、图4、图5、图6所示。在t时刻汽车四个角点的位置坐标分别为：In step S13, according to the centroid position (x _j , y _j ) of the car at time t+T _p obtained in step S12, the car is simplified into a rectangle with four corners, and the position coordinates of the four corners of the car are determined. Refer to Figure 2, Figure 4, Figure 5, and Figure 6. The position coordinates of the four corners of the car at time t are:

corner[0]：(f，w/2)；corner[0]: (f,w/2);

corner[1]：(-r，w/2)；corner[1]: (-r,w/2);

corner[2]：(-r，-w/2)；corner[2]: (-r,-w/2);

corner[3]：(f，-w/2)。corner[3]: (f, -w/2).

其中，f为汽车车头到质心的长度，r为车尾到质心的长度，w为汽车宽度。则在t+T_p时刻汽车四个角点坐标为：Among them, f is the length from the front of the car to the center of mass, r is the length from the rear of the car to the center of mass, and w is the width of the car. Then the coordinates of the four corners of the car at time t+T _p are:

corner[i]_Tp＝A*corner[i]^T+(x_j，y_j)^T corner[i] _Tp = A*corner[i] ^T + (x _j , y _j ) ^T

其中i＝0～3，分别代表四个角点，corner[i]_Tp为四个角点在t+T_p时刻的坐标向量。A为旋转矩阵，_j为t+T_p时刻汽车的横摆角。Wherein, i=0~3 represent four corner points respectively, and corner[i] _Tp is the coordinate vector of the four corner points at time t+T _p . A is the rotation matrix, and  _j is the yaw angle of the car at time t+T _p .

在步骤S20中，根据t+T_p时刻汽车的状态和四个角点位置坐标，以及车道标志线信息，计算车道偏离危险性评价的三个评价指标。In step S20, according to the state of the vehicle at time t+ _Tp , the position coordinates of the four corners, and the lane marking information, three evaluation indexes for lane departure risk evaluation are calculated.

在驾驶过程中，驾驶员判断汽车是否会偏离其行驶车道，是从横向和纵向两个方面进行的。在横向方面，驾驶员通常通过汽车至左侧以及右侧道路的横向相对距离进行判断评价，确保正常行驶时汽车与左侧和右侧车道标志线保持一定的横向距离，从而保证汽车不会从侧面偏离其行驶车道；在纵向方面，驾驶员会通过汽车与前方道路的纵向相对距离进行判断评价，确保正常行驶时汽车与前方道路保持一定的纵向距离，从而保证汽车不会从前方偏离其行驶车道。During the driving process, the driver judges whether the car will deviate from its driving lane, which is carried out from both horizontal and vertical aspects. In terms of lateral direction, the driver usually judges and evaluates the relative distance from the car to the left and right side of the road to ensure that the car maintains a certain lateral distance from the left and right lane markings during normal driving, so as to ensure that the car does not fall from the road. The side deviates from its driving lane; in the longitudinal aspect, the driver will judge and evaluate the longitudinal relative distance between the car and the road ahead to ensure that the car maintains a certain longitudinal distance from the road ahead during normal driving, so as to ensure that the car will not deviate from its driving direction from the front Lane.

因此，本发明建立了基于横向和纵向距离的偏离危险性评价指标，包括：车辆至左侧车道标志线距离的偏离危险性评价指标IDS1、车辆至右侧车道标志线距离的偏离危险性评价指标IDS2，以及车辆至前方车道标志线距离的偏离危险性评价指标IDS3三个基本的评价指标，并采用单极性Sigmoid函数将评价指标表征为0～1之间的数值，参考图3所示。Therefore, the present invention establishes a deviation risk evaluation index based on lateral and longitudinal distances, including: the deviation risk evaluation index IDS1 of the distance from the vehicle to the left lane marking line, the deviation risk evaluation index of the distance from the vehicle to the right lane marking line IDS2, and the deviation risk evaluation index IDS3 of the distance from the vehicle to the front lane marking line are three basic evaluation indicators, and the evaluation index is represented as a value between 0 and 1 by using a unipolar Sigmoid function, as shown in Figure 3.

${r r}_{IDS IDS 11} = = \frac{11}{11 + + {e e}^{- - a a 11 (({x x}_{si the si 11} - - c c 11))}}$

${r r}_{IDS IDS 22} = = \frac{11}{11 + + {e e}^{- - a a 22 (({x x}_{si the si 22} - - c c 22))}}$

${r r}_{IDS IDS 33} = = \frac{11}{11 + + {e e}^{- - a a 33 (({x x}_{si the si 33} - - c c 33))}}$

其中r_IDS1，r_IDS2，r_IDS3为三个评价指标值；x_si1，x_si2，x_si3为三个评价指标的特征值，分别表征了汽车与左侧、右侧和前方车道标志线的距离关系。a1，c1，a2，c2为与横向安全距离S_a有关的常数；a3，c3为与纵向安全距离S_f有关的常数。Among them, r _IDS1 , r _IDS2 , and r _IDS3 are the three evaluation index values; x _si1 , x _si2 , and x _si3 are the characteristic values of the three evaluation indexes, which respectively represent the distance between the car and the left, right and front lane marking lines relation. a1, c1, a2, c2 are constants related to the lateral safety distance S _a ; a3, c3 are constants related to the longitudinal safety distance S _f .

三个评价指标的计算可以通过以下四个步骤完成，参考附图3：The calculation of the three evaluation indicators can be completed through the following four steps, refer to Figure 3:

在步骤S21中，根据步骤S13获得汽车角点坐标以及步骤S00中获得的车道标志线信息，参考附图4，计算汽车角点3-0连线与车道标志线的交点A的坐标、角点2-1连线与车道标志线的交点B的坐标、角点0-3连线与车道标志线的交点C的坐标、角点1-2连线与车道标志线的交点D的坐标、角点1-0连线与车道标志线的交点E的坐标，以及角点2-3连线与车道标志线的交点F的坐标。In step S21, obtain the vehicle corner point coordinates and the lane marking line information obtained in step S00 according to step S13, with reference to accompanying drawing 4, calculate the coordinates of the intersection A of automobile corner point 3-0 connecting line and lane marking line, corner point The coordinates of the intersection point B of the line 2-1 and the lane marking line, the coordinates of the intersection point C of the corner point 0-3 line and the lane marking line, the coordinates of the intersection point D of the corner point 1-2 line and the lane marking line, and the angle The coordinates of the intersection point E of the line connecting point 1-0 and the lane marking line, and the coordinates of the intersection point F of the line connecting point 2-3 and the lane marking line.

在步骤S22中，计算汽车角点0与交点A之间的距离l0，角点1与交点B之间的距离l₁，角点3与交点C之间的距离l₂，角点2与交点D之间的距离l₃，角点0与交点E之间的距离l₄，以及角点3与交点F之间的距离l₅。In step S22, calculate the distance l0 between corner 0 and intersection A of the car, the distance l 1 between corner 1 and intersection B, the distance l ₂ between corner 3 and intersection C, and the distance l ₂ between corner 2 and intersection The distance l ₃ between D, the distance l ₄ between corner 0 and intersection E, and the distance l ₅ between corner 3 and intersection F.

在步骤S23中，取距离l₀和l₁中的最小值作为评价指标IDS1的特征值x_si1，距离l₂和l₃中的最小值作为评价指标IDS2的特征值x_si2，距离l₄和l₅中的最小值作为评价指标IDS3的特征值x_si3，即In step S23, the minimum value of the distance l ₀ and l ₁ is taken as the characteristic value x _si1 of the evaluation index IDS1, the minimum value of the distance l ₂ and l ₃ is taken as the characteristic value x _si2 of the evaluation index IDS2, the distance l ₄ and The minimum value among l ₅ is used as the characteristic value x _si3 of the evaluation index IDS3, namely

x_si1＝min{l₀，l₁}，x _si1 = min{l ₀ , l ₁ },

x_si2＝min{l₂，l₃}，x _si2 = min{l ₂ , l ₃ },

x_si3＝min{l₄，l₅}。x _si3 = min{l ₄ , l ₅ }.

在步骤S24中，根据车道偏离危险性评价指标计算公式以及步骤S23获得的三个评价指标的特征值x_si1，x_si2，x_si3，计算三个评价指标的值r_IDS1，r_IDS2以及r_IDS3。In step S24, calculate the values _{rIDS1, rIDS2 and rIDS3 of the three evaluation indexes according to the calculation formula of the lane departure risk evaluation index and the characteristic values xsi1} _{, xsi2, and xsi3} _of _the _three evaluation indexes obtained _in step S23 .

车道偏离危险性评价指标计算公式中的常数a1，c1，a2，c2，a3，c3进一步确定如下：参考图7所示，由于单极性Sigmoid函数具有非线性增益的功能，即中间部分为高增益区，两边部分时低增益区。在确定隶属度函数的系数时，主要是根据高增益区的自变量以及对应的值域范围求取系数。参考图6所示，以评价指标IDS1为例，它是根据估计获得的预期轨迹点与道路左侧车道标志线的相对位置建立的安全性指标，其特征值为汽车左侧的两个角点到道路左侧边界的最小距离。通常情况下，特征值小于0将是极度危险的，这就意味着汽车已经偏离车道线，因此低增益区的边界取为0。由此可以通过计算获得各个系数如下：The constants a1, c1, a2, c2, a3, and c3 in the calculation formula of the lane departure risk evaluation index are further determined as follows: Referring to Figure 7, since the unipolar Sigmoid function has a nonlinear gain function, that is, the middle part is high Gain area, low gain area on both sides. When determining the coefficients of the membership function, the coefficients are mainly obtained according to the independent variables in the high-gain region and the corresponding value ranges. Referring to Figure 6, taking the evaluation index IDS1 as an example, it is a safety index established based on the relative position of the estimated expected trajectory point and the lane marking line on the left side of the road, and its characteristic values are the two corner points on the left side of the car The minimum distance to the left border of the road. Usually, an eigenvalue less than 0 is extremely dangerous, which means that the car has deviated from the lane line, so the boundary of the low gain area is taken as 0. From this, the various coefficients can be obtained by calculation as follows:

$a a 11 = = a a 22 = = \frac{{S S}_{a a}}{22},, {a a}_{33} = = \frac{{S S}_{f f}}{22}$

c1＝c2＝S_a，c3＝S_f c1=c2=S _a , c3=S _f

如前所述，常数a1，c1，a2，c2，a3，c3与横向安全距离S_a和纵向安全距离S_f有关，因此道偏离危险性评价指标中的横、纵向安全距离进一步确定如下：由于安全距离的取值直接影响评价方法的可靠性和稳健性，如果安全距离取为固定值，则很难适应不同的车辆行驶工况的要求。很显然，汽车行驶速度较低时，安全性较好，其偏离危险性也较低，因此安全距离可以取较小值；而行驶速度较高时，汽车的安全性变差，稍有不慎就可能发生车道偏离事故，从而对生命和财产安全造成威胁，因此安全距离应该取较大值。也就是说，安全距离应该与车速保持一定的关系，随着的车速的不同而变化。这样才能保证在各种车速下都能准确判断汽车是否会偏离其行驶车道，既不会误报警(在不会发生车道偏离的情况下发出报警)，也不会遗漏报警(在即将发生车道偏离时没有报警)。根据试验结果，得到安全距离如下：As mentioned above, the constants a1, c1, a2, c2, a3, c3 are related to the transverse safety distance S _a and the longitudinal safety distance S _f , so the transverse and longitudinal safety distances in the evaluation index of road deviation risk are further determined as follows: The value of the safety distance directly affects the reliability and robustness of the evaluation method. If the safety distance is taken as a fixed value, it is difficult to adapt to the requirements of different vehicle driving conditions. Obviously, when the driving speed of the car is low, the safety is better, and the risk of deviation is also low, so the safety distance can be taken as a smaller value; when the driving speed is high, the safety of the car becomes worse, and a little carelessness Lane departure accidents may occur, thereby threatening life and property safety, so the safety distance should be taken as a larger value. That is to say, the safety distance should maintain a certain relationship with the speed of the vehicle, and change with the speed of the vehicle. Only in this way can it be ensured that whether the car will deviate from its driving lane can be accurately judged at various speeds, and there will be neither false alarms (alarms are issued when lane departures will not occur), nor missed alarms (alarms are issued when lane departures are about to occur). no alarm was reported). According to the test results, the safe distance is obtained as follows:

横向安全距离：S_a＝0.011758+0.002932v+0.000019v²；Lateral safety distance: S _a =0.011758+0.002932v+0.000019v ² ;

纵向安全距离：S_f＝0.0034v+0.0045v²。Longitudinal safety distance: S _f =0.0034v+0.0045v ² .

其中，S_a为横向安全距离，S_f为纵向安全距离，v为汽车在t时刻的速度(km/s)。Among them, S _a is the lateral safety distance, S _f is the longitudinal safety distance, and v is the speed of the car at time t (km/s).

在步骤S30中，根据步骤S20获得的车道偏离危险性评价的三个评价指标值，判断汽车是否即将发生车道偏离事故。In step S30, according to the three evaluation index values of the lane departure risk evaluation obtained in step S20, it is judged whether the vehicle is about to have a lane departure accident.

如前所述，将安全阈值设定为三个级别：安全，危险，特别危险。当IDS1、IDS2和IDS3中任何一个评价指标都大于安全阈值TH1(0.6)时，认为汽车不会发生车道偏离事故，即处于安全级别，从而不报警；当所有的评价指标都大于安全阈值TH2(0.4)并且至少有一个评价指标处于TH1和TH2之间时，认为汽车处于危险的级别，此时发出较为舒缓的报警声音，并点亮黄色警示灯；而三个评价指标中只要任何一个小于安全阈值TH2，则认为汽车处于特别危险的级别，此时发出非常急促的报警声音，并点亮红色警示灯。As mentioned earlier, set the safety threshold to three levels: Safe, Dangerous, and Extremely Dangerous. When any evaluation index in IDS1, IDS2 and IDS3 is greater than the safety threshold TH1 (0.6), it is considered that the car will not have a lane departure accident, that is, it is at a safe level, so no alarm is given; when all evaluation indexes are greater than the safety threshold TH2 ( 0.4) and at least one of the evaluation indicators is between TH1 and TH2, it is considered that the car is at a dangerous level. At this time, a relatively soothing alarm sound is issued, and the yellow warning light is lit; and as long as any one of the three evaluation indicators is less than the safe level Threshold TH2, it is considered that the car is in a particularly dangerous level, at this time a very rapid alarm sound is issued, and the red warning light is turned on.

最后，在步骤S40中，根据评价结果通过声音报警器和车载LCD警示灯箱驾驶员发出相应的声音和灯光信号，实现车道偏离预警功能。Finally, in step S40, according to the evaluation result, the driver sends corresponding sound and light signals through the sound alarm and the vehicle-mounted LCD warning light box to realize the lane departure warning function.

实施例2：Example 2:

如图8流程图所示：As shown in the flowchart in Figure 8:

首先，利用图像传感器采集道路图像，对其进行处理，获得前方道路的车道标志线信息。同时，通过车载传感器采集车辆的状态参数，如速度，换道信号，方向盘转角，油门踏板或者制动踏板的位置等，获得必要的输入信息。First, use the image sensor to collect road images, process them, and obtain the lane marking information of the road ahead. At the same time, the state parameters of the vehicle, such as speed, lane change signal, steering wheel angle, position of the accelerator pedal or brake pedal, etc., are collected through the on-board sensors to obtain the necessary input information.

然后，根据换道信号，判断汽车是否正在换道，如果汽车正在换道，系统停止工作；如果汽车没有处于换道过程，系统会根据上述状态参数，利用汽车的稳态响应特性对汽车的预期行驶轨迹进行预估，获得预期轨迹点的汽车状态参数，并计算汽车四个角点的位置坐标。Then, according to the lane-changing signal, it is judged whether the car is changing lanes. If the car is changing lanes, the system will stop working; if the car is not in the process of changing lanes, the system will use the steady-state response characteristics of the car to estimate The driving trajectory is estimated, the vehicle state parameters of the expected trajectory points are obtained, and the position coordinates of the four corners of the vehicle are calculated.

其次，根据汽车预期轨迹点处汽车角点的位置坐标，以及所获得的车道标志线信息，计算三个偏离危险性评价指标值。Secondly, according to the position coordinates of the car corners at the expected track point of the car and the obtained lane marking information, calculate the three departure risk evaluation index values.

最后根据所获得的评价指标值以及设定的安全阈值，判断汽车的偏离危险性级别，并据此发出相应的声音报警信号和LCD警示灯信号，提醒驾驶员进行矫正操作，避免汽车偏离其行驶车道。Finally, according to the obtained evaluation index value and the set safety threshold, the vehicle’s deviation risk level is judged, and the corresponding sound alarm signal and LCD warning light signal are issued accordingly to remind the driver to perform corrective operations to prevent the vehicle from deviating from its driving position. Lane.

Claims

1. A vehicle trajectory estimation and lane departure evaluation method, characterized in that it comprises the following steps:

Information perception step: provide the vehicle’s current motion state and driving environment information through the on-board sensor, where the driving environment information refers to the lane marking information of the road ahead obtained from the road image, and the vehicle motion state information includes the vehicle’s longitudinal speed

lateral velocity

Steering wheel angle δ _sw , accelerator opening or brake pedal stroke ratio α;

Trajectory estimation step: By simulating the forward-looking effect of the driver, the vehicle's steady-state response characteristics are used to estimate the expected driving trajectory of the car in a certain period of time in the future;

Lane departure risk assessment steps: According to the expected driving trajectory of the car and the lane markings of the road ahead, establish the lateral safety distance from the car to the left road and the right road at the expected trajectory point and the longitudinal safety distance from the car to the road ahead. Lane departure risk evaluation index, through these three indicators to analyze the departure risk of the expected track point.

2. A vehicle trajectory estimation and evaluation method according to claim 1, characterized in that the trajectory estimation step adopts the following method:

Step S11, according to the steering wheel angle δ _sw and the accelerator opening or brake pedal travel ratio α, obtain the steady-state value of the longitudinal acceleration of the car at the current moment t and lateral acceleration steady-state value

Step S12, dividing the look-ahead time T _p into equal discrete time slices, using the principle of rigid body kinematics to calculate the state of the vehicle, and then gradually accumulating to obtain the state of the vehicle at time t+T _p ;

Step S13, according to the centroid position (x _J , y _J ) of the car at time t+T _p obtained in step S12, simplify the car into a rectangle with four corners, and determine the position coordinates of the four corners of the car.

3. A vehicle trajectory estimation and evaluation method according to claim 2, characterized in that the steady-state value of the longitudinal acceleration at the previous moment t

and lateral acceleration steady-state value Obtained by the following method:

First, establish two two-dimensional numerical tables of longitudinal acceleration steady-state gain G _ax corresponding to different vehicle speeds, different accelerator openings or brake pedal travel ratios, and vehicle lateral acceleration steady-state gains G _ay corresponding to different vehicle speeds and different steering wheel angles ;

Then, through the two-dimensional interpolation method, determine the vehicle longitudinal acceleration steady-state gain G _ax and lateral acceleration steady-state gain _Gay corresponding to the current moment accelerator opening or brake pedal travel ratio α and steering wheel angle δ _sw , thereby determining the longitudinal The steady-state values of acceleration and lateral acceleration, calculated as follows:

\overset{\cdot \cdot \cdot &Center Dot;}{x x} = = {G G}_{ax ax} \cdot &Center Dot; α α

\overset{\cdot \cdot \cdot &Center Dot;}{y the y} = = {G G}_{ay ay} {\cdot &Center Dot; δ δ}_{sw sw} . .

4. A vehicle trajectory estimation and evaluation method according to claim 2, characterized in that said

The state of the vehicle at any time is obtained by the following method:

[\begin{matrix} {\overset{\cdot &Center Dot; \cdot &Center Dot;}{x x}}_{j j} \\ {\overset{\cdot &Center Dot; \cdot \cdot}{y the y}}_{j j} \end{matrix}] = = \begin{matrix} [\begin{matrix} {\overset{\cdot &Center Dot; \cdot &Center Dot;}{x x}}_{00} \\ {\overset{\cdot \cdot \cdot &Center Dot;}{y the y}}_{00} \end{matrix}] \end{matrix}

[\begin{matrix} {\overset{\cdot &Center Dot;}{x x}}_{j j} \\ {\overset{\cdot &Center Dot;}{y the y}}_{j j} \end{matrix}] = = \begin{matrix} [\begin{matrix} {\overset{\cdot \cdot}{x x}}_{j j - - 11} \\ {\overset{\cdot \cdot}{y the y}}_{j j - - 11} \end{matrix}] + + [\begin{matrix} {\overset{\cdot \cdot \cdot &Center Dot;}{x x}}_{j j} \\ {\overset{\cdot &Center Dot; \cdot &Center Dot;}{y the y}}_{j j} \end{matrix}] \times \times \end{matrix} {Δt Δt}_{p p}

[\begin{matrix} {x x}_{j j} \\ {y the y}_{j j} \end{matrix}] = = [\begin{matrix} {x x}_{j j - - 11} \\ {y the y}_{j j - - 11} \end{matrix}] + + {A A}_{j j - - 1,0 1,0} (([\begin{matrix} {\overset{\cdot &Center Dot;}{x x}}_{j j - - 11} \\ {\overset{\cdot \cdot}{y the y}}_{j j - - 11} \end{matrix}] \times \times Δ Δ {t t}_{p p} + + \frac{11}{22} \times \times [\begin{matrix} {\overset{\cdot &Center Dot; \cdot &Center Dot;}{x x}}_{j j} \\ {\overset{\cdot \cdot \cdot \cdot}{y the y}}_{j j} \end{matrix}] \times \times {Δt Δt}_{p p}^{22}))

In the formula, j=1, 2, ..., J, (J is the number of equally divided time slices), where x ₀ and y ₀ represent the initial vehicle position;

denote the initial longitudinal and lateral velocities;

Indicates the steady-state value of the initial longitudinal and lateral acceleration,  ₀ is the initial vehicle heading angle, Δt _p =T _p //J, T _p is the look-ahead time, is the transformation matrix from the coordinate system (x _j , y _j ) to the coordinate system (x ₀ , y ₀ ).

5. A vehicle trajectory estimation and evaluation method according to claim 2, characterized in that the position coordinates of the four corners of the vehicle are obtained by the following method:

corner[i] _Tp ＝A*corner[i] ^T +(x _j , y _j ) ^T , where corner[i]T _p is the coordinate vector of the four corner points at time t+T _p ,

is the rotation matrix, i=0~3, representing the four corner points respectively,  _j is the yaw angle of the car at the time t+T _p .

6. A vehicle trajectory estimation and lane departure evaluation method according to claim 1, characterized in that the lane departure risk assessment adopts the following method:

Step S20, establish the deviation risk evaluation index based on the horizontal and vertical distances, including: the deviation risk evaluation index IDS1 of the distance from the vehicle to the left lane marking line, the deviation risk evaluation index IDS2 of the distance from the vehicle to the right lane marking line, And the three basic evaluation indicators of the deviation risk evaluation index IDS3 of the distance from the vehicle to the lane marking line in front, and use the unipolar Sigmoid function to characterize the evaluation index as a value between 0 and 1,

{r r}_{IDS IDS 11} = = \frac{11}{11 + + {e e}^{- - al al (({x x}_{sil sil} - - c c 11))}};;

{r r}_{IDS IDS 22} = = \frac{11}{11 + + {e e}^{- - a a 22 (({x x}_{si the si 22} - - c c 22))}};;

{r r}_{IDS IDS 33} = = \frac{11}{11 + + {e e}^{- - a a 33 (({x x}_{si the si 33} - - c c 33))}};;

Among them, x _si1 , x _si2 , and x _si3 are the characteristic values of the three evaluation indicators, which are determined by the following method:

Step S21, calculating the coordinates of the intersection of the car corner line and the lane marking line;

Step S22, calculating the distance between the car corner and the intersection;

Step S23, respectively taking the minimum value in the distance as the feature value of the evaluation index;

a1, c1, a2, c2; a3, c3 are constants, determined by the following method:

a a 11 = = a a 22 = = \frac{{S S}_{a a}}{22},,

{a a}_{33} = = \frac{{S S}_{f f}}{22}

c1=c2=S _a , c3=S _f

Among them, S _a is the lateral safety distance, S _f is the longitudinal safety distance, which is further determined as:

S _a =0.011758+0.002932v+0.000019v ² ;

S _f ＝0.0034v+0.0045v ² , where v is the speed of the car at time t;

Step S30, according to the three evaluation index values of the lane departure risk evaluation obtained in step S20, judge whether the car is about to have a lane departure accident, and when any one of the evaluation indexes in IDS1, IDS2 and IDS3 is greater than the threshold value, it is considered that the car will not happen A lane departure accident means that it is at a safe level; and as long as any one of the three evaluation indicators is less than the threshold, the car is considered to be in a dangerous state.

7. A vehicle trajectory estimation and evaluation method according to claim 6, characterized in that the safety threshold can be set according to different types of lane departure warning modes.

8. A vehicle trajectory estimation and evaluation method according to claims 1 to 7, characterized in that if the risk of departure exceeds a certain threshold, a lane departure warning message will be sent to the driver through sound and on-board LCD display.

9. A vehicle trajectory estimation and evaluation method according to claims 1 to 7, characterized in that before the trajectory estimation step, it is judged whether the vehicle is changing lanes according to the lane changing signal, and if the vehicle is changing lanes, the trajectory prediction is stopped. estimate work.

10. A vehicle trajectory estimation and evaluation method according to claim 8, characterized in that before the trajectory estimation step, it is judged whether the vehicle is changing lanes according to the lane change signal, and if the vehicle is changing lanes, the trajectory estimation work is stopped .