CN103699717A

CN103699717A - Complex road automobile traveling track predication method based on foresight cross section point selection

Info

Publication number: CN103699717A
Application number: CN201310632659.4A
Authority: CN
Inventors: 徐进; 邵毅明; 杨奎; 罗庆; 毛嘉川
Original assignee: Chongqing Jiaotong University
Current assignee: China Railway Eryuan Engineering Group Co Ltd CREEC; Chongqing Jiaotong University
Priority date: 2013-12-03
Filing date: 2013-12-03
Publication date: 2014-04-02

Abstract

The invention discloses a complex road vehicle trajectory prediction method based on forward-looking cross-section selection points, which analyzes and simulates the driver's trajectory decision-making behavior; abstracts and evolves the calculation strategy for the driver's trajectory selection behavior, that is, "previous According to the calculation strategy of "choice point"; the motives behind five typical driving modes are studied and expressed mathematically, and finally a vehicle trajectory decision-making method that can adapt to complex roads with any mileage length and can simulate typical driving modes is obtained. By applying the technology of the present invention, the desired trajectories of the following five driving modes can be obtained for mountain roads and complex race tracks, namely the mode with the shortest track length, the mode with the smallest curvature of the track (track mode), and the mode with the smallest rate of curvature change (the most comfortable driving mode). mode), trajectory centering mode (centering driving mode) and mixed mode, where the mixed mode is a combination of the first four typical modes.

Description

Prediction method of vehicle trajectory on complex roads based on point selection of forward-looking section

技术领域technical field

本发明涉及的是一种基于前视断面选点的复杂道路汽车行驶轨迹预测方法。The invention relates to a method for predicting vehicle running tracks on complex roads based on forward-looking cross-section point selection.

背景技术Background technique

在开展“车辆-驾驶人-道路”系统的闭环行驶仿真时，或是在汽车自动驾驶时，需要在路面可行驶范围内事先决策出一条期望轨迹(目标轨迹)，以供车辆在行驶过程中跟随，从而实现车辆的自动行进。When carrying out the closed-loop driving simulation of the "vehicle-driver-road" system, or when the car is driving automatically, it is necessary to determine a desired trajectory (target trajectory) in advance within the driving range of the road surface for the vehicle to drive during driving. Follow, so as to realize the automatic driving of the vehicle.

现有的技术手段通常是把道路中线或是行车道中线作为期望轨迹，即假设驾驶人采用居中行驶的驾驶模式(方向控制模式)。但在实际公路行驶中，驾驶人是在可使用路面宽度内自由选择行驶轨迹，以双车道公路为例，可使用路面宽度除了包含一个车道之外，还包括右侧的硬路肩，和一部分左侧的对向行车道。因此与车身宽度相比，可使用路面宽度有很大的盈余，特别是对于小客车来讲更是如此。因此，汽车驾驶员在轨迹选择和速度选择上有很大的自由，比如在山区公路的曲线路段能观察到多种类型的方向控制习惯：切弯(进弯时从曲线外侧切向内侧，出弯时再切回外侧)、行车道居中、曲线外侧行驶、曲线内侧行驶以及侵占路肩，等等。为此，在开展山区公路或是赛道的行驶仿真时，需要针对以上驾驶模式开展轨迹的预测(决策)，以获得目标轨迹。Existing technical means usually take the centerline of the road or the centerline of the traffic lane as the expected trajectory, that is, it is assumed that the driver adopts a driving mode (direction control mode) driving in the middle. However, in actual road driving, the driver is free to choose the driving trajectory within the available road width. Taking a two-lane road as an example, the available road width includes not only one lane, but also a hard shoulder on the right side, and a part of the left side. side opposite traffic lane. The usable road width therefore has a large surplus compared to the body width, especially for small passenger cars. Therefore, car drivers have a lot of freedom in trajectory selection and speed selection. For example, in the curved section of mountainous roads, various types of directional control habits can be observed: cutting curves (cutting from the outside of the curve to the inside when entering the curve, exiting the curve, etc.) turning back to the outside), centering the carriageway, driving on the outside of the curve, driving on the inside of the curve, encroaching on the shoulder, etc. For this reason, when carrying out driving simulation on mountainous roads or race tracks, it is necessary to carry out trajectory prediction (decision-making) for the above driving modes to obtain the target trajectory.

根据研究手段的不同，可以将现有的轨迹预测技术分为三类，分别是基于模糊规则的预测技术、基于数学优化的预测技术、以及基于多项式回归方法的预测技术。现有轨迹预测(决策)技术的缺点如下：According to the different research methods, the existing trajectory prediction technology can be divided into three categories, which are prediction technology based on fuzzy rules, prediction technology based on mathematical optimization, and prediction technology based on polynomial regression method. The disadvantages of existing trajectory prediction (decision-making) techniques are as follows:

参考图1，在使用模糊规则手段来开展轨迹预测的技术中，Lauffanburgar的预测方法具有代表性，他通过观测实际的弯道行驶轨迹，得到了轨迹与道路边界之间在弯道进口、曲中以及出口等特征位置的侧向距离，然后以弯道半径为变量建立了6个隶属度子集，用以确定与给定弯道曲度对应的轨迹控制参数d1～d3，再用4次polar polynomials样条进行逼近来得到连续期望轨迹。缺点是d1～d3对路宽具有明显的依赖性，当路宽发生大幅度的改变时，轨迹会越出路面边界或是过于居中。并且，由于研究对象是偏爱内切的驾驶人，该方法不能满足对驾驶行为多样化的刻画要求。Referring to Fig. 1, in the technology of using fuzzy rules to carry out trajectory prediction, Lauffanburgar's prediction method is representative. By observing the actual curve driving trajectory, he obtained the distance between the trajectory and the road boundary at the entrance of the curve and in the middle of the curve. and the lateral distance of characteristic positions such as exits, and then 6 membership degree subsets are established with the curve radius as a variable to determine the trajectory control parameters d1~d3 corresponding to the given curve curvature, and then use 4 polar Polynomials splines are approximated to obtain continuous desired trajectories. The disadvantage is that d1~d3 have obvious dependence on the road width. When the road width changes greatly, the trajectory will go out of the road boundary or be too centered. Moreover, since the research object is a driver who prefers introversion, this method cannot meet the requirements for describing the diversification of driving behavior.

高振海和管欣等使用粒子群优化算法预测汽车行驶轨迹，他们的方法在稳定性、收敛性、全局优化能力等方面较差，仅能对几十米长度的道路开展轨迹决策，超过这个长度则无法得到稳定有效的计算结果，因此不能满足长距离行驶仿真的要求。并且，在适用场合方面，他们的研究主要是针对车辆换道和城市道路跟车行驶场合，而非复杂线形条件下的公路行驶轨迹决策。Gao Zhenhai and Guan Xin used the particle swarm optimization algorithm to predict the driving trajectory of the car. Their method is poor in terms of stability, convergence, and global optimization capabilities. It can only make trajectory decisions for roads with a length of tens of meters. Stable and effective calculation results cannot be obtained, so it cannot meet the requirements of long-distance driving simulation. Moreover, in terms of applicable occasions, their research is mainly aimed at vehicle lane changing and urban road following, rather than highway trajectory decision-making under complex linear conditions.

任园园根据弯道范围内轨迹与行车道中心线之间的横向位置变化，将轨迹分为理想、正常、摇摆、矫正、漂移和切弯6类，针对每一类行驶轨迹，使用回归分析方法建立了基于最大偏移值和曲线起点偏移值的极坐标方程，使用这些方程式能够得到单个弯道的轨迹曲线，而遇到连续弯道则会出现预测失效现象。由于实际的山区公路都是连续弯道占主导，此种技术无法用于山区公路的行驶仿真。According to the lateral position change between the trajectory and the centerline of the roadway within the curve range, Ren Yuanyuan divides the trajectory into six categories: ideal, normal, swing, correction, drift and cut-off curve. For each type of trajectory, regression analysis is used The method establishes polar coordinate equations based on the maximum offset value and the offset value of the starting point of the curve. Using these equations, the trajectory curve of a single curve can be obtained, but the prediction failure phenomenon will appear when encountering continuous curves. Since the actual mountain roads are dominated by continuous curves, this technology cannot be used for driving simulation on mountain roads.

发明内容Contents of the invention

本发明所要解决的技术问题是针对现有技术的不足提供一种基于前视断面选点的复杂道路汽车行驶轨迹预测方法。The technical problem to be solved by the present invention is to provide a complex road vehicle trajectory prediction method based on forward-looking cross-section point selection for the deficiencies of the prior art.

本发明的技术方案如下：Technical scheme of the present invention is as follows:

一种基于前视断面选点的复杂道路汽车行驶轨迹预测方法，包括以下步骤：A method for predicting vehicle trajectories on complex roads based on forward-looking section selection, comprising the following steps:

(1)从车载电子地图、在线道路数据库中提取道路的几何线形数据，通过几何解析计算求解出道路几何边界的平面坐标；(1) Extract the geometric linear data of the road from the on-board electronic map and the online road database, and solve the plane coordinates of the geometric boundary of the road through geometric analysis calculation;

(2)根据驾驶习惯，确定出驾驶人可使用的路面宽度，即设定一个路面宽度利用系数λ，然后进行坐标变换，确定出可使用路幅边界的平面坐标；(2) According to driving habits, determine the road width that the driver can use, that is, set a road width utilization coefficient λ, and then perform coordinate transformation to determine the plane coordinates of the usable road width boundary;

(3)在车辆前方的前视路面区域内按一定间距划分前视断面，间距根据“前视断面选点”的轨迹预测策略”设置；(3) Divide the front-sight section at a certain interval in the front-sight road surface area in front of the vehicle, and the interval is set according to the trajectory prediction strategy of "selection point of the front-sight section";

(4)从预设置的5种驾驶模式中选择一种，将其对应的目标函数作为迭代计算时的优化目标；(4) Select one of the five preset driving modes, and use its corresponding objective function as the optimization target during iterative calculation;

(5)读取车辆当前的行驶速度、侧向加速度参量，判断前方有无障碍物，如有障碍物计算出剩余的路面宽度，根据这些参数再结合车辆尺寸参数，完成约束条件设置，所述约束条件包括边界内行驶、障碍避绕、弯道通过性、行驶稳定性4种；(5) Read the current driving speed and lateral acceleration parameters of the vehicle, judge whether there is an obstacle ahead, calculate the remaining road width if there is an obstacle, and complete the constraint condition setting according to these parameters combined with the vehicle size parameters, the above Constraint conditions include driving within the boundary, obstacle avoidance, curve passability, and driving stability;

(6)由于轨迹的选择是在驾驶人视窗范围内进行，而视窗又随车辆的行驶而向前移动，因此将长里程道路分割为若干前后衔接的短路段；然后，采用优化求解器LINGO11.0沿行驶方向逐步推进求解各短路段的决策变量S_i的值；(6) Since the trajectory selection is carried out within the range of the driver's window, and the window moves forward with the driving of the vehicle, the long-mileage road is divided into several short-circuit sections connected front and back; then, the optimization solver LINGO11 is used. 0 step by step along the driving direction to solve the value of the decision variable S _i of each short-circuit section;

(7)根据比例系数S_i的值，按式(1)沿行驶方向逐一计算出每个前视断面的轨迹点平面坐标；(7) According to the value of the proportional coefficient S _i , calculate the plane coordinates of the track points of each forward-looking section one by one along the driving direction according to formula (1);

x_pti＝x_pri+w_di·S_i·cos α_i，x _pti = x _pri + w _di S _i cos α _i ,

y_pti＝y_pri-w_di·S_i·sin α_i (1)y _pti = y _pri -w _di S _i sin α _i (1)

其中α_i为前视断面i即线段P_liP_ri与大地坐标系X轴的夹角，P_li、P_ri分别是前视断面i的左右两侧端点，候选轨迹点P_ti在线段P_liP_ri之上；x_pti，y_pti为P_ti的平面坐标；x_pri，y_pri为P_ri的平面坐标；Among them, α _i is the angle between the front-sight section i, that is, the line segment P _li P _ri and the X-axis of the geodetic coordinate system, P _li and P _ri are the endpoints on the left and right sides of the front-sight section i respectively, and the candidate trajectory point P _ti is on the line segment P _li Above P _ri ; x _pti , y _pti are the plane coordinates of P _ti ; x _pri , y _pri are the plane coordinates of P _ri ;

(8)连接相邻轨迹点，得到连续的轨迹线，即决策得到的行驶轨迹，对于显示精度要求高的场合，可使用三次样条插值来得到平滑的轨迹曲线。(8) Connect adjacent trajectory points to obtain a continuous trajectory line, that is, the driving trajectory obtained by decision-making. For occasions that require high display accuracy, cubic spline interpolation can be used to obtain a smooth trajectory curve.

所述的方法，所述的5种驾驶模式包括行驶轨迹最短模式、轨迹曲率最小模式、轨迹曲率变化率最小模式、行车道轨迹居中模式以及前述四中模式的混合模式。In the method, the five driving modes include the shortest driving trajectory mode, the minimum trajectory curvature mode, the minimum trajectory curvature change rate mode, the lane centering mode and the mixed mode of the aforementioned four modes.

所述的方法，所述的行驶轨迹最短模式的目标函数，如下式：The method, the objective function of the shortest mode of the traveling trajectory is as follows:

$Min Min {f f}_{0101} = = {Σ Σ}_{i i = = 11}^{n no - - 11} {P P}_{ti ti} {P P}_{ti ti + + 11} = = {Σ Σ}_{i i = = 11}^{n no - - 11} {L L}_{ti ti} - - - - - - ((22))$

其中L_ti＝((x_pti-x_pti+1)²+(y_pti-y_pti+1)²)^0.5。where L _ti =((x _pti −x _pti+1 ) ² +(y _pti −y _pti+1 ) ² ) ^0.5 .

所述的方法，所述的轨迹曲率最小模式的目标函数表示成：In the method, the objective function of the minimum mode of trajectory curvature is expressed as:

$min min {f f}_{0202} = = {Σ Σ}_{i i = = 22}^{n no - - 11} {K K}_{i i} = = {Σ Σ}_{i i = = 22}^{n no - - 11} \frac{{α α}_{i i}}{{L L}_{i i}} - - - - - - ((55));;$

轨迹偏转角

L_i＝((x_pti-x_pti+1)²+(y_Pti-y_Pti+1)²)^0.5。track deflection angle

L _i =((x _pti −x _pti+1 ) ² +(y _Pti −y _Pti+1 ) ² ) ^0.5 .

所述的方法，所述的轨迹曲率变化率最小模式的目标函数表示为：In the described method, the objective function of the minimum mode of the trajectory curvature change rate is expressed as:

$min min {f f}_{0303} = = {Σ Σ}_{i i = = 33}^{n no - - 22} | | {K K}_{i i + + 11} - - {K K}_{i i} | | - - - - - - ((66))$

点P_ti处的曲率为K_i。The curvature at point P _ti is K _i .

所述的方法，所述的行车道轨迹居中模式的目标函数为：Described method, the objective function of described traffic lane trajectory centering pattern is:

$min min {f f}_{0404} = = {Σ Σ}_{i i = = 11}^{n no} | | Δ Δ {w w}_{i i} | | = = {Σ Σ}_{i i = = 11}^{n no} | | 0.5 0.5 {w w}_{Li Li} - - {w w}_{tri tri} | | - - - - - - ((77))$

w_tri为轨迹点P_ti与前视断面右侧端点P_ri之间的距离，w_Li是车道宽度。w _tri is the distance between the trajectory point P _ti and the right end point P _ri of the front view section, and w _Li is the width of the lane.

所述的方法，所述的混合模式目标函数表示为：In the described method, the mixed mode objective function is expressed as:

Min f₀₅＝β₁f′₀₁+β₂f′₀₂+β₃f′₀₃+β₄f′₀₄ (9)Min f ₀₅ ＝β ₁ f′ ₀₁ +β ₂ f′ ₀₂ +β ₃ f′ ₀₃ +β ₄ f′ ₀₄ (9)

式中，β1～β4≥0，是权重系数，需满足β1+β2+β3+β4＝1。In the formula, β1～β4≥0, which are weight coefficients, and must satisfy β1+β2+β3+β4=1.

本发明所进行的工作是对驾驶人的轨迹决策行为进行分析和模拟；对驾驶人的轨迹选择行为进行抽象并演化计算策略，即“前视选点”的计算策略；研究5种典型驾驶模式的背后动机并进行数学表示，最终得到能够适应任意里程长度复杂道路的、能够模拟典型驾驶模式的汽车行驶轨迹决策方法。应用本发明的技术，可以针对山区公路以及复杂赛道得到以下5种驾驶模式的期望轨迹，分别是轨迹长度最短模式、轨迹曲率最小模式(赛道模式)、曲率变化率最小模式(驾驶最舒适模式)、轨迹居中模式(居中行驶模式)和混合模式，其中混合模式为前4种典型模式的综合。The work carried out by the present invention is to analyze and simulate the driver's trajectory decision-making behavior; abstract and evolve the calculation strategy for the driver's trajectory selection behavior, that is, the calculation strategy of "forward-looking point selection"; study 5 typical driving modes The motivation behind it is expressed mathematically, and finally a vehicle trajectory decision-making method that can adapt to complex roads with any mileage length and can simulate typical driving patterns is obtained. By applying the technology of the present invention, the desired trajectories of the following five driving modes can be obtained for mountain roads and complex race tracks, namely the mode with the shortest track length, the mode with the smallest curvature of the track (track mode), and the mode with the smallest rate of curvature change (the most comfortable driving mode). mode), trajectory centering mode (centering driving mode) and mixed mode, where the mixed mode is a combination of the first four typical modes.

附图说明Description of drawings

图1为模糊规则方法；Figure 1 is the fuzzy rule method;

图2为“前视选点”计算策略；Figure 2 is the calculation strategy of "forward-looking point selection";

图3为前视断面两侧端点的坐标计算；Figure 3 is the coordinate calculation of the endpoints on both sides of the front view section;

图4为方向控制模式之一：轨迹长度最短；Figure 4 is one of the direction control modes: the track length is the shortest;

图5为方向控制模式之二：轨迹曲率最小；Figure 5 is the second direction control mode: the minimum trajectory curvature;

图6为方向控制模式之四：行车道内居中行驶；Figure 6 shows the fourth direction control mode: driving in the middle of the traffic lane;

图7为汽车行驶轨迹优化的计算过程；Fig. 7 is the calculation process of vehicle trajectory optimization;

图8为铃鹿赛道上的轨迹优化结果；Fig. 8 is the trajectory optimization result on the Suzuka track;

具体实施方式Detailed ways

以下结合具体实施例，对本发明进行详细说明。The present invention will be described in detail below in conjunction with specific embodiments.

1.“前视断面选点”的轨迹预测策略1. Trajectory prediction strategy of "point selection in front-sight section"

为了控制车辆安全行进，驾驶人需时刻注意前方道路的状况，以获得行驶通道的可使用宽度、曲度等信息。较早的研究通常假设驾驶人视线落在车辆前方路面的某一点，但后来发现，使用视窗模型得到的结果更合理，即驾驶人的视线是落在前方的一段路面上，随着车辆的行驶，落在视窗中的路面区域快速向前移动。在视窗假设基础上，本发明设计了“前视断面选点”策略，如图2。虽然车辆在视窗区域内的行驶是一连续过程，但在实际处理中却可以将其离散化，本发明是按一定的间隔将视窗路面进行分割，将每一个横割线(前视断面)看作是候选轨迹点的集合，驾驶人要做的是在每个前视断面上选择一个点P_ti，作为车辆驶过该断面时的横向期望位置。In order to control the safe driving of the vehicle, the driver needs to always pay attention to the condition of the road ahead in order to obtain information such as the usable width and curvature of the driving channel. Earlier studies usually assumed that the driver's line of sight fell on a certain point on the road in front of the vehicle, but it was later found that the results obtained using the window model were more reasonable, that is, the driver's line of sight fell on a section of the road ahead, and as the vehicle moved , and the road area falling in the window moves forward rapidly. On the basis of the assumption of the window, the present invention designs the strategy of "selection of front-view section", as shown in Fig. 2 . Although the driving of the vehicle in the viewing window area is a continuous process, it can be discretized in actual processing. The present invention divides the viewing window road surface at a certain interval, and regards each transverse secant line (front view section) as is a set of candidate trajectory points, what the driver needs to do is to select a point P _ti on each front-sight section as the expected lateral position of the vehicle when passing through the section.

前视断面的间隔设置要考虑视距、路线平均曲率和行驶通道宽度因素，而这3个因素都与道路设计速度密切相关。具体设置方法如下：The interval setting of the forward-looking section should consider the sight distance, the average curvature of the route and the width of the driving channel, and these three factors are closely related to the road design speed. The specific setting method is as follows:

a.在设计速度V_d较低时，比如V_d＝20～30km/h，曲率半径可以低至15～30m，弯道长度也会很短，不高于5m的断面间隔才能对弯道几何特性进行准确刻画。a. When the design speed V _d is low, such as V _d = 20 ~ 30km/h, the radius of curvature can be as low as 15 ~ 30m, and the length of the curve will be very short. properties are accurately described.

b.设计速度为100～120km/h时，弯道半径一般600m以上，曲率缓和并且弯道足够长，20～30m的间隔已经能够充分描述弯道特性。b. When the design speed is 100-120km/h, the radius of the curve is generally more than 600m, the curvature is mild and the curve is long enough, and the interval of 20-30m can fully describe the characteristics of the curve.

c.至于直道，可以使用与弯道相同的处理办法，这是因为山区公路线形比较琐碎，弯间直线通常也不会很长。c. As for the straight road, you can use the same treatment method as the curve, because the alignment of the mountainous road is relatively trivial, and the straight line between the curves is usually not very long.

在图2中，P_li、P_ri分别是前视断面i的左右两侧端点，候选轨迹点P_ti是在线段P_liP_ri之上。因此，可以用P_ti在P_liP_ri上的滑动来描述不同的轨迹选择行为，比如P_ti滑至中间可表示居中行驶，P_ti滑动至弯道内侧表示切弯道行驶，等等。In Fig. 2, P _li and P _ri are the left and right endpoints of the front view section i respectively, and the candidate trajectory point P _ti is on the line segment P _li P _ri . Therefore, the sliding of P _ti on P _li P _ri can be used to describe different trajectory selection behaviors. For example, sliding P _ti to the middle means driving in the middle, sliding P _ti to the inside of the curve means driving on a curve, and so on.

作为输入的道路几何信息可以使用以下几种手段获得：比如，以线形要素形式或以三维/二维坐标形式存储于在线的“数字公路库”、车载电子地图、导航地图、或是基于GPS/IMU的道路参数拟合方法，这几种手段已经比较成熟，如图3。由此，任意断面两个端点P_li和P_ri的平面坐标可以由线形要素结合路宽计算得到，或是对路边界坐标插值得到。The road geometry information as input can be obtained by the following means: for example, in the form of linear features or in the form of three-dimensional/two-dimensional coordinates stored in the online "digital road library", on-board electronic maps, navigation maps, or based on GPS/ The road parameter fitting method of IMU, these methods are relatively mature, as shown in Figure 3. Therefore, the plane coordinates of the two endpoints P _li and P _ri of any section can be calculated from the linear elements combined with the road width, or obtained by interpolating the coordinates of the road boundary.

因此，P_ti的滑动行为和滑动后的位置可以用比例系数S_i唯一的确定下来，S_i表示为S_i＝w_tri/w_di，其中w_tri为轨迹点P_ti与前视断面右侧端点P_ri之间的距离，w_di为前视断面i的宽度，即该断面所在位置的可使用路幅宽度。那么轨迹点P_ti的平面坐标x_pri和y_pri可以由下式计算得到：Therefore, the sliding behavior and post-sliding position of P _ti can be uniquely determined by the proportional coefficient S _i , and S _i is expressed as S _i =w _tri /w _di , where w _tri is the trajectory point P _ti and the right side of the front view section. The distance between the endpoints P _ri , w _di is the width of the front-sight section i, that is, the usable road width at the location of the section. Then the plane coordinates x _pri and y _pri of the trajectory point P _ti can be calculated by the following formula:

x_pti＝x_pri+w_di·S_i·cos α_i，x _pti = x _pri + w _di S _i cos α _i ,

y_pti＝y_pri-w_di·S_i·sin α_i (1)y _pti = y _pri -w _di S _i sin α _i (1)

其中α_i为前视断面i(即线段P_liP_ri)与大地坐标系X轴的夹角。因此，只要视窗内各断面的比例系数S_i确定下来，期望轨迹便随之确定下来。Among them, α _i is the angle between the front-sight section i (that is, the line segment P _li P _ri ) and the X-axis of the earth coordinate system. Therefore, as long as the scale coefficient S _i of each section in the window is determined, the expected trajectory will be determined accordingly.

2模拟典型驾驶模式的技术方案2 Technical solutions for simulating typical driving patterns

在驾驶过程中，轨迹点的确定首先是驾驶人的决策行为，而在背后支配这种决策行为的必然是使其受益的目标。很多驾驶人在选择轨迹时目标单一明确，但也有一些驾驶人需要在多个目标之间进行折中，为此，本发明建立了能够对4种典型驾驶行为(驾驶模式)进行描述的目标函数，然后阐述在将多个目标进行组合时目标函数的归一化方法以及权重系数的确定，从而得到混合模式(最后1种)。In the process of driving, the determination of the track point is firstly the decision-making behavior of the driver, and the goal behind this decision-making behavior must be the benefit. Many drivers have a single and clear goal when choosing a trajectory, but some drivers need to make a compromise between multiple goals. For this reason, the present invention establishes an objective function that can describe four typical driving behaviors (driving modes) , and then explain the normalization method of the objective function and the determination of the weight coefficient when combining multiple objectives, so as to obtain the mixed mode (the last one).

2.1行驶轨迹最短(距离最优模式)2.1 The shortest driving track (distance optimal mode)

“抄近路”是人们在趋利原则驱使下的一种本能表现，对于公路上的汽车驾驶而言，抄近都是发生在弯道上，表现为轨迹紧贴弯道内侧，并形成了“切弯”效果，如图4。根据图2和图4，能够得到行驶轨迹最短的目标函数，如下式："Cut-cutting" is an instinctive performance driven by the principle of profit-seeking. For car driving on the road, taking shortcuts always occurs on curves, and the performance is that the trajectory is close to the inner side of the curve and forms a "cutting curve". ” effect, as shown in Figure 4. According to Figure 2 and Figure 4, the objective function of the shortest driving trajectory can be obtained, as follows:

其中L_ti＝((x_pti-x_pti+1)²+(y_pti-y_pti+1) ²)^0.5。where L _ti =((x _pti −x _pti+1 ) ² +(y _pti −y _pti+1) ² ) ^0.5 .

2.2轨迹曲率最小(时间最优、制动最优、发动机转速最优和加速度最优模式)2.2 Minimal trajectory curvature (optimal time, optimal braking, optimal engine speed and optimal acceleration mode)

对于赛道工况，赛车手总是尝试将道路几何特性使用到极限，比如遇到急弯时，车手通过轨迹优化可以大幅提高过弯速度，缩短行驶时间(时间最优)。而商用车辆驾驶人以及一部分比较爱惜车辆的普通驾驶人，则希望尽量减少车辆元件的磨损，过弯时如能降低轨迹曲率可以减少进弯时制动器的使用，从而延长摩擦衬片的寿命(制动最优)。大型车辆在过弯时，车身的平转会增加额外的功率消耗，因此若要维持一定的过弯速度，必须提高发动机转速，增加了发动机部件的磨耗，而降低轨迹曲率则可以有效控制发动机的转速变化(发动机转速最优)。此外，如果驾驶人选择一个较大的轨迹半径，会降低曲线行驶的进弯减速度、出弯加速度以及侧向加速度，可以明显提高行驶舒适性(加速度最优)。For track conditions, racers always try to use the geometric characteristics of the road to the limit. For example, when encountering a sharp bend, the driver can greatly increase the cornering speed and shorten the driving time (time is optimal) through trajectory optimization. Drivers of commercial vehicles and some ordinary drivers who cherish their vehicles hope to reduce the wear of vehicle components as much as possible. If the curvature of the trajectory can be reduced when cornering, the use of brakes when cornering can be reduced, thereby prolonging the life of friction linings (manufacturing optimal motion). When a large vehicle is cornering, the flat rotation of the body will increase additional power consumption. Therefore, to maintain a certain cornering speed, the engine speed must be increased, which increases the wear of engine components, and reducing the trajectory curvature can effectively control the engine. RPM variation (optimum engine RPM). In addition, if the driver chooses a larger trajectory radius, the deceleration, acceleration, and lateral acceleration of curves will be reduced, which can significantly improve driving comfort (optimal acceleration).

轨迹曲率最小主要是通过放缓弯道行驶时的轨迹半径来实现，如图5。曲率最优与长度最优两种模式下的轨迹都是在曲中位置向弯道内侧靠拢，但曲率最优时，轨迹在弯道进口端和出口端都是尽量靠近外缘，只有临近曲中时才切向弯道内侧，而长度最优时的轨迹在进弯时是直接切向弯道内侧，没有向外侧靠拢的过程。任意轨迹点P_ti对应的曲率，可用图5中的方法计算得到。首先，按下式求出轨迹点P_ti位置的轨迹偏转角α_i：The minimum trajectory curvature is mainly achieved by slowing down the trajectory radius when driving on a curve, as shown in Figure 5. The trajectories under the two modes of optimal curvature and optimal length are close to the inside of the curve in the middle of the curve, but when the curvature is optimal, the trajectory is as close to the outer edge as possible at the entrance and exit of the curve, and only the adjacent It only cuts to the inside of the curve when it is in the middle, and the trajectory when the length is optimal is directly tangential to the inside of the curve when entering the bend, without the process of moving closer to the outside. The curvature corresponding to any trajectory point P _ti can be calculated by the method in Figure 5. First, calculate the trajectory deflection angle α _i at the position of the trajectory point P _ti according to the following formula:

${α α}_{i i} = = \frac{{L L}_{i i - - 11}^{22} + + {L L}_{i i}^{22} - - {L L}_{i i - - 11,, i i}^{22}}{22 \times \times {L L}_{i i - - 11} \times \times {L L}_{i i}} - - - - - - ((33))$

L_i＝((x_pti-x_pti+1)²+(y_Pti-y_Pti+1)²)^0.5，L _i =((x _pti -x _pti+1 ) ² +(y _Pti -y _Pti+1 ) ² ) ^0.5 ,

L_i-1，i＝((x_pti-1-x_pti+1)²+(y_pti-1-y_pti+1)²)^0.5 (4)L _{i-1, i} ＝((x _pti-1 -x _pti+1 ) ² +(y _pti-1 -y _pti+1 ) ² ) ^0.5 (4)

由于α_i是轨迹在P_ti处的偏转所引起的，P_ti点处的曲率K_i＝dθ/dL＝α_i/L₂因此，轨迹总曲率最小的目标函数可以表示成：Since α _i is caused by the deflection of the trajectory at P _ti , the curvature at P _ti point K _i =dθ/dL=α _i /L ₂ Therefore, the objective function of the minimum total curvature of the trajectory can be expressed as:

$min min {f f}_{0202} = = {Σ Σ}_{i i = = 22}^{n no - - 11} {K K}_{i i} = = {Σ Σ}_{i i = = 22}^{n no - - 11} \frac{{α α}_{i i}}{{L L}_{i i}} - - - - - - ((55))$

2.3轨迹曲率变化率最小(转向操纵最优模式)2.3 Minimum change rate of trajectory curvature (optimal mode for steering manipulation)

转动方向盘将车辆控制在可行驶的路面宽度之内，是驾驶人在行车时的一个主要工作内容。如果能尽量维持方向盘不动或是少转动，显然能够降低驾驶负荷。驾驶人转动方向盘的目的是调整轨迹曲率，因此，将相邻轨迹点的曲率变化控制在最小，自然能够减少对方向盘角输入的需求(转向操纵最优)。前面已经给出了任意轨迹点曲率的计算方法，因此，曲率变化率最小的目标函数可表示为：Turning the steering wheel to control the vehicle within the drivable road width is one of the main tasks of the driver when driving. If you can keep the steering wheel still or turn less as much as possible, it will obviously reduce the driving load. The purpose of the driver turning the steering wheel is to adjust the curvature of the trajectory. Therefore, controlling the curvature change of adjacent trajectory points to a minimum can naturally reduce the demand for steering wheel angle input (steering manipulation is optimal). The calculation method of the curvature of any trajectory point has been given above, so the objective function with the minimum curvature change rate can be expressed as:

2.4行车道轨迹居中(安全最优模式)2.4 Centering of lane trajectory (safety optimal mode)

对于一部分驾驶人来讲，安全在任何时候都是第一位的，因此在公路上行驶时他们最愿意将车辆控制在行车道中间，这样可同时保持与右侧路缘和左侧对向来车的侧向安全距离。根据图6，当车辆居中行驶时，轨迹点位于路中线和右侧路缘线的中间，而偏离该位置时，式Δw_i＝0.5(w_Li-w_tri)必然不为零，式中w_Li是车道宽度，因此可用下式作为车道中间行驶模式的目标函数：For some drivers, safety is the first priority at all times, so they are most willing to control the vehicle in the middle of the lane when driving on the road, so that they can keep oncoming traffic on the right side of the road and on the left side at the same time. lateral safety distance. According to Figure 6, when the vehicle is driving in the middle, the track point is located in the middle of the road center line and the right curb line, and when it deviates from this position, the formula Δw _i =0.5(w _Li -w _tri ) must be non-zero, where w _Li is the width of the lane, so the following formula can be used as the objective function of the driving mode in the middle of the lane:

需要说明的是，在使用前4种驾驶模式时，驾驶人可占用的路面宽度可以根据实际情况自行设置。在车流稀少的山区公路上，可以取整幅路面宽度；也可以取车道宽度加上一侧硬路肩宽度再加上一部分对向车道宽度(0.5～1m)；当然，也可以只是车道宽度。It should be noted that when using the first four driving modes, the road width that the driver can occupy can be set according to the actual situation. On mountainous roads with little traffic flow, the entire width of the road surface can be taken; the width of the lane plus the width of the hard shoulder on one side plus a part of the width of the opposite lane (0.5-1m) can also be taken; of course, it can also be just the width of the lane.

2.5目标函数归一化以及多目标组合策略2.5 Objective function normalization and multi-objective combination strategy

在实际的驾驶行为观测中，还有一部分驾驶人操纵车辆时的行驶轨迹具有混合特征，而多目标决策正好可以描述这种行为。由于执行各目标时得到的函数值具有不同的量纲，在组合前应进行归一化处理，使目标值在同一个数量级之内。使用下式对f₀₁～f₀₄归一化，之后得到 In the actual driving behavior observation, there are still some drivers whose driving trajectories have mixed characteristics when manipulating the vehicle, and multi-objective decision-making can just describe this kind of behavior. Since the function values obtained when executing each target have different dimensions, they should be normalized before combining so that the target values are within the same order of magnitude. Use the following formula to normalize f ₀₁ ~ f ₀₄ , and then get

${f f}_{0101}^{n no} = = \frac{{f f}_{0101}}{{Σ Σ}_{i i = = 11}^{n no - - 11} {L L}_{ci ci}},, {f f}_{0202}^{n no} = = \frac{{f f}_{0202}}{{Σ Σ}_{i i = = 22}^{n no - - 11} {K K}_{ci ci}},, {f f}_{0303}^{n no} = = \frac{{f f}_{0303}}{{Σ Σ}_{i i = = 22}^{n no - - 22} | | {K K}_{ci ci} - - {K K}_{ci ci + + 11} | |},, {f f}_{0404}^{n no} = = \frac{{f f}_{0404}}{0.1 0.1 \times \times N N} - - - - - - ((88))$

式中L_ci为相邻前视断面之间的道路中线长度，如前文图3所示；K_ci为道路中线与前视断面P_liP_ri相交位置的曲率；N是前视断面数量。式中0.1N的含义是，驾驶人在判断车道中间位置时存在0.2m左右的误差，取其中间值，N个前视断面上的总的误差大致为0.1N。对归一化后的目标函数赋予一定的权重，多目标轨迹优化问题可以表示为：In the formula, L _ci is the length of the road centerline between adjacent front-sight sections, as shown in Figure 3 above; K _ci is the curvature of the intersection position between the road centerline and the front-sight section P _li P _ri ; N is the number of front-sight sections. The meaning of 0.1N in the formula is that there is an error of about 0.2m when the driver judges the middle position of the lane. Taking the median value, the total error on the N front-view sections is roughly 0.1N. Given a certain weight to the normalized objective function, the multi-objective trajectory optimization problem can be expressed as:

式中，β₁～β₄≥0，是权重系数，需满足β₁+β₂+β₃+β₄＝1。In the formula, β ₁ ~ β ₄ ≥ 0 are weight coefficients, and β ₁ + β ₂ + β ₃ + β ₄ = 1 must be satisfied.

3轨迹预测的工作流程3 Workflow of Trajectory Prediction

整个轨迹决策的主要环节一共有四个，分别是“前视断面选点”策略、典型驾驶行为模拟(建立与典型驾驶模式对应的优化目标函数)、约束条件设置、以及求解算法。本发明所涉及的重点是“前视断面选点”方法策略和典型驾驶行为的模拟实现。图7是行驶轨迹的预测流程，具体的步骤如下：There are four main links in the entire trajectory decision-making process, which are the strategy of "selection of forward-looking sections", simulation of typical driving behavior (establishment of an optimization objective function corresponding to a typical driving mode), setting of constraints, and a solution algorithm. The emphases involved in the present invention are the method strategy of "selecting points in front-view section" and the simulation realization of typical driving behavior. Figure 7 is the prediction process of the driving trajectory, and the specific steps are as follows:

(3)在车辆前方的前视路面区域内按一定间距划分前视断面，间距设置参见“1.“前视断面选点”的轨迹预测策略”中的内容；(3) Divide the front-sight section at a certain interval in the front-sight road surface area in front of the vehicle. For the spacing setting, refer to the content in "1. Trajectory prediction strategy of "selection point of the front-sight section"";

(4)从前面设定的5种驾驶模式中选择一种，将其对应的目标函数作为迭代计算时的优化目标；(4) Select one of the five driving modes set earlier, and use its corresponding objective function as the optimization target during iterative calculation;

(5)读取车辆当前的行驶速度、侧向加速度参量，判断前方有无障碍物，如有障碍物计算出剩余的路面宽度，根据这些参数再结合车辆尺寸参数，完成约束条件设置(边界内行驶、障碍避绕、弯道通过性、行驶稳定性等4种)；(5) Read the current driving speed and lateral acceleration parameters of the vehicle, judge whether there are obstacles ahead, and calculate the remaining road width if there are obstacles, and complete the constraint condition setting (within the boundary) according to these parameters and combined with the vehicle size parameters Driving, obstacle avoidance, curve passability, driving stability, etc.);

(6)由于轨迹的选择是在驾驶人视窗范围内进行，而视窗又随车辆的行驶而向前移动，因此将长里程道路分割为若干前后衔接的短路段。然后，采用优化求解器LINGO11.0沿行驶方向逐步推进求解各短路段的决策变量S_i的值；(6) Since the trajectory selection is carried out within the range of the driver's window, and the window moves forward with the driving of the vehicle, the long-distance road is divided into several short-circuit sections connected front and back. Then, the optimization solver LINGO11.0 is used to gradually advance along the driving direction to solve the value of the decision variable S _i of each short-circuit section;

(7)根据S_i的值，按式(1)沿行驶方向逐一计算出每个前视断面的轨迹点平面坐标；(7) According to the value of S _i , calculate the plane coordinates of the track points of each front-sight section one by one along the driving direction according to formula (1);

4计算实例4 Calculation examples

使用本发明的技术，进行铃鹿赛道(位于日本三重县铃鹿市的F1赛道)的汽车行驶轨迹仿真。对于F1赛车手而言，在最短时间内驶完规定圈数是其唯一的目标，缩短行驶时间最有效的手段是高速过弯，而选手只有将弯道宽度使用到极限，最大程度地降低轨迹曲率，才能减少弯道引起的速度折损，因此，本发明中的轨迹曲率最小模式(2.2节)最适宜用来描述车手的这种行为。将轨迹曲率最小模式对应的目标函数作为决策目标，使用本发明的轨迹预测流程进行轨迹预测，计算结果如图8所示(图中的C2-C20为弯道编号)。如果经常观看F1比赛的话，能发现图8中的轨迹与真实比赛中的赛车轨迹是非常一致的，以行驶轨迹为中介，可以对赛车手的驾驶行为进行解读，进而了解车手为了获得最大的通过半径是如何选择行驶轨迹的，这能够为赛车手训练、驾驶模式、以及赛道设计提供科学根据。Use the technology of the present invention to carry out the simulation of the car running track of the Suzuka Circuit (the F1 circuit located in Suzuka City, Mie Prefecture, Japan). For F1 racers, the only goal is to complete the specified number of laps in the shortest time. The most effective way to shorten the driving time is to corner at high speed, and the players can only use the width of the curve to the limit to minimize the trajectory. Curvature can reduce the speed loss caused by the curve. Therefore, the minimum trajectory curvature mode (section 2.2) in the present invention is most suitable for describing this behavior of the driver. The objective function corresponding to the mode with the minimum trajectory curvature is used as the decision target, and the trajectory prediction process of the present invention is used for trajectory prediction, and the calculation results are shown in Figure 8 (C2-C20 in the figure are curve numbers). If you often watch F1 games, you can find that the trajectory in Figure 8 is very consistent with the racing trajectory in the real race. Using the driving trajectory as an intermediary, you can interpret the driving behavior of the driver, and then understand that the driver is in order to get the maximum pass. How the radius selects the driving trajectory can provide a scientific basis for racer training, driving patterns, and track design.

应当理解的是，对本领域普通技术人员来说，可以根据上述说明加以改进或变换，而所有这些改进和变换都应属于本发明所附权利要求的保护范围。It should be understood that those skilled in the art can make improvements or changes based on the above description, and all these improvements and changes should fall within the protection scope of the appended claims of the present invention.

Claims

1. the complicated road traval trace Forecasting Methodology based on the reconnaissance of forward sight section, is characterized in that, comprises the following steps:

(1) from electronic map of automobile navigation, online transportation database, extract the geometry linear data of road, by geometry analytical Calculation, solve the planimetric coordinates of road geometrical boundary;

(2) according to driving habits, determine the spendable width of roadway of driver, set a width of roadway usage factor λ, then carry out coordinate transform, determine the planimetric coordinates that can use road breadths circle;

(3) in the region, forward sight road surface of vehicle front, divide at a certain distance forward sight section, spacing is according to the trajectory predictions strategy of " reconnaissance of forward sight section " " arrange;

(4) from 5 kinds of driving models that pre-set, select a kind of, the optimization aim using its corresponding objective function during as iterative computation;

(5) read the current travel speed of vehicle, side acceleration parameter, there is clear in judgement the place ahead, if any barrier, calculate remaining width of roadway, according to these parameters again in conjunction with vehicle dimension parameter, complete constraint condition setting, described constraint condition comprise in border travel, obstacle keeps away around 4 kinds of, bend trafficability characteristic, riding stabilities;

(6) selection due to track is to carry out within the scope of driver's form, and form is with the travelling and move forward of vehicle, and therefore long mileage road is divided into some successive shorted segments; Then, adopt Optimization Solution device LINGO11.0 along travel direction iterative method, to solve the decision variable S of each shorted segment _ivalue;

(7) according to scale-up factor S _ivalue, by formula (1), along travel direction, calculate one by one the tracing point planimetric coordinates of each forward sight section;

x _pti＝x _pri+w _di·S _i·cos α _i，

y _pti＝y _pri-w _di·S _i·sin α _i (1)

α wherein _ifor forward sight section i is line segment P _lip _riwith the angle of earth coordinates X-axis, P _li, P _rirespectively the left and right sides end points of forward sight section i, candidate's tracing point P _tiat line segment P _lip _rion; x _pti, y _ptifor P _tiplanimetric coordinates; x _pri, y _prtfor P _riplanimetric coordinates;

(8) connect adjacent track point, obtain continuous trajectory, i.e. the driving trace that decision-making obtains, requires high occasion for display precision, can obtain level and smooth geometric locus by cubic spline interpolation.

2. method according to claim 1, it is characterized in that, 5 kinds of described driving models comprise the mixed mode of pattern in the shortest pattern of driving trace, track curvature minimal mode, track curvature variation minimal mode, runway track pattern placed in the middle and aforementioned four.

3. method according to claim 2, is characterized in that, described driving trace is the objective function of short pattern, as shown in the formula:

Min f_{01} = Σ_{i = 1}^{n - 1} P_{ti} P_{ti + 1} = Σ_{i = 1}^{n - 1} L_{ti} - - - (2)

L wherein _ti=((x _pti-x _pti+1) ²+ (y _pti-y _pti+1) ²) ^0.5.

4. method according to claim 2, is characterized in that, the objective function of described track curvature minimal mode is expressed as:

\min f_{02} = Σ_{i = 2}^{n - 1} K_{i} = Σ_{i = 2}^{n - 1} \frac{α_{i}}{L_{i}} - - - (5);

Trajectory deflection angle

l _i=((x _pti-x _pti+1) ²+ (y _pti-y _pti+1) ²) ^0.5.

5. method according to claim 2, is characterized in that, the objective function of described track curvature variation minimal mode is expressed as:

\min f_{03} = Σ_{i = 3}^{n - 2} | K_{i + 1} - K_{i} | - - - (6)

Point P _tithe curvature at place is K _i.

6. method according to claim 2, is characterized in that, the objective function of described runway track pattern placed in the middle is:

\min f_{04} = Σ_{i = 1}^{n} | Δ w_{i} | = Σ_{i = 1}^{n} | 0.5 w_{Li} - w_{tri} | - - - (7)

W _trifor tracing point P _tiwith forward sight section right side end points P _ribetween distance, w _liit is lane width.

7. method according to claim 2, is characterized in that, described mixed mode objective function is expressed as:

Minf ₀₅＝β ₁f′ ₀₁+β ₂f′ ₀₂+β ₃f′ ₀₃+β ₄f′ ₀₄ (9)

In formula, β 1～β 4 >=0, is weight coefficient, need meet β 1+ β 2+ β 3+ β 4=1.