CN115938106A - Online Verification Method for Autonomous Driving Decision Based on Accessibility Analysis of Traffic Participants - Google Patents

Abstract

The invention belongs to the technical field of automatic driving, in particular to an automatic driving decision online verification method based on reachability analysis of traffic participants, which comprises the following steps: safety verification of the expected track of the self vehicle based on reachability analysis; step two: the method comprises the steps of generating a standby safety track based on a legal reachable area, if an expected track is safe, firstly calculating the legal reachable area of a self vehicle within a specific time, further calculating the end point position of the standby safety track according to the legal reachable area of the self vehicle, judging whether lane change is needed, formally predicting all legal evolution of a traffic scene by combining a visible and invisible traffic regulation, calculating the reachable area of each traffic participant, further verifying whether the expected track meets legal safety, enabling the self vehicle to cope with all possible traffic scenes, avoiding driving on unsafe tracks, and meanwhile dynamically outputting the standby safety track in real time when the expected track fails, so as to ensure that the self vehicle is in a safe state.

Description

Online verification method for autonomous driving decision-making based on traffic participant accessibility analysis

Technical Field

The present invention relates to the field of autonomous driving technology, and specifically to an online verification method for autonomous driving decisions based on traffic participant accessibility analysis.

Background Art

Safety is the main challenge facing the popularization of autonomous driving technology. Unsafe driving strategies seriously threaten the lives and property safety of drivers and passengers, and trigger a crisis of public trust. In order to ensure that autonomous driving vehicles have the ability to cope with any complex traffic scenarios in the real world and ensure that autonomous driving vehicles do not actively cause accidents, it is necessary to conduct safety verification of the decisions made by autonomous driving.

At present, the safety verification of decision-making systems is mostly carried out offline. The main methods include mileage, disengagement rate, simulation and scenario-based test verification. The mileage-based safety verification method requires more than 100 million kilometers of mileage to obtain sufficient statistical data to prove the safety of autonomous driving decisions. The annual verification cycle is difficult to match the efficient iteration of autonomous driving methods; the disengagement rate-based safety verification method has a single dimension and cannot ensure whether the number of disengagements obtained from the test truly represents the decision safety under driving scenarios of different complexities; the simulation-based safety verification method is difficult to prove that the established virtual environment can realistically and accurately restore the actual traffic scene; the scenario-based safety verification method has the risk of "overfitting". More importantly, the above-mentioned offline verification methods all evaluate the decision-making performance of the self-vehicle when passing through the current scene based on offline data such as the position and speed of the self-vehicle and other vehicles. It is impossible to realize online correction of the decision-making algorithm and ensure the safety of the self-vehicle in real time online.

As a supplement to offline verification, online safety verification methods can verify the safety of autonomous driving vehicles in real time, and are a necessary means to ensure that vehicles implement safe decisions. Existing online verification methods mainly include the RSS model proposed by Mobileye and the SFF model proposed by NVIDIA. The above methods are based on pre-set driving rules such as safe distance maintenance to conduct online safety verification of autonomous driving systems. However, such models can only verify whether the state of the vehicle at a certain moment meets the established rules, and cannot guarantee the safety of the vehicle's continuous trajectory. At the same time, they do not provide alternative strategies for bringing the vehicle into a safe state.

Therefore, it is necessary to design a method that can verify the safety of the trajectory online and provide a backup safe trajectory based on the expected trajectory output by the autonomous driving decision-making layer and combined with explicit and implicit traffic regulations, so that the autonomous driving vehicle always has a safe driving trajectory to choose from and meets the safety needs of future smart cars and smart transportation.

In view of the above application requirements and technical problems faced, the online safety verification method of autonomous driving decision-making based on legal accessibility analysis of traffic participants proposed in the present invention, as the safety base of the existing motion planning layer, calculates the legally accessible area of each traffic participant, and then verifies whether the expected trajectory is safe and provides a backup safety trajectory. This method corrects the autonomous driving decision in real time through online verification, has the generalization ability to adapt to all traffic scenes, and overcomes the shortcomings of offline verification methods; by combining explicit and implicit traffic regulations to formally predict all legal evolutions of traffic scenes, verify the safety of the expected trajectory of the vehicle and generate a backup safety trajectory, ensure the safety of the vehicle on a continuous trajectory in the future, and overcome the shortcomings of the existing online verification RSS model and SFF model. The method disclosed in the present invention can verify the expected trajectory of the vehicle in real time and generate a backup safety trajectory in a variety of typical traffic scenes, so that the vehicle is always on a safe trajectory, effectively respond to dangerous conditions, strictly ensure the legal safety of the vehicle, and at the same time have the generalization ability to adapt to all traffic scenes, easy to add new rules or delete original rules, and then impose more constraints on the vehicle or relax constraints.

Summary of the invention

The purpose of this section is to summarize some aspects of the embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the specification abstract and invention title of this application to avoid blurring the purpose of this section, specification abstract and invention title, and such simplifications or omissions cannot be used to limit the scope of the present invention.

In view of the problems existing in the prior art, the present invention is proposed.

Therefore, the purpose of the present invention is to provide an online verification method for autonomous driving decisions based on traffic participant accessibility analysis. Based on traffic participant accessibility analysis, an online safety verification method for autonomous driving decisions that takes into account both safety and real-time performance is designed and implemented. It effectively combines explicit and implicit traffic regulations to formally predict all legal evolutions of traffic scenes, calculates the reachable areas of each traffic participant, and then verifies whether the expected trajectory meets legal safety and generates a backup safety trajectory, so that the vehicle can cope with all possible traffic scenarios. Under the premise that other traffic participants do not violate traffic regulations, the vehicle can be guaranteed to be on a safe trajectory. At the same time, this method is independent of the trajectory planning algorithm, has low computing resource requirements, and is highly real-time.

To solve the above technical problems, according to one aspect of the present invention, the present invention provides the following technical solutions:

The online verification method of autonomous driving decision-making based on traffic participant accessibility analysis includes the following steps:

Step 1: Safety verification of the expected trajectory of the ego vehicle based on accessibility analysis. First, the relevant vehicles, relevant roads and reachable areas related to vehicle collaboration of other traffic participants are calculated based on the collected traffic environment and ego vehicle status information. At the same time, the occupied area of the ego vehicle is calculated based on the expected trajectory information given by the trajectory planner and the physical model of the ego vehicle. This is used to determine whether the occupied area of the ego vehicle intersects with the reachable areas of other traffic participants, and then verify whether the expected trajectory is safe, which is used as the input of step 2.

Step 2: Generate an alternative safety trajectory based on the legally reachable area. If the expected trajectory is safe, first calculate the legally reachable area of the vehicle within a specific time, and then calculate the end position of the alternative safety trajectory based on the legally reachable area of the vehicle to determine whether a lane change is required, and then choose to use the quintic polynomial method or generate an alternative safety trajectory along the center line of the lane; based on whether the area occupied by the vehicle along the alternative safety trajectory is within the legally reachable area of the vehicle, determine whether the alternative safety trajectory is generated successfully. If the alternative safety trajectory is generated successfully, execute the successfully verified expected trajectory at the corresponding time. Otherwise, if the expected trajectory is not safe or the generation of the alternative safety trajectory fails, it is necessary to determine whether the executable time of the alternative safety trajectory is insufficient to decide whether to generate a new alternative safety trajectory and execute the previous successfully verified alternative safety trajectory.

As a preferred solution of the online verification method of autonomous driving decision based on traffic participant accessibility analysis described in the present invention, the step 1 specifically includes:

(1) Vehicle and road modeling

This sub-step models vehicles and roads, divides vehicles into the self-vehicle and other traffic participating vehicles, and further divides other traffic participating vehicles into motor vehicles and non-motor vehicles; all vehicles are dynamically modeled as a point mass model, and the kinematic model adopts a second-order integrator model. All roads are divided into non-overlapping roads and overlapping roads. The road consists of a motor vehicle lane, a non-motor vehicle lane and a shoulder. The lane line boundary is the road boundary. The center point of the overlapping area is defined as the intersection of the center lines of each branch road, and the road is modeled using the Frenet coordinate system;

(2) Formal expression of traffic rules

This sub-step formalizes the constraints imposed by explicit and implicit traffic regulations on each vehicle participating in the traffic, and divides the constraints of each vehicle participating in the traffic into constraints related to road traffic, constraints related to vehicle motion, and constraints related to cooperative driving, and further subdivides these three categories of constraints;

(3) Prediction of accessible areas for participating vehicles

This sub-step divides the reachable areas of other traffic participating vehicles into reachable areas related to road traffic, reachable areas related to vehicle kinematics, and reachable areas related to vehicle cooperative driving according to the three major types of constraints of traffic regulations, and defines the intersection of the three reachable areas as the reachable area of the traffic participating vehicles for a period of time in the future; the road-related reachable area is calculated through lane following, road signs, and V2X warning information; the vehicle kinematics-related reachable area considers the vehicle's kinematic model and size, and the reachable area related to vehicle kinematics is super-approximated by dynamically divergent hexagons; the vehicle cooperative driving-related reachable area considers the vehicle's road right enjoyment and safety distance keeping rules, and calculates the relevant reachable area;

(4) Safety verification of the expected trajectory of the vehicle

This step calculates the area occupied by the vehicle based on the size of the vehicle and the expected trajectory, and determines whether the area occupied by the vehicle intersects with the reachable areas of all other traffic participants to verify whether the expected trajectory of the vehicle is safe. If there is no intersection, the expected trajectory of the vehicle is legal and safe. Otherwise, the expected trajectory of the vehicle is unsafe and needs to drive a verified alternative safe trajectory.

As a preferred solution of the online verification method of autonomous driving decision based on traffic participant accessibility analysis described in the present invention, the step 2 specifically includes:

(1) Selection of the endpoint of the vehicle's backup safety trajectory

This sub-step selects the end point of the backup safety trajectory of the ego vehicle. The selection of the trajectory end point takes into account the following factors: a. The backup safety trajectory smoothly connects to the corresponding expected trajectory; b. The backup safety trajectory ensures the legal safety of the ego vehicle (i.e., the area occupied by the ego vehicle and the accessible area of other participating vehicles do not intersect at the same time); c. The length of the backup safety trajectory meets the normal driving requirements and can transition the ego vehicle to the safe area;

(2) Generation of backup safe trajectories for the ego vehicle

This sub-step generates an alternative safety trajectory according to the trajectory endpoint. When lane changing is required, the quintic polynomial method is used to generate the alternative safety trajectory for the ego vehicle to change lanes. When lane changing is not required, the lane centerline method is used to generate the alternative safety trajectory. The alternative safety trajectory is successfully generated by checking whether the area occupied by the ego vehicle along the alternative safety trajectory is included in the legally reachable area of the ego vehicle.

Compared with the prior art, the beneficial effects of the present invention are: a method for online safety verification of autonomous driving decisions based on traffic participant accessibility analysis is disclosed, which is used to verify the safety of the expected trajectory of the self-vehicle and generate a backup safety trajectory, so as to ensure that the self-vehicle is always on a safe trajectory. The proposed method can serve as a safety layer of the existing motion planning framework, and by combining explicit and implicit traffic regulations to formally predict all legal evolutions of traffic scenes, calculate the reachable areas of each traffic participant, and then verify whether the expected trajectory meets legal safety, so that the self-vehicle can cope with all possible traffic scenes and avoid driving on unsafe trajectories. At the same time, when the expected trajectory fails, it can dynamically output a backup safety trajectory in real time to ensure that the self-vehicle is in a safe state. Under the premise that other traffic participants do not violate traffic regulations, it can ensure that the self-vehicle is always on a safe trajectory. The method is independent of the trajectory planning algorithm, has low computing resource requirements, strong real-time performance, can add rules or delete original rules, and then add constraints to the self-vehicle or relax constraints. It is easy to update and iterate, and easy to apply and deploy in practice.

BRIEF DESCRIPTION OF THE DRAWINGS

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the present invention will be described in detail below in combination with the accompanying drawings and detailed embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor. Among them:

FIG1 is a flow chart of an online security verification method of the present invention;

FIG2 is a schematic diagram of a vehicle and road model of the present invention;

FIG3 is a schematic diagram of a lane model of the present invention;

FIG4 is a schematic diagram of a reachable area related to lane following according to the present invention;

(a)不考虑车辆尺寸，(b)考虑车辆尺寸；FIG. 5 is a diagram of other accessible areas related to the movement of traffic participating vehicles in the present invention.

(a) without considering vehicle size, (b) considering vehicle size;

FIG6 is a schematic diagram of a reachable area related to a safety distance of the present invention;

FIG7 is a schematic diagram of a timing sequence for generating a backup safety trajectory according to the present invention;

FIG8 is a diagram showing different situations of generating backup safety trajectories according to the present invention;

FIG9 is a schematic diagram of lane change trajectory prediction according to the present invention;

FIG10 is a one-way three-lane curve scene diagram of the present invention;

FIG11 is a diagram showing the safety verification results of a one-way three-lane curve scenario according to the present invention;

FIG12 is a scene diagram of a two-way six-lane straight road according to the present invention;

FIG13 is a diagram showing the safety verification results of a two-way six-lane straight road scenario according to the present invention;

FIG14 is a scene diagram of a two-lane intersection according to the present invention;

FIG. 15 is a diagram showing the safety verification results of a two-lane intersection scenario of the present invention.

DETAILED DESCRIPTION

In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

Secondly, the present invention is described in detail with reference to schematic diagrams. When describing the embodiments of the present invention in detail, for the sake of convenience, the cross-sectional diagrams showing the device structure will not be partially enlarged according to the general scale, and the schematic diagrams are only examples, which should not limit the scope of protection of the present invention. In addition, in actual production, the three-dimensional dimensions of length, width and depth should be included.

To make the objectives, technical solutions and advantages of the present invention more clear, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

The present invention provides the following technical solutions: an online verification method for autonomous driving decisions based on traffic participant accessibility analysis. Based on traffic participant accessibility analysis, an online safety verification method for autonomous driving decisions that takes into account both safety and real-time performance is designed and implemented. It effectively combines explicit and implicit traffic regulations to formally predict all legal evolutions of traffic scenes, calculates the reachable areas of each traffic participant, and then verifies whether the expected trajectory meets legal safety and generates backup safety trajectories, so that the vehicle can cope with all possible traffic scenarios. Under the premise that other traffic participants do not violate traffic regulations, it can ensure that the vehicle is always on a safe trajectory. At the same time, this method is independent of the trajectory planning algorithm, has low computing resource requirements, and is highly real-time.

Example 1

This section will further explain the specific implementation method of the proposed online safety verification method for autonomous driving decision-making based on traffic participant accessibility analysis in detail in conjunction with the attached drawings. The steps of the method are as follows:

Step 1: Safety verification of the expected trajectory of the ego vehicle based on accessibility analysis. First, the relevant vehicles, relevant roads and reachable areas related to vehicle collaboration of other traffic participants are calculated based on the collected traffic environment and ego vehicle status information. At the same time, the occupied area of the ego vehicle is calculated based on the expected trajectory information given by the trajectory planner and the physical model of the ego vehicle. This is used to determine whether the occupied area of the ego vehicle intersects with the reachable areas of other traffic participants, and then verify whether the expected trajectory is safe, which is used as the input of step 2.

Step 2: Generate an alternative safety trajectory based on the legally accessible area. If the expected trajectory is safe, first calculate the legally accessible area of the vehicle within a specific time, and then calculate the end position of the alternative safety trajectory based on the legally accessible area of the vehicle to determine whether a lane change is required, and then choose to use the quintic polynomial method or generate an alternative safety trajectory along the center line of the lane; based on whether the area occupied by the vehicle along the alternative safety trajectory is within the legally accessible area of the vehicle, determine whether the alternative safety trajectory is generated successfully. If the alternative safety trajectory is generated successfully, execute the expected trajectory that has been successfully verified at the corresponding time. On the contrary, if the expected trajectory is not safe or the generation of the alternative safety trajectory fails, it is necessary to determine whether the executable time of the alternative safety trajectory is insufficient, and decide whether to generate a new alternative safety trajectory and execute the previous successfully verified alternative safety trajectory. The alternative safety trajectory can guide the vehicle to a safe area in dangerous situations, ensuring that the vehicle is always driving on a safe trajectory.

1. Safety verification of the expected trajectory of the ego vehicle based on reachability analysis

The safety verification of the expected trajectory of the ego vehicle based on accessibility analysis mainly includes four sub-steps: vehicle and road modeling, formal expression of traffic regulations, prediction of the reachable area of traffic participating vehicles and safety verification of the expected trajectory of the ego vehicle. Among them, the vehicle and road modeling constructs a model of the vehicle and the road, the traffic regulations formally express the constraints that each traffic participating vehicle should abide by as stipulated by explicit and implicit traffic regulations, the reachable area of traffic participating vehicles predicts the reachable area of traffic participating vehicles with collision risks around the ego vehicle at future times, and the safety verification of the expected trajectory of the ego vehicle judges the safety of the expected trajectory of the ego vehicle based on whether the occupied area of the ego vehicle overlaps with the reachable areas of other traffic participating vehicles at the same time.

(1) Vehicle and road modeling

矩形后端到参考点距离为

将第i(i＝1,2……N_obj)个其他交通参与车建模为长为l_obj，i宽为w_obj，i的矩形，参考点为矩形中心。车辆运动学模型定义为二阶积分器模型：This sub-step classifies and models vehicles and roads. As shown in FIG2 , the present invention divides vehicles into the ego vehicle and other traffic participating vehicles, and further divides traffic participating vehicles into motor vehicles and non-motor vehicles. All vehicles are modeled as a point mass model. Considering the uncertainty of measurement, the ego vehicle is modeled as a rectangle with a length of l _ego and a width of w _ego . The vehicle reference point is the center of the rear axle, and the distance from the front end of the rectangle to the reference point is

The distance from the back end of the rectangle to the reference point is

The i-th (i＝1,2……N _obj ) other traffic participant vehicle is modeled as a rectangle with a length of l _{obj and} a width of w _obj, and the reference point is the center of the rectangle. The vehicle kinematic model is defined as a second-order integrator model:

Where _px and _py are the x-axis and y-axis coordinates of the vehicle reference point at time t, _px,0 and _py,0 are the x-axis and y-axis coordinates of the vehicle reference point at time _t0 , _vx,0 and _vy,0 are the velocities of the vehicle reference point in the x-axis and y-axis directions at _time t0 _{, ax,0} and _ay,0 are the accelerations of the vehicle reference point in the x-axis and y-axis directions at time _t0 ;

第i个交通参与车到重叠区域中心的距离为

本发明中道路由机动车道、非机动车道与路肩组成，对于任意车道，假设车道中心线与车道左右边界为光滑的有向曲线，车道中心线与车道左右边界线平行，均由n_arc条圆弧组成。如图3所示，本发明将车道中心线表示为

车道左边界表示为

车道右边界表示为

The present invention divides roads into non-overlapping roads and overlapping roads: non-overlapping roads have only one main road, no branches and overlapping areas; overlapping roads have branches and overlapping areas. The overlapping area is the overlapping area of each branch, and the center point of the overlapping area is defined as the intersection of the center lines of each branch. The distance from the vehicle to the center of the overlapping area is

The distance from the i-th participating vehicle to the center of the overlapping area is

In the present invention, the road consists of a motor vehicle lane, a non-motor vehicle lane and a shoulder. For any lane, it is assumed that the lane centerline and the left and right boundaries of the lane are smooth directed curves, and the lane centerline is parallel to the left and right boundaries of the lane and is composed of n _arc arcs. As shown in FIG3 , the present invention represents the lane centerline as

The left boundary of the lane is represented by

The right lane boundary is represented by

In the Frenet coordinate system, the center line of the lane is the reference line, the lane width is w _lane , the unit normal vector at the reference line position s is represented by r ^⊥ (s), the tangent angle corresponding to the reference line point F is α _F , and the set of all points R within the lane boundary is:

Among them, r(s) is the point corresponding to the reference line position s, γ is a coefficient with a certain value range, which is used to ensure that all points are located within the lane boundary, s _min and s _max are the minimum and maximum values of the lane reference line length, and the road modeling of the present invention complies with the road network standard of OpenDRIVE.

(2) Formal expression of traffic rules

This sub-step formally expresses the constraints of explicit and implicit traffic regulations on each traffic participant vehicle. As shown in Table 1, the present invention divides the constraints of each traffic participant vehicle into constraints related to road traffic, constraints related to vehicle motion, and constraints related to cooperative driving, and further subdivides these three categories of constraints:

Table 1. Constraints of traffic participants

(3) Prediction of accessible areas for traffic participants

车辆运动相关的可达区域

以及协同驾驶相关的可达区域

并给出了各个可达区域的具体算法，交通参与者的总体可达区域

为以上三种可达区域的交集。本发明提出的三种可达区域计算方法能够保证结果的超近似性，即计算得到的可达区域大于真实的可达区域，因此能够严格验证自动驾驶车辆是否可能与其他交通参与者发生碰撞，不存在漏检可能，保障自动驾驶车辆预期轨迹的安全性。This sub-step divides the reachable area of the ith traffic participant vehicle in the time interval [t _k , t _k+1 ] into the reachable area related to road traffic

Accessible area related to vehicle movement

and accessible areas related to collaborative driving

The specific algorithms for each accessible area and the overall accessible area for traffic participants are given.

The three reachable area calculation methods proposed in the present invention can ensure the super-approximation of the results, that is, the calculated reachable area is larger than the actual reachable area, so it can strictly verify whether the autonomous driving vehicle is likely to collide with other traffic participants, and there is no possibility of missed detection, thus ensuring the safety of the expected trajectory of the autonomous driving vehicle.

a. Accessible areas related to road traffic

由道路条件决定，主要考虑了车道跟随、路标指示以及V2X信息等因素的影响。如式4所示，对有关车道跟随的可达区域

有关路标的可达区域

以及有关V2X信息的不可达区域的补集

取交集可得与第i个交通参与车在时间间隔[t_k，t_k+1]内考虑道路交通约束，能够到达可达区域

Accessible areas related to road traffic

Determined by road conditions, the influence of factors such as lane following, road signs and V2X information is mainly considered. As shown in Formula 4, the reachable area of lane following

Accessible areas with road signs

and the complement of the unreachable area for V2X information

Taking the intersection, we can get the reachable area that the ith traffic participant vehicle can reach within the time interval [t _k , t _k+1 ] considering the road traffic constraints.

主要考虑了车道边界约束R_{road_boundary}和车速R_velocity的影响。考虑到车辆在车道上行驶时，存在无限多的路径，本发明沿车道的最短路径来执行车道边界的约束，保证车辆沿此路径行驶的最远位置大于等于其实际能够行驶到的最远位置；当确定了车辆沿某条路径可行驶的最远位置时，认为车辆能够到达车道边界内垂直于该路径的任何位置，以实现有关车道跟随的可达区域的超近似估计。本发明在遵循车道网络前提下计算通过某一路段最短路径的下限。在Frenet坐标系下，参考线为车道的中心线，

为参考线长度s_F处对应的最短路径，s_F为点F处对应的参考线长度，|Δα(s)|为参考线切向角随参考线长度s变化量的绝对值，则有：Reachable area for lane following

The influence of lane boundary constraint R _{road_boundary} and vehicle speed R _velocity is mainly considered. Considering that there are infinite paths when a vehicle is traveling on a lane, the present invention executes lane boundary constraints along the shortest path of the lane to ensure that the farthest position of the vehicle along this path is greater than or equal to the farthest position it can actually travel to; when the farthest position that the vehicle can travel along a certain path is determined, it is considered that the vehicle can reach any position within the lane boundary that is perpendicular to the path, so as to achieve a super-approximate estimate of the reachable area of the relevant lane following. The present invention calculates the lower limit of the shortest path through a certain road section under the premise of following the lane network. In the Frenet coordinate system, the reference line is the center line of the lane,

is the shortest path corresponding to the reference line length s _F , s _F is the reference line length corresponding to point F, |Δα(s)| is the absolute value of the change in the reference line tangent angle with the reference line length s, then:

s corresponds to the shortest path ξ one by one, let s = f(ξ). Considering the constraint of R _velocity , assume that the farthest distance that other traffic participants can reach within the time interval [t _k , t _k+1 ] is d _max , let the current position of other traffic participants s ₀ be the corresponding shortest path ξ ₀ , and the farthest position s _max that can be reached corresponds to the shortest path ξ _max should satisfy:

ξ _max =ξ ₀ +d _max (6)

令s_min＝s₀，考虑到自车的尺寸，如图4所示，取车道内的

与

间的连线构建垂直于相应车道边界的多边形可得到有关车道跟随的可达区域

So let s _max = f(ξ _max ), in order to overestimate

Let s _min = s ₀ , considering the size of the vehicle, as shown in Figure 4, take the

and

The polygon perpendicular to the lane boundary can be constructed by connecting the lines between the lanes to obtain the reachable area of the lane.

考虑了约束条件R_{road_permit}，例如道路的车道线、转向路标或路障等标识时，各个交通参与者应遵循道路标志的指示在指定区域中行驶，基于路标的指示获取可行驶的车道信息，根据可行驶车道的边界构建相应的多边形生成有关路标指示的可达区域。Accessible areas indicated by road signs

Taking into account the constraint condition R _{road_permit} , such as lane lines, turn signs or roadblocks on the road, each traffic participant should follow the instructions of the road signs to drive in the designated area, obtain the drivable lane information based on the instructions of the road signs, and construct the corresponding polygon according to the boundary of the drivable lane to generate the reachable area indicated by the relevant road signs.

的坐标信息，将其投影到位置域中，可得到有关V2X信息的不可达区域

在位置域上的投影。V2X warning information can indicate the inaccessible areas on the road (such as areas where traffic accidents have occurred or areas under construction). Obtaining the inaccessible area polygon through V2X technology

By projecting the coordinate information of V2X into the location domain, we can get the unreachable area of V2X information.

Projection onto the position domain.

b. Accessible area related to vehicle movement

主要考虑了R_acceleration约束，即绝对加速度a_max的限制，假设其他交通参与者的p_x，0＝0，p_y，0＝0，v_x，0＝v₀，v_y，0＝0，参考点在时间t的可达区域可由如图5(a)所示的一个中心为c(t)、半径为r(t)的圆表示：Accessible area related to vehicle kinematics

The R _acceleration constraint, i.e., the limit of absolute acceleration a _max , is mainly considered. Assuming that p _x,0 = 0, _py,0 = 0, v _x,0 = v ₀ , _vy,0 = 0 of other traffic participants, the reachable area of the reference point at time t can be represented by a circle with a center of c(t) and a radius of r(t) as shown in Figure 5(a):

The boundary of the reachable region is described by a two-dimensional function [b _x (t), _by (t)] ^T , where:

The reachable area of other participating vehicles at a given time interval [t _k , t _k+1 ] is constrained by two circles and a concave boundary at t _k and t _k+1 . Regardless of the occupied area of the vehicle size, this reachable area can be overestimated by a convex hexagon M (m ₁ , ..., m ₆ ) as shown in Figure 5(a):

可由图5(b)中的一个六边形N(n₁，...，n₆)来过度近似，其中：When considering the reachable area of the vehicle size, the exact reachable area

It can be over-approximated by a hexagon N (n ₁ , ..., n ₆ ) in FIG. 5( b ), where:

与绝对加速度值

The above derivation assumes a specific relative position and orientation of other traffic vehicles. In order to obtain the reachable area based on acceleration in any case, the reachable area based on acceleration must be rotated and transformed according to the actual initial position and orientation of other traffic participants. When the weather conditions are bad, the constraint R _{cautious_driving} should be followed and the maximum speed of the vehicle should be adjusted to

With absolute acceleration

c. Accessible areas related to collaborative driving

是由交通参与者共同决定的，主要考虑有关安全距离不可达区域

以及有关路权的可达区域

如式11所示，有关路权的可达区域与有关安全距离不可达区域的补集取交集即可得到与车辆协同相关的可达区域。Accessible areas related to vehicle cooperative driving

It is jointly determined by traffic participants, mainly considering the unreachable areas related to safety distance

and accessible areas related to the right of way

As shown in Formula 11, the reachable area related to vehicle collaboration can be obtained by taking the intersection of the complement of the reachable area related to road rights and the unreachable area related to safety distance.

为超近似值，应使有关安全距离的不可达区域

为欠近似值。考虑到自动驾驶车辆的反应时间小于人类的反应时间，为了适当的低估安全距离，本发明令驾驶员的反应时间δ_human＝0.5s，自动驾驶车辆反应时间δ_vehicle∈(0，δ_human)，并假设后车在反应时间内减速制动，并且制动时保持最大制动减速度，则不考虑车辆占用区域的安全距离：Traffic regulations require that two adjacent vehicles should maintain a sufficient safety distance and should not endanger any traffic participants behind them when changing lanes. Let v _r,0 and v _f,0 be the initial speeds of the rear and front vehicles respectively, δ is the reaction delay time, that is, the time between the front vehicle fully braking in the initial state and the rear vehicle fully braking. Referring to the longitudinal safety distance in the RSS model, in order to obtain the accessible area

For over-approximation, the inaccessible area of the relevant safety distance should be

is an under-approximation. Considering that the reaction time of an autonomous vehicle is shorter than that of a human, in order to appropriately underestimate the safety distance, the present invention assumes that the driver's reaction time δ _human = 0.5s, the autonomous vehicle's reaction time δ _vehicle ∈ (0, δ _human ), and assumes that the following vehicle decelerates and brakes within the reaction time, and maintains the maximum braking deceleration during braking, then the safety distance of the vehicle occupied area is not considered:

As shown in Figure 6, when other traffic participants are traveling in the same direction as the ego vehicle, but not in the same lane as the ego vehicle, or in the same lane as the ego vehicle and behind the ego vehicle, other traffic participants need to maintain a certain safety distance from the ego vehicle. In order to make an under-approximation calculation of the safety distance and take into account the influence of the occupied areas of the ego vehicle and other traffic participants on the safety distance, for the time interval [t _k , t _k+1 ], the traffic participants in different lanes at t _k are considered to be the front vehicles of the autonomous driving vehicle, and the safety distance is:

At t _k+1 , if the traffic participant vehicle is considered to be the rear vehicle of the autonomous driving vehicle, the safety distance is:

Where v _{ego, 0} is the initial speed of the ego vehicle, and v _{obj, 0} is the initial speed of other traffic participants.

The safety distance derived above is the value in the Frenet coordinate system. Taking the vehicle reference point as the starting point, combined with the approximate safety distance derived above, the overall approximate safety interval within a certain period of time can be obtained. Constructing a polygon perpendicular to the corresponding lane boundary can obtain the unreachable area of the relevant safety distance.

主要针对重叠道路，考虑了约束条件R_{road_right}，即车辆必须避让具有优先通行权的其他车辆。重叠区域没有交通信号灯时，距离重叠区域更近的车辆享有优先通行的权利。重叠区域设有交通信号灯时，有关路权的可达区域

需考虑车辆能否在剩余可通行时间内进入道路重叠区域。令第i个交通参与车到重叠区域中心的距离为

当前速度为v_obj，0，至t_k+1剩余信号灯指示可通行时间为t_passable，不同情况下第i个交通参与车在[t_k，t_k+1]与路权相关可达区域如表2所示。Accessible area related to right of way

For overlapping roads, the constraint R _{road_right} is considered, that is, vehicles must give way to other vehicles with right of way. When there is no traffic light in the overlapping area, vehicles closer to the overlapping area have the right of way. When there is a traffic light in the overlapping area, the accessible area of the right of way

It is necessary to consider whether the vehicle can enter the road overlap area within the remaining passable time. Let the distance from the i-th traffic participant vehicle to the center of the overlap area be

The current speed is v _{obj, 0} , and the remaining traffic light-indicated passable time to t _k+1 is t _passable . The accessible areas related to the right of way of the i-th traffic participant vehicle at [t _k , _{t k+1} ] under different conditions are shown in Table 2 .

Table 2. Accessible areas of other traffic participants [t _k , t _k+1 ] corresponding to the right of way

(4) Safety verification of the expected trajectory of the vehicle

This sub-step calculates the area occupied by the vehicle based on the size of the vehicle and the expected trajectory, and determines whether the area occupied by the vehicle intersects with the reachable areas of all other traffic participants to verify whether the expected trajectory of the vehicle is safe. If there is no intersection, the expected trajectory of the vehicle is legal and safe. Otherwise, the expected trajectory of the vehicle is unsafe and needs to travel a verified alternative safe trajectory.

a. Trajectory safety verification method

According to the expected trajectory generated by the ego vehicle decision system, the occupied area in the time period [t _k , t _k+1 ] is calculated. Considering the size of the vehicle, the occupied area of the ego vehicle at a certain moment is a rectangle Q (q ₁ , q ₂ , q ₃ , q ₄ ), the coordinates of the ego vehicle reference point are (p _x , p _y ), and the angle between the longitudinal axis of the vehicle body and the x-axis is β. According to the rotation axis formula, we can get:

如果一条轨迹产生的自车占用区域

与其他交通参与者在此时间内的所有合法行为对应的可达区域

没有交集，则这条轨迹可被验证为符合合法安全；反之，验证结果为不安全。The occupied area of the ego vehicle at each point of the expected trajectory corresponding to [t _k , t _k+1 ] can be calculated, and the reachable area of the ego vehicle A _ego (t _k , t _k+1 ) can be obtained by taking the union. In order to verify whether a continuous expected trajectory is safe, it is necessary to check whether there is an intersection between the occupied area of the ego vehicle and the reachable areas of other traffic participants in the continuous time interval. As shown in formula (16), taking the union of the reachable areas of each other traffic participant vehicle can obtain the reachable areas of all other traffic participants

If a trajectory generates an ego vehicle occupying area

The reachable area corresponding to all legal actions of other traffic participants during this time

If there is no intersection, the trajectory can be verified as legal and safe; otherwise, the verification result is unsafe.

b. Online security verification timing

若

在t_k之前被成功验证为符合合法安全，则将这段轨迹记为

终止时间为t_k+1，为避免自动驾驶车辆运行不安全轨迹，为其提供一条连续的备用安全轨迹

并将

与备用安全轨迹

的串联表示为

如图7中验证周期i＝1与i＝2所示，若在t_k之前，

被成功验证为

且成功生成了备用安全轨迹

则可认为

验证成功，允许自动驾驶车辆在t_k时刻进入自动驾驶模式，开始执行已验证的轨迹

若直到t_k时刻，轨迹安全验证结果仍不成功，则在t_k时刻仍然执行前一个已成功验证的

直到再次成功验证一个新的预期轨迹；若当前行驶的备用安全轨迹的可执行时长t_e满足t_a≤t_e≤t_b(t_a与t_b为两个常量，且满足0＜t_a≤t_b≤T_alter)，则计算特定时间内自车合法可达区域，继续生成对应时长为T_alter的备用安全轨迹；若t_e＜t_a仍无法成功生成备用安全轨迹，考虑到约束R_{safe_distance}，车辆在其对应的车道上全力减速行驶对应的轨迹必定满足合法安全，因此沿当前车道中心线生成备用安全轨迹并规定在此轨迹上全力减速行驶，仍将后续生成的备用安全轨迹记为

Let the expected trajectory corresponding to the time interval t _k to t _k+1 be

like

If it is successfully verified as legal and safe before t _k , this trajectory is recorded as

The termination time is tk ₊₁ . To avoid the unsafe trajectory of the autonomous driving vehicle, a continuous backup safe trajectory is provided for it.

and will

With alternate safety track

The series representation of

As shown in FIG7 for verification cycles i=1 and i=2, if before t _k ,

Successfully verified as

And successfully generated a backup safety trajectory

It can be considered

The verification is successful, allowing the autonomous vehicle to enter the autonomous driving mode at time _tk and start executing the verified trajectory.

If the trajectory safety verification result is still unsuccessful at time _tk , the previous successfully verified step will be executed at time _tk.

Until a new expected trajectory is successfully verified again; if the executable time _{t e} of the current backup safety trajectory satisfies _ta ≤t _e ≤t _b ( _ta and t _b are two constants, and satisfy 0＜ _ta ≤t _b ≤T _alter ), then calculate the legally reachable area of the vehicle within a specific time, and continue to generate backup safety trajectories with a corresponding duration of T _alter ; if t _e ＜t _{a and} the backup safety trajectory still cannot be successfully generated, considering the constraint R _{safe_distance} , the trajectory corresponding to the vehicle's full deceleration in its corresponding lane must meet the legal safety requirements, so the backup safety trajectory is generated along the center line of the current lane and it is stipulated that the vehicle should decelerate at full speed on this trajectory, and the subsequently generated backup safety trajectory is still recorded as

2. Generation of alternative safe trajectories based on legally accessible areas

The generation of backup safety trajectories mainly includes two sub-steps: the selection of the endpoint of the backup safety trajectory of the ego vehicle and the generation of the backup safety trajectory of the ego vehicle. The selection of the endpoint of the backup safety trajectory of the ego vehicle is used to select the endpoint of the backup safety trajectory that meets the conditions. The generation of the backup safety trajectory of the ego vehicle uses the quintic polynomial method or the lane centerline method to generate the backup safety trajectory according to whether the ego vehicle needs to change lanes.

(1) Selection of the endpoint of the vehicle's backup safety trajectory

首先需要确定备用安全轨迹

的起点(x_s，y_s)和终点(x_e，y_e)。针对

应满足的三个条件：首先，

需平滑连接

因此将

的起点(x_s，y_s)设定为

的终点；其次，

需要保证自车的合法安全且沿整条备用安全轨迹自车的占用区域与其他交通参与车的可达区域不产生交集，因此

应在自车合法可达区域

中：This sub-step defines the conditions that the backup safety trajectory should meet and gives the principles for selecting the endpoint of the backup safety trajectory.

First, we need to determine the alternative safe trajectory

The starting point ( _xs , _ys ) and the end point ( _xe , _ye ).

Three conditions should be met: First,

Smooth connection required

Therefore,

The starting point (x _s , y _s ) is set to

The end point; secondly,

It is necessary to ensure the legal safety of the vehicle and the occupied area of the vehicle along the entire backup safety trajectory does not intersect with the accessible areas of other traffic participating vehicles.

The vehicle should be in an area that is legally accessible by car.

middle:

是自车有关路标的可达区域，

是有关V2X信息的不可达区域，

为所有其他交通参与者有关车辆和道路的可达区域

的补集，式(17)所有区域对应时间为[t_k+1，t_k+1+T_alter]，由于自车的轨迹尚未确定，因此仅计算其他交通参与车有关车辆和道路的可达区域

需在

中选择备用安全轨迹

的终点(x_e，y_e)；最后，

的长度应满足正常行驶需求，可将自车过渡到安全区域中，令自车的速度为v_ego，0，最大制动减速度为a_max，brake，为保证备用安全轨迹

足以使自车的减速至停止，在Frenet坐标系中，假定备用安全轨迹

的起点(x_s，y_s)对应参考线的长度为s_start，终点(x_e，y_e)对应参考线的长度为s_end，s_end应满足：in

is the accessible area of the vehicle related to the road signs,

It is an unreachable area for V2X information.

For all other traffic participants regarding the accessible area of the vehicle and the road

The complement of all regions in formula (17) corresponds to the time [t _k+1 , t _k+1 +T _alter ]. Since the trajectory of the ego vehicle has not been determined, only the reachable areas of other traffic participating vehicles and roads are calculated.

Need to

Select an alternate safe trajectory

The end point (x _e , y _e ); finally,

The length of should meet the normal driving requirements, and the ego vehicle can be transferred to the safe area. Let the ego vehicle's speed be v _ego,0 and the maximum braking deceleration be a _max,brake . To ensure the backup safety trajectory

Sufficient to slow the vehicle down to a stop, assuming an alternative safety trajectory in the Frenet coordinate system

The length of the reference line corresponding to the starting point ( _xs , _ys ) is _sstart , and the length of the reference line corresponding to the _end point ( _xe , _ye ) is send. _send should satisfy:

Considering that lane changing is more risky than staying in the original lane, the present invention stipulates that lane changing is not allowed unless necessary. When lane changing is necessary, the adjacent lane with a larger legally accessible area of the vehicle is selected for lane changing. First, determine whether the point with the longest reference line length corresponding to the center line of the lane where the vehicle is currently located in the legally accessible area of the vehicle satisfies formula (18). If so, select the point with the longest reference line length on the center line of the lane as the _end point of the backup safety trajectory; otherwise, determine the point with the longest reference line length corresponding to the center line of the adjacent lane in the legally accessible area of the vehicle and determine whether it satisfies formula (18). If so, select the point with the _longest reference line length on the center line of the lane as the end point of the backup safety trajectory for the next step of calculation; if none of the above conditions are met, the generation of the backup safety trajectory fails.

(2) Generation of backup safe trajectories for the ego vehicle

的终点后，需要根据备用安全轨迹的起点和终点选择适当的轨迹规划方法生成备用安全轨迹参考线。本发明根据是否需要换道来讨论备用安全轨迹参考线的生成方法。无需换道的情况如图8(a)所示，备用安全轨迹的起点和终点在同一车道中心线上，自车无需换道；需要换道的情况如图8(b)、(c)所示，备用安全轨迹的起点和终点不在同一车道中心线上，自车需要进行换道行驶。换道行驶的起点为备用安全轨迹的起点(x_s，y_s)，需要在目标车道中心线上选取一个换道完成的终点(x_c，y_c)。在Frenet坐标系下，为了给换道预留足够的距离，令车辆从开始换道到完成换道需要的距离为d_change，换道终点(x_c，y_c)对应的参考线长度s_change应满足s_change≥s_start+d_change，本发明取s_change＝s_start+d_change，在目标车道中心线上检索，可得到满足条件的换道终点(x_c，y_c)。当给出的备用安全轨迹终点(x_e，y_e)对应的参考线长度s_end大于(x_c，y_c)对应的参考线长度s_change，如图8(b)所示，自车在完成换道后，可继续沿着(x_c，y_c)到(x_e，y_e)的目标车道中心线继续行驶；备用安全轨迹的参考线即为曲线(x_s，y_s)-(x_c，y_c)-(x_e，y_e)；若(x_e，y_e)对应的参考线长度s_end小于(x_c，y_c)对应的参考线长度s_change，如图8(c)所示，则仅生成部分换道轨迹，在生成的参考线上选取终点

使其对应的参考线长度为s_end。When the alternate safety trajectory is selected

After reaching the end point of the backup safety trajectory, it is necessary to select an appropriate trajectory planning method according to the starting point and end point of the backup safety trajectory to generate a backup safety trajectory reference line. The present invention discusses the method for generating the backup safety trajectory reference line based on whether a lane change is required. The situation where no lane change is required is shown in Figure 8(a). The starting point and end point of the backup safety trajectory are on the same lane center line, and the vehicle does not need to change lanes; the situation where lane change is required is shown in Figures 8(b) and (c). The starting point and end point of the backup safety trajectory are not on the same lane center line, and the vehicle needs to change lanes. The starting point of the lane change is the starting point of the backup safety trajectory ( _xs , _ys ), and it is necessary to select an end point ( _xc , _yc ) on the center line of the target lane where the lane change is completed. In the Frenet coordinate system, in order to reserve enough distance for lane changing, the distance required for the vehicle from the start of lane changing to the completion of lane changing is d _change , and the reference line length s _change corresponding to the lane changing end point (x _c , y _c ) should satisfy s _change ≥s _start +d _change . The present invention takes s _change =s _start +d _change , and searches on the center line of the target lane to obtain the lane changing end point (x _c , y _c ) that meets the conditions. When the reference line length s _end corresponding to the given backup safety trajectory end point (x _e , _ye ) is greater than the reference line length s _change corresponding to (x _c , y _c ), as shown in Figure 8(b), the vehicle can continue to drive along the center line of the target lane from (x _c , y _c ) to (x _e , _ye ) after completing the lane change; the reference line of the backup safety trajectory is the curve (x _s , y _s )-(x _c , y _c )-(x _e , _ye ); if the reference line length s _end corresponding to (x _e , _ye ) is less than the reference line length s _change corresponding to (x _c , y _c ), as shown in Figure 8(c), only a partial lane change trajectory is generated, and the end point is selected on the generated reference line.

Let the corresponding reference line length be s _end .

The process of generating an alternative safety trajectory based on the centerline method is as follows: as shown in Figure 8(a), the starting point and end point of the alternative safety trajectory are on the same lane centerline, and the ego vehicle does not need to change lanes. The lane centerline in the range of ( _xs , _ys ) to ( _xe , _ye ) is selected as the reference line of the alternative safety trajectory.

The process of generating an alternative safety trajectory based on the quintic polynomial method is as follows: As shown in Figure 9, the initial position of the ego vehicle reference point is used as the coordinate origin, the velocity direction of the ego vehicle is used as the x-axis to establish a coordinate system, and it is assumed that the lateral velocity of the ego vehicle is very small and its influence is ignored. The lane change trajectory reference line equation of the ego vehicle is:

y＝a ₀ x ⁵ +a ₁ x ⁴ +a ₂ x ³ +a ₃ x ² +a ₄ x+a ₅ (19)

Assume that the current speed of the ego vehicle is v _ego,0 , the yaw rate is ω _ego,0 , the point of tangency between the ego vehicle and the center line of the target lane after the lane change is completed is P _c , and the lane line sensor can measure the coordinates of the target point of the lane change center line as (x _c , y _c ), slope _rc and curvature K _c . In the current coordinate system, the coordinates of the initial position (x s , y _s ) of the lane change are (0, 0), and the slope of the point (x _s , y _s ) is 0, and the curvature of the reference line corresponding to the point (x _s , y _s ) can be obtained _from the speed and yaw rate of the ego vehicle at this point. Therefore, the reference line equation at the point (x _s , y _s ) should satisfy the following conditions:

At the lane change endpoint (x _c , y _c ), the slope and curvature of the backup safety trajectory reference line should be consistent with the slope and curvature of the lane centerline here. Therefore, the reference line equation at (x _c , y _c ) should satisfy the following relationship:

According to the above two equations, the fifth-order polynomial reference line equation when the vehicle changes lanes can be solved using information such as the initial state of the vehicle, the position of the lane change end point (x _c , y _c ) and the center line parameters of the lane change target lane:

According to the coefficient matrix, a ₃ , a ₄ , and a ₅ can be solved, and the calculation of the backup safety trajectory reference line is completed.

The present invention uses Prescan and Matlab joint simulation to test and verify the proposed online safety verification method for autonomous driving decision-making, where Prescan is used to build a variety of typical traffic scenarios, including the construction of road networks and the establishment of the road environment, parameter settings and trajectory planning of each traffic participant, and the construction of the perception layer; Matlab is used for data transmission and processing, and the online safety verification algorithm is implemented through the m language.

为4m，

为1.3m，w_ego为2m的矩形中；同时自车及其他交通参与者的最大加速度a_max，accel及最大减速度a_max，brake设置为8m/s²，备用安全轨迹终点对应的参考线长度s_end为

在线安全验证再规划周期为0.1s，每段预期轨迹

对应的时长为0.3s，备用安全轨迹的时长T_alter为0.6s，t_a为0.1s，t_b为0.3s。本发明假定预期轨迹在0.3s内不发生变化，由于在线安全验证再规划周期为0.1s，预期轨迹

对应的时长为[t_k，t_k+0.3s]，对每一段预期轨迹都需要验证3次，3次中若有一次验证成功，则判定

验证成功，在t_k时开始执行

只有当3次验证全都失败时，

判定为验证失败，在t_k时继续执行上一个验证成功的备用安全轨迹

可得到每个验证周期内应执行的轨迹。The present invention encloses the point mass model of other traffic participants in a rectangle where l _obj,i is 4m and w _obj,i is 2m; the point mass model of the vehicle is enclosed in

is 4m,

is 1.3m, w _ego is 2m; at the same time, the maximum acceleration a _{max, accel} and the maximum deceleration a _{max, brake} of the ego vehicle and other traffic participants are set to 8m/s ² , and the reference line length s _end corresponding to the end point of the backup safety trajectory is

The online safety verification replanning cycle is 0.1s, and each expected trajectory

The corresponding duration is 0.3s, the duration of the backup safety trajectory T _alter is 0.6s, t _a is 0.1s, and t _b is 0.3s. The present invention assumes that the expected trajectory does not change within 0.3s. Since the online safety verification replanning cycle is 0.1s, the expected trajectory

The corresponding duration is [t _k , t _k +0.3s]. Each expected trajectory needs to be verified three times. If one of the three verifications succeeds, it is determined

Verification is successful, and execution starts at t _k

Only when all three verifications fail,

The verification is judged as failed, and the backup safety trajectory that was successfully verified is continued at t _k.

The trace that should be executed in each verification cycle can be obtained.

Based on the test specifications, common test scenarios in autonomous driving research and development, and typical scenarios that need to be dealt with when driving a vehicle, the present invention creates three typical scenarios of one-way multi-lane, two-way multi-lane, and crossroad in Prescan for testing. The parameters of each simulation scenario are shown in Table 3:

Table 3. Simulation scenario parameter settings

The online safety verification method proposed in the present invention is verified under dangerous conditions in the typical traffic scenarios defined above, and the analysis of the test results is mainly carried out from the two perspectives of safety and real-time performance. Safety means that the technology can adapt to a large number of traffic scenarios, accurately judge the safety of the expected trajectory, and the generated backup safety trajectory can guide the vehicle to a safe state, does not actively cause traffic accidents and does not violate traffic regulations, and strictly guarantees the legal safety of the vehicle. Real-time performance means that the technology can complete the calculation and processing within the specified time, can timely verify whether the expected trajectory is safe, and generate a backup safety trajectory.

(1) Simulation verification for one-way multi-lane scenarios

The initial speeds and relative positions of the traffic participants in the one-way three-lane curve scene constructed by the present invention are shown in FIG10 :

The online safety verification results of the one-way three-lane curve scenario are shown in Figure 11:

At t = 4.5s, the expected trajectory of the ego vehicle failed to be verified because the vehicle ahead was preparing to change lanes. The vehicle executed the latest successfully verified backup safety trajectory within 3.9s-4.2s and began to change lanes to the right adjacent lane. At 6.0s, the expected trajectory of the ego vehicle failed to be verified. The vehicle executed the latest successfully verified backup safety trajectory within 5.7s-6.0s. At 5.1s, 6.6s, 7.5s, and 8.7s, the expected trajectory of the ego vehicle was successfully verified and executed. The ego vehicle executed the backup safety trajectory within 4.5s-5.1s, 6.0s-6.6s, 7.2s-7.5s, and 7.8s-8.7s, and the successfully verified expected trajectory was executed at other times.

(2) Simulation verification for two-way multi-lane scenarios

The initial speeds and relative positions of each traffic participant in the two-way six-lane straight road scene constructed by the present invention are shown in FIG12:

The online safety verification results of the two-way six-lane straight road scenario are shown in Figure 13:

At t = 4.2s, the expected trajectory of the ego vehicle failed to be verified because the vehicle in front of it on the right was preparing to change lanes. The vehicle executed the latest successfully verified backup safety trajectory within 3.9s-4.2s and began to change lanes to the left adjacent lane. At 5.7s, the expected trajectory of the ego vehicle failed to be verified. The latest successfully verified backup safety trajectory within 5.4s-5.7s was executed. At 3.5s, 5.4s, 6.6s, and 8.7s, the expected trajectory of the ego vehicle was successfully verified and the expected trajectory was executed. The ego vehicle executed the backup safety trajectory within 4.2s-5.4s and 5.7s-6.6s, and the successfully verified expected trajectory was executed at other times.

(3) Simulation verification for intersection scenarios

The initial speeds and relative positions of the traffic participants in the two-lane intersection scene constructed by the present invention are shown in FIG14 :

The online safety verification results of the two-lane intersection scenario are shown in Figure 15:

At t = 4.8s, the expected trajectory of the ego vehicle failed to be verified because a vehicle was passing through the intersection in front of the left. The vehicle executed the latest successfully verified backup safety trajectory within 3.5s-4.8s and decelerated. At 5.7s, the vehicle on the right violated the right of way and passed through the intersection. The expected trajectory of the ego vehicle failed to be verified. The vehicle executed the latest successfully verified backup safety trajectory within 5.4s-5.7s and decelerated. The expected trajectory of the ego vehicle was successfully verified at 2.7s, 5.4s, 8.1s, and 10.2s, and the expected trajectory was executed. The ego vehicle executed the backup safety trajectory within 4.8s-5.4s and 5.7s-8.1s, and the successfully verified expected trajectory was executed at the rest of the time.

The present invention measured the calculation time of the expected trajectory safety verification and the backup safety trajectory generation of the three scenarios 10 times, and obtained the average time of the expected trajectory safety verification and the backup safety trajectory generation of the three scenarios, as well as the average value of the total time and its overall variance, as shown in Table 4:

Table 4. Simulation verification calculation time

Through the simulation analysis of the above three scenarios, it can be proved that the online safety verification technology proposed by the present invention can adapt to a variety of complex traffic scenarios, accurately judge the safety of the expected trajectory, ensure the legal safety of the vehicle under dangerous conditions, and not actively cause traffic accidents. The generated backup safety trajectory can guide the vehicle to a safe state. In addition, the calculation time of each scenario can be controlled within the replanning cycle of 100ms, indicating that the online safety verification technology proposed by the present invention can complete the expected trajectory safety verification and the generation of backup safety trajectories in a timely manner.

Although the present invention has been described above with reference to the embodiments, various modifications may be made thereto and parts thereof may be replaced by equivalents without departing from the scope of the present invention. In particular, as long as there is no structural conflict, the various features in the embodiments disclosed in the present invention may be used in combination with each other in any manner, and the fact that these combinations are not exhaustively described in this specification is only for the sake of omitting space and saving resources. Therefore, the present invention is not limited to the specific embodiments disclosed herein, but includes all technical solutions falling within the scope of the claims.

Claims (3)

1. An automatic driving decision online verification method based on traffic participant reachability analysis is characterized in that: the method comprises the following steps:

the method comprises the following steps: the safety verification of the expected track of the self-vehicle based on reachability analysis comprises the steps of firstly, calculating relevant vehicles, relevant roads and reachable areas which are cooperatively related with the vehicles of other traffic participants according to collected traffic environment and self-vehicle state information, and meanwhile, calculating the occupied areas of the self-vehicle according to the expected track information given by a track planner and a physical model of the self-vehicle; the system is used for judging whether the intersection exists between the occupied area of the self vehicle and the reachable area of other traffic participants so as to verify whether the expected track is safe or not as the input of the second step;

step two: generating a standby safe track based on a legal reachable area, if the expected track is safe, firstly calculating the legal reachable area of the self-vehicle within a specific time, further calculating the end point position of the standby safe track according to the legal reachable area of the self-vehicle, judging whether the lane needs to be changed or not, and further selecting a quintic polynomial method or generating the standby safe track along the center line of the lane; and judging whether the standby safety track is successfully generated or not based on whether the occupied area of the self vehicle is located in the legal reachable area of the self vehicle or not along the standby safety track, if the standby safety track is successfully generated, executing an expected track which is successfully verified at the corresponding time, otherwise, if the expected track is unsafe or the standby safety track is unsuccessfully generated, judging whether the executable time of the standby safety track is insufficient or not, and determining whether to generate a new standby safety track and executing the previous standby safety track which is successfully verified.

2. The automated driving decision online verification method based on traffic participant reachability analysis according to claim 1, wherein: the step 1 specifically comprises:

(1) Vehicle and road modeling

The substep models the vehicle and the road, divides the vehicle into a self vehicle and other traffic participating vehicles, and further divides the other traffic participating vehicles into a motor vehicle and a non-motor vehicle; all vehicles are dynamically modeled into a point mass model, a kinematics model adopts a second-order integrator model, all roads are divided into non-overlapping roads and overlapping roads, each road consists of a motor lane, a non-motor lane and a road shoulder, a lane line boundary is a road boundary, a central point of an overlapping area is defined as an intersection point of central lines of all branches, and a Frenet coordinate system is adopted to model the roads;

(2) Cross-regular formalized expression

The sub-step formally expresses the constraint of the implicit traffic regulation on each traffic participating vehicle, divides the constraint of each traffic participating vehicle into the constraint related to road traffic, the constraint related to vehicle movement and the constraint related to cooperative driving, and further subdivides the three types of constraints;

(3) Traffic participant vehicle reachable area prediction

The sub-step divides the reachable areas of other traffic-participating vehicles into reachable areas related to road traffic, reachable areas related to vehicle kinematics and reachable areas related to vehicle cooperative driving according to three kinds of cross-regulation constraints, and defines the intersection of the three reachable areas as the reachable areas of the traffic-participating vehicles at a future moment; calculating the road related reachable area through lane following, road signs and V2X warning information; the accessible area related to the vehicle kinematics takes the kinematic model and the size of the vehicle into consideration, and the accessible area related to the vehicle kinematics is subjected to super-approximate calculation through a dynamically divergent hexagon; the vehicle cooperative driving related reachable area takes the right of way of the vehicle and the safe distance keeping rule into consideration, and the related reachable area is calculated;

(4) Safety verification of expected track of self-vehicle

The method comprises the steps of calculating the occupied area of the self vehicle according to the size of the vehicle and the expected track, judging whether the occupied area of the self vehicle is intersected with the accessible areas of all other traffic participants to verify whether the expected track of the self vehicle is safe or not, if not, the expected track of the self vehicle is legal and safe, otherwise, the expected track of the self vehicle is unsafe, and a verified standby safe track needs to be driven.

3. The automated driving decision online validation method based on traffic participant reachability analysis of claim 1, wherein: the step 2 specifically comprises:

(1) Backup safe trajectory end point selection for a bicycle

The substep selects a safe track end point for the vehicle standby, and the selection of the track end point considers the following factors: a. smoothly connecting the standby safe track with the corresponding expected track; b. the standby safety track ensures the legal safety of the self-vehicle (namely, the occupied area of the self-vehicle and the accessible area of other traffic participating vehicles do not generate intersection at the same time); c. the length of the standby safe track meets the requirement of normal driving, and the self vehicle can be transited to a safe area;

(2) Safe trajectory generation for self-vehicle standby

The sub-step generates a standby safety track according to the track end point, adopts a quintic polynomial method to generate the standby safety track when the own vehicle changes the lane when the lane needs to be changed, adopts a lane central line method to generate the standby safety track when the lane does not need to be changed, and judges whether the standby safety track is successfully generated by checking whether the area occupied by the own vehicle along the standby safety track is included in the legal reachable area of the own vehicle.

Images