Trajectory planning method, system, vehicle, equipment and medium based on grid search

Through the method based on grid search, the problem of high computing power requirements and poor anti-interference ability of driving trajectory planning in complex environments is solved, and obstacle point cloud rasterization and dynamic trajectory planning under small computing power are realized. Simplified engineering applications.

CN115077553BActive Publication Date: 2024-11-22CHONGQING CHANGAN AUTOMOBILE CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210754909.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-06-30
Publication Date
2024-11-22
Estimated Expiration
2042-06-30

AI Technical Summary

Technical Problem

The driving trajectory planning in complex environments requires high computing power, weak anti-interference ability, inconvenient engineering application, and difficult to achieve dynamic and time-lapse planning.

Method used

Adopt obstacle avoidance trajectory planning method based on grid search, detect obstacle target categories, speeds, outlines and other information through sensors, generate raster diagrams and search graphs to find the best paths, and use spline fitting for path smoothing and optimization , outputs executable planning paths.

Benefits of technology

The gridding of obstacle point clouds is achieved with a smaller computing power, which improves the efficiency and anti-interference ability of path planning, simplifies engineering implementation, and is suitable for dynamic trajectory planning in complex environments.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115077553B_ABST
    Figure CN115077553B_ABST
Patent Text Reader

Abstract

The present invention discloses an obstacle avoidance trajectory planning method based on grid search, which relates to the technical field of autonomous driving. A grid is generated according to a reference baseline and perception information, a high-precision map boundary is added to the grid, obstacles are projected in the grid and expanded, the grid is processed according to the current vehicle position and kinematic constraints, and it is judged whether there is a completely passable "column". If there is a completely passable column, an anti-shake process is performed on the safety line, and a passable path is generated, and the passable path is smoothed. If there is no completely passable column, it is found whether there is a left or right obstacle avoidance safety line. If there is a left or right obstacle avoidance safety line, the safety line is anti-shake processed, a passable path is generated and smoothed. If there is no left or right obstacle avoidance safety line, the path at the previous moment is output as the current path. The method is suitable for dynamic trajectory planning of low-speed driving in complex environments.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to automobile automatic driving technology, and specifically to dynamic trajectory planning for low-speed driving in complex environments, including real-time path planning such as obstacle avoidance and lane deviation. Background Art

[0002] As the foundation and technical core of the autonomous driving system, the trajectory planning algorithm is used to generate a set of paths connecting the starting point and the end point of the vehicle on the basis of complying with the spatiotemporal constraints such as vehicle dynamics and kinematics, driving road environment, and traffic regulations. Among them, graph search is a commonly used method in trajectory planning. Graph search usually generates a grid through high-precision map information and environmental information recognized by sensors, and then uses a search algorithm to search for paths on the generated grid. The invention patent application with the invention name of "A method and device for unmanned vehicle path planning" with the publication number CN111857160A introduces an artificial potential field method to obtain the repulsion coefficient of each grid, uses a hybrid A star for path search, and uses a gradient descent algorithm to smooth the trajectory. The core of the patent is still based on a hybrid A star. The introduction of an artificial potential field can indeed improve the efficiency of the search, but in a complex environment, the repulsion coefficient of each grid needs to be calculated, which requires high computing power, and the hybrid A star algorithm itself also requires high computing power. It is difficult to achieve dynamic real-time planning in engineering for the entire scheme. The Chinese invention patent application with publication number CN113267199A and title “Driving trajectory planning method and device” generates multiple paths by offsetting the baseline, and then calculates the cost function of each candidate path, and finally selects the optimal path, which can realize the planning of dynamic scenes. The Chinese invention patent with publication number CN108241370B and title “Method and device for obtaining obstacle avoidance path through grid map” discloses a method for obtaining obstacle avoidance path through grid map, by obtaining a static global map of the obstacle avoidance area; taking each grid in the static grid map as a grid point; determining the obstacle area in the static grid map; determining the obstacle cost value of each grid point in the reachable area; and determining the obstacle avoidance path according to the position of the starting point and the end point in the static grid map and the obstacle cost value of each grid point in the reachable area. However, the core idea of ​​the above method is closer to sampling optimization, and it is only based on the grid map for planning, and no search algorithm is used, so the computing power is relatively low. The invention patent application with publication number CN113561968A, "A vehicle horizontal parking trajectory planning algorithm and device, vehicle, and storage medium", is a parking planning method based on geometry, which is not suitable for planning driving scenes. The invention patent application with publication number CN 113467456 A, "A path planning algorithm for searching for specific targets in an unknown environment", also proposes a new path planning algorithm. The difference is that the core algorithm of the path planning is the RRT planning algorithm guided by the artificial potential field.

[0003] Therefore, it is necessary to develop a trajectory planning method that can make full use of the grid to conveniently represent the surrounding environment, while having low computing power requirements, being easy to engineer, having strong anti-interference ability, and being able to meet the requirements of dynamic planning in real time during driving. Summary of the invention

[0004] The present invention aims to solve the problems in the prior art that the driving trajectory planning in complex environments requires high computing power, has weak anti-interference ability, and is inconvenient for engineering application, and proposes an obstacle avoidance trajectory planning method based on grid search. In the scenarios of autonomous driving in urban areas and underground garages, due to the presence of various irregular obstacles such as walls, pillars, and curbs, these obstacles are difficult to represent using conventional attributes of obstacle targets. At this time, the obstacles are usually represented by obstacle point clouds. The method proposed in the present invention can achieve rasterization of obstacle point clouds with less computing power.

[0005] The technical solution of the present invention to solve the above technical problems is: using sensors to detect information such as obstacle target categories, speeds, and contours, locating the reference baseline required for the current local path planning through a positioning system and global path planning, integrating the reference baseline and obstacle target information, using a grid-based search method to find the best path, and smoothing and optimizing the path through spline fitting and other methods, and finally outputting an executable planning path, and then realizing the tracking control of the trajectory through a control module. Mainly based on the reference baseline, a grid is generated in the Frenet coordinate system, wherein the direction along the reference baseline is defined as a "row" and the direction perpendicular to the reference baseline is defined as a "column", and the map boundary information and sensor perception information are projected into the grid to form a planning space. When performing a graph search, no search method such as A* is used to search nodes one by one. In order to improve efficiency, the "columns" of the planning space are directly searched to find the optimal "column" to form a safety baseline, and then the optimal value of each "row" grid is calculated according to the current vehicle position and the grid passable area, and finally the searched path is smoothly optimized. Due to the improvement in the search method, the search can be completed in a shorter period of time. In the dynamic planning of vehicle driving paths, simpler algorithms can be used, which saves computing power and is easy to implement in engineering, thus achieving the purpose of dynamic planning of vehicle paths under complex paths. Specifically, it is a method for obstacle avoidance trajectory planning based on grid search. According to the current position of the vehicle, the points on the global path are matched, and the path points at a fixed distance from the current vehicle position in the forward direction are selected as the reference baseline for local path planning. The trajectory planning is performed according to the reference baseline and the vehicle perception information, and the planned trajectory is output to the lateral and longitudinal control modules, and then the whole vehicle is controlled by the whole vehicle actuator. The trajectory planning is specifically as follows: a grid is generated according to the reference baseline and the perception information, a high-precision map boundary is added to the grid, obstacles are projected in the grid and expanded, and the grid is processed according to the current vehicle position and kinematic constraints. It is determined whether there is a completely passable "column". If there is a completely passable column, the safety line is de-jittered, and a passable path is generated, and the passable path is smoothed. If there is no completely passable column, it is found whether there is a left or right obstacle bypassing safety line. If there is a left or right obstacle bypassing safety line, the safety line is de-jittered, and a passable path is generated and smoothed. If there is no left or right obstacle bypassing safety line, the path at the previous moment is output as the current path.

[0006] Further preferably, the sensor module detects obstacle information, performs perception fusion and environmental cognition processing, and outputs it to the local path planning module. The local path planning module generates a grid in the Frenet coordinate system according to the reference trajectory, wherein the direction along the reference baseline is defined as a "row", the "row" spacing is the distance between two points on the reference baseline, and the direction perpendicular to the reference baseline is defined as a "column". The "row" spacing is determined according to the vehicle speed and the computing power of the controller, and the "column" spacing is determined according to the width of the road. The ratio of the "row" spacing to the "column" spacing meets certain constraints, and this constraint comes from the vehicle kinematics requirements.

[0007] Further preferably, the "rows" of the grid gradually increase from the current position of the vehicle to the reference line direction, the "columns" of the grid gradually increase from left to right, the value of the passable grid is set to 1, and the value of the impassable grid is set to 0.

[0008] Further preferably, the adding of the high-precision map boundary in the grid is specifically to project the boundary of the high-precision map in the grid, the grid outside the boundary is an impassable area, the grid inside the boundary is a passable area, the perceived obstacle and the obstacle point cloud are expanded and then put into the grid, and the grid blocked by the obstacle is an impassable area. The safety line is a line obtained by translating the reference baseline in the "column" direction after searching and determining the translation distance.

[0009] Further preferably, the generating of the passable path specifically includes: generating a passable grid based on the current vehicle posture according to the current vehicle position in the grid and the vehicle kinematic constraints; calculating the grid coordinates with the minimum cost from each "row" in the passable grid to the safety line based on the current vehicle posture, and forming the searched path as the passable path.

[0010] Further preferably, projecting the obstacle in the grid includes: obtaining the type, dynamic and static attributes, and vertex positions of the target obstacle from the sensor module; converting the obstacle vertices to the coordinate system of the reference baseline; finding the grid corresponding to the coordinates of each vertex in the obstacle; and expanding the grid corresponding to the coordinate vertex, and the grids within the expansion range are all inaccessible areas.

[0011] Further preferably, projecting obstacles in the grid includes: converting the obstacle point cloud to the coordinate system of the reference baseline; selecting rows and columns of the grid to generate a closed area according to the autonomous driving scenario; the obstacle point cloud in the closed area is a valid point cloud, and calculating the vertical distance and the corresponding longitudinal distance of each valid point cloud to the reference baseline; classifying the valid point clouds according to the vertical distance values; sorting the point clouds according to the corresponding longitudinal distances on the reference baseline; and merging all point clouds and projecting them into the grid according to the distance, classification, and sorting of the valid point clouds to the reference baseline.

[0012] Further preferably, judging whether there is a completely passable "column" specifically includes: if the rows corresponding to the current column in the grid are all 1, it is a passable row and column, forming a completely passable row and column set; forming a search list with the "column" where the reference baseline is located, sorted by vertical distance to the left and right sides; according to the search list, searching in the completely passable row and column set according to the limited search width.

[0013] More preferably, finding whether there is a safety line for bypassing obstacles on the left or right side further includes: selecting an area with a "column" value smaller than a reference baseline, judging and finding the minimum passable "column" value in each "row" in the area, and this "column" is the safety line for bypassing obstacles on the left side. Selecting an area with a "column" value larger than a reference baseline, judging and finding the maximum passable "column" value in each "row" in the area, and this "column" is the safety line for bypassing obstacles on the right side.

[0014] The present invention also proposes an obstacle avoidance trajectory planning system based on grid search, comprising: a sensor module, a positioning module, a global path planning module, a local path planning module, a lateral and longitudinal control module, a reference baseline module, and a vehicle actuator. The reference baseline module matches the points on the global path determined by the global path planning module according to the current position of the vehicle provided by the positioning module, and selects the path points at a fixed distance from the current vehicle position in the forward direction as the reference baseline for local path planning. The local path planning module performs trajectory planning according to the reference baseline and the perception information provided by the sensor module, outputs the trajectory to the lateral and longitudinal control module, and then controls the whole vehicle through the vehicle actuator. The trajectory planning The method is divided into: generating a grid according to the reference baseline and perception information, adding a high-precision map boundary to the grid, projecting obstacles in the grid and expanding it, processing the grid according to the current vehicle position and kinematic constraints, judging whether there is a completely passable "column", if there is a completely passable column, stabilizing the safety line, generating a passable path, and smoothing the passable path; if there is no completely passable column, finding out whether there is a left or right obstacle avoidance safety line; if there is a left or right obstacle avoidance safety line, stabilizing the safety line, generating a passable path and smoothing it; if there is no left or right obstacle avoidance safety line, outputting the path at the previous moment as the current path.

[0015] Further preferably, the sensor module detects obstacle information, performs perception fusion and environmental cognition processing, and outputs it to the local path planning module. The local path planning module generates a grid in the Frenet coordinate system according to the reference trajectory, wherein the direction along the reference baseline is defined as a "row", the "row" spacing is the distance between two points on the reference baseline, and the direction perpendicular to the reference baseline is defined as a "column". The "row" spacing is determined according to the vehicle speed and the computing power of the controller, and the "column" spacing is determined according to the width of the road. The ratio of the "row" spacing to the "column" spacing meets certain constraints, and this constraint comes from the vehicle kinematics requirements.

[0016] Further preferably, projecting the obstacle in the grid includes: obtaining the type, dynamic and static attributes, and vertex positions of the target obstacle from the sensor module; converting the obstacle vertices to the coordinate system of the reference baseline; finding the grid corresponding to the coordinates of each vertex in the obstacle; and expanding the grid corresponding to the coordinate vertex, and the grids within the expansion range are all inaccessible areas.

[0017] Further preferably, projecting obstacles in the grid includes: converting the obstacle point cloud to the coordinate system of the reference baseline; selecting rows and columns of the grid to generate a closed area according to the autonomous driving scenario; the obstacle point cloud in the closed area is a valid point cloud, and calculating the vertical distance and the corresponding longitudinal distance of each valid point cloud to the reference baseline; classifying the valid point clouds according to the vertical distance values; sorting the point clouds according to the corresponding longitudinal distances on the reference baseline; and merging all point clouds and projecting them into the grid according to the distance, classification, and sorting of the valid point clouds to the reference baseline.

[0018] In a third aspect, the present invention further proposes a car, comprising the above-mentioned obstacle avoidance trajectory planning system based on grid search.

[0019] In a fourth aspect, the present invention also proposes a control device, comprising: a processor and a memory; wherein the processor runs a program corresponding to the executable program code by reading the executable program code stored in the memory, so as to implement the obstacle avoidance trajectory planning method based on grid search as described above.

[0020] In a fifth aspect, the present invention proposes a computer-readable storage medium, on which a program or instruction is stored, and the program or instruction can be loaded and run by a processor to execute the obstacle avoidance trajectory planning method based on grid search as described above.

[0021] In the process of realizing obstacle avoidance planning, the present invention no longer traverses and searches each node, but searches the "columns" to reduce the number of searches, finds the safety line, and then calculates the grid with the minimum cost from each "row" of passable grids to the safety line, thereby forming a planned path. It can save controller computing resources to the maximum extent and realize real-time planning of paths in dynamic scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 The present invention is a system architecture diagram of obstacle avoidance trajectory planning based on grid search;

[0023] Figure 2 The present invention provides a flow chart of obstacle avoidance trajectory planning based on grid search;

[0024] Figure 3 Schematic diagram of the traversable grid. DETAILED DESCRIPTION

[0025] To facilitate understanding of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0026] The present invention will be further described below in conjunction with the accompanying drawings.

[0027] like Figure 1 The present invention is a grid search-based obstacle avoidance trajectory planning system architecture diagram, comprising a sensing module (1), a positioning module (2), a global path planning module (3), a reference baseline module (4), a local path planning module (5), a transverse and longitudinal control module (6), and a vehicle actuator (7). The reference baseline module matches the point on the global path determined by the global path planning module according to the current position of the vehicle provided by the positioning module, thereby selecting a path point at a fixed distance from the current vehicle position in the forward direction as a reference baseline for local path planning. The local path planning module performs trajectory planning according to the reference baseline and the perception information provided by the sensing module, outputs the trajectory to the transverse and longitudinal control module, and then controls the vehicle through the vehicle actuator, thereby realizing the automatic driving function.

[0028] The sensing system is mainly based on visual sensors, laser radars, millimeter-wave radars, ultrasonic radars and other sensors, and is used to detect obstacles, lane lines and other information; the positioning system mainly provides the current vehicle position and heading information; the global path system mainly plans a path connecting the starting point to the end point based on the starting point, the end point and the map information; the reference baseline module matches the point on the global path according to the current position of the vehicle, so as to select the path point at a fixed distance from the current vehicle position in the forward direction as the reference baseline for local path planning; the local path planning module performs trajectory planning such as obstacle avoidance and longitudinal speed according to the sensing information and the reference baseline information; the lateral and longitudinal control performs lateral and longitudinal control according to the trajectory of the local path planning to ensure that the vehicle follows the trajectory within a certain error range; the whole vehicle actuator includes EMS, ESP, EPS, gear, BCM, etc., wherein EPS receives lateral control angle request information, EMS receives longitudinal control torque request information, and ESP receives longitudinal control deceleration request information.

[0029] Figure 2The figure shows the obstacle avoidance trajectory planning flow chart based on grid search of the present invention. Generate a grid based on the reference baseline and perception information, add a high-precision map boundary to the grid, project obstacles in the grid and expand, process the grid based on the current vehicle position and kinematic constraints, and check whether there is a completely passable "column". If there is a completely passable row and column, perform anti-shake processing on the safety line, generate a path, and smooth the path. If there is no completely passable row and column, find out whether there is a left or right safety line. If there is a left or right obstacle avoidance safety line, perform anti-shake processing on the safety line, generate a path and smooth it. If there is no left or right safety line, output the path of the previous moment.

[0030] When the vehicle is in the automatic driving mode; the sensor module detects the obstacle information, performs perception fusion and environmental cognition processing, and then outputs it to the local path planning module. At the same time, the reference baseline module outputs the reference trajectory to the local path planning module; the local path planning module generates a grid in the Frenet coordinate system, where the direction along the reference baseline is defined as a "row", the "row" spacing is the distance between two points on the reference baseline, and this distance is related to the vehicle speed and the computing power of the controller, and the direction perpendicular to the reference baseline is defined as a "column". The proportional relationship between the "column" spacing and the "row" spacing needs to consider the kinematic information of the vehicle. At the same time, the total number of "columns" takes into account the width of the vehicle's driving area. In this embodiment, the "row" of the grid gradually increases from the current position of the vehicle to the front of the reference line, and the "column" of the grid gradually increases from left to right. The value of the passable grid is set to 1, and the value of the impassable grid is set to 0;

[0031] The boundaries in the high-precision map are projected into the grid. The grid outside the boundary is the inaccessible area, and the grid inside the boundary is the accessible area.

[0032] The perceived obstacles and obstacle point clouds are expanded and then put into the grid. The grid blocked by obstacles is an inaccessible area.

[0033] The grid is processed according to the vehicle kinematic constraints. The main processing principle is based on the current passable area. If the vehicle meets the vehicle kinematic constraints, the grid is reachable. Otherwise, even if the grid is within the boundary of the high-precision map and there are no obstacles filled, it will still be processed as an impassable grid.

[0034] Search and determine the safety line. The safety line is a line obtained by translating the reference baseline in the "column" direction, and the specific value of the translation is achieved based on the search.

[0035] In order to prevent the instability of obstacles from causing fluctuations in the safety line, an anti-shake process is performed on the safety line. The processing method may adopt a conventional anti-shake processing method used by those skilled in the art.

[0036] According to the current vehicle position in the grid and the processing of vehicle kinematic constraints, a passable grid based on the current vehicle posture is generated.

[0037] The grid coordinates with the minimum cost from each "row" in the passable grid to the safety line based on the current vehicle posture are calculated to form the searched path.

[0038] The generated path may be smoothed by, but is not limited to, spline interpolation, least squares method, and other methods.

[0039] Since it is difficult to fully represent complex scenes with obstacles, point clouds are needed to supplement them. The grid projection of obstacles is divided into two parts. Specifically, the obstacles can be projected onto the grid in the following ways: Implementation method 1: vertex projection method:

[0040] (1) Obtain the type, dynamic and static properties, and vertex position of the target obstacle from the sensor module;

[0041] (2) Convert the target vertex to the coordinate system of the reference baseline;

[0042] (3) Find the grid corresponding to the coordinates of each vertex in each target;

[0043] (4) The grids corresponding to the coordinate vertices are expanded, and the grids within the expansion range are treated as inaccessible areas.

[0044] Implementation method 2 is point cloud projection method, in which the obstacle point cloud is projected onto a grid.

[0045] (1) Convert the obstacle point cloud to the coordinate system corresponding to the reference path;

[0046] (2) According to different autonomous driving scenarios, select rows and columns of the grid to generate a closed area. For example, select the 2nd to 30th columns and the 1st to 50th rows of the grid to form a closed area.

[0047] (3) The obstacle point cloud within the closed area is a valid point cloud. The perpendicular distance from each point cloud to the reference baseline and the corresponding longitudinal distance on the reference baseline are calculated. The perpendicular distance to the reference baseline can be defined as negative on the left and positive on the right.

[0048] (4) Classify the perceived point cloud left or right according to the positive and negative values ​​of the vertical distance;

[0049] (5) Sort the point clouds from near to far according to their longitudinal distance on the reference baseline, where the point clouds closer to the vehicle have a higher priority;

[0050] (6) Merge all point clouds according to the calculated distance, classification, and sorting;

[0051] (7) After collision processing, the point cloud is projected into the grid.

[0052] Regarding the method of searching and determining the safety line, this embodiment may specifically adopt the following search method:

[0053] (1) Determine whether there is a completely passable "column". A completely passable "column" means that every "row" of the grid on this "column" is passable. The specific judgment method is based on the reference baseline, forming a search list sorted by vertical distance to the left and right sides of the "column" where the reference baseline is located. Search according to the search list. If the search is successful, the safety line is de-shaked, otherwise the next search is performed. The detailed search steps for safe and passable "columns" are as follows:

[0054] Step 1: Find completely passable columns in the formed grid. The specific method is to check whether the rows corresponding to each column of the grid are all 1. If so, the column is a passable row and column, and finally a completely passable row and column set is formed.

[0055] Step 2: Generate a search list based on the reference baseline sorting. The principle of generating the list is to search on the left first, and sort from the column where the current reference baseline is located. For example, if the column number of the current reference baseline is 21, one of the generated search lists is {21, 20, 22, 19, 23, 17...}.

[0056] Step 3: According to the search list and the limited search width, search in the generated fully accessible row and column set. If the fully accessible row and column set exists in the column range of the current search, it means the search is successful and the safety line is de-shaked. If the search is unsuccessful, proceed to the next search. An example is given below to illustrate: the search width refers to the range of the current column extending to the left and right. For example, if the search width is 2 and the current column is 21, the column range to be searched is {19, 20, 21, 22, 23}. If there are 19, 20, 21, 22, 23 in the fully accessible row and column, it means the search is successful, otherwise it is unsuccessful.

[0057] (2) Prioritize the left bypass, that is, prioritize the search for a safety line on the left side of the reference baseline. The specific method is to select the left area of ​​interest, that is, the area where the "column" value is smaller than the reference baseline. Determine and find the minimum passable "column" value in each "row" in the area. This "column" is the safety line for bypassing obstacles on the left side. If the search is successful, the safety line is de-shaked. If the search is unsuccessful, the next step of the search is performed.

[0058] (3) Detour on the right, that is, search for a safety line on the right side of the reference baseline. The specific method is to select the area of ​​interest on the right side, that is, the area where the "column" value is greater than the reference baseline. Find the maximum passable "column" value in each "row" in the area. This "column" is the safety line for detour on the right side. If the search is successful, the safety line is de-shaked. Otherwise, the current area is an impassable area, the path at the previous moment is maintained, and the parking operation is performed.

[0059] The safety line is processed to prevent the instability of obstacles from causing fluctuations in the safety line. According to the current position of the vehicle in the grid and the processing of vehicle kinematic constraints, a passable grid based on the current vehicle posture is generated.

[0060] A method for calculating the traversable grid based on the current vehicle position and vehicle kinematic constraints is described as follows:

[0061] (1) According to the vehicle positioning information, match the position in the grid.

[0062] (2) Starting from the current grid, the vehicle can reach at most three grids on the left, middle, and right of the next “row”.

[0063] Figure 3 The figure shows a schematic diagram of a passable grid, where number 2 indicates the current position of the vehicle, number 3 indicates an inaccessible grid due to obstacle filling, number 1 indicates a passable grid, and number 0 indicates a different passable grid generated according to the vehicle kinematic constraints. When generating the final passable grid for driving, number 2 can only move to the grids in the left middle 3 positions in the next column, so only these 3 grids have values ​​of 1. Although the other grids in this column are not occupied by obstacles, they are still filled with 0 and treated as inaccessible grids.

[0064] Although only a limited number of embodiments are explained in detail, the present invention is not limited to the applications listed in the description and embodiments, and it can be fully applied to various fields suitable for the present invention. For engineers and technicians familiar with the field, modifications, supplements and substitutions that can be easily realized, therefore, the present invention should not be considered to be limited by the previous description without departing from the general concept defined by the claims and equivalent scope.

Claims

1. An obstacle avoidance trajectory planning method based on grid search, characterized in that: According to the current position of the vehicle, match the points on the global path, select the path points at a fixed distance from the current vehicle position in the forward direction as the reference baseline for local path planning, perform trajectory planning according to the reference baseline and vehicle perception information, output the planned trajectory to the lateral and longitudinal control modules, and then control the whole vehicle through the whole vehicle actuator. The trajectory planning is specifically as follows: generate a grid in the Frenet coordinate system according to the reference baseline and perception information, add a high-precision map boundary to the grid, project obstacles in the grid and expand, process the grid according to the current vehicle position and kinematic constraints, determine whether there is a completely passable "column", if there is a completely passable column as a safety line, perform anti-shake processing on the safety line, generate a passable path and perform smoothing, if there is no completely passable column, find out whether there is a left or right obstacle bypassing safety line, if there is a left or right obstacle bypassing safety line, perform anti-shake processing on the safety line, generate a passable path and perform smoothing, if there is no left or right obstacle bypassing safety line, output the path at the previous moment as the current path; The projecting of obstacles in the grid includes: converting the obstacle point cloud to the coordinate system of the reference baseline; selecting the rows and columns of the grid to generate a closed area according to the autonomous driving scenario; the obstacle point cloud in the closed area is a valid point cloud, and the vertical distance and the corresponding longitudinal distance of each valid point cloud to the reference baseline are calculated, wherein the vertical distance of the reference baseline is defined as negative on the left and positive on the right, and the sensed point clouds are classified left and right according to the positive and negative values ​​of the vertical distance; and the point clouds are sorted from near to far according to the longitudinal distance corresponding to the point clouds on the reference baseline, wherein the closer the point cloud is to the vehicle, the higher the priority; and all point clouds are merged according to the calculated distance, classification, and sorting; and the point clouds are projected into the grid after collision processing.

2. The method according to claim 1, characterized in that The detected obstacle information is subjected to perception fusion and environmental cognition processing to perform local path planning. Specifically, a grid is generated in the Frenet coordinate system based on the reference baseline, where the direction along the reference baseline is defined as a "row", the "row" spacing is the distance between two points on the reference baseline, and the direction perpendicular to the reference baseline is defined as a "column". The "row" spacing is determined according to the vehicle speed and the controller's computing power, and the "column" spacing is determined according to the width of the road.

3. The method according to claim 2, characterized in that The "rows" of the grid gradually increase from the current position of the vehicle to the reference line direction, and the "columns" of the grid gradually increase from left to right. The value of the passable grid is set to 1, and the value of the impassable grid is set to 0.

4. The method according to claim 1, characterized in that: The method of adding a high-precision map boundary to the grid specifically includes projecting the boundary of the high-precision map into the grid, the grid outside the boundary is an inaccessible area, the grid inside the boundary is a passable area, the obstacle point cloud is expanded and then put into the grid, and the grid blocked by the obstacle is an inaccessible area.

5. The method according to claim 1, characterized in that The safety line is a line obtained by translating the reference baseline in the "column" direction after searching and determining the translation distance.

6. The method according to claim 1, characterized in that The generating of the passable path specifically includes: generating a passable grid based on the current vehicle posture according to the current vehicle position in the grid and the vehicle kinematic constraints; calculating the grid coordinates with the minimum cost from each "row" in the passable grid to the safety line based on the current vehicle posture, so that the searched path is the passable path.

7. The method according to any one of claims 1 to 6, characterized in that: Projecting obstacles in the grid includes: obtaining the type, dynamic and static properties, and vertex positions of the obstacle; converting the obstacle vertices to the coordinate system of the reference baseline; finding the grid corresponding to the coordinates of each vertex in the obstacle; and expanding the grid corresponding to the coordinate vertex, and the grids within the expansion range are all inaccessible areas.

8. The method according to claim 3, characterized in that Determining whether there is a completely passable "column" specifically includes: if the rows corresponding to the current column in the grid are all 1, it is a passable row and column, forming a completely passable row and column set; forming a search list sorted by vertical distance to the left and right sides of the "column" where the reference baseline is located; based on the search list, searching in the completely passable row and column set according to the limited search width.

9. The method according to any one of claims 1 to 6 and 8, characterized in that: Finding whether there is a safety line for bypassing obstacles on the left or right side further includes selecting an area with a "column" value smaller than a reference baseline, judging and finding the minimum passable "column" value in each "row" in the area, and this "column" is the safety line for bypassing obstacles on the left side; selecting an area with a "column" value larger than a reference baseline, judging and finding the maximum passable "column" value in each "row" in the area, and this "column" is the safety line for bypassing obstacles on the right side.

10. An obstacle avoidance trajectory planning system based on grid search, characterized in that: include: The sensor module, the positioning module, the global path planning module, the local path planning module, the lateral and longitudinal control module, the reference baseline module, and the vehicle actuator. The reference baseline module matches the points on the global path determined by the global path planning module according to the current position of the vehicle provided by the positioning module, and selects the path points at a fixed distance from the current vehicle position in the forward direction as the reference baseline for local path planning. The local path planning module performs trajectory planning according to the reference baseline and the perception information provided by the sensor module, outputs the trajectory to the lateral and longitudinal control module, and then controls the vehicle through the vehicle actuator. The trajectory planning is as follows: according to the reference baseline and the perception information in Fren et coordinate system, add high-precision map boundaries to the grid, project obstacles in the grid and expand them, process the grid according to the current vehicle position and kinematic constraints, determine whether there is a completely passable "column", if there is a completely passable column, perform anti-shake processing on the safety line, generate a passable path, and smooth the passable path. If there is no completely passable column, find out whether there is a safety line for bypassing obstacles on the left or right side. If there is a safety line for bypassing obstacles on the left or right side, perform anti-shake processing on the safety line, generate a passable path and smooth it. If there is no safety line for bypassing obstacles on the left or right side, output the path at the previous moment as the current path; The projecting of obstacles in the grid includes: converting the obstacle point cloud to the coordinate system of the reference baseline; selecting the rows and columns of the grid to generate a closed area according to the autonomous driving scenario; the obstacle point cloud in the closed area is a valid point cloud, and the vertical distance and the corresponding longitudinal distance of each valid point cloud to the reference baseline are calculated, wherein the vertical distance of the reference baseline is defined as negative on the left and positive on the right, and the sensed point clouds are classified left and right according to the positive and negative values ​​of the vertical distance; and the point clouds are sorted from near to far according to the longitudinal distance corresponding to the point clouds on the reference baseline, wherein the closer the point cloud is to the vehicle, the higher the priority; and all point clouds are merged according to the calculated distance, classification, and sorting; and the point clouds are projected into the grid after collision processing.

11. The system according to claim 10, characterized in that The sensor module detects obstacle information, performs perception fusion and environmental cognition processing, and outputs it to the local path planning module. The local path planning module generates a grid in the Frenet coordinate system based on the reference baseline. The direction along the reference baseline is defined as a "row", and the "row" spacing is the distance between two points on the reference baseline. The direction perpendicular to the reference baseline is defined as a "column". The "row" spacing is determined according to the vehicle speed and the controller's computing power, and the "column" spacing is determined according to the width of the road.

12. The system according to claim 11, characterized in that The "rows" of the grid gradually increase from the current position of the vehicle to the reference line direction, and the "columns" of the grid gradually increase from left to right. The value of the passable grid is set to 1, and the value of the impassable grid is set to 0.

13. The system according to claim 10, characterized in that The method of adding a high-precision map boundary to the grid specifically includes projecting the boundary of the high-precision map into the grid, the grid outside the boundary is an inaccessible area, the grid inside the boundary is a passable area, the obstacle point cloud is expanded and then put into the grid, and the grid blocked by the obstacle is an inaccessible area.

14. The system according to claim 10, characterized in that The safety line is a line obtained by translating the reference baseline in the "column" direction after searching and determining the translation distance.

15. The system according to claim 10, characterized in that The generating of the passable path specifically includes: generating a passable grid based on the current vehicle posture according to the current vehicle position in the grid and the vehicle kinematic constraints; calculating the grid coordinates with the minimum cost from each "row" in the passable grid to the safety line based on the current vehicle posture, and forming the searched path as the passable path.

16. The system according to any one of claims 10 to 15, characterized in that: Projecting obstacles in the grid includes: obtaining the type, dynamic and static properties, and vertex positions of the obstacles from the sensor module; converting the vertices of the obstacles to the coordinate system of the reference baseline; finding the grid corresponding to the coordinates of each vertex in the obstacle; and expanding the grid corresponding to the coordinate vertex, and the grids within the expansion range are all inaccessible areas.

17. The system according to claim 12, characterized in that Determining whether there is a completely passable "column" specifically includes: if the rows corresponding to the current column in the grid are all 1, it is a passable row and column, forming a completely passable row and column set; forming a search list sorted by vertical distance to the left and right sides of the "column" where the reference baseline is located; based on the search list, searching in the completely passable row and column set according to the limited search width.

18. The system according to any one of claims 10-15 and 17, characterized in that: Finding whether there is a safety line for bypassing obstacles on the left or right side further includes selecting an area with a "column" value smaller than a reference baseline, judging and finding the minimum passable "column" value in each "row" in the area, and this "column" is the safety line for bypassing obstacles on the left side; selecting an area with a "column" value larger than a reference baseline, judging and finding the maximum passable "column" value in each "row" in the area, and this "column" is the safety line for bypassing obstacles on the right side.

19. A car, characterized in that: It comprises an obstacle avoidance trajectory planning system based on grid search as described in any one of claims 10 to 18.

20. A control device, characterized in that: include: A processor and a memory; wherein the processor runs a program corresponding to the executable program code by reading the executable program code stored in the memory, so as to implement the obstacle avoidance trajectory planning method based on grid search as described in any one of claims 1 to 9.

21. A computer-readable storage medium, characterized in that: The readable storage medium stores programs or instructions, and the programs or instructions can be loaded and executed by a processor to execute the obstacle avoidance trajectory planning method based on grid search as described in any one of claims 1 to 9.