A vehicle automatic driving obstacle avoidance method and system based on high-precision map lane lines

Abstract

The application relates to a high-precision map lane line-based automatic driving obstacle avoidance method and system for a vehicle, the method comprising the following steps: Q1: the vehicle travels on an open road, high-precision map data information is used to acquire obstacle data information outside and inside the road boundary of the road where the vehicle is located, two-dimensional coordinate transformation is carried out, and the obstacle data information outside and inside the road boundary in a two-dimensional coordinate system is output; and Q2: the obstacle data information outside and inside the road boundary in the two-dimensional coordinate system is used. The application not only reduces system resource occupation, improves the efficiency of a perception algorithm and a planning decision algorithm, reduces system resource occupation of a perception module and a planning decision module, but also guarantees the experience effect of an automatic driving system, the heat map of a planning track in a costmap is detected outside the lane, the distance between a lane line, a planning track and a polygon inside the lane is detected to make the last safety guarantee, and the safety performance of the vehicle is guaranteed.

Description

Technical Field

[0001] This invention relates to the field of autonomous driving technology, and in particular to a method and system for autonomous vehicle obstacle avoidance based on lane lines in high-precision maps. Background Technology

[0002] With the development of autonomous driving technology, the functions of onboard sensors are becoming increasingly powerful. Common sensors include cameras and LiDAR. Cameras are used to acquire image information of objects around the vehicle, while LiDAR is used to acquire point cloud information of objects around the vehicle. By fusing information from various sensing devices, autonomous vehicles can perceive their surroundings more accurately, ensuring safe driving. Other sensing sensors include IMUs (Integrated Device Units).

[0003] Autonomous vehicles rely on the collaborative efforts of artificial intelligence, computer vision, radar, control devices, and global positioning systems to enable computers to operate motor vehicles automatically and safely without any human intervention. The autonomous driving process places high demands on the vehicle's automatic obstacle avoidance and requires a large number of safety redundancy mechanisms to ensure the vehicle's safety. If the vehicle fails to react immediately to the appearance of obstacles due to errors, it will lead to serious traffic accidents.

[0004] Currently, sensors used in autonomous driving technology include cameras, radar sensors, and lidar sensors. Each sensor has different performance characteristics, and a single sensor cannot accurately perceive the environment; therefore, data from multiple sensors is fused. However, as the level of autonomous vehicles increases, traditional multi-sensor target fusion can no longer meet the perception requirements of decision-making, and the large amount of redundant perception information brings great difficulties to decision-making. Therefore, how to solve these problems has become an urgent issue to address. Summary of the Invention

[0005] In view of the shortcomings of the prior art, the present invention provides a vehicle autonomous driving obstacle avoidance method and system based on high-precision map lane lines. This method not only reduces system resource consumption and improves the efficiency of perception and planning decision algorithms, but also reduces the system resource consumption of perception and planning decision modules. Furthermore, it ensures the user experience of the autonomous driving system. Outside the lane, the method detects the heat map of the planned trajectory in the cost map, and inside the lane, it detects the distance between the lane lines and the planned trajectory and the polygons within the lane to provide a final safety guarantee, thus ensuring vehicle safety performance.

[0006] To achieve the above and other related objectives, the present invention provides the following technical solution:

[0007] An obstacle avoidance method for autonomous driving vehicles based on lane lines in high-precision maps includes:

[0008] Q1: When a vehicle is driving on a public road, based on high-precision map data, acquire obstacle data information outside and inside the road boundary of the road where the vehicle is located, perform vehicle coordinate transformation, and output obstacle data information outside and inside the road boundary in the same coordinate system;

[0009] Q2: Based on the obstacle data information outside and inside the road boundary under the same coordinate system, the obstacle data information outside the road boundary is mapped to a two-dimensional costmap grid map, and the grids containing obstacles are marked, outputting the grid map data information with annotations. The obstacle data information inside the road boundary under the same coordinate system is mapped to a three-dimensional obstacle polygon. The perception data information obtained by the vehicle multi-sensor is coupled with the three-dimensional obstacle polygon data information to output the polygon data information of the obstacle.

[0010] Q3: Based on the labeled grid map data and the polygon data of the obstacles, and according to the lane line data obtained from the high-precision map, the system traverses the lane line data to determine whether there is a collision risk in the current lane, the target lane after lane change, and the lane change path, and outputs vehicle planning decision data.

[0011] Q4: Based on the vehicle planning and decision-making data, fit the vehicle's trajectory using the third-order Bézier curve algorithm and output the vehicle obstacle avoidance motion control data.

[0012] Furthermore, the third-order Bézier curve algorithm includes:

[0013] Q41: Based on the vehicle planning and decision data, obtain the coordinate data of the starting point and the target point of the vehicle's trajectory;

[0014] Q42: Based on the coordinate data of the starting point and target point of the vehicle's trajectory, according to the third-order Bézier curve function:

[0015] p(t)=p0(1-t) 3 +3p1t(1-t) 2 +3p2t(1-t)+p3t 3 Where p0 is the starting point coordinate of the vehicle's trajectory, p3 is the target point coordinate of the vehicle's trajectory, p1 is the coordinate of the intersection of the vehicle's trajectory and the lane centerline, and p2 is the center point coordinate of the vehicle's trajectory. The vehicle's trajectory is corrected, and the fitted vehicle trajectory is output.

[0016] Q43: Based on the fitted vehicle trajectory, output the vehicle obstacle avoidance motion control data.

[0017] Furthermore, the labeled raster map data information includes raster cells with obstacles marked in black and raster cells without obstacles marked in white.

[0018] Furthermore, the lane line data information includes lane line coordinate information, lane center line coordinate information, lane line attributes, and intersection attributes.

[0019] Furthermore, the lane line attributes include vehicle left-turn markings, vehicle right-turn markings, and vehicle straight-ahead markings.

[0020] Furthermore, in step Q3, the output vehicle planning decision data information specifically includes:

[0021] Q31: Based on the lane line data information obtained from the high-precision map, determine the lane type of the vehicle, wherein the lane type includes the left lane, the middle lane, and the right lane;

[0022] Q32: Based on the vehicle's lane type, retrieve the labeled grid map data and the polygon data of the obstacle to determine whether there is an obstacle in the current lane and whether there is an obstacle in the adjacent lane;

[0023] Q33: Based on the data information of whether there are obstacles in the current lane and whether there are obstacles in the adjacent lane, set a preset threshold for the distance between the vehicle and the obstacle, and output vehicle planning decision data information.

[0024] Furthermore, the preset threshold for the distance between the vehicle and the obstacle is α, where 50≤α≤100. When the distance between the vehicle and the obstacle is greater than α, there is no risk of collision; otherwise, there is a risk of collision.

[0025] To achieve the above and other related objectives, the present invention also provides a vehicle autonomous driving obstacle avoidance system based on high-precision map lane lines, the system comprising:

[0026] The high-precision map module is used to obtain map data information of the lane where the vehicle is located;

[0027] The perception module is based on multiple onboard sensors to acquire perception data information of the vehicle.

[0028] The planning and decision-making module is connected to the high-precision map module and the perception module, and is used to plan the vehicle obstacle avoidance trajectory based on the information output by the high-precision map module and the perception module.

[0029] Furthermore, the vehicle-mounted multi-sensor includes a camera, lidar, ultrasonic sensors, and a gyroscope.

[0030] The present invention has the following positive effects:

[0031] 1. This invention simplifies the obstacle handling process, reduces system resource consumption, improves the efficiency of perception and planning decision algorithms, and reduces the system resource consumption of perception and planning decision modules.

[0032] 2. This invention reduces the system resource consumption of the planning and decision-making module, improves the efficiency of the software algorithm, and ensures the real-time performance of the system by traversing lane line points.

[0033] 3. This invention detects potential collisions within the lane by traversing lane lines, enabling advance planning of lane changes or obstacle avoidance to ensure the user experience of the autonomous driving system. Outside the lane, it detects the heat map of the planned trajectory in the cost map. Inside the lane, it provides a final safety guarantee by detecting the distance between the lane lines, the planned trajectory, and the polygons within the lane, ensuring vehicle safety performance.

[0034] 4. This invention constructs a lane-changing path using Bézier curves, thus achieving the function of smooth lane changing. Attached Figure Description

[0035] Figure 1 This is a schematic diagram of the method flow of the present invention;

[0036] Figure 2 This is a schematic diagram of the sensing information and map lane lines of the present invention. Detailed Implementation

[0037] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0038] Example 1: As Figure 1 As shown, an autonomous driving obstacle avoidance method for vehicles based on lane lines in high-precision maps includes:

[0039] Q1: When a vehicle is driving on a public road, based on high-precision map data, acquire obstacle data information outside and inside the road boundary of the road where the vehicle is located, perform vehicle coordinate transformation, and output obstacle data information outside and inside the road boundary in the same coordinate system;

[0040] Q2: Based on the obstacle data information outside and inside the road boundary under the same coordinate system, the obstacle data information outside the road boundary is mapped to a two-dimensional costmap grid map, and the grids containing obstacles are marked, outputting the grid map data information with annotations. The obstacle data information inside the road boundary under the same coordinate system is mapped to a three-dimensional obstacle polygon. The perception data information obtained by the vehicle multi-sensor is coupled with the three-dimensional obstacle polygon data information to output the polygon data information of the obstacle.

[0041] Q3: Based on the labeled grid map data and the polygon data of the obstacles, and according to the lane line data obtained from the high-precision map, the system traverses the lane line data to determine whether there is a collision risk in the current lane, the target lane after lane change, and the lane change path, and outputs vehicle planning decision data.

[0042] Q4: Based on the vehicle planning and decision-making data, fit the vehicle's trajectory using the third-order Bézier curve algorithm and output the vehicle obstacle avoidance motion control data.

[0043] In this embodiment, the third-order Bézier curve algorithm includes:

[0044] Q41: Based on the vehicle planning and decision data, obtain the coordinate data of the starting point and the target point of the vehicle's trajectory;

[0045] Q42: Based on the coordinate data of the starting point and target point of the vehicle's trajectory, according to the third-order Bézier curve function:

[0046] p(t)=p0(1-t) 3 +3p1t(1-t) 2 +3p2t(1-t)+p3t 3 Where p0 is the starting point coordinate of the vehicle's trajectory, p3 is the target point coordinate of the vehicle's trajectory, p1 is the coordinate of the intersection of the vehicle's trajectory and the lane centerline, and p2 is the center point coordinate of the vehicle's trajectory. The vehicle's trajectory is corrected, and the fitted vehicle trajectory is output.

[0047] Q43: Based on the fitted vehicle trajectory, output the vehicle obstacle avoidance motion control data.

[0048] In this embodiment, the labeled raster map data information includes raster cells with obstacles marked in black and raster cells without obstacles marked in white.

[0049] In this embodiment, the lane line data information includes lane line coordinate information, lane center line coordinate information, lane line attributes, and intersection attributes.

[0050] In this embodiment, the lane line attributes include vehicle left-turn markings, vehicle right-turn markings, and vehicle straight-ahead markings.

[0051] In this embodiment, in step Q3, the output vehicle planning decision data information specifically includes:

[0052] Q31: Based on the lane line data information obtained from the high-precision map, determine the lane type of the vehicle, wherein the lane type includes the left lane, the middle lane, and the right lane;

[0053] Q32: Based on the vehicle's lane type, retrieve the labeled grid map data and the polygon data of the obstacle to determine whether there is an obstacle in the current lane and whether there is an obstacle in the adjacent lane;

[0054] Q33: Based on the data information of whether there are obstacles in the current lane and whether there are obstacles in the adjacent lane, set a preset threshold for the distance between the vehicle and the obstacle, and output vehicle planning decision data information.

[0055] In this embodiment, the preset threshold for the distance between the vehicle and the obstacle is α, where 50≤α≤100. When the distance between the vehicle and the obstacle is greater than α, there is no risk of collision; otherwise, there is a risk of collision.

[0056] To achieve the above and other related objectives, the present invention also provides a vehicle autonomous driving obstacle avoidance system based on high-precision map lane lines, the system comprising:

[0057] The high-precision map module is used to obtain map data information of the lane where the vehicle is located;

[0058] The perception module is based on multiple onboard sensors to acquire perception data information of the vehicle.

[0059] The planning and decision-making module is connected to the high-precision map module and the perception module, and is used to plan the vehicle obstacle avoidance trajectory based on the information output by the high-precision map module and the perception module.

[0060] In this embodiment, the vehicle-mounted multi-sensor includes a camera, lidar, ultrasonic sensors, and a gyroscope.

[0061] Example 2: Based on the vehicle autonomous driving obstacle avoidance method and system based on high-precision map lane lines in Example 1, the present invention will be further described below.

[0062] The positioning module publishes the current global positioning information of the vehicle in the Gaussian coordinate system in real time for use by perception, mapping, and planning.

[0063] Based on the collected offline global map information and current coordinate data, the high-precision map transmits the lane lines, lane boundaries, and lane center lines within a range of 150 meters in front and 50 meters behind the current vehicle via Rostopic. This data is then processed by the perception and planning decision-making module. Each lane line, lane boundary, and lane center line is composed of a series of Gaussian coordinate points.

[0064] Obstacle classification method: Based on localization and lane boundary information, the point cloud information perceived outside the lane is mapped onto a two-dimensional costmap grid map for processing in a two-dimensional form, reducing the processing time of perception data. The data inside the lane is fused through perception to generate dynamic polygon obstacle information. The polygon obstacle information includes the relative coordinate information of each point of the polygon, the absolute velocity information of the obstacle, the direction information of the obstacle velocity in the Gaussian coordinate system, the obstacle acceleration information, etc., which are provided to the planning for decision-making reference.

[0065] After receiving the high-precision map lane line data, the planning and decision module first matches the center line of the lane where the current vehicle is located to confirm which lane the vehicle belongs to. Then, starting from the center of the rear axle of the vehicle's rectangular frame, with the vehicle's front facing the Y-axis and the right side the X-axis, it traverses all the shape points on the center line of the lane where the current vehicle is located. It calculates the distance between the vehicle and polygonal obstacles at this shape point position. It also traverses and calculates the distance between all obstacles and the two polygons when the vehicle is at this shape point. By checking if the distance is less than or equal to 0, it determines whether a collision will occur. Since the shape point interval is 2 meters, within a 200-meter range, only 200 traversals are needed in the current lane to determine whether a collision will occur. If the costmap is processed with a grid precision of 0.2 meters, it would require nearly 2000 calculations to complete one collision calculation. In planning and decision-making, this solution directly improves the computational efficiency by 10 times.

[0066] The costmap first generates a heatmap using the OpenCV open-source library cv::distanceTransform. It then iterates through the planned path of the current vehicle (composed of 50 points spaced 0.6 meters apart) and calculates the heatmap values ​​of the corresponding locations of all points in the costmap. When the vehicle is on the planned path with the rear axle center as the origin, the heatmap values ​​of the corresponding coordinates of the rear axle center, vehicle center, and front axle center are calculated. This represents the distance to the nearest obstacle at the current coordinates. Combining the vehicle's dimensions (length, width, and height) with the expansion distance (safe distance), it can determine whether a collision will occur with the obstacles in the costmap, thus achieving collision detection.

[0067] The planning and decision-making module defaults to driving on the currently matched lane. It searches forward using the base distance base_dis plus the distance traveled at the current speed over a specified time (v*t). If there is an obstacle with a collision risk, it first searches the left lane and then the right lane at the same distance to see if there is an obstacle with a collision risk. If there is a lane without an obstacle or collision risk, it changes lanes to the corresponding lane. If there is no lane to travel in the adjacent lane, it searches all adjacent lanes until there is a lane to travel in. If there is still no lane to travel in after traversing all lanes, it stops, slows down, and brakes until a lane to travel in is found before continuing.

[0068] like Figure 2 As shown, the planning and decision-making module receives lane line data sent from the map, combines it with obstacle information from the costmap and polygon obstacles sent from the perception system, and determines whether there is a collision risk in the current lane, the target lane after lane change, and the lane change path by traversing the lane lines. The collision risk with the polygon is determined by the distance between the position of the current vehicle's bounding box with the rear axis center as the origin in the lane line points and the polygon. If the distance is α, it indicates that there is a collision risk. The collision risk with static obstacles is determined by the distance between the position of the current vehicle's bounding box in the lane line points and the nearest point in the costmap. If there is a collision risk, i.e. the distance is less than or equal to α, then a corresponding decision is made to implement the autonomous driving obstacle avoidance function.

[0069] In summary, this invention not only reduces system resource consumption and improves the efficiency of perception and planning decision algorithms, but also reduces the system resource consumption of perception and planning decision modules, while ensuring the user experience of the autonomous driving system. Outside the lane, the invention detects the heat map of the planned trajectory in the cost map, and inside the lane, it detects the distance between the lane line and the planned trajectory and the polygon within the lane to provide a final safety guarantee, thus ensuring vehicle safety performance.

[0070] Although the invention has been described with reference to a limited number of embodiments, those skilled in the art will understand from the foregoing description that other embodiments are conceivable within the scope of the invention described herein. Furthermore, it should be noted that the language used in this specification has been chosen primarily for readability and instructional purposes, and not for the purpose of interpreting or limiting the subject matter of the invention. Therefore, many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the appended claims. The disclosure of the invention is illustrative and not restrictive, and the scope of the invention is defined by the appended claims.

Claims

1. A method for obstacle avoidance in autonomous driving based on lane lines in high-precision maps, characterized in that, include: Q1. When a vehicle is driving on a public road, based on high-precision map data, acquire obstacle data information outside and inside the road boundary of the road where the vehicle is located, perform vehicle coordinate transformation, and output obstacle data information outside and inside the road boundary in the same coordinate system; Q2. Based on the obstacle data information outside and inside the road boundary under the same coordinate system, the obstacle data information outside the road boundary is mapped to a two-dimensional costmap grid map, and the grids containing obstacles are marked, outputting the grid map data information with annotations. The obstacle data information inside the road boundary under the same coordinate system is mapped to a three-dimensional obstacle polygon. The perception data information obtained by the vehicle multi-sensor is coupled with the three-dimensional obstacle polygon data information to output the polygon data information of the obstacle. Q3. Based on the labeled grid map data and the polygon data of the obstacles, and according to the lane line data obtained from the high-precision map, determine whether there is a collision risk in the current lane, the target lane after lane change, and the lane change path by traversing the lane line data, and output vehicle planning decision data information; Q4. Based on the vehicle planning and decision data, fit the vehicle trajectory using the third-order Bézier curve algorithm and output the vehicle obstacle avoidance motion control data. The third-order Bézier curve algorithm includes: Q41. Based on the vehicle planning and decision data, obtain the coordinate data of the starting point and the target point of the vehicle's trajectory; Q42. Based on the coordinate data of the starting point and target point of the vehicle's trajectory, according to the third-order Bézier curve function, p(t) = p0(1-t). 3 +3p1t(1-t) 2 +3p2t(1-t)+p3t 3 Where p0 is the starting point coordinate of the vehicle's trajectory, p3 is the target point coordinate of the vehicle's trajectory, p1 is the coordinate of the intersection of the vehicle's trajectory and the lane centerline, and p2 is the center point coordinate of the vehicle's trajectory. The vehicle's trajectory is corrected, and the fitted vehicle trajectory is output. Q43. Based on the fitted vehicle trajectory, output the vehicle obstacle avoidance motion control data information; In step Q3, the output vehicle planning decision data information specifically includes: Q31. Based on the lane line data information obtained from the high-precision map, determine the lane type of the vehicle, wherein the lane type includes the left lane, the middle lane, and the right lane; Q32. Based on the vehicle's lane type, retrieve the labeled grid map data and the polygon data of the obstacle to determine whether there is an obstacle in the current lane and whether there is an obstacle in the adjacent lane; Q33. Based on the data information of whether there are obstacles in the current lane and whether there are obstacles in the adjacent lane, a preset threshold for the distance between the vehicle and the obstacle is set, and vehicle planning decision data information is output; the preset threshold for the distance between the vehicle and the obstacle is α, 50≤α≤100. When the distance between the vehicle and the obstacle is greater than α, there is no risk of collision; otherwise, there is a risk of collision.

2. The vehicle autonomous driving obstacle avoidance method based on high-precision map lane lines according to claim 1, characterized in that: The labeled raster map data information includes raster cells with obstacles marked in black and raster cells without obstacles marked in white.

3. The vehicle autonomous driving obstacle avoidance method based on high-precision map lane lines according to claim 1, characterized in that: The lane line data information includes lane line coordinate information, lane center line coordinate information, lane line attributes, and intersection attributes.

4. The vehicle autonomous driving obstacle avoidance method based on high-precision map lane lines according to claim 3, characterized in that: The lane line attributes include markings for vehicles turning left, turning right, and going straight.

5. A vehicle autonomous driving obstacle avoidance system based on high-precision map lane lines, characterized in that, The system is used to implement the vehicle autonomous driving obstacle avoidance method based on high-precision map lane lines as described in any one of claims 1-4, the system comprising: The high-precision map module is used to obtain map data information of the lane where the vehicle is located; The perception module is based on multiple onboard sensors to acquire perception data information of the vehicle. The planning and decision-making module is connected to the high-precision map module and the perception module, and is used to plan the vehicle obstacle avoidance trajectory based on the information output by the high-precision map module and the perception module.

6. The vehicle autonomous driving obstacle avoidance system based on high-precision map lane lines according to claim 5, characterized in that: The vehicle-mounted multi-sensor system includes cameras, lidar, ultrasonic sensors, and gyroscopes.

Images