Sidewalk full coverage path generation method applied to ackerman vehicle and vehicle
By acquiring obstacle information through high-definition maps and generating zigzag paths, this technology solves the problems of high energy cost and inapplicability of Ackerman vehicles in existing pedestrian path planning, achieving efficient coverage and efficient traversal.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- COWA TECHNOLOGY CO LTD
- Filing Date
- 2023-03-16
- Publication Date
- 2026-06-19
AI Technical Summary
Existing path planning algorithms generate too many cells in sidewalks, increasing energy costs, and are not suitable for Ackerman vehicles, especially in narrow areas.
By acquiring pedestrian boundary and obstacle information through high-definition maps and using collision detection to obtain key points between obstacles, a zigzag path is generated to avoid U-turns, which is suitable for Ackerman vehicles.
It increases sidewalk coverage, reduces energy costs, is suitable for path planning of Ackerman vehicles, and improves traversal efficiency.
Smart Images
Figure CN116295488B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of robot or vehicle path planning, and more specifically, to a method and vehicle for generating full-coverage sidewalk paths for Ackerman vehicles. Background Technology
[0002] In the field of autonomous driving technology for robots or vehicles, path planning is crucial. Traversal path planning (CPP) has become a popular research topic in robotic applications such as autonomous cleaning, lawn mowing, and agriculture, as well as exploration, mapping, search and rescue. Robotic end effectors can also benefit from CPP, for example, in surface treatment applications (milling, laser cleaning, painting, cladding, modeling, printing, and manufacturing inspection). CPP determines a path covering all points from an initial state to a final state while detecting and avoiding obstacles in the target environment. Generally, the goal of traversal path generation methods is to find an optimal coverage path, generating collision-free trajectories by reducing travel time, processing speed, energy cost, number of turns along the path length, and overlap rate.
[0003] In existing technologies, the paper "Revisiting Boustrophed on Coverage Path Planning as a Generalized Traveling Salesman Problem" proposes an algorithm that combines boustrophed decomposition with the traveling salesman problem. For example... Figure 1 and Figure 2 As shown. The specific approach is to abstract the coverage area and obstacles into polygons, and then decompose the traversal area into individual cell units through a "plowing" process, and execute a "plowing" path for each cell. Figure 2 This demonstrates the generation of line segments in a monotonic polygon along the x-axis. We initialize the first line segment at the leftmost vertex, parallel to the y-axis, which we call the scan direction. Generally, we restrict the scan direction to be collinear with one of the polygon's boundaries, as these directions have been shown to cover the polygon with the fewest number of line segments. Individual line segments are generated by alternating between lines intersecting the polygon in the y-direction and offsetting the lines from the leftmost vertex to the rightmost vertex. Figure 3 As shown, the final step is to connect the individual units. This method is formulated as a traveling salesman problem, with the goal of finding the shortest path that visits exactly one scan pattern (gray point) in each polygonal unit (point ellipse) from the starting node ns to the target node ng. Figure 4 It demonstrates a scenario where it generates a fully covered path for the lawn.
[0004] The above technology has the following drawbacks:
[0005] (1) The algorithm is only suitable for large, regular areas such as squares, farmland, and lawns. It is not suitable for long, narrow areas such as sidewalks. In sidewalks, due to the presence of many obstacles, the algorithm will be divided into too many units, resulting in too many U-turn paths and greatly increasing energy costs.
[0006] (2) The algorithm finds the shortest path between two points using Euclidean distance. Therefore, this path is only applicable to more flexible robots such as drones and omnidirectional robots, but not to Ackerman vehicles.
[0007] (3) There are often narrow areas in the sidewalk, and in this cell decomposition method, there will be situations where it cannot be handled in narrow areas. Summary of the Invention
[0008] To address the problems of the existing technology, this patent proposes a method and vehicle for generating full-coverage pedestrian walkways for Ackerman vehicles. This method generates walkways with multiple obstacles that have high coverage and are suitable for Ackerman vehicles. The generated path can solve the problem of traversing and covering areas between obstacles, providing a reference for the vehicle to complete the traversal. To achieve the above objectives, this invention adopts the following technical solution: a method for generating full-coverage pedestrian walkways for Ackerman vehicles, used for planning vehicle travel paths, including the following steps:
[0009] S1, Input the ID of the sidewalk to be traversed;
[0010] S2, obtain data from HDMap based on the ID and perform sampling processing;
[0011] S3, offset to the opposite side with one boundary as a reference line;
[0012] S4, check if the distance between the current path and the opposite boundary is less than the safe distance. If it is greater than the safe distance, save the current path and continue to offset to the opposite side. If it is less than the safe distance, obtain the initial path.
[0013] S5: Obtain key points between adjacent obstacles by determining whether a collision occurs with an obstacle;
[0014] S6, obtain the intersection path based on the key points; obtain the zigzag path based on the outermost collision path and the innermost collision path; where the outermost collision path is the path that collides with the obstacle but is farthest from the road boundary on the side of the obstacle, and the innermost collision path is the shortest path.
[0015] S7 selects a starting point based on the vehicle's location and connects the intersecting path and the non-collision path, controlling the vehicle to run along the path; where the non-collision path is the path along which the vehicle will not collide with obstacles.
[0016] Furthermore, in S2, the pedestrian boundary information and obstacle information to be traversed are obtained through a high-definition map. Then in S3, the pedestrian boundary is first filtered by the slope k. When the slope difference between two points is greater than 1, it is added to the set of pedestrian boundary vertices.
[0017] Furthermore, in S2, the sampling process is as follows: referencing the current sidewalk on the S-axis in the SL coordinate system, sampling is performed once every first preset length to obtain the left_point and right_point of the sampling point on the current S-axis, until the end of the sidewalk, and all left_point and right_point are saved. The left_point and right_point are extended to both sides by a second preset length with the S-axis as the center. The second preset length is equal to half the width of the vehicle plus a safety distance.
[0018] Furthermore, the first preset length is 0.3m, and the safety distance is 0.15m.
[0019] Furthermore, in S3 and S4, a relatively smooth line is selected based on the curvature of the left and right boundaries of the sidewalk, and this line is used as a reference line to translate towards the other boundary. The distance of each translation is determined by the width of the vehicle. The initial translation distance leaves a safe distance, and the translation continues until it exceeds the safe distance of the opposite lane. Then the translation stops. Finally, the opposite boundary is used as a reference line to translate in the opposite direction by a safe distance. The purpose of this is mainly to ensure that the edge of the opposite side is covered.
[0020] Furthermore, in S3 and S4, the maximum width of the current sidewalk is first obtained, and then the number of traversals required is calculated based on the maximum width. Specifically, the safe width on both sides of the traversal path is first set, and then twice the safe width is subtracted from the maximum width of the current sidewalk. The remaining width is divided by the vehicle width, which is the number of traversals required for the current sidewalk.
[0021] Furthermore, the safe width is equal to half the vehicle width plus a safe distance.
[0022] Furthermore, in S5, collision detection is performed between the vehicle's Box and the Polygon of obstacles (such as tree pits). The self-generated path is used as a reference line. First, assuming the vehicle is at the starting point of the path, it is determined whether it will collide. If a collision occurs, the vehicle is shifted forward 0.3m along the path until a collision is avoided. This point is then used as a key point between obstacles. The shift continues until a collision occurs with an obstacle, and the previous point is added to the set as a non-collision key point. This process is repeated until the end of the road. In this way, non-collision key points between two obstacles or between an obstacle and its two boundary points can be obtained, and the IDs of the obstacles before and after the collision point, as well as the driving direction, can be saved.
[0023] Furthermore, in S6, collision-free paths that do not intersect with obstacles are retained and directly added to the traversal path. Collision-free paths that intersect with obstacles are only retained when the two obstacles are far apart; when the obstacles are close, they are generated through key points. Here, the distance is set to 10m. It is assumed that obstacles intersect with at most the first two paths. First, the first key point of the second path is the starting point for entering the obstacle. Its next target is the key point of the first layer on the opposite side, which is then connected to the first key point of the first path. It is important to note that the journey from the key point of the first layer on the opposite side to this side involves reversing, and a buffer point is added before connecting the key point on the opposite side to allow the vehicle to adjust its posture, thus better achieving reversing. The same method is used to process the area between all obstacles. After traversing half of the area, it is traversed again in the opposite direction, entering the area between obstacles twice in total, thus traversing the entire area. The path through the obstacle is obtained by offsetting outwards from the line segment connecting the key points on both sides obtained from the collision detection of the outermost collision path and the obstacle.
[0024] The present invention also provides a vehicle, including a control system and a sensing system, wherein the control system uses any of the above-described methods for generating full-coverage pedestrian paths for Ackerman vehicles to plan the driving path.
[0025] Compared with existing technologies, this invention obtains key points between obstacles through collision and connects these key points in a regular manner to form a cross traversal. The traversal is completed by entering the area between adjacent obstacles twice, thus effectively covering the area between obstacles. This invention avoids U-turns by reversing, which only occur at both ends of the road, improving traversal efficiency. This invention generates a coverage path with high coverage for sidewalks with multiple obstacles (such as tree pits) that is suitable for Ackerman vehicles, providing an application reference for area traversal coverage for autonomous sanitation vehicles and other vehicles. Attached Figure Description
[0026] Figure 1-4 This is a schematic diagram of existing technology;
[0027] Figure 5 This is a schematic diagram of the path planning of the present invention. Detailed Implementation
[0028] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments applied to the embodiments of the present invention, and all other embodiments obtained by those skilled in the art without creative effort, are within the scope of protection of the present invention.
[0029] A method for generating full-coverage pedestrian paths for Ackerman vehicles, used for planning vehicle travel routes, includes the following steps:
[0030] S1, Input the ID of the sidewalk to be traversed;
[0031] S2, obtain data from HDMap based on the ID and perform sampling processing;
[0032] S3, offset to the opposite side with one boundary as a reference line;
[0033] S4. Check if the distance between the current path and the opposite boundary is less than the safe distance. If it is greater than the safe distance, save the current path and continue to offset to the opposite side. If it is less than the safe distance, obtain the initial path. The initial path is the first parallel path generated based on one side boundary. The 'path' mentioned in subsequent steps includes the initial path and each parallel path generated by offset.
[0034] S5 obtains key points between adjacent obstacles by determining whether a collision occurs; the initial path and other offset paths are used as reference paths to determine whether the vehicle collides with an obstacle.
[0035] S6, obtain intersecting paths based on key points; obtain zigzag paths based on the outermost and innermost collision paths; where the outermost collision path is the path that collides with an obstacle but is farthest from the road boundary on the obstacle's side, and the innermost collision path is the shortest path; the method for obtaining intersecting paths based on key points is to select key points between adjacent obstacles and connect these key points in a preset order to form a path that intersects within the area. Specifically, traveling from a key point on one path to a corresponding key point on another path achieves intersecting coverage within the area. Figure 5As shown. For each path that collides with an obstacle, the distance to the road boundary on the side where the obstacle is located is calculated. The path with the largest distance is taken as the outermost collision path, and the path with the smallest distance is taken as the innermost collision path. The zigzag path is a zigzag path formed by alternately connecting key points between the outermost and innermost collision paths, used to achieve efficient coverage in narrow areas between obstacles.
[0036] S7 selects a starting point based on the vehicle's location and connects the intersecting path and the non-collision path, controlling the vehicle to run along the path; where the non-collision path is the path along which the vehicle will not collide with obstacles.
[0037] This invention utilizes high-definition maps to number all pedestrian walkway segments that need to be traversed, facilitating path generation, optimization, management, and rapid path retrieval based on these numbers. During initial path planning, a high-definition map of the target segment is quickly obtained using the pedestrian walkway ID, revealing its shape, size, and the location and size of obstacles (such as tree pits). This basic data is then used to plan the traversal path according to rules. First, a path is planned along one boundary. Then, when turning back, a second path is created by offsetting a certain distance to the opposite side, and so on, until the entire pedestrian walkway is covered. During path planning, when obstacles obstruct the path, collision points are calculated to obtain key points. Based on these key points, a zigzag driving path is planned for the vehicle to enter the road surface between adjacent obstacles. Finally, a complete traversal path is obtained and provided to the vehicle. The vehicle selects its starting point based on its location and travels along the path to complete the task.
[0038] This invention generates a coverage path with high coverage for pedestrian walkways with multiple obstacles, suitable for Ackerman vehicles, and provides a zigzag coverage path as an alternative route for actual situations. The generated path effectively solves the problem of traversing and covering areas between obstacles, providing a good reference for vehicles to complete traversal.
[0039] Compared with existing technologies, this invention obtains key points between obstacles through collision and connects these key points in a regular manner to form a cross traversal. The traversal is completed by entering the area between adjacent obstacles twice, thus effectively covering the area between obstacles. This invention avoids U-turns by reversing, which only occur at both ends of the road, improving traversal efficiency. This invention generates a coverage path with high coverage for sidewalks with multiple obstacles (such as tree pits) that is suitable for Ackerman vehicles, providing an application reference for area traversal coverage for autonomous sanitation vehicles and other vehicles.
[0040] In some embodiments, in S2, the pedestrian walkway boundary information and obstacle information to be traversed are obtained through a high-definition map. Here, based on the scale of the high-definition map and image analysis, an SL coordinate system for the pedestrian walkway is established to obtain parameters such as the size and coordinates of pedestrian walkway-level obstacles. Then, in S3, the pedestrian walkway boundaries are first filtered by slope k. If the slope difference between two consecutive points is greater than 1, it is added to the pedestrian walkway boundary vertex set. In practical applications, the average traversal method is usually used, and the distance between adjacent paths is the same.
[0041] In a practical application example, in S2, the corresponding processing is as follows: referring to the S-axis of the current sidewalk in the SL coordinate system, the first preset sampling length is set to 0.3m (this value can be determined according to actual needs). Sampling is performed every 0.3m to obtain the left_point and right_point of the current sampling point on the S-axis. This process continues until the end of the sidewalk, where all left_points and right_points are saved. The left_points and right_points are extended to both sides by a second preset length centered on the S-axis. The second preset length is equal to half the width of the vehicle plus a safety distance. For example, if the width of the vehicle is 0.9m and the safety distance is 0.15m, then the second preset length is 0.6m.
[0042] In S3 and S4, a smoother line is selected based on the curvature of the left and right boundaries of the sidewalk. This line is used as a reference line to translate towards the other boundary. The distance of each translation is determined by the width of the vehicle. A safe distance is left for the first translation. The translation continues until it exceeds the safe distance of the opposite lane. Then the translation stops. Finally, a safe distance is translated in the opposite direction using the opposite boundary as a reference line. The purpose of this is mainly to cover the edge of the opposite side.
[0043] In S3 and S4, the maximum width of the current sidewalk is first obtained, and then the number of traversals required is calculated based on the maximum width. Specifically, the safe width on both sides of the traversal path is first set, and then twice the safe width is subtracted from the maximum width of the current sidewalk. The remaining width is divided by the vehicle width, which is the number of traversals required for the current sidewalk.
[0044] Specific example: In S3 and S4, the safe width is equal to half the width of the vehicle plus the safe distance. Currently, the width of the traversed vehicle is 0.9m, and the safe distance is set to 0.15m. Therefore, the safe width on both sides of the two traversal paths is 0.6m. Subtract 1.2m (0.9 + 0.15 x 2 = 1.2m, the distance between the two sides of the traversal path) from the maximum width of the current sidewalk, and divide the remaining width by 0.9 to get the number of times the current sidewalk needs to be traversed.
[0045] In S5, collision detection is performed by comparing the vehicle's Box with the Polygon (outer frame) of an obstacle (usually a tree pit). The generated path is used as a reference line. First, assuming the vehicle is at the beginning of the path, it is determined whether a collision will occur. If a collision occurs, the vehicle is shifted forward 0.3m along the path until a collision is avoided. This point is then used as a key point between obstacles. The shift continues until a collision occurs with an obstacle, and the previous point is added to the set as a non-collision key point. This process is repeated until the end of the road. In this way, non-collision key points between two obstacles or between an obstacle and its two boundary points can be obtained, and the IDs of the obstacles before and after the collision point, as well as the driving direction, can be saved.
[0046] In S6, collision-free paths that do not intersect with obstacles are retained and directly added to the traversal path. Collision-free paths that intersect with obstacles are only retained when the two obstacles are far apart. When the obstacles are close, they are generated through key points; for example, this distance can be set to 10m. Assuming that an obstacle intersects with at most the first two paths, the first key point of the second path is the starting point for entering the obstacle. Its next target is the key point of the first layer on the opposite side, which is then connected to the first key point of the first path. It is important to note that the journey from the key point of the first layer on the opposite side to this side involves reversing, and a buffer point is added before connecting the key point on the opposite side to allow the vehicle to adjust its posture, thus better achieving reversing. The same method is used to process the area between all obstacles. After traversing half of the area, it is traversed again in the opposite direction, entering the area between obstacles twice in total, thus traversing the entire area. The path through the obstacle is obtained by offsetting outwards from the line segment connecting the key points on both sides obtained from the collision detection of the outermost collision path and the obstacle.
[0047] like Figure 5As shown in the example, in the high-definition map, pedestrian crossing ID00100 is as follows: First, using the lower edge (R side) of the pedestrian crossing as a reference line, the planned path starts from point 1. First, it shifts towards the R side to point 2, then moves forward to key point 3 (this point is as close as possible to the obstacle boundary without collision). Then, it reverses to point 4, completing the traversal of the path intersecting the obstacle from the starting point. Next, it shifts towards the L side to point 5. At this point, the vehicle can be reversed back to point 1, completing the journey. The traversal between points 1 and 5 can be skipped depending on the situation. Then, proceed to point 6, shift to the R side to point 7, adjust the vehicle's position to point 8, reverse to point 9, shift to the left from point 9 to point 10, reverse to point 6, then proceed to point 11, and so on, following the numerical sequence to the final point 32. When reaching point 20, reverse to point 16, then proceed to point 20, turn and proceed to point 21, then turn around and proceed to point 22, finally completing the traversal of the entire pedestrian walkway. The pedestrian walkway between obstacles T1, T2, and T3 uses the zigzag path shown in the diagram to achieve a highly suitable operational path.
[0048] The present invention also provides a vehicle, including a control system and a sensing system. The control system can automatically plan a driving path using any of the above-described methods for generating full-coverage pedestrian paths for Ackerman vehicles, or the control system can download a pre-planned path from a server, and then set the starting and ending positions according to the vehicle's location to complete the operation according to the path.
[0049] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.
Claims
1. A method for generating full-coverage pedestrian paths for Ackerman vehicles, used for planning vehicle travel paths, characterized in that, Includes the following steps: S1, Input the ID of the sidewalk to be traversed; S2, obtain data from HDMap based on the ID and perform sampling processing; S3, offset to the opposite side with one boundary as a reference line; S4, check if the distance between the current path and the opposite boundary is less than the safe distance. If it is greater than the safe distance, save the current path and continue to offset to the opposite side. If it is less than the safe distance, obtain the initial path. The initial path is the first parallel path generated based on one side boundary; S5, the initial path and its parallel paths generated by offset are used as reference paths to determine whether the vehicle collides with an obstacle, and key points between adjacent obstacles are obtained by determining whether a collision occurs with an obstacle. S6, obtain the intersection path based on the key points; obtain the zigzag path based on the outermost collision path and the innermost collision path; where the outermost collision path is the path that collides with the obstacle but is farthest from the road boundary on the side of the obstacle, and the innermost collision path is the shortest path. S7 selects a starting point based on the vehicle's location and connects the intersecting path and the non-collision path, controlling the vehicle to run along the path; where the non-collision path is the path along which the vehicle will not collide with obstacles.
2. The method for generating full-coverage pedestrian paths for Ackerman vehicles according to claim 1, characterized in that, In S2, the pedestrian boundary information and obstacle information to be traversed are obtained through a high-definition map. Then in S3, the pedestrian boundary is filtered by the slope k. When the slope difference between two points is greater than 1, it is added to the set of pedestrian boundary vertices.
3. The method for generating full-coverage pedestrian paths for Ackerman vehicles according to claim 2, characterized in that, In S2, the sampling process is as follows: referencing the current sidewalk on the S-axis in the SL coordinate system, sampling is performed once every first preset length to obtain the left_point and right_point of the sampling point on the current S-axis, until the end of the sidewalk, and all left_point and right_point are saved. The left_point and right_point are extended to both sides by a second preset length with the S-axis as the center. The second preset length is equal to half the width of the vehicle plus a safety distance.
4. The method for generating full-coverage pedestrian paths for Ackerman vehicles according to claim 3, characterized in that, The first preset length is 0.3m, and the safety distance is 0.15m.
5. The method for generating full-coverage pedestrian paths for Ackerman vehicles according to claim 2, characterized in that, In S3 and S4, a smoother line is selected based on the curvature of the left and right boundaries of the sidewalk. This line is used as a reference line to translate towards the other boundary. The distance of each translation is determined by the width of the vehicle. A safe distance is left for the first translation. The translation continues until it exceeds the safe distance of the opposite lane. Then the translation stops. Finally, a safe distance is translated in the opposite direction using the opposite boundary as a reference line. The purpose of this is mainly to cover the edge of the opposite side.
6. The method for generating full-coverage pedestrian paths for Ackerman vehicles according to claim 5, characterized in that, In S3 and S4, the maximum width of the current sidewalk is first obtained, and then the number of traversals required is calculated based on the maximum width. Specifically, the safe width on both sides of the traversal path is first set, and then twice the safe width is subtracted from the maximum width of the current sidewalk. The remaining width is divided by the vehicle width, which is the number of traversals required for the current sidewalk.
7. The method for generating full-coverage pedestrian paths for Ackerman vehicles according to claim 6, characterized in that, The safe width is equal to half the vehicle width plus the safe distance.
8. The method for generating full-coverage pedestrian paths for Ackerman vehicles according to claim 2, characterized in that, In S5, collision detection is performed between the vehicle's Box and the obstacle's Polygon. The self-generated path is used as a reference line. First, assuming the vehicle is at the starting point of the path, it is determined whether a collision will occur. If a collision occurs, the vehicle is shifted forward 0.3m along the path until a collision is avoided. This point is then used as a key point between obstacles. The shifting continues until a collision occurs with an obstacle, and the previous point is added to the set as a non-collision key point. This process is repeated until the end of the road. In this way, non-collision key points between two obstacles or between an obstacle and its two boundary points can be obtained, and the obstacle IDs before and after the collision point, as well as the driving direction, can be saved.
9. The method for generating full-coverage pedestrian paths for Ackerman vehicles according to claim 2, characterized in that, In S6, collision-free paths that do not intersect with obstacles are retained and directly added to the traversal path. Collision-free paths that intersect with obstacles are only retained when the two obstacles are far apart. When the obstacles are close together, they are generated through key points. The distance set here is 10m.
10. The method for generating full-coverage pedestrian paths for Ackerman vehicles according to claim 9, characterized in that, In S6, it is assumed that the obstacle intersects with the first two paths at most. First, the first key point of the second path is the starting point for entering the obstacle, and its next target is the key point of the first layer on the opposite side, and then connects to the first key point of the first path. The same method is used to process the area between all obstacles. After traversing half of the area, it is traversed once in the opposite direction. In total, the area between obstacles is entered twice, and the entire area can be traversed. The path through the obstacle is obtained by offsetting outward from the line segment formed by the two key points on the outermost collision path and the obstacle obtained by collision judgment.
11. A vehicle, characterized in that, The system employs an Ackerman architecture, including a control system and a sensing system. The control system uses the pedestrian walkway full-coverage path generation method for Ackerman vehicles as described in any one of claims 1-10 to plan the driving path.