A Global Path Planning Method and System under Autonomous Driving

Abstract

The present invention discloses a global path planning method and system under autonomous driving. The method includes the following steps: Step 1: Establish a class-structured topological map; Step 2: Nearest neighbor matching of the starting point and the ending point; Step 3: Path search; Step 4: Path smoothing. The beneficial effects of the present invention are as follows: 1. The nearest neighbor matching idea in the present invention avoids the detour caused by direct nearest matching by expanding the matching range; 2. The vehicle model and kinematic constraints are considered in the path search, ensuring the rationality of the path; 3. The path smoothing idea largely considers the curvature problem of the path, ensuring that the path can be normally driven.

Description

Technical Field

[0001] The present invention is applied to the technical field of autonomous driving, and particularly relates to a global path planning method and system under autonomous driving. Background Art

[0002] In recent years, with the development of science and technology, driverless technology is gradually changing people's travel modes. The entire driverless technology can be simply divided into several modules such as positioning, perception, decision-making, planning, and control. Among them, the work that the planning module needs to do is to plan a reasonable path for the vehicle to travel throughout the process.

[0003] Planning is divided into two parts: global path planning and local path planning. The problem solved by global path planning is how to generate a drivable path from the starting point to the ending point in a given environment. This path can meet the kinematic constraints of the vehicle to a certain extent and is collision-free in the entire environment.

[0004] Current global path planning methods generally include search-based path planning and sampling-based path planning, and there are also some optimization methods based on both, such as topology-based path search. The idea of the search-based global path planning method is: in the searchable area, search around with the grid as the step size, and at the same time establish a cost function to evaluate each search result, and select a better result for search until the end point is found. The idea of the sampling-based global path planning algorithm is: in the passable area, sample in space according to a certain sampling function, after evaluating the sampled points, select better sampled points for the next iteration until near the end point is sampled. The idea of topology-based path search is: establish a topological environment model in the searchable area, in such an environment model, find the nearest topological points of the starting point and the ending point, and then perform step-by-step path search from the starting point. After evaluating the distance of each search result, reach the end point by choosing the better one.

[0005] The search-based path planning method has a strong correlation between its operation efficiency and the map size. In the case of a large map size, the search efficiency will be significantly reduced. At the same time, in the case of complex map terrain, the search-based path planning method may have a slow search efficiency due to the problem of only being locally optimal. Although the sampling-based path planning method avoids the problem of slow operation efficiency caused by the map to a certain extent, it has a certain degree of randomness, making it difficult to ensure the rationality of the final path and also unable to guarantee the kinematic constraints of vehicle driving. The topology-based path search method can avoid the problem of slow search efficiency in terms of the model, but since it is generally applied to mobile robots and does not pay attention to the vehicle model, the planned path will have some spikes, which cannot ensure the smoothness of the path and the motion constraints of the vehicle. At the same time, simply finding the nearest topological points of the starting and ending points is likely to cause the problem of a detoured path. Summary of the Invention

[0006] To solve the above problems, the purpose of this application is to provide a global path planning method and system for autonomous driving.

[0007] To achieve the purpose of the present invention, the present invention provides a global path planning method for autonomous driving, including the following steps:

[0008] Step 1: Establish a class-structured topological map;

[0009] Step 2: Neighbor matching of the starting point and the ending point;

[0010] Step 3: Path search;

[0011] Step 4: Path smoothing;

[0012] Among them, Step 2 specifically includes: taking the starting point and the ending point as the centers, respectively searching whether there are corresponding topological points within a finite distance threshold L. Among the searched topological points, according to their connection relationships, filter out unnecessary topological points, so as to screen out the matching topological points with a relatively large distance as the neighbor points.

[0013] Among them, Step 3 specifically includes: directly using the A* search algorithm for path planning between topological points. When designing the cost function, consider the width information and turning information of the vehicle, and establish the optimized cost function as follows:

[0014] G = Func(Dis history , Dir, Dis topo , Err angle )

[0015] Where Dis historyis the length of the searched path from the starting point to the point to be searched; Dir is the direction value between the point to be searched and its parent node. If it is in the forward direction, Dir is 0; if it is in the reverse direction, Dir is 1; Dis topo is the distance attribute of the point to be searched. This attribute is compared with the width of the vehicle to ensure that the vehicle can pass through the topological point smoothly; Err angle is the angular deviation between the line connecting the point to be searched and its parent node and the line connecting the parent node and its grandparent node. If the angular deviation exceeds the turning threshold of the vehicle itself, it is considered that there is a certain degree of difficulty when the vehicle turns.

[0016] Among them, step four specifically includes: taking each waypoint in the path as the point to be smoothed, taking the perpendicular direction of the line connecting the previous waypoint and the next waypoint as the smoothing direction, and moving the point to be smoothed in the smoothing direction by a single distance D; according to the distance attributes of each topological point, the smoothed distance attribute can be obtained as Dis topo -D. If the new distance has reached the distance safety margin, there is no need to smooth this point anymore; at the same time, calculate whether the smoothness of the sections before and after this point meets the set threshold; in this way, after a limited number of iterations, until a relatively smooth path is obtained.

[0017] Among them, the class-structured topological map is obtained in the following way:

[0018] The overall map includes two parts. The black grid area is the non-passable area, which can be considered as an obstacle, and the white grid area is the passable area. Starting from the black area and extending in all directions until two different obstacles extend to the same grid, the skeleton map of the entire passable area is obtained. At the same time, considering the requirement of the vehicle to drive on the right, during the extension process, the grids at a certain distance from the obstacle are recorded, and these grid points are combined into a forward and reverse map according to the positional relationship between the grids and the obstacle. After synthesizing the skeleton map and the forward and reverse map, the class-structured topological map is obtained.

[0019] Corresponding to the above method, the present invention also provides a global path planning system for autonomous driving, including the following units:

[0020] A unit for establishing a class-structured topological map, a unit for near-neighbor matching of the starting point and the ending point, a path search unit, and a path smoothing unit;

[0021] Among them, the unit for near-neighbor matching of the starting point and the ending point is used to: respectively search whether there are corresponding topological points within a limited distance threshold L centered on the starting point and the ending point. Among the searched topological points, according to their connection relationships, filter out unnecessary topological points, so as to screen out the relatively distant matching topological points as near-neighbor points.

[0022] Among them, the path search unit is used to: directly perform path planning between topological points using the A* search algorithm. When designing the cost function, the width information and turning information of the vehicle are taken into account, and the optimized cost function is established as follows:

[0023] G = Func(Dis history , Dir, Dis topo , Err angle )

[0024] Among them, Dis history is the length of the searched path from the starting point to the point to be searched; Dir is the direction value between the point to be searched and its parent node. If it is in the forward direction, Dir is 0; if it is in the reverse direction, Dir is 1; Dis topo is the distance attribute of the point to be searched, and this attribute is compared with the width of the vehicle to ensure that the vehicle can pass through the topological point smoothly; Err angle is the angle deviation between the line connecting the point to be searched and its parent node and the line connecting the parent node and its grandparent node. If the angle deviation exceeds the turning threshold of the vehicle itself, it is considered that there is a certain difficulty when the vehicle turns.

[0025] Among them, the path smoothing unit is used to: take each waypoint in the path as the point to be smoothed, take the perpendicular direction of the line connecting the previous waypoint and the next waypoint as the smoothing direction, and move the point to be smoothed in the smoothing direction by a single distance D; according to the distance attributes of each topological point, the smoothed distance attribute can be obtained as Dis topo -D. If the new distance has reached the distance safety margin, there is no need to smooth this point anymore; at the same time, calculate whether the smoothing degree of the sections before and after this point meets the set threshold; in this way, after a limited number of iterations, until a relatively smooth path is obtained.

[0026] Among them,

[0027] the unit for establishing a class-structured topological map establishes a class-structured topological map in the following way:

[0028] The overall map includes two parts. The black grid area is the non-passable area, which can be regarded as an obstacle, and the white grid area is the passable area. Starting from the black area and extending in all directions until two different obstacles extend to the same grid, the skeleton map of the entire passable area is obtained. At the same time, considering the requirement that the vehicle drives on the right, during the extension process, the grids at a certain distance from the obstacle are recorded, and these grid points are grouped into a forward and reverse map according to the position relationship between the grids and the obstacle. After synthesizing the skeleton map and the forward and reverse map, a class-structured topological map is obtained.

[0029] Compared with the prior art, the beneficial effects of the present invention are

[0030] 1. The nearest neighbor matching idea in the present invention avoids the detour caused by direct nearest matching by expanding the matching range;

[0031] 2. Considering the vehicle model and kinematic constraints in path search ensures the rationality of the path;

[0032] 3. The path smoothing idea largely considers the curvature problem of the path to ensure that the path can be driven normally. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 The class-structured generation relationship diagram in the present application is shown;

[0034] Figure 2 The class-structured topological map in the present application is shown;

[0035] Figure 3 The nearest neighbor matching schematic diagram in the present application is shown;

[0036] Figure 4 The Dis topo schematic diagram in the present application is shown;

[0037] Figure 5 The Err angle schematic diagram in the present application is shown;

[0038] Figure 6 The schematic diagram of the smoothing direction when the path is smoothed in the present application is shown;

[0039] Figure 7 The schematic diagram of the path smoothing effect in the present application is shown. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0040] It should be noted that, without conflict, the embodiments in the present application and the features in the embodiments may be combined with each other.

[0041] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0042] The embodiment of the present application discloses a global path planning method for autonomous driving,

[0043] The entire global path planning method mainly includes four steps, namely, establishing a class-structured topological map, nearest neighbor matching of the starting point and the ending point, path search, and path smoothing.

[0044] Each step is as follows:

[0045] Step 1: Establish a class-structured topological map

[0046] AsFigure 1 , Figure 2 as shown.

[0047] The overall map consists of two parts. The black grid area is the non-passable area, which can be regarded as an obstacle, and the white grid area is the passable area. Starting from the black area and extending in all directions until two different obstacles extend to the same grid, the skeleton map of the entire passable area is obtained. At the same time, considering the need for vehicles to drive on the right, during the extension process, the grids at a certain distance from the obstacle are recorded, and these grid points are grouped into a forward and reverse map according to the position relationship between the grids and the obstacle. After synthesizing the skeleton map and the forward and reverse map, a quasi-structured topological map is obtained.

[0048] Step 2: Proximity matching of the starting point and the ending point

[0049] As Figure 3 shown, on a topological map, each topological point has a distance attribute (the closest distance to the obstacle). Taking the starting point and the ending point as the centers, search respectively whether there are corresponding topological points within a finite distance threshold L. Among these topological points, according to their connection relationships, filter out unnecessary topological points, so as to screen out the matching topological points with a relatively large distance as the neighboring points.

[0050] Step 3: Search-based path planning

[0051] After obtaining the matching topological points of the starting point and the ending point, since there are connection relationships between the various topological points in the map, the A* search algorithm can be directly used for path planning between the topological points. Considering the vehicle model and kinematic constraints, when designing the cost function, the width information and turning information of the vehicle are taken into account. The optimized cost function is established as follows:

[0052] G = Func(Dis history , Dir, Dis topo , Err angle )

[0053] where Dis history is the length of the searched path from the starting point to the point to be searched; Dir is the direction value of the point to be searched and its parent node. If going forward, Dir is 0, and if going backward, Dir is 1; Dis topo is the distance attribute of the point to be searched, and this attribute is compared with the width of the vehicle to ensure that the vehicle can pass through the topological point smoothly; Err angle is the angle deviation between the line connecting the point to be searched and its parent node and the line connecting the parent node and its grandparent node. If the angle deviation exceeds the turning threshold of the vehicle itself, it is considered that there is a certain difficulty for the vehicle to turn. Among them, Figure 4 is the schematic diagram of Dis topo , Figure 5 is the schematic diagram of Errabgle Schematic diagram.

[0054] Step Four: Path smoothing

[0055] As Figure 7 shown, after obtaining the path, the path needs to be smoothed so that the path can have a certain safety margin from obstacles while also having a certain degree of smoothness. Here, each path point in the path is taken as a point to be smoothed, the perpendicular direction of the line connecting the previous path point and the next path point is taken as the smoothing direction, and the point to be smoothed is moved a single distance D in the smoothing direction.

[0056] According to the distance attributes of each topological point, the smoothed distance attribute can be obtained as Dis topo -D. If the new distance has reached the distance safety margin, then this point does not need to be smoothed anymore; at the same time, calculate whether the smoothness of the sections before and after this point meets the set threshold.

[0057] In this way, after a finite number of iterations, until a relatively smooth path is obtained.

[0058] The present invention provides an optimized global path planning method. By optimizing the matching of topological points and searching and smoothing the path, a better path is obtained. Its main advantages are:

[0059] 1. The optimized strategy of near-neighbor point matching can, to a certain extent, avoid the vehicle taking a detour.

[0060] 2. Design a cost function that better conforms to the vehicle motion constraints, which better meets the vehicle kinematic constraints.

[0061] 3. By means of path smoothing, the smoothness of the path is ensured.

[0062] Corresponding to the above method, the present invention also provides a global path planning system for autonomous driving, including the following units:

[0063] A unit for establishing a class-structured topological map, a near-neighbor matching unit for the starting point and the ending point, a path search unit, and a path smoothing unit;

[0064] Among them, the near-neighbor matching unit for the starting point and the ending point is used to: respectively search whether there are corresponding topological points within a finite distance threshold L centered on the starting point and the ending point. Among the searched topological points, according to their connection relationships, filter out unnecessary topological points, so as to screen out the matching topological points that are farther away as near-neighbor points.

[0065] It should be noted that for the technical solutions not detailed in this application, well-known technologies are adopted.

[0066] The above are only the preferred embodiments of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and modifications can be made, and these improvements and modifications should also be regarded as the protection scope of the present invention.

Claims

1. A global path planning method under autonomous driving, characterized in that, It includes the following steps: Step 1: Establish a class-structured topological map; Step 2: Neighbor matching of the starting point and the ending point; Step 3: Path search; Step 4: Path smoothing; Among them, step four specifically includes: taking each path point in the path as a point to be smoothed, taking the perpendicular direction of the line connecting the previous path point and the next path point as the smoothing direction, and moving the point to be smoothed in the smoothing direction by a single distance D; according to the distance attributes of each topological point, the smoothed distance attribute can be obtained as Dis topo -D. If the new distance attribute has reached the distance safety margin, there is no need to smooth this point anymore; at the same time, calculate whether the smoothing degree of the road sections before and after this point meets the set threshold; in this way, after a limited number of iterations, until a relatively smooth path is obtained; Among them, Step 2 specifically includes: Taking the starting point and the ending point as the centers, respectively search whether there are corresponding topological points within a finite distance threshold L. Among these searched topological points, according to their connection relationships, filter out unnecessary topological points, so as to screen out the matching topological points with a relatively large distance as the neighbor points.

2. The global path planning method under autonomous driving according to claim 1, characterized in that Among them, Step 3 specifically includes: Directly use the A* search algorithm for path planning between topological points. When designing the cost function, take into account the vehicle width information and turning information, and establish the optimized cost function as follows: G = Func(Dis history , Dir, Dis topo , Err angle ) Among them, Dis histcry is the length of the searched path from the starting point to the point to be searched; Dir is the direction value between the point to be searched and its parent node. If it is in the forward direction, Dir is 0; if it is in the reverse direction, Dir is 1; Dis topo is the distance attribute of the point to be searched. This attribute is compared with the width of the vehicle to ensure that the vehicle can pass through the topological point smoothly; Err angle is the angular deviation between the line connecting the point to be searched and its parent node and the line connecting the parent node and its grandparent node. If the angular deviation exceeds the turning threshold of the vehicle itself, it is considered that there is a certain degree of difficulty when the vehicle is turning.

3. The global path planning method under autonomous driving according to claim 1, characterized in that The class-structured topological map is obtained through the following method: The overall map includes two parts. The black grid area is an impassable area, which can be considered as an obstacle, and the white grid area is a passable area. Starting from the black area, extend it in all directions until two different obstacles extend to the same grid, so as to obtain the skeleton map of the entire passable area. At the same time, considering the need for the vehicle to drive on the right, during the extension process, record the grids at a certain distance from the obstacle, and form a clockwise and counterclockwise map according to the position relationship between the grids and the obstacle. After synthesizing the skeleton map and the clockwise and counterclockwise map, the class-structured topological map is obtained.

4. A global path planning system under autonomous driving, characterized in that, It includes the following units: A unit for establishing a class-structured topological map, a unit for neighbor matching of the starting point and the ending point, a path search unit, and a path smoothing unit; Among them, the path smoothing unit is used to: take each path point in the path as a point to be smoothed, take the perpendicular direction of the line connecting the previous path point and the next path point as the smoothing direction, and move the point to be smoothed in the smoothing direction by a single distance D; according to the distance attributes of each topological point, the smoothed distance attribute can be obtained as Dis topo -D. If the new distance attribute has reached the distance safety margin, there is no need to smooth this point anymore; at the same time, calculate whether the smoothing degree of the sections before and after this point meets the set threshold; in this way, after a limited number of iterations, until a relatively smooth path is obtained; Among them, the unit for neighbor matching of the starting point and the ending point is used to: Taking the starting point and the ending point as the centers, respectively search whether there are corresponding topological points within a finite distance threshold L. Among these searched topological points, according to their connection relationships, filter out unnecessary topological points, so as to screen out the matching topological points with a relatively large distance as the neighbor points.

5. The global path planning system under autonomous driving according to claim 4, characterized in that Among them, The path search unit is used to: Directly use the A* search algorithm for path planning between topological points. When designing the cost function, take into account the vehicle width information and turning information, and establish the optimized cost function as follows: G = Func(Dis history , Dir, Dis topo， Err angle ) Among them, Dis history is the length of the searched path from the starting point to the point to be searched; Dir is the direction value between the point to be searched and its parent node. If it is in the forward direction, Dir is 0. If it is in the reverse direction, Dir is 1; Dis topo is the distance attribute of the point to be searched. This attribute is compared with the width of the vehicle to ensure that the vehicle can pass through the topological point smoothly; Err angle is the angular deviation between the line connecting the point to be searched and its parent node and the line connecting the parent node and its grandparent node. If the angular deviation exceeds the turning threshold of the vehicle itself, it is considered that there is a certain difficulty when the vehicle turns.

6. The global path planning system under autonomous driving according to claim 4, characterized in that The unit for establishing a class-structured topological map establishes a class-structured topological map through the following method: The overall map consists of two parts. The black grid area is an impassable area, which can be regarded as an obstacle, and the white grid area is a passable area. Starting from the black area, it extends in all directions until two different obstacles extend to the same grid, thus obtaining the skeleton map of the entire passable area. At the same time, considering the need for vehicles to drive on the right, during the extension process, the grids at a certain distance from the obstacle are recorded, and these grid points are combined into a clockwise and counterclockwise map according to the positional relationship between the grid and the obstacle. After synthesizing the skeleton map and the clockwise and counterclockwise map, a quasi-structured topological map is obtained.

Images