Method for planning a path to be traversed

Abstract

The invention relates to a method for planning a path to be traversed from a defined vehicle position to a target position, having the steps of: discretizing the surroundings of the vehicle between the defined vehicle position and the target position using a grid, ascertaining path segment models using vehicle kinematics, wherein the path segment models represent a drive of the vehicle between two nodes of the grid, and determining a path from the defined vehicle position to the target position by connecting nodes of the grid using path segments selected from the ascertained path segment models.

Description

[0001] Description

[0002] title

[0003] Method for path planning of a traversable path

[0004] State of the art

[0005] The present invention relates to a method for path planning of a traversable path, as well as a data processing device, a computer program and a computer-readable storage medium for carrying out the path planning method.

[0006] Path planning methods are known in various forms from the state of the art. To calculate a path from a current vehicle pose (location and orientation) to a target pose, behavioral rules are often used to influence path planning. The behavioral rules include criteria that are known to be comfortable and goal-oriented for the driver. Since the rules are linked to the respective parking situation, a new set of behavioral rules must be used for each parking situation.

[0007] It would be desirable to have a generic path planning method that finds a collision-free path from a defined vehicle pose to a target pose, independent of the parking situation.

[0008] Disclosure of the invention

[0009] The path planning method according to the invention provides an efficient generic path planning method which quickly determines a traversable path from a defined vehicle pose to a target pose, taking into account the vehicle kinematics, the required continuity requirements, and other comfort- and acceptance-based optimization criteria. This is achieved according to the invention in that the method discretizes a vehicle environment between the defined vehicle pose and the target pose using a grid in one step. In a further step, path piece templates are defined based on vehicle kinematics, whereby the path piece templates represent a journey of the vehicle between two nodes of the grid. Subsequently, in a step, a path from the defined vehicle pose to the target pose is determined by connecting nodes of the grid with path pieces selected from the defined path piece templates.The selected path piece templates can be stored in a memory device and used for further path planning processes. The path piece templates can also be determined by offline pre-calculation. The vehicle environment is preferably determined using a 3D camera and / or a 3D scanner on the vehicle and can be present, for example, as a point cloud. The vehicle environment is preferably selected to be so extensive that it includes the best path from the vehicle pose to the target pose with a high probability. The path pieces can connect two adjacent nodes. However, they can also connect any two nodes of the grid. The path piece templates are determined in particular by parameters of the vehicle kinematics, such as a minimum steering radius, a steering speed and / or a vehicle acceleration.

[0010] The subclaims show preferred developments of the invention.

[0011] Preferably, the path is determined using a search algorithm that minimizes a cost function between the defined vehicle pose and the target pose. The search algorithm can be, for example, a Monte Carlo Tree Search or an A* algorithm. This allows for efficient and rapid discovery of suitable path segments that form the best possible path from the vehicle pose to the target pose.

[0012] More preferably, the cost function considers the distance to objects and / or the minimum number of path segments and / or user comfort. This allows the quality of the path to be further improved, and it can be ensured that the path is safe, fast, and meets the comfort requirements of the vehicle occupants. Grid nodes whose approach would cause the vehicle to collide with objects in the environment are preferably excluded from the method. By excluding nodes whose approach would cause a collision, the complexity of the path planning problem can be reduced and the efficiency of the path planning method can be increased.

[0013] Preferably, the path is checked for changes in the vehicle's environment. Changes in the vehicle's environment may require an adjustment of the path. For example, a change in the vehicle's environment may be accompanied by a change in the target pose, which is subsequently adjusted. Thus, by checking the vehicle's environment, changes can be responded to and the path adjusted.

[0014] Further preferably, the collision-free nature of the path is checked. A change in the vehicle's surroundings can result in a potential collision with the calculated path. The collision-free nature of the path is checked, in particular, after a new path section has been determined. Furthermore, the path can be checked for collision-free nature before departure and / or during travel of the pad. If a collision is detected, the path can be adjusted to prevent the collision. Thus, potential collisions along the path can be reliably avoided.

[0015] Particularly preferably, when a potential collision is detected along the path, only those path segments that absolutely require a change are modified. According to the invention, this can be achieved simply by selecting new path segments from the defined path segment templates. By changing only the absolutely necessary path segments, the complexity of the calculation effort can be reduced and the path can be recalculated quickly.

[0016] The method preferably comprises a further step in which the determined path is optimized with respect to predefined optimization criteria. Thus, the path can be optimized for the respective situation easily and quickly from defined path segment templates, independent of the grid. For example, the path can be optimized with respect to its continuity, curviness, and / or user comfort. In particular, an attempt is made to minimize steering angle changes.

[0017] Furthermore, the invention describes a device for data processing, comprising means for carrying out a method described above.

[0018] Furthermore, the invention describes a computer program comprising instructions which, when the program is executed by a computer, cause the computer to carry out the method described above.

[0019] Furthermore, the invention describes a computer-readable storage medium comprising instructions which, when executed by a computer, cause the computer to carry out the method described above.

[0020] Short description of the drawings

[0021] Embodiments of the invention are described in detail below with reference to the accompanying drawings. In the drawing:

[0022] Figure 1 is a flowchart of a path planning method according to a preferred embodiment.

[0023] Embodiments of the invention

[0024] A method for path planning of a drivable path from a defined vehicle pose to a target pose according to a preferred embodiment of the invention is described in detail below with reference to Figure 1.

[0025] The path planning method relates in particular to the path planning of a vehicle for an assisted parking maneuver. In particular, the method is applicable to the tasks of path planning for parking and reversing a vehicle. The vehicle is in particular a car or a trailer combination, such as a truck or a car with a trailer. In a step S1, the vehicle environment between a defined vehicle pose and a target pose of a vehicle is discretized by a grid. The target pose is, for example, a parking space. The defined vehicle pose is preferably the current position of the vehicle. The grid comprises nodes which are connected to one another by lines. The grid can be spanned in a two- or three-dimensional space. The vehicle environment is preferably detected by sensors of the vehicle.Information about the vehicle's surroundings can also be received from other road users or stored in a memory.

[0026] In a further step S2, path segment templates are defined based on the vehicle kinematics. The path segment templates represent a vehicle journey between two nodes of the grid. The path segment templates are preferably precalculated offline and preferably include the most probable paths a vehicle would take between two nodes of a grid. Additional path templates can be added subsequently.

[0027] Based on the path templates from step S2 and the grid from step S1, a path from the defined vehicle pose to the target pose is determined in step S3. In the preferred embodiment shown in Figure 1, step S3 is divided into the substeps S3.1, S3.2, S3.3, and S3.4.

[0028] In step S3.1, path pieces are selected from the path piece templates that connect nodes along the grid to determine a path from the defined vehicle pose to the target pose. A search algorithm is used to select the next path piece. By selecting path pieces, the search algorithm attempts to minimize a cost function for the path between the defined vehicle pose and the target pose. The cost function of the search algorithm takes into account, for example, the distance of the path from objects and / or the minimum number of path pieces and / or the user comfort of the vehicle occupants. Furthermore, the cost function can take into account the path length, the change in angle, the change in curvature and / or the number of moves. High user comfort occurs, for example, when the vehicle moves continuously without strong acceleration or rapid changes of direction. In a step S3.2, the collision-free nature of the path is checked. For example, the vehicle's surroundings may change during the process, posing a risk of collision on the path determined in step S3.1. Step S3.2 is preferably performed after each newly selected path segment. However, step S3.2 can also be performed before and / or during travel of the path. This ensures that no collision occurs even if the vehicle's surroundings change.

[0029] If a collision is detected in step S3.2, the path segments that absolutely require a change to prevent the collision are modified in step S3.3. To do this, new path segments are selected from the specified path segment templates that enable a collision-free path in the area of ​​the detected collision.

[0030] After ensuring that a collision-free path exists, step S3.4 checks whether the target pose has been reached or whether additional path segments need to be added to reach the target pose. The target pose can include a tolerance range to ensure that the target pose can be reached if it was determined inaccurately or does not lie exactly on a node of the grid.

[0031] Once path planning has determined a path from the defined vehicle pose to the target pose, the path planning from step S3 is complete. Subsequently, in step S4, the determined path can be optimized with respect to predefined optimization criteria. For example, the path can be optimized with respect to continuity criteria, a reduction in curvature, the elimination of unnecessary path sections, an increase in the distance to objects, and / or a low radial acceleration, in particular to ensure improved user comfort. In the subsequent optimization, in contrast to steps S1 to S3, one is no longer bound to the nodes of the grid. Therefore, the path can be optimized here with respect to the same or additional criteria.Because the optimization is no longer performed in a discrete grid but in continuous space, an improvement of the path can be expected again in step S4 - even if the same optimization criteria are used.

Claims

Claims 1 . A method for planning a drivable path from a defined vehicle pose to a target pose, comprising the steps: Discretizing a vehicle environment (S1) between the defined vehicle pose and the target pose by a grid, Defining path piece templates (S2) based on vehicle kinematics, whereby the path piece templates represent a journey of the vehicle between two nodes of the grid, Determining a path (S3) from the defined vehicle pose to the target pose by connecting nodes of the grid with path pieces selected from the defined path piece templates.

2. The method according to claim 1, wherein the path is determined by a search algorithm which minimizes a cost function between the defined vehicle pose and the target pose.

3. The method according to claim 2, wherein the cost function takes into account the distance to objects and / or the minimum number of path pieces and / or user comfort.

4. Method according to one of the preceding claims, wherein nodes of the grid are excluded which, when approaching, would cause the vehicle to collide with objects in the environment.

5. Method according to one of the preceding claims, wherein the path is checked for changes in the vehicle environment.

6. The method according to claim 5, wherein the collision freedom of the path is checked.

7. The method according to claim 6, wherein, when a possible collision is detected along the path, only those path sections are changed which absolutely require a change.

8. Method according to one of the preceding claims, comprising the step of: optimizing (S4) the determined path with respect to predefined optimization criteria.

9. A data processing device comprising means for carrying out the method according to any one of the preceding claims.

10. A computer program comprising instructions which, when executed by a computer, cause the computer to carry out the method according to any one of claims 1 to 8.

11. A computer-readable storage medium comprising instructions which, when executed by a computer, cause the computer to carry out the method according to any one of claims 1 to 8.

Images