Methods, devices, vehicles and storage media for planning parking routes
By generating multiple parking paths and combining grid resolution and multi-dimensional evaluation metrics, the problem of insufficient adaptability of single-path algorithms is solved, achieving more efficient and safer automatic parking path planning.
Patent Information
- Application Number
- CN202510844743.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-23
- Publication Date
- 2026-11-17
- Estimated Expiration
- 2045-06-23
AI Technical Summary
Existing automated parking systems use a single-path algorithm, which results in insufficient adaptability of path planning and an inability to find the optimal path in complex environments.
Multiple parking paths are generated based on the vehicle's initial pose, target parking space, and preset parking operations. These paths are then discretized using a grid resolution. The path endpoint angle deviation, the number of short paths, and the proportion of path nodes that meet preset conditions are combined to evaluate and select the highest quality path for control.
It improves the adaptability and efficiency of parking paths, ensures the accuracy and safety of paths, adapts to the needs of different driving styles, and enhances parking quality.
Smart Images

Figure CN120668169B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of vehicle technology, and more specifically, to methods, apparatus, vehicles, and storage media for planning parking routes in the field of vehicle technology. Background Technology
[0002] With the development of autonomous driving technology, automated parking systems have become an important component of intelligent vehicles. By integrating perception, decision-making, and control subsystems, automated parking systems enable fully automated parking of intelligent vehicles without the driver needing to directly control the steering wheel, accelerator pedal, and brake pedal.
[0003] In related technologies, automated parking systems employ a single-path algorithm to plan parking routes. However, due to local optima, environmental complexity, or algorithmic limitations, the single-path algorithm can only find a path feasible under specific assumptions. Therefore, the adaptability of the planned path needs to be improved.
[0004] Therefore, there is an urgent need for a solution that can overcome the limitations of single-path planning. Summary of the Invention
[0005] This application provides a method, apparatus, vehicle, and storage medium for planning parking routes. This method can effectively overcome the limitations of single-path planning and improve the adaptability of parking routes.
[0006] Firstly, a method for planning parking paths is provided. This method includes: responding to a parking instruction from a vehicle, determining multiple parking paths based on the vehicle's initial pose at its current position, the target pose of the target parking space, and preset parking operations performed by the vehicle during parking; determining a quality score for each parking path on a first planning objective, the first planning objective including the angular deviation between the vehicle's frontal orientation and the orientation of the target parking space in the last segment of the parking operation corresponding to the parking path, the number of short paths in the parking path, and the proportion of path nodes in the parking path that meet preset conditions, wherein a short path is a path whose length corresponding to the parking operation is less than a preset length, and the preset conditions indicate that the curvature of the path nodes corresponding to the pose is less than a preset curvature; and controlling the vehicle to park based on the parking path with the highest quality score among the multiple parking paths.
[0007] In the aforementioned technical solution, after determining multiple parking paths based on the vehicle's initial pose, the target pose of the target parking space, and the vehicle's preset parking operations during the parking process, this method performs a quality assessment on each parking path. During the quality assessment, the following three planning indicators are introduced: the angular deviation between the vehicle's frontal orientation and the target parking space's orientation in the final segment of the parking operation (which can be mapped to the parking accuracy at the path's endpoint); the number of short paths in the parking path (which can be mapped to the operational efficiency corresponding to the parking path); and the proportion of path nodes in the parking path that meet preset conditions (which can be mapped to the safety of the parking path). These three planning indicators can effectively and comprehensively quantify multiple aspects of the parking path, effectively overcoming the limitations of single-path planning and improving the adaptability of the parking path. Furthermore, by improving the comprehensiveness of the parking path score, controlling vehicle parking based on the parking path with the highest quality score can significantly improve parking efficiency and parking quality.
[0008] In conjunction with the first aspect, in some possible implementations, multiple parking paths are determined based on the vehicle's initial pose at its current position, the target pose of the target parking space, and the vehicle's preset parking operations during parking. This includes: starting from the initial pose, determining multiple first predicted poses using the preset parking operations; using any one of the first predicted poses as a second predicted pose, discretizing the second predicted pose based on a preset grid resolution to obtain the grid coordinates corresponding to the second predicted pose; determining whether the second predicted pose should be discarded based on whether the grid coordinates corresponding to the second predicted pose are the same as the grid coordinates corresponding to the first pose, thereby correcting the multiple first predicted poses, where the path node corresponding to the first pose is the parent node of the path node corresponding to the second predicted pose; and determining the multiple parking paths based on the initial pose, the target pose, and the corrected multiple first predicted poses.
[0009] In the above technical solution, multiple first predicted poses are generated based on the discrete actions of a preset parking operation, ensuring that the parking path conforms to the vehicle's kinematic constraints and avoiding the generation of invalid paths that exceed the vehicle's physical limits. Secondly, a grid resolution is introduced to discretize the first predicted poses, mapping continuous pose states to grid coordinates. By comparing the grid coordinates corresponding to any first predicted pose (second predicted pose) with the grid coordinates corresponding to the first pose of its parent node, it can be determined whether the second predicted pose and the first pose are mapped to the same grid. If they are the same, the second predicted pose is discarded, which effectively eliminates redundant states, compresses the pose space, and improves the search efficiency of path nodes. Furthermore, multiple parking paths are constructed based on the corrected first predicted poses, initial pose, and target pose, which retains physically feasible and discretely independent candidate paths. This provides a rich and efficient source of path input for parking path quality evaluation, assisting the automatic parking system in determining parking paths that are excellent across various planning objectives.
[0010] In combination with the first aspect and the above implementation methods, in some possible implementation methods, the preset grid resolution includes position resolution and orientation resolution. Based on the preset grid resolution, the second predicted pose is discretized to obtain the grid coordinates corresponding to the second predicted pose, including: determining the ratio between the first horizontal axis coordinate in the second predicted pose and the first length corresponding to the position resolution as the target first horizontal axis coordinate in the grid coordinates; determining the ratio between the first vertical axis coordinate in the second predicted pose and the first length as the target first vertical axis coordinate in the grid coordinates; and determining the ratio between the first orientation in the second predicted pose and the first orientation corresponding to the orientation resolution as the target first orientation in the grid coordinates.
[0011] In the above technical solution, the mesh coordinates corresponding to the second predicted pose are determined by a rigid ratio mapping between the predicted pose and the resolution. This eliminates errors during floating-point operations and ensures the accuracy of the mesh coordinates. The dimensional decoupling process described above allows the horizontal axis coordinates to be independently discrete, avoiding mesh distortion caused by joint quantization.
[0012] In conjunction with the first aspect and the above implementation methods, in some possible implementation methods, determining the multiple parking paths based on the initial pose, the target pose, and the corrected multiple first predicted poses includes: determining the second pose corresponding to the target pose from the corrected multiple first predicted poses, and connecting the path node corresponding to the target pose with the path node corresponding to the second pose, wherein the parent node of the path node corresponding to the target pose is the path node corresponding to the second pose; deleting the second pose from the corrected multiple first predicted poses to obtain at least one third predicted pose, determining the third pose corresponding to each third predicted pose, and connecting the path node corresponding to each third predicted pose with the path node corresponding to the third pose. The path nodes are connected, and the parent node of the path node corresponding to the third predicted pose is the path node corresponding to the third pose, to determine each fourth predicted pose corresponding to the initial pose. The path nodes corresponding to each fourth predicted pose should be connected to the path node corresponding to the initial pose, and the parent node of each fourth predicted pose is the path node corresponding to the initial pose. Based on the connection relationship between the multiple path nodes formed by the initial pose, the target pose, and the multiple corrected first predicted poses, starting from the path node corresponding to the target pose, passing through the path nodes corresponding to the multiple corrected first predicted poses, and sequentially tracing back to the path node corresponding to the initial pose, the multiple parking paths are obtained.
[0013] In the above technical solution, the second pose corresponding to the target pose is identified and a parent-child relationship is established, that is, the path node corresponding to the target pose is connected to the path node corresponding to the second pose. This ensures that each parking path terminates precisely at the target parking space, avoiding endpoint deviation. Furthermore, after deleting the second pose, at least one remaining third predicted pose is processed, and the parent node is recursively connected to the path node corresponding to the initial pose. This can cover multiple first predicted poses corrected after discretization, quickly generating multiple parking paths. In addition, the complete connection relationship tracing back from the target pose to the initial pose can directly output multiple physically feasible parking paths, providing a low-redundancy and highly diverse path input source for parking path quality assessment, significantly improving the generation efficiency and completeness of multiple parking paths.
[0014] Combining the first aspect and the above implementation methods, in some possible implementation methods, determining the quality score of each parking path on the first planning objective includes: outputting a weight configuration reminder, which is used to remind the target object in the vehicle to configure the weights corresponding to each of the first planning objectives; upon receiving the weights corresponding to each of the first planning objectives, weighting and fusing the first score, the second score, and the third score based on the first configuration weight, the second configuration weight, and the third configuration weight to obtain the quality score, where the first score is based on the angle deviation, the second score is based on the number of short paths, and the third score is based on the proportion of path nodes that meet preset conditions; if the weights corresponding to each of the first planning objectives are not received, outputting a target configuration reminder, which is used to remind the target object to use driving style as the planning objective; upon receiving a confirmation operation for the target configuration reminder, determining the fourth configuration weight, the fifth configuration weight, and the sixth configuration weight based on the driving style for weighting and fusing the first score, the second score, and the third score to determine the quality score.
[0015] In the aforementioned technical solution, the weighted configuration reminder allows the target user to adjust the angular deviation between the vehicle's heading and the target parking space's orientation during the final parking maneuver, the number of short paths, and the proportion of path nodes meeting preset conditions in emphasizing path quality assessment. This enables the quality score (scoring result) to be personalized to match the driver's parking preferences. When not configured, intelligent guidance for driving style selection transforms abstract driving styles into specific weight parameters, addressing the lack of universality caused by default weights. Furthermore, during manual configuration, multiple scores are weighted and fused based on the configured weights; during unmanned configuration, dynamic weights are derived from driving styles. This two-stage scoring mechanism covers various weight determination scenarios, significantly improving the accuracy of parking path selection.
[0016] In conjunction with the first aspect and the above implementation methods, in some possible implementation methods, based on the driving style, a fourth configuration weight, a fifth configuration weight, and a sixth configuration weight are determined for weighted fusion of the first score, the second score, and the third score. This includes: when the driving style is aggressive, determining the fourth configuration weight as a first preset weight, the fifth configuration weight as a second preset weight, and the sixth configuration weight as a third preset weight, wherein the first preset weight is less than the second preset weight, and the third preset weight is less than the second preset weight; when the driving style is conservative, determining the fourth configuration weight as a fourth preset weight, the fifth configuration weight as a fifth preset weight, and the sixth configuration weight as a sixth preset weight, wherein the fifth preset weight is less than the fourth preset weight, and the fifth preset weight is less than the sixth preset weight; when the driving style is balanced, determining the fourth configuration weight, the fifth configuration weight, and the sixth configuration weight as target preset weights.
[0017] In the above technical solution, under the aggressive style, the configuration weight corresponding to angle deviation (first preset weight) and the configuration weight corresponding to the aforementioned proportion (third preset weight) are reduced, while the weight corresponding to the number of short paths is increased (second preset weight). These configuration weight settings are more suitable for targets prioritizing parking efficiency. Under the robust style, the curvature safety of path nodes is strengthened (sixth preset weight is larger) and the importance of angle deviation is emphasized, ensuring smoothness during vehicle turns. Under the balanced style, the three planning objectives are balanced with equal weights. This solution, through a directional mapping mechanism between driving style and configuration weights, can significantly optimize the scenario adaptability of the parking path quality score.
[0018] In conjunction with the first aspect and the above implementation methods, in some possible implementation methods, determining the quality score of each parking path on the first planning objective includes: determining the target path cost of each parking path, wherein the parking path is determined based on at least one first predicted pose related to the initial pose, the target pose, and the preset parking operation; the target path cost is the sum of the first path cost of the vehicle traveling from the initial pose to any first predicted pose and the second path cost of traveling from any first predicted pose to the target pose; the first path cost is the sum of the cumulative path cost of the vehicle traveling from the initial pose to the fifth predicted pose and the operation cost of traveling from the fifth predicted pose to any first predicted pose; the path node corresponding to the fifth predicted pose is the parent node of the path node corresponding to any first predicted pose; determining the target path cost, total path length, and number of paths of each parking path as the second planning objective when planning the parking path; and determining the quality score of each parking path on the first planning objective and the second planning objective.
[0019] In the above technical solution, the total path cost (target path cost) of each parking path is determined. The target path cost starts from the initial pose, goes to the first predicted pose corresponding to the intermediate path nodes, and finally to the target pose, integrating all historical and future costs. This avoids the limitations of only evaluating the endpoint state (such as angle deviation), accurately quantifies the cost of path segments, and improves the accuracy of quality score determination. Furthermore, incorporating the total path length and the number of paths into the second planning objective increases the comprehensiveness of the planning objective, making the quality score more accurate. This solution, through the hierarchical path cost calculation and the fusion mechanism of multiple planning objectives, can significantly improve the comprehensiveness and practicality of parking path scoring.
[0020] Secondly, an apparatus for planning parking paths is provided. The apparatus includes: a determining module, configured to: in response to a parking instruction from a vehicle, determine multiple parking paths based on the initial pose of the vehicle at its current position, the target pose of the target parking space, and preset parking operations performed by the vehicle during parking; determine a quality score for each parking path on a first planning objective, the first planning objective including the angular deviation between the vehicle's front orientation and the orientation of the target parking space in the last segment of the parking operation corresponding to the parking path, the number of short paths in the parking path, and the proportion of path nodes in the parking path that meet preset conditions, wherein the short path is a path whose length corresponding to the parking operation is less than a preset length, and the preset conditions are used to indicate that the curvature of the path nodes corresponding to the pose is less than a preset curvature; and a control module, configured to control the vehicle to park based on the parking path with the highest quality score among the multiple parking paths.
[0021] In conjunction with the second aspect, in some possible implementations, the determining module is specifically used for: starting from the initial pose, using the preset parking operation to determine multiple first predicted poses; taking any one of the first predicted poses as a second predicted pose, discretizing the second predicted pose based on a preset grid resolution to obtain the grid coordinates corresponding to the second predicted pose; determining whether the second predicted pose should be discarded based on whether the grid coordinates corresponding to the second predicted pose are the same as the grid coordinates corresponding to the first pose, so as to correct the multiple first predicted poses, wherein the path node corresponding to the first pose is the parent node of the path node corresponding to the second predicted pose; and determining the multiple parking paths based on the initial pose, the target pose, and the corrected multiple first predicted poses.
[0022] In conjunction with the second aspect and the above implementation methods, in some possible implementation methods, the preset grid resolution includes position resolution and orientation resolution. The device further includes: a processing module, configured to: determine the ratio between the first horizontal axis coordinate in the second predicted pose and the first length corresponding to the position resolution as the target first horizontal axis coordinate in the grid coordinates; determine the ratio between the first vertical axis coordinate in the second predicted pose and the first length as the target first vertical axis coordinate in the grid coordinates; and determine the ratio between the first orientation in the second predicted pose and the first orientation corresponding to the orientation resolution as the target first orientation in the grid coordinates.
[0023] In conjunction with the second aspect and the above implementation methods, in some possible implementation methods, the determining module is further configured to: determine the second pose corresponding to the target pose from the modified plurality of first predicted poses, and connect the path node corresponding to the target pose to the path node corresponding to the second pose, wherein the parent node of the path node corresponding to the target pose is the path node corresponding to the second pose; delete the second pose from the modified plurality of first predicted poses to obtain at least one third predicted pose, determine the third pose corresponding to each third predicted pose, and connect the path node corresponding to each third predicted pose to the path node corresponding to the third pose, wherein the third predicted pose is further configured to: determine the second pose corresponding to the target pose from the modified plurality of first predicted poses, and determine the third pose corresponding to each third predicted pose, and connect the path node corresponding to each third predicted pose to the path node corresponding to the third pose, wherein the parent node of the path node corresponding to the target pose is the path node corresponding to the second pose; delete the second pose from the modified plurality of first predicted poses to obtain at least one third predicted pose, and determine the third pose corresponding to each third predicted pose, and connect the path node corresponding to each third predicted pose to the path node corresponding to the third pose, wherein the parent node of the path node corresponding to the target pose is the path node corresponding to the second pose; and ... determine the third pose corresponding to each third predicted pose, wherein the parent node of the path node corresponding to the target pose is the path node corresponding to the second pose; and determine the third pose corresponding to each third predicted pose, and determine the third pose corresponding to each third predicted pose, wherein the parent node of the path node corresponding to the target pose is the path node corresponding to the second pose; and determine the third pose corresponding to each third predicted The parent node of the path node corresponding to the measured pose is the path node corresponding to the third pose, so as to determine each fourth predicted pose corresponding to the initial pose. The path node corresponding to each fourth predicted pose should be connected to the path node corresponding to the initial pose, and the parent node of each fourth predicted pose is the path node corresponding to the initial pose. Based on the connection relationship between the initial pose, the target pose and the multiple path nodes formed by the multiple corrected first predicted poses, starting from the path node corresponding to the target pose, passing through the path nodes corresponding to the multiple corrected first predicted poses, and tracing back to the path node corresponding to the initial pose in sequence, multiple parking paths are obtained.
[0024] In conjunction with the second aspect and the above implementation methods, in some possible implementation methods, the device further includes: an output module, used to output a weight configuration reminder, which reminds the target object in the vehicle to configure the weights corresponding to each first planning target; a fusion module, used to, upon receiving the weights corresponding to each first planning target, perform weighted fusion of the first score, the second score, and the third score based on the first configuration weight, the second configuration weight, and the third configuration weight to obtain the quality score, wherein the first score is obtained based on the angle deviation, the second score is obtained based on the number of short paths, and the third score is obtained based on the proportion of path nodes that meet preset conditions; the output module is also used to output a target configuration reminder when the weights corresponding to each first planning target are not received, which reminds the target object to use driving style as the planning target; the determining module is specifically used to, upon receiving a confirmation operation for the target configuration reminder, determine a fourth configuration weight, a fifth configuration weight, and a sixth configuration weight based on the driving style for weighted fusion of the first score, the second score, and the third score to determine the quality score.
[0025] In conjunction with the second aspect and the above implementation methods, in some possible implementation methods, the determining module is further configured to: determine the fourth configuration weight as the first preset weight, the fifth configuration weight as the second preset weight, and the sixth configuration weight as the third preset weight when the driving style is aggressive; determine the fourth configuration weight as the fourth preset weight, the fifth configuration weight as the fifth preset weight, and the sixth configuration weight as the sixth preset weight when the driving style is conservative; determine the fifth configuration weight as the fourth preset weight, the fifth configuration weight as the fifth preset weight, and the sixth configuration weight as the sixth preset weight when the driving style is balanced; and determine the fourth, fifth, and sixth configuration weights as target preset weights when the driving style is balanced.
[0026] In conjunction with the second aspect and the above implementation methods, in some possible implementation methods, the determining module is further specifically used for: determining the target path cost of each parking path, wherein the parking path is determined based on at least one first predicted pose related to the initial pose, the target pose, and the preset parking operation; the target path cost is the sum of the first path cost of the vehicle traveling from the initial pose to any first predicted pose and the second path cost of traveling from any first predicted pose to the target pose; the first path cost is the sum of the cumulative path cost of the vehicle traveling from the initial pose to the fifth predicted pose and the operation cost of traveling from the fifth predicted pose to any first predicted pose; the path node corresponding to the fifth predicted pose is the parent node of the path node corresponding to any first predicted pose; determining the target path cost, total path length, and number of paths of each parking path as the second planning objective when planning the parking path; and determining the quality score of each parking path on the first planning objective and the second planning objective.
[0027] Thirdly, a vehicle is provided, including a memory and a processor. The memory is used to store executable program code, and the processor is used to call and run the executable program code from the memory, causing the vehicle to perform the methods described in the first aspect or any possible implementation thereof.
[0028] Fourthly, a computer-readable storage medium is provided that stores executable program code, which, when run on a computer, causes the computer to perform the methods described in the first aspect or any possible implementation thereof. Attached Figure Description
[0029] Figure 1 This is a schematic diagram of a parking scenario provided in an embodiment of this application;
[0030] Figure 2 This is a schematic flowchart illustrating a method for planning parking routes provided in an embodiment of this application;
[0031] Figure 3 This is a schematic diagram illustrating the determination of an initial pose and a target pose according to an embodiment of this application;
[0032] Figure 4 This is a schematic diagram illustrating the determination of multiple parking paths provided in an embodiment of this application;
[0033] Figure 5 This is a schematic diagram of the structure of a device for planning parking routes provided in an embodiment of this application;
[0034] Figure 6 This is a schematic diagram of the structure of a vehicle provided in an embodiment of this application. Detailed Implementation
[0035] The technical solutions in this application will be clearly and thoroughly described below with reference to the accompanying drawings. In the description of the embodiments of this application, unless otherwise stated, " / " means "or," for example, A / B can mean A or B. "And / or" in the text is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Furthermore, in the description of the embodiments of this application, "multiple" refers to two or more than two.
[0036] Hereinafter, the terms "first" and "second" are used for descriptive purposes only and should not be construed as implying or suggesting relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature.
[0037] Figure 1 This is a schematic diagram of a parking scenario provided in an embodiment of this application. The following will be combined with... Figure 1 Describe the problems in parking scenarios and related technologies.
[0038] For example, such as Figure 1 The parking lot shown contains multiple parking spaces. Vehicle A drives to... Figure 1 After determining the location shown, the vehicle user observes and confirms that there is an available parking space B to the right front of vehicle A. At this point, the user triggers vehicle A's automatic parking function, meaning the automatic parking system begins operation. Through its integrated perception, decision-making, and control subsystems, the system plans a parking path for vehicle A to move from its current position into parking space B. Subsequently, based on the parking path, the system automatically controls vehicle A to drive into parking space B. During this process, the user does not need to directly control the steering wheel, accelerator pedal, or brake pedal, achieving a fully automated parking operation for vehicle A.
[0039] However, in the above process, the automatic parking system in the relevant technology uses a single-path algorithm to plan the parking path, that is, it can only find one parking path that is feasible under specific assumptions, such as... Figure 1 The parking path shown is 'a'. This parking path is an exemplary parking path that takes into account the impact of obstacles. However, this parking path does not balance other path planning objectives, such as the number of paths, and often only focuses on one aspect or makes simple compromises. Therefore, the adaptability of the planned path needs to be improved.
[0040] To address the aforementioned problems, this application proposes a method for planning parking routes, as detailed below. Figure 2 The methods and steps in the text.
[0041] Figure 2This is a schematic flowchart illustrating a method for planning parking routes provided in an embodiment of this application.
[0042] It should be understood that the method for planning parking routes provided in this application embodiment can be applied to, for example, Figure 1 The vehicle shown (e.g., vehicle A) can also be used in the automatic parking system of vehicle A.
[0043] For example, such as Figure 2 As shown, the method 200 includes the following steps 201 to 203.
[0044] Step 201: In response to the vehicle's parking instruction, determine multiple parking paths based on the vehicle's initial pose at the current position, the target pose of the target parking space, and the vehicle's preset parking operations during the parking process.
[0045] It should be understood that the initial pose in step 201 above can be characterized by the vehicle's position coordinates and heading at its current position, i.e., by (x0, y0, theta0), where the position coordinates (x0, y0) can be the coordinates of a first feature point on the vehicle. Optionally, the first feature point is the vehicle center (the geometric center of the vehicle body), and the position coordinates (x0, y0) are coordinates in a two-dimensional coordinate system with the target point as the origin. Optionally, the target point is the center point of the parking lot entrance. Furthermore, the heading is specifically characterized by degrees (°), with a heading of 0° indicating due east and a heading of 90° indicating due north.
[0046] It should also be understood that the target pose in step 201 above can be characterized by the position coordinates and orientation of the target parking space, i.e., by (x g ,y g theta g ) represents, where the position coordinates (x g ,y g The coordinates (x, y) can be the coordinates of a second feature point on the target parking space. Optionally, the second feature point is the center of the target parking space, and its position coordinates (x, y) are... g ,y g The coordinates are in a two-dimensional coordinate system with the target point as the origin. The orientation of the target parking space refers to the direction the front of the vehicle should face when parked in the target parking space, and the orientation of the target parking space is parallel to the long side of the target parking space. In this application, there is one orientation of the parking space. In addition, x0, y0, x g and y g The unit for all values is meters (m). Furthermore, the target parking space must be an available parking space.
[0047] It should be noted that the preset parking operation of the vehicle during the parking process in step 201 above refers to the pre-set basic movement command, which guides the vehicle to complete the parking process from the starting position (current position) to the target parking space. The preset parking operation is used to control the vehicle's driving direction and the driving step length in the driving direction. Optionally, the preset parking operation includes forward straight (steering angle of 0°, moving forward 1m), forward left turn (steering angle of 30°, moving forward 1m), forward right turn (steering angle of -30°, moving forward 1m), backward straight (steering angle of 0°, reversing 1m), backward left turn (steering angle of 30°, reversing 1m), and backward right turn (steering angle of -30°, reversing 1m). Among them, turning left forward (steering angle of 30°, moving forward 1m) means driving 1m when the current gear is in forward gear and the steering wheel is turned 30° to the left, and turning right backward (steering angle of -30°, moving backward 1m) means driving 1m when the current gear is in reverse gear and the steering wheel is turned 30° to the right.
[0048] It should also be noted that each parking path in step 201 above is a path connecting the starting path node, at least one first prediction node, and the target path node. The starting path node corresponds to the starting pose and can be represented as S0 = Node(x0, y0, theta0), and the target path node corresponds to the target pose and can be represented as S0 = Node(x0, y0, theta0). g =Node(x) g ,y g theta g At least one first prediction node corresponds to a first prediction pose, and the first prediction pose is determined based on an initial pose and a preset parking operation, such as the first prediction pose being represented as (x 11 ,y 11 theta 11 When ), the corresponding first prediction node is represented as S. 11 =Node(x) 11 ,y 11 theta 11 ).
[0049] Optionally, a parking command is triggered when the automatic parking button in the vehicle is pressed.
[0050] Figure 3 This is a schematic diagram illustrating the determination of an initial pose and a target pose according to an embodiment of this application. The following will be combined with... Figure 3 Taking the target point as the center point of the parking lot entrance, the first feature point as the center of the vehicle, the second feature point as the center of the target parking space, and the direction of 0° indicating due east as an example, the process of determining the initial pose and the target pose is described.
[0051] For example, such as Figure 3 As shown, location b indicates the parking lot entrance, which is also the parking lot exit. c is the center point of the parking lot entrance. A two-dimensional coordinate system is constructed with c as the origin. At this time, a vehicle C travels to location D. The coordinates of the center d of vehicle C are (x...). d ,y d Vehicle C's front is facing due east. Therefore, the initial pose of vehicle C after triggering the parking command can be determined as (x...). d ,y d At this moment, there is an empty parking space E to the right front of vehicle C. The coordinates of the center e of parking space E are (x, 0°). e ,y e When vehicle C is parked in parking space E, the direction in which the front of the vehicle should be facing is due north. At this time, the target pose of parking space E can be determined as (x e ,y e (90°).
[0052] It should be noted here that, as Figure 3 As shown, when vehicle C is parked in parking space E, the front of the car can face either due north or due south. However, it is more convenient for vehicle C to leave parking space E if the front of the car faces due north. Therefore, the front of the car should face due north.
[0053] Furthermore, if the car is facing due north, a more convenient parking scenario for vehicle C when leaving parking space E is when there is an obstacle to the south of parking space E, such as a wall or guardrail. Here, "convenient" means that when vehicle C leaves parking space E, it moves forward rather than backward. Moving forward provides a wider field of vision, allowing the driver to directly see the road conditions and pedestrian movements ahead, making the forward movement simple and smooth.
[0054] Furthermore, if the vehicle's front is facing due north or due south, it is convenient for vehicle C to leave parking space E in either direction. In this case, a prompt message is displayed on the vehicle's screen, indicating that the target user can choose either a due north or due south orientation when parking vehicle C into parking space E. If the prompt message instructs the target user to choose a direction, the selected orientation is determined as the correct orientation for the vehicle. The selected orientation can be either due north or due south. If the prompt message instructs the target user not to choose a direction, the automatic parking system arbitrarily selects one of the two orientations as the correct orientation. Additionally, a parking scenario where it is convenient for vehicle C to leave parking space E in either due north or due south is when there are no obstacles to the south or north of parking space E.
[0055] In one possible implementation, step 201, which determines multiple parking paths based on the vehicle's initial pose at its current position, the target pose of the target parking space, and the vehicle's preset parking operations during parking, includes: starting from the initial pose, determining multiple first predicted poses using the preset parking operations; using any one of the first predicted poses as a second predicted pose, discretizing the second predicted pose based on a preset grid resolution to obtain the grid coordinates corresponding to the second predicted pose; determining whether the second predicted pose should be discarded based on whether the grid coordinates corresponding to the second predicted pose are the same as the grid coordinates corresponding to the first pose, to correct the multiple first predicted poses, wherein the path node corresponding to the first pose is the parent node of the path node corresponding to the second predicted pose; and determining the multiple parking paths based on the initial pose, the target pose, and the corrected multiple first predicted poses.
[0056] It should be understood that in the above scheme, the vehicle's pose is continuous during parking, and continuous poses have infinite possibilities, which computer equipment cannot handle. Therefore, this method uses a discretization strategy to divide the pose space into grid cells to manage the state of the path nodes corresponding to the first predicted pose. The discretization strategy divides the pose space into grid cells of a fixed size using a preset grid resolution. Optionally, the preset grid resolution includes a position resolution d1m and an orientation resolution θ°. That is, if two first predicted poses are within the same d1m×d1m grid and the orientation deviation is less than θ°, the two first predicted poses can be mapped to the same grid, and are therefore considered to correspond to the same path node. When two first predicted poses are mapped to the same grid, the grid coordinates corresponding to the two first predicted poses are the same.
[0057] In the above technical solution, multiple first predicted poses are generated based on the discrete actions of a preset parking operation, ensuring that the parking path conforms to the vehicle's kinematic constraints and avoiding invalid paths. Secondly, a grid resolution is introduced to discretize the first predicted poses, mapping continuous pose states to grid coordinates. By comparing the grid coordinates corresponding to any first predicted pose (second predicted pose) with the grid coordinates corresponding to the first pose of its parent node, it can be determined whether the second predicted pose and the first pose are mapped to the same grid. If they are the same, the second predicted pose is discarded, which effectively eliminates redundant states, compresses the pose space, and improves the search efficiency of path nodes. Furthermore, based on the corrected multiple first predicted poses, the initial pose, and the target pose, multiple parking paths are constructed, which retains physically feasible and discretely independent candidate paths. This provides a rich and efficient source of path input (multiple candidate paths) for parking path quality evaluation, assisting the automatic parking system in determining parking paths that are excellent across various planning objectives.
[0058] In some embodiments, starting from the initial pose, a plurality of first predicted poses are determined using the preset parking operation, including: determining at least one first candidate pose based on the initial pose, the driving step size of the preset parking operation, and the corresponding steering angle, wherein the at least one first candidate pose is a predicted pose corresponding to a child node of the path node corresponding to the initial pose; taking each of the at least one first candidate poses as a target candidate pose, and determining at least one second candidate pose based on the target candidate pose, the driving step size of the preset parking operation, and the corresponding steering angle, wherein the at least one second candidate pose is a predicted pose corresponding to a child node of the path node corresponding to the target candidate pose. The predicted pose is obtained by taking each of the at least one second candidate pose as a reference candidate pose, and determining at least one third candidate pose based on the reference candidate pose, the driving step length of the preset parking operation and the corresponding steering angle. The at least one third candidate pose is the predicted pose corresponding to the child node of the path node corresponding to the reference candidate pose. This process continues until the predicted pose corresponding to the child node of the path node corresponding to the determined at least one fourth candidate pose is the target pose. The plurality of first predicted poses include the at least one first candidate pose, the at least one second candidate pose, the at least one third candidate pose and the at least one fourth candidate pose.
[0059] It should be understood that the above scheme is merely an example of reaching the target pose after four rounds of the first predicted pose. In reality, it typically takes more than four rounds to reach the target pose, depending on the specific circumstances.
[0060] It should also be understood that the predicted pose of the child node corresponding to at least one of the path nodes of the fourth candidate pose is the target pose means that the fourth candidate pose meets the termination condition, specifically, the position error between the fourth candidate pose and the target pose is <0.5m, and the orientation error between the fourth candidate pose and the target pose is <10°.
[0061] In one possible implementation, the preset grid resolution includes position resolution and orientation resolution. Based on the preset grid resolution, the second predicted pose is discretized to obtain the grid coordinates corresponding to the second predicted pose. This includes: determining the ratio between the first horizontal axis coordinate in the second predicted pose and the first length corresponding to the position resolution as the target first horizontal axis coordinate in the grid coordinates; determining the ratio between the first vertical axis coordinate in the second predicted pose and the first length as the target first vertical axis coordinate in the grid coordinates; and determining the ratio between the first orientation in the second predicted pose and the first orientation corresponding to the orientation resolution as the target first orientation in the grid coordinates.
[0062] It should be understood that the position resolution in the above scheme refers to the spatial size of the mesh cell, specifically including the size in the x-direction and the size in the y-direction, while the orientation resolution refers to the precision of the mesh cell division in the orientation dimension. Optionally, the position resolution is 0.5m and the orientation resolution is 15°.
[0063] In the above technical solution, the mesh coordinates corresponding to the second predicted pose are determined by a rigid ratio mapping between the predicted pose and the resolution. This eliminates errors during floating-point operations and ensures the accuracy of the mesh coordinates. The dimensional decoupling process described above allows the horizontal axis coordinates to be independently discretized, avoiding mesh distortion caused by joint quantization. Furthermore, the specific parameters in the preset mesh resolution are strongly correlated with the vehicle's physical characteristics. This allows the spatial size of the mesh cells to accurately match parking requirements, distinguishing differences between key poses while significantly compressing the search space of path nodes, generating efficient multiple parking paths.
[0064] In some embodiments, after determining a plurality of first predicted poses, the method 200 further includes: acquiring the coordinates of obstacles around the vehicle and the target parking space; determining whether the second predicted pose should be discarded based on whether the position coordinates in the second predicted pose fall within the obstacle area formed by the obstacle coordinates, so as to correct the plurality of first predicted poses.
[0065] Optionally, the obstacle coordinates include a first coordinate of a static obstacle and a second coordinate of a dynamic obstacle. The first coordinate is obtained through a high-precision map around the target parking space, and the second coordinate is obtained through a target sensor in the vehicle.
[0066] Optionally, the target sensors include LiDAR and vehicle-mounted cameras.
[0067] Optionally, static obstacles include stationary vehicles and railings, while dynamic obstacles include pedestrians and animals.
[0068] In some embodiments, determining whether the second predicted pose should be discarded based on whether the position coordinates in the second predicted pose fall within the obstacle area formed by the obstacle coordinates includes: determining that the second predicted pose should be discarded if the position coordinates in the second predicted pose fall within the obstacle area formed by the obstacle coordinates; and determining that the second predicted pose should not be discarded if the position coordinates in the second predicted pose do not fall within the obstacle area formed by the obstacle coordinates.
[0069] In some embodiments, determining whether the second predicted pose should be discarded based on whether the mesh coordinates corresponding to the second predicted pose are the same as the mesh coordinates corresponding to the first pose includes: determining that the second predicted pose should be discarded if the mesh coordinates corresponding to the second predicted pose are the same as the mesh coordinates corresponding to the first pose; and determining that the second predicted pose should not be discarded if the mesh coordinates corresponding to the second predicted pose are different from the mesh coordinates corresponding to the first pose.
[0070] It should be understood that once it is determined that the second predicted pose should be discarded, the second predicted pose is not included among the multiple first predicted poses.
[0071] In one possible implementation, determining the multiple parking paths based on the initial pose, the target pose, and a plurality of corrected first predicted poses includes: determining a second pose corresponding to the target pose from the plurality of corrected first predicted poses, and connecting the path nodes corresponding to the target pose with the path nodes corresponding to the second pose, wherein the parent node of the path node corresponding to the target pose is the path node corresponding to the second pose; deleting the second pose from the plurality of corrected first predicted poses to obtain at least one third predicted pose, determining the third pose corresponding to each third predicted pose, and connecting the path nodes corresponding to each third predicted pose with the path nodes corresponding to the third pose. Next, the parent node of the path node corresponding to the third predicted pose is the path node corresponding to the third pose, so as to determine each fourth predicted pose corresponding to the initial pose. The path node corresponding to each fourth predicted pose should be connected to the path node corresponding to the initial pose, and the parent node of each fourth predicted pose is the path node corresponding to the initial pose. Based on the connection relationship between the initial pose, the target pose and the multiple path nodes formed by the multiple corrected first predicted poses, starting from the path node corresponding to the target pose, passing through the path nodes corresponding to the multiple corrected first predicted poses, and tracing back to the path node corresponding to the initial pose in sequence, the multiple parking paths are obtained.
[0072] It should be understood that the above scheme describes starting from the target pose, exploring the parent nodes of the path nodes corresponding to the target pose to determine the corrected first predicted pose (i.e., the second pose) corresponding to the parent node; deleting the second pose from the multiple corrected first predicted poses to obtain at least one third predicted pose; for each of the at least one third predicted pose, exploring the parent nodes of the path nodes corresponding to the third predicted pose to determine the corrected first predicted pose (i.e., the third pose) corresponding to the parent node, and so on, through multiple rounds of exploration until the pose corresponding to the parent node is determined as the initial pose. Furthermore, path nodes corresponding to poses with parent-child relationships are connected to determine multiple parking paths through backtracking.
[0073] In the above technical solution, the second pose corresponding to the target pose is identified and a parent-child relationship is established, that is, the path node corresponding to the target pose is connected to the path node corresponding to the second pose. This ensures that each parking path terminates precisely at the target parking space, avoiding endpoint deviation. Furthermore, after deleting the second pose, at least one remaining third predicted pose is processed, and the parent node is recursively connected to the path node corresponding to the initial pose. This can cover multiple first predicted poses corrected after discretization, quickly generating multiple parking paths. In addition, the complete connection relationship tracing back from the target pose to the initial pose can directly output multiple physically feasible parking paths, providing a low-redundancy and highly diverse path input source for parking path quality assessment, significantly improving the generation efficiency and completeness of multiple parking paths.
[0074] Step 202: Determine the quality score of each parking path on the first planning objective. The first planning objective includes the angular deviation between the vehicle's front orientation and the orientation of the target parking space in the last parking operation corresponding to the parking path, the number of short paths in the parking path, and the proportion of path nodes in the parking path that meet preset conditions. The short path is the path whose length corresponding to the parking operation is less than a preset length. The preset conditions are used to indicate that the curvature of the path node corresponding to the pose is less than a preset curvature.
[0075] It should be understood that the vehicle's heading in the final segment of the parking operation in step 202 refers to the vehicle's heading at the start of the final segment of the parking path (usually the last reverse parking maneuver). Step 202 uses angular deviation to measure the difficulty and accuracy of parking the vehicle into the target space. A larger angular deviation indicates that the vehicle requires a greater or more urgent steering adjustment in the final segment of the parking operation to straighten the vehicle and park in the target space. At the same time, a high-quality parking path should minimize the aforementioned angular deviation, ensuring that the vehicle is close to the ideal parking posture at the start of the final segment of the parking operation. Angular deviation is used as an indicator to measure the "readiness" of the parking path.
[0076] For example, the target parking space is a standard perpendicular parking space with an orientation of 90°. In the planned parking path, when the vehicle performs the last parking operation (reversing operation), the front of the vehicle is not completely aligned with the parking space direction, but deviates 10° to the right, resulting in a 10° angle between the vehicle body and the parking space line. At this point, the front of the vehicle is 80°. Therefore, the angle deviation is 10° (90° - 80°).
[0077] It should also be understood that the number of short paths in the parking path in step 202 above refers to the number of segments in a single complete parking path where the length of consecutive forward or backward operation segments (i.e., the path segment corresponding to one operation) is less than a preset length. The preset length can be determined based on the vehicle's length or the vehicle's minimum safe moving distance; optionally, the preset length is 0.35 times the vehicle's length or 1 meter. The number of short paths is used to measure the redundancy of the parking path operation. A higher number of short paths indicates that the parking path planning is inefficient or involves unnecessary fine-tuning. Automatic parking systems require frequent gear shifting but only move a very short distance. At the same time, a high-quality parking path should minimize the number of short paths (even to zero). Frequent short-distance forward / short-distance backward movements can cause discomfort (a sense of jerking) for the target user and make the parking operation appear unskilled or clumsy.
[0078] The aforementioned "one-shot operation" refers to the continuous movement of a vehicle in the same gear (such as forward or reverse) without shifting gears, and its core lies in the continuity of gears.
[0079] It should also be understood that the percentage of path nodes meeting the preset conditions in the parking path in step 202 above refers to the percentage of path nodes with curvature less than the preset curvature among the N path nodes in a single complete parking path. The curvature of a path node is the reciprocal of the radius of curvature R at the path node, and the radius of curvature is the radius of the best-fit circle tangent to the parking path at the path node. The smaller the radius of curvature, the greater the curvature at that path node, the more severe the curvature of the parking path at that path node, and the sharper the turn, i.e., a sharp turn. The curvature at the path node is used to measure the driving smoothness of the parking path. At the same time, a high-quality parking path should control the above percentage as large as possible to avoid large-angle steering when the vehicle makes a sharp turn during parking. In addition, the larger the percentage corresponding to the above parking path, the better the comfort of the target object. Therefore, the above percentage can measure the comfort of the parking path.
[0080] In some embodiments, the method for determining the number of short paths in a parking path includes: dividing the parking path into continuous forward path segments and backward path segments; taking each path segment in the forward and backward path segments as a target path segment, accumulating the Euclidean distances between all adjacent path nodes included in the target path segment to obtain the path length of the target path segment; if the path length is less than a preset length, determining the target path segment as a short path; and counting all short paths in the parking path to obtain the number of short paths.
[0081] In some embodiments, the method for determining the proportion of path nodes in a parking path that meet preset conditions includes: determining a first total number of path nodes in the parking path; determining a first number of path nodes in the parking path whose curvature is less than a preset curvature; and determining the ratio between the first number and the first total number as the proportion of path nodes in the parking path that meet the preset conditions.
[0082] Optionally, the first planning objective also includes the maximum curvature of the path nodes in the parking path and the monotonicity of the orientation during single-handle operation.
[0083] It should be understood that, generally, for a single complete parking path, the maximum curvature of the path nodes should be as small as possible. The monotonicity of the steering wheel's orientation during a single steering wheel operation in the above scheme refers to whether the trend of change in the vehicle's orientation is singular and non-repetitive on the path segment corresponding to a single consecutive forward or backward operation (i.e., one operation). Monotonic increase means the vehicle's orientation continuously increases within the path segment (e.g., continuously turning left); monotonic decrease means the vehicle's orientation continuously decreases within the path segment (e.g., continuously turning right); non-monotonic: the vehicle's orientation first increases and then decreases within the path segment (e.g., first turning left, then turning right). Monotonic orientation changes indicate smoother steering wheel rotation, eliminating the need for significant corrections during forward / backward movements and reducing the number of steering wheel operations. Non-monotonic orientation changes resemble hesitation or repeated corrections, making the experience feel abrupt, disjointed, and uncomfortable for the target audience. In evaluating parking paths, parking paths with a singular trend of orientation change during a single steering wheel operation should be selected whenever possible.
[0084] It should be noted that, for any path segment corresponding to an operation, the orientation of all path nodes within the path segment is extracted in chronological order and an orientation sequence is generated; the amount of orientation change between adjacent path nodes is determined, resulting in multiple orientation change amounts; if all multiple orientation change amounts are positive or negative, it is determined that the orientation change trend of all path nodes within the path segment is unidirectional; if both positive and negative orientation change amounts exist among the multiple orientation change amounts, it is determined that the orientation change trend of all path nodes within the path segment is not unidirectional.
[0085] In one possible implementation, step 202, determining the quality score of each parking path on the first planning objective, includes: outputting a weight configuration reminder, which reminds the target object in the vehicle to configure the weights corresponding to each of the first planning objectives; upon receiving the weights corresponding to each of the first planning objectives, weighting and fusing the first score, second score, and third score based on the first configuration weight, second configuration weight, and third configuration weight to obtain the quality score, wherein the first score is based on angle deviation, the second score is based on the number of short paths, and the third score is based on the proportion of path nodes that meet preset conditions; if the weights corresponding to each of the first planning objectives are not received, outputting a target configuration reminder, which reminds the target object to use driving style as the planning objective; upon receiving a confirmation operation for the target configuration reminder, determining a fourth configuration weight, fifth configuration weight, and sixth configuration weight based on the driving style for weighting and fusing the first score, second score, and third score to determine the quality score.
[0086] It should be understood that the sum of the first configuration weight, the second configuration weight, and the third configuration weight in the above scheme is a preset value, which is 1. Optionally, the first configuration weight is 0.35, the second configuration weight is 0.35, and the third configuration weight is 0.3.
[0087] It should also be understood that the driving style in the above scheme refers to the target audience's expectations of the vehicle's operation and behavior during the parking process. Optionally, driving styles include aggressive, conservative, and balanced styles.
[0088] Specifically, when the driving style is aggressive, the target audience often allows for sharp turns, large angle deviations, and is less tolerant of short path segments corresponding to single-operation maneuvers. Therefore, the fourth configuration weight corresponding to the first score can be configured with a smaller weight, indicating that the target audience does not place much importance on the alignment of the car's front-end orientation with the target parking space orientation in the last segment of the parking operation corresponding to a single parking path (each parking path). The fifth configuration weight corresponding to the second score can be configured with a larger weight, meaning that the target audience prefers long-path operations and places more importance on the proportion of non-short paths in the total number of single parking paths. The sixth configuration weight corresponding to the third score can be configured with a smaller weight, indicating that the target audience does not pay much attention to the proportion of path nodes in a single parking path that meet the preset conditions and can tolerate larger curvature and more aggressive steering angles at path nodes.
[0089] When the driving style is conservative, the target audience tends to dislike sharp turns, large angle deviations, and shorter path lengths corresponding to single maneuvers. Therefore, the fourth configuration weight corresponding to the first score can be assigned a larger weight, the fifth configuration weight corresponding to the second score a smaller weight, and the sixth configuration weight corresponding to the third score a larger weight. When the driving style is balanced, the target audience tends to maintain a neutral attitude towards various planning objectives. Therefore, the fourth configuration weight corresponding to the first score, the fifth configuration weight corresponding to the second score, and the sixth configuration weight corresponding to the third score can be assigned the same weight. Furthermore, after determining the fourth, fifth, and sixth configuration weights, the first, second, and third scores can be weighted and fused based on these weights to obtain the quality score of a single parking path.
[0090] Furthermore, the above scheme of appropriately configuring the fifth configuration weight to a smaller / larger value, and appropriately configuring the sixth configuration weight to a smaller / larger value, are all relative to each configuration weight.
[0091] In the above technical solution, the weight configuration reminder allows the target user to adjust the angular deviation between the vehicle's orientation and the target parking space's orientation, the number of short paths, and the proportion of path nodes meeting preset conditions in the evaluation of path quality as needed. This enables the quality score (scoring result) to be personalized to match the parking preferences of the driving target user. When not configured, intelligent guidance is provided for driving style selection (e.g., for an aggressive style, the penalty weight for short paths can be automatically reduced), transforming the abstract driving style into specific weight parameters and addressing the pain point of insufficient universality caused by default weights. Furthermore, during manual configuration, multiple scores are weighted and fused based on the configured weights; during unmanned configuration, dynamic weights are derived from the driving style. This two-stage scoring mechanism can cover various weight determination scenarios, significantly improving the accuracy of parking path selection. Therefore, the above solution, through dynamic weight configuration and user interaction mechanisms, can significantly optimize the adaptability and accuracy of parking paths in quality assessment.
[0092] In addition to outputting weight configuration reminders, it can also output various first planning objectives and their corresponding specific meanings. For example, the first reference planning objective (the angular deviation between the vehicle's front orientation in the last parking operation of the parking path and the orientation of the target parking space) is used to measure the difficulty and accuracy of parking the vehicle into the target parking space; the second reference planning objective (the number of short paths in the parking path) is used to measure the redundancy of the parking path operation; and the third reference planning objective (the proportion of path nodes in the parking path that meet preset conditions) is used to measure the smoothness and comfort of the parking path, in order to assist in configuring weights for the target object.
[0093] In some embodiments, after outputting the target configuration reminder, the method 200 further includes: if no confirmation operation is received for the target configuration reminder, determining the sum of the first score, the second score, and the third score as the total score, and determining the product between the total score and the target preset weight as the quality score.
[0094] It should be understood that the "target preset weight" in the above scheme is the ratio between the preset value and the preset number. The preset number is 3, and the target preset weight is 1 / 3.
[0095] In some embodiments, the quality score is obtained by weighting and fusing the first score, the second score, and the third score based on the first configuration weight, the second configuration weight, and the third configuration weight, including: weighting the first score based on the first configuration weight to obtain the first quality score; weighting the second score based on the second configuration weight to obtain the second quality score; weighting the third score based on the third configuration weight to obtain the third quality score; and determining the sum of the first quality score, the second quality score, and the third quality score as the quality score.
[0096] In some embodiments, the method for determining the first score includes: determining the angle difference between the maximum angle deviation and the angle deviation, wherein the maximum angle deviation is the maximum angle deviation between the vehicle's front orientation and the target parking space orientation of the last parking operation corresponding to the parking path; and determining the ratio between the angle difference and the maximum angle deviation as the first score.
[0097] It should be understood that the maximum angular deviation mentioned above is 180°, which is the maximum theoretical angular deviation. This angular deviation can be considered as the actual angular deviation. The first score reflects the degree of alignment (the degree to which the orientations are the same) between the vehicle's front-end orientation and the target parking space in the last segment of a parking path. The maximum first score is 1 point, at which point the angular deviation is 0°; the minimum first score is 0 points, at which point the angular deviation is 180°. Furthermore, the greater the angular deviation between the vehicle's front-end orientation and the target parking space in the last segment of a parking path, the lower the first score for that parking path.
[0098] In some embodiments, the method for determining the second score includes: dividing a single parking path into continuous forward path segments and backward path segments; determining a second total number of forward path segments and backward path segments; determining the quantity deviation between the second total number and the number of short paths in the single parking path; and determining the ratio between the quantity deviation and the second total number as the second score.
[0099] It should be understood that the aforementioned quantity deviation refers to the number of non-short paths (not short paths) in a single parking path. The second score reflects the proportion of non-short paths in a single parking path within the total second quantity. The maximum second score is 1 point, at which point there are no short paths in the single parking path. The fewer the number of non-short paths in a single parking path, the lower the second score for that single parking path.
[0100] In some embodiments, the method for determining the third score includes: determining the proportion of path nodes in a single parking path that meet the preset conditions as the third score.
[0101] It should be understood that the greater the proportion of the above in a single parking path, the higher the third score of that single parking path.
[0102] In some embodiments, the method for determining the driving style includes: if the target object is a driver, analyzing the driving style of the target object based on the driving operation data of the target object in the historical driving process; or, if the target object is not a driver, analyzing the behavioral changes or external clothing of the target object to determine the desired driving style of the target object.
[0103] In some embodiments, driving operation data includes the total number of vehicle accelerations and the first number of rapid accelerations within a historical preset time period, as well as the peak value of the steering angular velocity. Based on the driving operation data of the target object during historical driving, the driving style of the target object is analyzed, including: determining the ratio between the first number and the total number of accelerations as the rapid acceleration frequency; determining the driving style as an aggressive style when the rapid acceleration frequency is greater than a first preset frequency; determining whether the peak value of the steering angular velocity is within a first range less than or equal to a preset peak value when the rapid acceleration frequency is less than or equal to the first preset frequency, the first range being used to indicate that the target object turns the steering wheel slowly when turning; determining the driving style as a conservative style when the peak value of the steering angular velocity is within the first range; and determining the driving style as a balanced style when the peak value of the steering angular velocity is within a second range greater than a preset peak value, the second range being used to indicate that the target object turns the steering wheel at a medium speed when turning.
[0104] It should be understood that "slowly" in the above refers to the steering wheel's angular velocity being less than the first steering angular velocity, such as 80° / s; "medium speed" in the above refers to the steering wheel's angular velocity being greater than the first steering angular velocity, such as 80° / s, and can also be between 80° / s and 120° / s.
[0105] In some embodiments, analyzing the behavioral changes or external attire of the target object to determine the target object's desired driving style includes: determining whether the target object exhibits stressful behavior when the vehicle is under rapid acceleration; determining an aggressive driving style when the target object does not exhibit stressful behavior; determining the duration of stressful behavior when the target object exhibits stressful behavior; determining a conservative driving style when the duration is greater than or equal to a preset duration, where the preset duration is the product of the total duration corresponding to the rapid acceleration condition and a preset coefficient; and determining a stable driving style when the duration is greater than or equal to a preset duration. If the driving duration is less than the preset duration, the driving style will be determined as a balanced style; or, if the external clothing is brightly colored, the style of the external clothing will be determined; if the style of the external clothing is bold and avant-garde, the driving style will be determined as an aggressive style; if the external clothing is brightly colored but not bold and avant-garde, the driving style will be determined as a stable style; if the external clothing is neither brightly colored nor bold and avant-garde, the driving style will be determined as a balanced style.
[0106] Optionally, the preset coefficient is 0.5.
[0107] In one possible implementation, based on the driving style, a fourth, fifth, and sixth configuration weight are determined for weighted fusion of the first, second, and third scores. This includes: when the driving style is aggressive, determining the fourth configuration weight as a first preset weight, the fifth configuration weight as a second preset weight, and the sixth configuration weight as a third preset weight, wherein the first preset weight is less than the second preset weight, and the third preset weight is less than the second preset weight; when the driving style is conservative, determining the fourth configuration weight as a fourth preset weight, the fifth configuration weight as a fifth preset weight, and the sixth configuration weight as a sixth preset weight, wherein the fifth preset weight is less than the fourth preset weight, and the fifth preset weight is less than the sixth preset weight; when the driving style is balanced, determining the fourth, fifth, and sixth configuration weights as target preset weights.
[0108] In the above technical solution, under the aggressive style, the configuration weight corresponding to angle deviation (first preset weight) and the configuration weight corresponding to the aforementioned proportion (third preset weight) are reduced, while the weight corresponding to the number of short paths is increased (second preset weight). These configuration weight settings are more suitable for targets prioritizing parking efficiency. Under the robust style, the curvature safety of path nodes is strengthened (sixth preset weight is larger) and the importance of angle deviation is emphasized, ensuring smoothness during vehicle turns. Under the balanced style, the three planning objectives are balanced with equal weights. This solution, through a directional mapping mechanism between driving style and configuration weights, can significantly optimize the scenario adaptability of the parking path quality score.
[0109] Optionally, the first preset weight is 0.3, the second preset weight is 0.5, and the third preset weight is 0.2.
[0110] Optionally, the fourth preset weight is 0.4, the fifth preset weight is 0.2, and the sixth preset weight is 0.4.
[0111] In one possible implementation, determining the quality score of each parking path on the first planning objective in step 202 includes: determining the target path cost of each parking path, wherein the parking path is determined based on at least one first predicted pose related to the initial pose, the target pose, and the preset parking operation; the target path cost is the sum of the first path cost of the vehicle traveling from the initial pose to any first predicted pose and the second path cost of traveling from any first predicted pose to the target pose; the first path cost is the sum of the cumulative path cost of the vehicle traveling from the initial pose to a fifth predicted pose and the operation cost of traveling from the fifth predicted pose to any first predicted pose; the path node corresponding to the fifth predicted pose is the parent node of the path node corresponding to any first predicted pose; determining the target path cost, total path length, and number of paths of each parking path as the second planning objective when planning the parking path; and determining the quality score of each parking path on the first planning objective and the second planning objective.
[0112] It should be understood that the target path cost f in the above scheme includes three costs: the first cost is the cumulative path cost g.acc of the vehicle traveling from the initial pose to the fifth predicted pose; the second cost is the operation cost g.oper of the vehicle traveling from the fifth predicted pose to a certain first predicted pose; and the third cost is the second path cost h of the vehicle traveling from a certain first predicted pose to the target pose. The sum of the cumulative path cost and the operation cost is the first path cost g. Furthermore, the cumulative path cost of the vehicle traveling from the initial pose to the fifth predicted pose can be further obtained by summing the cumulative path cost of the vehicle traveling from the initial pose to the sixth predicted pose and the operation cost of traveling from the sixth predicted pose to the fifth predicted pose. Here, the path node corresponding to the sixth predicted pose is the parent node of the path node corresponding to the fifth predicted pose.
[0113] It should also be understood that the number of path shifts in the above scheme refers to the total number of times a single parking path requires switching between forward and reverse gears during the parking process. Similarly, when adding the second planning objective, it is necessary to determine the configuration weights corresponding to the target path cost, total path length, and number of path shifts. For example, the configuration weight corresponding to the target path cost is the seventh configuration weight, the configuration weight corresponding to the total path length is the eighth configuration weight, and the configuration weight corresponding to the number of path shifts is the ninth configuration weight. When determining the quality score of each parking path on the first and second planning objectives, the sum of the first, second, third, seventh, eighth, and ninth configuration weights is a preset value, or the sum of the fourth, fifth, sixth, seventh, eighth, and ninth configuration weights is a preset value.
[0114] In the above technical solution, the cumulative path cost of each parking path from the initial pose to the fifth predicted pose, the operational cost of the vehicle moving from the fifth predicted pose to a certain first predicted pose, and the second path cost of the vehicle moving from a certain first predicted pose to the target pose are determined. This determines the total path cost (target path cost) of each parking path. The target path cost starts from the initial pose, extends to the first predicted pose corresponding to intermediate path nodes, and finally reaches the target pose, integrating all historical and future costs. This avoids the limitations of only evaluating the endpoint state (such as angle deviation), accurately quantifies the cost of path segments, and improves the accuracy of quality score determination. Furthermore, incorporating the total path length and the number of paths into the second planning objective increases the comprehensiveness of the planning objective, making the quality score more accurate. This solution, through the hierarchical path cost calculation and the fusion mechanism of multiple planning objectives, significantly improves the comprehensiveness and practicality of parking path scoring.
[0115] Optionally, the operating costs include a forward cost of 1.0 / m, a reverse cost of 1.5 / m (reversing penalty), and a steering cost of 0.2 / 30° (steering change penalty).
[0116] Figure 4 This is a schematic diagram illustrating the determination of multiple parking paths provided in an embodiment of this application. The following will be combined with... Figure 4 Taking an automatic parking scenario as an example, this application describes the process of determining multiple parking paths.
[0117] For example, such as Figure 4 The scene setting shown in (a) is as follows:
[0118] The origin of the two-dimensional coordinate system is the center point of the parking lot entrance;
[0119] The vehicle's initial pose is (1,4,0°), which means that the horizontal axis coordinate of the vehicle's center is 1, the vertical axis coordinate of the vehicle's center is 4, and the vehicle's orientation (i.e., the direction the front of the vehicle is facing) theta is due east and the orientation is 0°.
[0120] The target position of the parking space is (4,2,90°), which means that the horizontal axis coordinate of the center of the parking space is 4, the vertical axis coordinate of the center of the parking space is 2, the orientation of the parking space is due north, that is, the direction that the front of the car should face when the vehicle is parked in the parking space, and the orientation of the parking space is 90°.
[0121] Vehicle parameters: The vehicle has a wheelbase L of 2.5m, a length of 4.8m, and a width of 1.8m;
[0122] Operation set (including the following parking operations): step size d is 1m;
[0123] Forward: Go straight (turning angle 0°), forward: turn left (turning angle +30°), forward: turn right (turning angle -30°); Reverse: Go straight (turning angle 0°), reverse: turn left (turning angle +30°), reverse: turn right (turning angle -30°).
[0124] Operating costs: Forward cost is 1.0 / m, reverse cost is 1.5 / m (reversing penalty), and steering cost is 0.2 / 30° (steering change penalty);
[0125] Termination conditions: Positional error relative to the target pose < 0.5m, orientation error < 10°;
[0126] Grid resolution: Position resolution is 0.5m, orientation resolution is 15°;
[0127] The following is the process of determining multiple path nodes in the parking path based on the above known conditions:
[0128] 1. Initialization process: Construct the starting path node based on the initial pose.
[0129] Starting path node: S0 = Node(1,4,0°);
[0130] The path cost of the vehicle traveling from the initial pose to the current pose (from the starting path node to the current path node (i.e., the starting path node)); g = 0; specifically: the sum of the cumulative cost g.acc from the starting path node to the parent node of the current path node, and the operation cost from the parent node to the current path node. Where the cumulative cost is 0, and the operation cost is 0;
[0131] The second path cost for the vehicle to travel from the initial pose to the target pose (from the current path node to the target path node): h = ReedsShepp(1,4,0°; 4,2,90°) = 5.2;
[0132] Cost f: f = g + h = 5.2;
[0133] The starting path node has no parent node;
[0134] Initialize the data structure: Open list: Add the starting path node and its corresponding total cost; Closed list: Initially empty.
[0135] The discretized grid coordinates of the starting path node are: (2, 8, 0);
[0136] 2. First round of iteration: Expand sub-path nodes based on the starting path node.
[0137] 1) Retrieve any path node from the previous round (i.e., the starting path node) from the open list, i.e., S0 = Node(1,4,0°); Based on the starting path node and the target pose corresponding to the target path node S... g =Node(4,2,90°) determines whether the termination condition is met: If 90°-0° is greater than 10°, the termination condition is not met, and the path node expansion continues.
[0138] 2) Using the starting path node S0 as a reference, expand the sub-path nodes based on the operation set:
[0139] Moving forward in a straight line: new_x = 1 + 1 * cos(0°) = 2; new_y = 4 + 1 * sin(0°) = 4; new_theta = 0°; S 11 =Node(2,4,0°); Operation cost: g.oper = 1 * 1 = 1; Total cost: f 11 =g.acc+g.oper+h=0+1+4.1=5.1; where h=ReedsShepp(1,4,0°; 2,4,0°), which sets the path node S11 The discretized grid coordinates are: (4, 8, 0);
[0140] Forward left turn (turning angle +30°): Turning radius: R = L / tan(30°) = 4.33m; New orientation change: Δθ = d / R rad = 13.2°; new_theta = 0° + Δθ = 13.2°; 13.2° / 2 = 6.6°; new_x = 1 + 1*cos(6.6°) = 1.99; new_y = 4 + 1*sin(6.6°) = 4.12; S 12 =Node(1.99,4.12,13.2°); Operation cost: g.oper = 1 * 1 + 0.2 = 1.2; Total cost: f 12 =g.acc+g.oper+h=0+1.2+4.3=5.5; h=ReedsShepp(1,4,0°;1.99,4.12,13.2°), setting the path node S 12 The discretized grid coordinates are: (3,8,0);
[0141] Forward right turn (turning angle -30°): new_theta = 0° - Δθ = -13.2°; -13.2° / 2 = -6.6°; new_x = 1 - 1 * cos(-6.6°) = 0.01; new_y = 4 - 1 * sin(-6.6°) = 3.88; S 13 =Node(0.01,3.88,-13.2°); Operation cost: g.oper = 1 * 1.5 + 0.2 = 1.7; Total cost: f 13 =g.acc+g.oper+h=0+1.7+5.8=7.5; h=ReedsShepp(1,4,0°; 0.01,3.88,-13.2°), setting the path node S 13 The discretized grid coordinates are: (0, 7, -1);
[0142] Backwards in a straight line: new_x = 1 - 1 * cos(0°) = 0; new_y = 4 - 1 * sin(0°) = 4; new_theta = 0°; S 14 =Node(0,4,0°); Operation cost: g.oper = 1 * 1.5 = 1.5; Total cost: f 14 =g.acc+g.oper+h=0+1.5+5.2=6.7; h=ReedsShepp(1,4,0°; 0,4,0°), which sets the path node S 14 The discretized grid coordinates are: (0,8,0);
[0143] Turn left (turning angle +30°): new_theta=(d*tan(30°)) / L*(-1)=-13.24°;-13.24° / 2=-6.62°;new_x=1+1*cos(-6.62°)*(-1)=0.007;new_y=4+1*sin(-6.62°)*(-1)=4.115;S 15 =Node(0.007,4.115,-13.24°); Operation cost: g.oper = 1 * 5 + 0.2 = 1.7; Total cost: f 15 =g.acc+g.oper+h=0+1.7+5.5=7.2; where h=ReedsShepp(1,4,0°;0.007,4.15,-13.24°), which sets the path node S 15 The discretized grid coordinates are: (0, 8, -1);
[0144] Turn right (turning angle -30°): new_theta=(d*tan(-30°)) / L*(-1)=13.24°; 13.24° / 2=6.62°; new_x=1+1*cos(6.62°)*(-1)=0.007; new_y=4+1*sin(6.62°)*(-1)=3.885; S 16 =Node(0.007,3.885,13.24°); Operation cost: g.oper = 1 * 1.5 + 0.2 = 1.7; Total cost: f 16 =g.acc+g.oper+h=0+1.7+5.8=7.5; where h=ReedsShepp(1,4,0°;0.007,3.885,13.24°), which sets the path node S 16 The discretized grid coordinates are: (0,7,0);
[0145] Path node S 11 The grid coordinates are (4,8,0); the path node S 12 The grid coordinates are (3,8,0); the path node S 13 The grid coordinates are (0, 7, -1); the path node S 14 The grid coordinates are (0,8,0); the path node S 15 The grid coordinates are (0, 8, -1), and the path node is S. 16 The grid coordinates are (0,7,0); these 6 path nodes S 11 S 12 S 13 S 14 S 15 and S 16All of these are different from the grid coordinates (2,8,0) of the starting path node S0, meaning that the path node S0 can be... 11 Path node S 12 Path node S 13 Path node S 14 Path node S 15 Path node S 16 Add the corresponding total cost f; Close the list: Add the starting path node S0.
[0146] 3. Second round of loop: Continue to expand the sub-path nodes.
[0147] 1) Take any path node from the open list from the previous round (you can start by taking the path node with the smallest f), i.e., S 11 =Node(2,4,0°); Based on path node S 11 Target path node S corresponding to the target pose g =Node(4,2,90°) determines whether the termination condition is met: If 90°-0° is greater than 10°, the termination condition is not met, and the path node expansion continues.
[0148] 2) Using path node S 11 Based on the baseline, sub-path nodes are expanded according to the operation set. The specific implementation process can be found in steps 2 and 2) above. After determining the path node S 11 After the child path node, similarly, any path node from the previous round can be retrieved from the open list (we can start by retrieving the path node with the smallest f), i.e., S. 12 =Node(1.99,4.12,13.2°); Based on path node S 12 With the target path node S g =Node(4,2,90°) determines whether the termination condition is met: If 90° - 13.2° is greater than 10°, the termination condition is not met, and the expansion of the path node continues; this process continues until the path node S is... 11 Path node S 12 Path node S 13 Path node S 14 Path node S 15 Path node S 16 Once the child path nodes of each path node in the path are determined, the above steps can be repeated to determine their child path nodes until the determined path nodes match the target path node S. g The termination conditions have been met.
[0149] Furthermore, when considering obstacle collisions, path nodes that overlap with the location of obstacles can be discarded.
[0150] Assume that after multiple path node expansions, the resulting path nodes include: path node S0, and the child path nodes of path node S0 include path node S... 11 Path node S 12 and path node S 13 Path node S 11 The child path nodes include path node S 21 Path node S 12 The child path nodes include path node S 31 and path node S 32 Path node S 13 The child path nodes include path node S 41 Path node S 21 The child path nodes include path node S 51 Path node S 31 The child path nodes include path node S 61 Path node S 41 The child path nodes include path node S 71 Path node S 51 The child path nodes include path node S 81 Path node S 61 The child path nodes include path node S 91 and path node S 92 Path node S 71 The child path nodes include path node S 101 Path node S 81 child path nodes, path node S 91 and path node S 92 The child path nodes, and the path node S 101 All sub-path nodes are target path nodes S g Then, connect the path nodes that have a parent-child relationship among the above multiple path nodes.
[0151] Starting from the path node corresponding to the target pose (target path node), and passing through multiple expanded path nodes, we sequentially backtrack to the starting path node corresponding to the initial pose to obtain... Figure 4 The multiple parking paths shown in (b) include: S0->S 11 ->S 21 ->S 51 ->S 81 ->S g ;S0->S 12 ->S 31 ->S61 ->S 91 ->S g ;S0->S 12 ->S 31 ->S 61 ->S 92 ->S g ;S0->S 13 ->S 41 ->S 71 ->S 101 ->S g That is, we get 4 parking routes.
[0152] It should be understood that the above Figure 4 In the embodiment shown, the automatic parking system can be regarded as using the Hybrid A* algorithm, specifically utilizing the vehicle motion model (the aforementioned preset parking operation) and the Reeds Shepp heuristic algorithm (the ReedsShepp() function mentioned above) to search for possible path nodes (or multiple first predicted poses corresponding to path nodes).
[0153] The Hybrid A* algorithm described above is an improvement on the traditional A* algorithm, which is a grid-based heuristic search algorithm. It evaluates the cost of each grid node and prioritizes searching for the grid node with the lowest overall cost, thereby efficiently planning the shortest travel path in discrete space. It is widely used in path planning scenarios for robots on static maps.
[0154] The Hybrid A* algorithm introduces vehicle kinematic constraints (such as forward straight movement, forward right turn, and backward left turn) to the traditional A* algorithm, ensuring that the generated driving path conforms to the vehicle's dynamic characteristics. In path planning, the Hybrid A* algorithm is widely used in complex environments (such as narrow garages) in fields such as autonomous driving and robotics to generate smooth, executable, and obstacle-avoiding driving paths.
[0155] Step 203: Control the vehicle to park based on the parking path with the highest quality score among the multiple parking paths.
[0156] It should be understood that in step 203 above, the parking path with the highest quality score also needs to be converted into vehicle control actions, and the vehicle is then controlled to park in the target parking space based on these vehicle control actions. These vehicle control actions include steering actions such as the angle at which the steering wheel is turned, and state actions such as the vehicle speed.
[0157] In some embodiments, the method for determining the parking path with the highest quality score among the multiple parking paths includes: sorting the parking paths corresponding to the quality scores in descending order to obtain a sorted set of multiple parking paths; and determining the first parking path among the sorted multiple parking paths as the parking path with the highest quality score.
[0158] In some embodiments, step 203 includes: smoothing the parking path with the highest quality score to obtain a first target path; converting the first target path into multiple candidate vehicle control actions; adjusting the multiple candidate vehicle control actions based on preset dynamic control constraints to obtain multiple target vehicle control actions; and controlling the vehicle to park based on the multiple target vehicle control actions.
[0159] It should be understood that the smoothing of the parking path with the highest quality score mentioned above refers to processing the parking path itself to optimize its geometry and output a smoother first target path; adjusting the control actions of multiple candidate vehicles based on preset dynamic control constraints to obtain multiple target vehicle control actions refers to the process of solving the execution problem of path tracking to ensure that the vehicle smoothly tracks the first target path.
[0160] Furthermore, the above scheme can smooth the parking path with the highest quality score using a preset gradient descent method. Model predictive control algorithms can be used to adjust the control actions of multiple candidate vehicles based on preset dynamic control constraints.
[0161] Optionally, the dynamic control constraints include limiting the steering wheel's angular velocity to ≤70° / s.
[0162] In some embodiments, converting the first target path into multiple candidate vehicle control actions includes: decomposing the first target path into multiple consecutive path segments based on adjacent path nodes, each path segment being defined by two adjacent path nodes; for each path segment, extracting the vehicle state corresponding to the adjacent first node and the vehicle state corresponding to the adjacent second node, the vehicle state including candidate position coordinates and candidate orientation; determining a candidate speed for traveling from the first candidate position coordinate to the second candidate position coordinate based on the first candidate position coordinate and the second candidate position coordinate, and a preset time step; determining the first candidate steering wheel angle for traveling from the first candidate position coordinate to the second candidate position coordinate based on the first candidate orientation and the second candidate orientation; and determining the candidate speed and steering wheel angle corresponding to each path segment as multiple candidate vehicle control actions.
[0163] Figure 5 This is a schematic diagram of the structure of a device for planning parking routes provided in an embodiment of this application.
[0164] For example, such as Figure 5 As shown, the device 500 includes:
[0165] Determine module 501, used for:
[0166] In response to the vehicle's parking command, multiple parking paths are determined based on the vehicle's initial pose at the current location, the target pose of the target parking space, and the vehicle's preset parking operations during the parking process.
[0167] Determine the quality score of each parking path on the first planning objective. The first planning objective includes the angular deviation between the vehicle's front orientation and the orientation of the target parking space in the last parking operation of the parking path, the number of short paths in the parking path, and the proportion of path nodes in the parking path that meet the preset conditions. The short path is the path whose length corresponding to the parking operation is less than the preset length. The preset conditions are used to indicate that the curvature of the path node corresponding to the pose is less than the preset curvature.
[0168] The control module 502 is used to control the parking of the vehicle based on the parking path with the highest quality score among the multiple parking paths.
[0169] Optionally, the determining module 501 is specifically configured to: starting from the initial pose, determine multiple first predicted poses using the preset parking operation; take any one of the first predicted poses as a second predicted pose, and discretize the second predicted pose based on a preset grid resolution to obtain the grid coordinates corresponding to the second predicted pose; determine whether the second predicted pose should be discarded based on whether the grid coordinates corresponding to the second predicted pose are the same as the grid coordinates corresponding to the first pose, so as to correct the multiple first predicted poses, wherein the path node corresponding to the first pose is the parent node of the path node corresponding to the second predicted pose; and determine the multiple parking paths based on the initial pose, the target pose, and the corrected multiple first predicted poses.
[0170] Optionally, the preset grid resolution includes position resolution and orientation resolution. The device 500 further includes a processing module, configured to: determine the ratio between the first horizontal axis coordinate in the second predicted pose and the first length corresponding to the position resolution as the target first horizontal axis coordinate in the grid coordinates; determine the ratio between the first vertical axis coordinate in the second predicted pose and the first length as the target first vertical axis coordinate in the grid coordinates; and determine the ratio between the first orientation in the second predicted pose and the first orientation corresponding to the orientation resolution as the target first orientation in the grid coordinates.
[0171] Optionally, the determining module 501 is further configured to: determine the second pose corresponding to the target pose from the modified plurality of first predicted poses, and connect the path node corresponding to the target pose with the path node corresponding to the second pose, wherein the parent node of the path node corresponding to the target pose is the path node corresponding to the second pose; delete the second pose from the modified plurality of first predicted poses to obtain at least one third predicted pose, determine the third pose corresponding to each third predicted pose, and connect the path node corresponding to each third predicted pose with the path node corresponding to the third pose, wherein the parent node of the path node corresponding to the third predicted pose is the path node corresponding to the second pose; The parent node is the path node corresponding to the third pose, to determine each fourth predicted pose corresponding to the initial pose. The path nodes corresponding to each fourth predicted pose should be connected to the path node corresponding to the initial pose. The parent node of each fourth predicted pose is the path node corresponding to the initial pose. Based on the connection relationship between the initial pose, the target pose, and the multiple path nodes formed by the multiple corrected first predicted poses, starting from the path node corresponding to the target pose, passing through the path nodes corresponding to the multiple corrected first predicted poses, and sequentially backtracking to the path node corresponding to the initial pose, multiple parking paths are obtained.
[0172] Optionally, the device 500 further includes: an output module for outputting a weight configuration reminder, which reminds the target object in the vehicle to configure the weights corresponding to each first planning target; a fusion module for, upon receiving the weights corresponding to each first planning target, weighting and fusing the first score, second score, and third score based on the first configuration weight, second configuration weight, and third configuration weight to obtain the quality score, wherein the first score is based on angle deviation, the second score is based on the number of short paths, and the third score is based on the proportion of path nodes that meet preset conditions; the output module is also used to output a target configuration reminder when the weights corresponding to each first planning target are not received, which reminds the target object to use driving style as the planning target; and the determination module 501 is specifically used to, upon receiving a confirmation operation for the target configuration reminder, determine the fourth configuration weight, fifth configuration weight, and sixth configuration weight for weighting and fusing the first score, second score, and third score based on the driving style to determine the quality score.
[0173] Optionally, the determining module 501 is further configured to: when the driving style is aggressive, determine the fourth configuration weight as the first preset weight, the fifth configuration weight as the second preset weight, and the sixth configuration weight as the third preset weight, wherein the first preset weight is less than the second preset weight, and the third preset weight is less than the second preset weight; when the driving style is conservative, determine the fourth configuration weight as the fourth preset weight, the fifth configuration weight as the fifth preset weight, and the sixth configuration weight as the sixth preset weight, wherein the fifth preset weight is less than the fourth preset weight, and the fifth preset weight is less than the sixth preset weight; when the driving style is balanced, determine the fourth configuration weight, the fifth configuration weight, and the sixth configuration weight as target preset weights.
[0174] Optionally, the determining module 501 is further configured to: determine the target path cost of each parking path, wherein the parking path is determined based on at least one first predicted pose related to the initial pose, the target pose, and the preset parking operation; the target path cost is the sum of the first path cost of the vehicle traveling from the initial pose to any first predicted pose and the second path cost of traveling from any first predicted pose to the target pose; the first path cost is the sum of the cumulative path cost of the vehicle traveling from the initial pose to a fifth predicted pose and the operation cost of traveling from the fifth predicted pose to any first predicted pose; the path node corresponding to the fifth predicted pose is the parent node of the path node corresponding to any first predicted pose; determine the target path cost, total path length, and number of paths of each parking path as the second planning objective when planning the parking path; and determine the quality score of each parking path on the first planning objective and the second planning objective.
[0175] Figure 6 This is a schematic diagram of the structure of a vehicle provided in an embodiment of this application.
[0176] For example, such as Figure 6 As shown, the vehicle 600 includes a memory 601 and a processor 602. The memory 601 stores executable program code 603, and the processor 602 is used to call and execute the executable program code 603 to perform a method for planning a parking path.
[0177] Furthermore, embodiments of this application also protect an apparatus that may include a memory and a processor, wherein the memory stores executable program code, and the processor is used to call and execute the executable program code to perform a method for planning parking routes provided in embodiments of this application.
[0178] This embodiment can divide the device into functional modules according to the above method example. For example, each module can correspond to a separate function, or two or more functions can be integrated into one processing module. The integrated module can be implemented in hardware. It should be noted that the module division in this embodiment is illustrative and only represents one logical functional division. In actual implementation, there may be other division methods.
[0179] When each functional module is divided according to its corresponding function, the device may further include a determining module, a controlling module, a processing module, an output module, and a fusion module. It should be noted that all relevant content in the above method embodiments can be referenced from the functional descriptions of the corresponding functional modules, and will not be repeated here.
[0180] It should be understood that the apparatus provided in this embodiment is used to execute the above-described method for planning parking routes, and therefore can achieve the same effect as the above-described implementation method.
[0181] When using an integrated unit, the device may include a processing module and a storage module. When the device is applied to a vehicle, the processing module can be used to control and manage the vehicle's movements. The storage module can be used to support the vehicle in executing relevant executable program code.
[0182] The processing module may be a processor or a controller, which can implement or execute various exemplary logic blocks, modules, and circuits shown in conjunction with the disclosure of this application. The processor may also be a combination of functions that implement computing capabilities, such as a combination of one or more microprocessors, a combination of digital signal processing (DSP) and a microprocessor, etc., and the storage module may be a memory.
[0183] In addition, the device provided in the embodiments of this application may specifically be a chip, component or module. The chip may include a connected processor and a memory. The memory is used to store instructions. When the processor calls and executes the instructions, the chip can execute a parking path planning method provided in the above embodiments.
[0184] This embodiment also provides a computer-readable storage medium storing executable program code. When the executable program code is run on a computer, the computer performs the aforementioned method steps to implement the parking path planning method provided in the above embodiment.
[0185] This embodiment also provides a computer program product that, when run on a computer, causes the computer to perform the aforementioned steps to implement a method for planning parking routes provided in the above embodiment.
[0186] In this embodiment, the device, computer-readable storage medium, computer program product, or chip are all used to execute the corresponding methods provided above. Therefore, the beneficial effects they can achieve can be referred to the beneficial effects in the corresponding methods provided above, and will not be repeated here.
[0187] Through the above description of the embodiments, those skilled in the art will understand that, for the sake of convenience and brevity, only the division of the above functional modules is used as an example. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above.
[0188] In the embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.
[0189] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.
Claims
1. A method of planning a parking path, characterized in that, The method includes: In response to a parking command from a vehicle, multiple parking paths are determined based on the initial pose of the vehicle at its current position, the target pose of the target parking space, and the preset parking operations performed by the vehicle during the parking process. Determine the quality score of each parking path on the first planning objective. The first planning objective includes the angular deviation between the vehicle's front orientation and the orientation of the target parking space in the last parking operation corresponding to the parking path, the number of short paths in the parking path, and the proportion of path nodes in the parking path that meet preset conditions. The short path is the path whose length corresponding to the parking operation is less than a preset length. The preset conditions are used to indicate that the curvature of the path node corresponding to the pose is less than a preset curvature. The vehicle is parked based on the parking path with the highest quality score among the multiple parking paths; The determination of the quality score of each parking path on the first planning objective includes: Output weight configuration reminder, which is used to remind the target objects in the vehicle to configure the weights corresponding to each first planning target; Upon receiving the weights corresponding to each of the first planning objectives, the first score, the second score, and the third score are weighted and fused based on the first configuration weight, the second configuration weight, and the third configuration weight to obtain the quality score. The first score is obtained based on the angle deviation, the second score is obtained based on the number of short paths, and the third score is obtained based on the proportion of path nodes that meet preset conditions. If the weights corresponding to each of the first planning objectives are not received, a target configuration reminder is output, which is used to remind the target object to take driving style as the planning objective; Upon receiving a confirmation operation for the target configuration reminder, based on the driving style, a fourth configuration weight, a fifth configuration weight, and a sixth configuration weight are determined for weighted fusion of the first score, the second score, and the third score, in order to determine the quality score.
2. The method of claim 1, wherein, The method determines multiple parking paths based on the vehicle's initial pose at its current location, the target pose of the target parking space, and the vehicle's preset parking operations during the parking process, including: Starting from the initial pose, multiple first predicted poses are determined using the preset parking operation; Any first predicted pose is used as the second predicted pose. Based on a preset grid resolution, the second predicted pose is discretized to obtain the grid coordinates corresponding to the second predicted pose. Based on whether the mesh coordinates corresponding to the second predicted pose are the same as the mesh coordinates corresponding to the first pose, it is determined whether the second predicted pose should be discarded in order to correct the plurality of first predicted poses. The path node corresponding to the first pose is the parent node of the path node corresponding to the second predicted pose. Based on the initial pose, the target pose, and the corrected multiple first predicted poses, the multiple parking paths are determined.
3. The method of claim 2, wherein, The preset grid resolution includes position resolution and orientation resolution. The step of discretizing the second predicted pose based on the preset grid resolution to obtain the grid coordinates corresponding to the second predicted pose includes: The ratio between the first horizontal axis coordinate in the second predicted pose and the first length corresponding to the position resolution is determined as the target first horizontal axis coordinate in the grid coordinates; The ratio between the first vertical axis coordinate in the second predicted pose and the first length is determined as the target first vertical axis coordinate in the grid coordinates; The ratio between the first orientation in the second predicted pose and the first orientation corresponding to the orientation resolution is determined as the target first orientation in the grid coordinates.
4. The method according to claim 2, characterized in that, The determination of the multiple parking paths based on the initial pose, the target pose, and the corrected multiple first predicted poses includes: The second pose corresponding to the target pose is determined from the multiple corrected first predicted poses, and the path node corresponding to the target pose is connected to the path node corresponding to the second pose, wherein the parent node of the path node corresponding to the target pose is the path node corresponding to the second pose. The second pose is deleted from the multiple first predicted poses after correction to obtain at least one third predicted pose. The third pose corresponding to each third predicted pose is determined, and the path nodes corresponding to each third predicted pose are connected to the path nodes corresponding to the third pose. The parent node of the path node corresponding to the third predicted pose is the path node corresponding to the third pose. The fourth predicted pose corresponding to each fourth predicted pose is determined, and the path nodes corresponding to each fourth predicted pose are connected to the path nodes corresponding to the initial pose. The parent node of the path nodes corresponding to each fourth predicted pose is the path node corresponding to the initial pose. Based on the connection relationship between the initial pose, the target pose, and the multiple path nodes formed by the corrected multiple first predicted poses, starting from the path node corresponding to the target pose, passing through the path nodes corresponding to the multiple corrected first predicted poses, and sequentially tracing back to the path node corresponding to the initial pose, the multiple parking paths are obtained.
5. The method according to claim 1, characterized in that, The determination of the fourth, fifth, and sixth configuration weights based on the driving style for weighted fusion of the first, second, and third scores includes: When the driving style is aggressive, the fourth configuration weight is determined as the first preset weight, the fifth configuration weight is determined as the second preset weight, and the sixth configuration weight is determined as the third preset weight. The second preset weight is greater than the first preset weight, and the second preset weight is greater than the third preset weight. When the driving style is a stable style, the fourth configuration weight is determined as the fourth preset weight, the fifth configuration weight is determined as the fifth preset weight, and the sixth configuration weight is determined as the sixth preset weight. The fifth preset weight is less than the fourth preset weight, and the fifth preset weight is less than the sixth preset weight. When the driving style is a balanced style, the fourth configuration weight, the fifth configuration weight, and the sixth configuration weight are determined as the target preset weights.
6. The method according to claim 1, characterized in that, The determination of the quality score of each parking route on the first planning objective includes: The target path cost of each parking path is determined, wherein the parking path is determined based on at least one first predicted pose related to the initial pose, the target pose, and the preset parking operation. The target path cost is the sum of the first path cost of the vehicle traveling from the initial pose to any first predicted pose and the second path cost of traveling from any first predicted pose to the target pose. The first path cost is the sum of the cumulative path cost of the vehicle traveling from the initial pose to the fifth predicted pose and the operation cost of traveling from the fifth predicted pose to any first predicted pose. The path node corresponding to the fifth predicted pose is the parent node of the path node corresponding to any first predicted pose. The target path cost, total path length, and number of paths for each parking path are determined as the second planning objective when planning parking paths. Determine the quality score of each parking path on the first planning objective and the second planning objective.
7. A device for planning parking routes, characterized in that, The device includes: The determination module is used for: In response to a parking command from a vehicle, multiple parking paths are determined based on the initial pose of the vehicle at its current position, the target pose of the target parking space, and the preset parking operations performed by the vehicle during the parking process. Determine the quality score of each parking path on the first planning objective. The first planning objective includes the angular deviation between the vehicle's front orientation and the orientation of the target parking space in the last parking operation corresponding to the parking path, the number of short paths in the parking path, and the proportion of path nodes in the parking path that meet preset conditions. The short path is the path whose length corresponding to the parking operation is less than a preset length. The preset conditions are used to indicate that the curvature of the path node corresponding to the pose is less than a preset curvature. The control module is used to control the parking of the vehicle based on the parking path with the highest quality score among the multiple parking paths; The determining module is specifically used to output a weight configuration reminder, which reminds the target object in the vehicle to configure the weights corresponding to each first planning target; upon receiving the weights corresponding to each of the first planning targets, the module performs a weighted fusion of the first score, the second score, and the third score based on the first configuration weight, the second configuration weight, and the third configuration weight to obtain the quality score, where the first score is based on the angle deviation, the second score is based on the number of short paths, and the third score is based on the proportion of path nodes that meet preset conditions; if the weights corresponding to each of the first planning targets are not received, the module outputs a target configuration reminder, which reminds the target object to use driving style as the planning target; upon receiving a confirmation operation for the target configuration reminder, the module determines a fourth configuration weight, a fifth configuration weight, and a sixth configuration weight based on the driving style for weighted fusion of the first score, the second score, and the third score to determine the quality score.
8. A vehicle, characterized in that, The vehicles include: Memory, used to store executable program code; A processor for calling and running the executable program code from the memory, causing the vehicle to perform the method as described in any one of claims 1 to 6.
9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores executable program code that, when executed, implements the method as described in any one of claims 1 to 6.
Citation Information
Patent Citations
Parking control method and device, computer equipment and storage medium
CN111806431A
Automatic parking path planning method and system
CN118244274A
Path planning method and device, equipment and storage medium
CN118597114A
