Parking path planning method and device, storage medium, vehicle-mounted equipment and vehicle

Abstract

The invention discloses a parking path planning method and device, a storage medium, vehicle-mounted equipment and a vehicle. The parking path planning method comprises the following steps: identifying a U-turn point candidate area; selecting a target U-turn point from the U-turn point candidate area; searching a first effective path from the parking space point to the target turning point based on the narrow channel parameters; searching a second effective path from the target turning point to the vehicle point based on the broken road parameters; and performing path fusion on the first effective path and the second effective path to obtain a parking path. Through the method, high-precision and high-efficiency path planning in a narrow channel and broken road combined scene is realized, the success rate, the search efficiency and the robustness of path planning can be remarkably improved, and meanwhile, the method has relatively high scene adaptability and expandability; and the landing application of the intelligent driving vehicle in the narrow channel and broken road combined scene can be promoted.

Description

Technical Field

[0001] This application relates to the field of intelligent driving technology, and in particular to a parking path planning method, device, storage medium, vehicle-mounted equipment, and vehicle. Background Technology

[0002] In intelligent driving parking scenarios, parking path planning is one of the core aspects of achieving autonomous parking. It requires generating a smooth and continuous parking trajectory that meets vehicle kinematic constraints, obstacle avoidance safety requirements, and directly impacts the efficiency and safety of the parking process. Currently, the A* search algorithm is widely used in parking path planning due to its high search efficiency and strong path controllability. In the path planning process of different parking scenarios, this algorithm needs to adapt to the scenario requirements by setting specific search parameters to balance path feasibility and search efficiency. For example, in narrow passage scenarios, a smaller grid size and step size are used to fully utilize space to obtain feasible paths, ensuring the accuracy of path planning. Conversely, in dead-end road scenarios, a larger grid size and step size are used to quickly generate U-turn trajectories, improving search speed. Meanwhile, during the path search process, the path cost function Cost(n) is generally defined as the sum of the actual cost g(n) from the starting point to the current node and the heuristic cost h(n) from the current node to the target point. The heuristic cost must satisfy the constraint that it is less than or equal to the true optimal cost (Cost-to-Go) to ensure the optimality of the search results.

[0003] However, in complex parking scenarios involving narrow passages and dead-end roads (such as...) Figure 1 The example shows a narrow passage, only 30-50cm wider than the vehicle, connecting to the target parking space. The end of the passage is a wall or fixed obstacle, creating a "dead end." The raised trapezoidal area in the vehicle graphic represents the front of the vehicle. The open area in the image provides space for turning around; see subsequent sections for details. Figure 3The existing technology has significant limitations in the following aspects: First, a single parameter setting cannot simultaneously meet the dual requirements of the combined scenario. Using fine-grid modeling leads to excessively slow search speeds, potentially causing timeouts before generating a U-turn trajectory. Conversely, using coarse-grid modeling makes it difficult to generate accurate and feasible paths near parking space entrances. Second, heuristic function design struggles to effectively guide the vehicle to find a reasonable path, easily causing it to repeatedly explore near obstacles and become trapped in local optima, failing to guide the vehicle to reverse into an open area, make a U-turn, and then enter the parking space. Third, in the path cost function, the actual cost increment from ancestor nodes to child nodes on the potentially optimal path often far exceeds the heuristic cost reduction. This leads to a preference for expanding nearby low-cost nodes due to the lower cost of ancestor nodes, rather than advancing along the potentially optimal path, resulting in low overall search efficiency. These problems severely impact the robustness and real-time performance of parking path planning for intelligent driving vehicles in complex and constrained scenarios, resulting in poor path feasibility. Summary of the Invention

[0004] This application provides a parking path planning method, device, storage medium, vehicle-mounted equipment, and vehicle, aiming to solve the problem of poor path feasibility in traditional parking path planning methods in scenarios involving narrow passages and dead-end roads.

[0005] To address the aforementioned technical problems, this application provides a parking path planning method. This method includes: identifying candidate U-turn point areas; selecting a target U-turn point from the candidate U-turn point areas; searching a first effective path from the parking spot to the target U-turn point based on narrow channel parameters; searching a second effective path from the target U-turn point to the vehicle's starting point based on dead-end road parameters; and fusing the first and second effective paths to obtain a parking path.

[0006] In some embodiments, identifying the candidate area for U-turn points includes: identifying the area within a preset drivable range whose area is not less than the sum of the vehicle's minimum U-turn space area and the safety redundancy area, and whose distance from each obstacle is not less than the vehicle's maximum outline radius, as the candidate area for U-turn points.

[0007] In some embodiments, selecting a target U-turn point from the candidate U-turn point region includes: filtering valid U-turn points in the candidate U-turn point region based on preset kinematic constraint indicators and preset accessibility indicators; scoring each valid U-turn point based on U-turn space sufficiency indicators and proximity to parking spaces indicators, and selecting the valid U-turn point with the highest score as the target U-turn point.

[0008] In some embodiments, searching for a first effective path from the parking spot to the target U-turn point based on narrow channel parameters includes: searching for a plurality of nodes with the lowest obstacle avoidance priority cost from the parking spot to the target U-turn point based on narrow channel parameters to obtain the first effective path; wherein the obstacle avoidance priority cost includes a first actual cost and a first heuristic cost, the first actual cost includes the steering angle cost from the parking spot to the corresponding node and the narrow channel boundary distance penalty cost, and the first heuristic cost includes the motion curve length penalty cost from the corresponding node to the target U-turn point and the number of gear shifts cost.

[0009] In some embodiments, searching for a second effective path from the target U-turn point to the vehicle's starting point based on dead-end road parameters includes: searching for a plurality of nodes with the lowest efficiency-priority cost from the target U-turn point to the vehicle's starting point based on dead-end road parameters to obtain the second effective path; wherein the efficiency-priority cost includes a second actual cost and a second heuristic cost, the second actual cost includes the path length cost from the target U-turn point to the corresponding node and the steering smoothness penalty cost, and the second heuristic cost includes the motion curve length penalty cost from the corresponding node to the vehicle's starting point and the number of gear shifts cost.

[0010] In some embodiments, the path fusion of the first effective path and the second effective path to obtain a parking path includes: smoothing the first effective path, the second effective path, and the path connection segments using B-spline curves to obtain the parking path; wherein the smoothing process constrains the continuity index of the path connection segments for adapting to U-turns and constrains the curvature of the parking path to be continuous and the maximum curvature not greater than the reciprocal of the vehicle's minimum turning radius.

[0011] To address the aforementioned technical problems, another technical solution adopted in this application is: providing a parking path planning device, which includes a U-turn point preprocessing module for identifying candidate U-turn point areas and selecting a target U-turn point from the candidate U-turn point areas; a hierarchical path search module for searching a first effective path from the parking spot to the target U-turn point based on narrow channel parameters, and searching a second effective path from the target U-turn point to the vehicle's location based on dead-end road parameters; and a path fusion module for fusing the first effective path and the second effective path to obtain a parking path.

[0012] To solve the above-mentioned technical problems, another technical solution adopted in this application is: to provide a storage medium storing program data, characterized in that the program data, when executed by a processor, implements the steps of the parking path planning method as described above.

[0013] To solve the above-mentioned technical problems, another technical solution adopted in this application is: to provide an in-vehicle device, which includes a processor and a memory connected to each other, wherein the memory stores a computer program, and when the processor executes the computer program, it implements the steps of the parking path planning method described above.

[0014] To solve the above-mentioned technical problems, another technical solution adopted in this application is to provide a vehicle that includes at least one of the parking path planning device, storage medium and vehicle-mounted equipment as described above.

[0015] The beneficial effects of this application are as follows: Unlike existing technologies, this application discloses a parking path planning method, device, storage medium, in-vehicle equipment, and vehicle. This application provides clear guidance for path search by identifying candidate U-turn point areas and filtering target U-turn points, avoiding the vehicle falling into local optima traps due to heuristic irrationality in scenarios involving narrow passages and dead-end roads. Through a hierarchical path search strategy, the complex scenario of narrow passages and dead-end roads is decomposed into two independent paths: "from parking spot to target U-turn point" and "from target U-turn point to vehicle's starting point." These paths can be searched using search methods adapted to narrow passages and dead-end roads, respectively, balancing path planning accuracy and search speed, and ensuring a balance between feasibility and efficiency in different scenarios. By fusing the paths found in the two layers, the path curvature is ensured to be continuous and meets vehicle kinematic constraints, improving path executability and eliminating abrupt changes at path junctions to ensure the smoothness of the overall path. This application effectively solves the problems of poor path feasibility, low search efficiency, and easy getting trapped in local optima in traditional path planning methods in scenarios with narrow passages and dead ends. It achieves high-precision and high-efficiency path planning in complex and restricted scenarios, which can significantly improve the success rate, search efficiency and robustness of path planning. At the same time, it has strong scenario adaptability and scalability, which is conducive to promoting the application of intelligent driving vehicles in scenarios with narrow passages and dead ends. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort, wherein: Figure 1 This is a schematic diagram of a scenario combining narrow passages and dead-end roads; Figure 2 This is a flowchart illustrating an embodiment of the parking path planning method provided in this application; Figure 3 yes Figure 1 A schematic diagram of a parking method in a scenario combining narrow passages and dead-end roads; Figure 4 This is a schematic diagram of the structure of an embodiment of the parking path planning device provided in this application; Figure 5 This is a schematic diagram of the structure of an embodiment of the storage medium provided in this application; Figure 6 This is a schematic diagram of the structure of an embodiment of the vehicle-mounted device provided in this application; Figure 7 This is a structural schematic diagram of an embodiment of the vehicle provided in this application. Detailed Implementation

[0017] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0018] The terms "first," "second," and "third" used in the embodiments of this application are for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first," "second," or "third" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or devices.

[0019] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a mutually exclusive, independent, or alternative embodiment. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0020] This application provides a parking path planning method, see reference. Figure 2 , Figure 2 This is a flowchart illustrating an embodiment of the parking path planning method provided in this application. The parking path planning method includes: Step 10: Identify candidate regions for U-turn points.

[0021] In this embodiment, the candidate area for a U-turn point refers to a drivable area within a certain range where vehicles can complete a U-turn. (See also...) Figure 3 Understandably, in such cases Figure 1 Based on the scenario of a combination of narrow passages and dead ends, if a vehicle needs to be parked in a parking space at the end of a narrow passage, it can be reversed to an open area via path S1 (the diagram uses this open area at the end of the narrow passage as an example; this open area needs to meet certain space requirements to ensure the vehicle can smoothly complete the U-turn; see the subsequent constraints on this space for details). Then, it can make a U-turn in this area via path S2, and after the U-turn, reverse into the parking space via path S3. Therefore, this open area, or a portion of it, can serve as a candidate area for identifying U-turn points for subsequent parking path planning. It is important to understand that... Figure 1 and Figure 3 The above examples are merely illustrative and do not represent the actual recognition results of autonomous vehicles. The candidate U-turn area is an area outside the narrow passage definition defined in this application that meets the vehicle U-turn kinematic constraints and obstacle avoidance safety requirements described later. For example, an open area is a relatively open part in the middle of a road that includes the narrow passage. This part does not meet the definition of a narrow passage in this method (e.g., its width is at least 50cm longer than the vehicle length), and therefore is not considered a narrow passage in this application but is considered part of the open area to determine whether this part meets the criteria for a candidate U-turn area. If this part still does not meet the criteria for a candidate U-turn area, the search for a candidate U-turn area that meets the criteria will continue in a more distant open area until a candidate U-turn area that meets the criteria is identified, all potential open areas are traversed, or the identification process of the candidate area is actively / passively terminated.

[0022] In this embodiment, the identification of U-turn candidate regions can specifically employ methods such as obstacle boundary extraction and dilation calculation based on perceptual point clouds, semantic information extraction and region filtering based on high-precision maps, or obstacle detection and region segmentation based on vehicle-mounted camera images. For example, in... Figure 3In cases where an open area exists at the end of a narrow passage, obstacle point clouds in the end and surrounding areas can be identified first using LiDAR, surround-view cameras, etc. Semantic segmentation or clustering of these obstacle point clouds can then be performed to determine the location of the open area. Based on this, an area satisfying the U-turn spatial conditions can be identified using dilation algorithms or pre-defined geometric model matching methods as a candidate U-turn point area for further processing. Identifying this candidate U-turn point area allows for rapid location of effective areas suitable for vehicle U-turns, providing a reliable foundation for subsequent selection of effective U-turn points. This prevents vehicles from getting trapped in local optima within narrow passages, significantly improving the efficiency and feasibility of path planning.

[0023] Optionally, in some embodiments, identifying candidate U-turn point regions includes: identifying regions within a preset drivable range whose area is not less than the sum of the vehicle's minimum U-turn space area and the safety redundancy area, and whose distance from each obstacle is not less than the vehicle's maximum outline radius as candidate U-turn point regions.

[0024] In this optional embodiment, a method for identifying candidate U-turn point areas in the vehicle's environment is specified. The preset drivable range refers to the area within which the vehicle can safely drive in the current environment. This range can be determined using vehicle sensor data, high-precision map information, or pre-set safety boundaries to exclude areas the vehicle cannot enter or areas outside its perception range. Alternatively, areas such as the end of narrow passages or other potentially open areas can be directly used as the preset drivable range, and spatial aggregator analysis can be performed to determine whether the corresponding area is sufficiently open and meets the conditions for being identified as a candidate U-turn point area.

[0025] In this optional embodiment, candidate regions for U-turn points are identified based on constraints in terms of both area and distance. Regarding area, the minimum U-turn space area and the safety redundancy area can be predefined spatial thresholds based on the vehicle's kinematic parameters (such as wheelbase and minimum turning radius). The minimum U-turn space area refers to the minimum closed area required for the vehicle to complete a U-turn without obstacle interference. This can be determined and preset based on the minimum U-turn circle area calculated from the vehicle's wheelbase and minimum turning radius. The safety redundancy area, on the other hand, refers to the additional space area reserved to offset uncertainties such as sensor measurement errors and dynamic changes in obstacles. This can be preset based on factors such as sensor accuracy, obstacle type, or scene complexity to ensure obstacle avoidance safety during the U-turn process. When verifying the area of ​​a candidate region, spatial geometric calculations (such as polygon area integration) can be used to determine whether the area of ​​the candidate region is not less than the sum of the minimum turning space area of ​​the vehicle and the safety redundancy area. If the area of ​​the candidate region is less than the sum, it means that the space in the region is insufficient and cannot meet the kinematic requirements of the vehicle turning around; if the area of ​​the candidate region is not less than the sum, it means that the space in the region is sufficient and can be used as a candidate region to enter the subsequent verification stage.

[0026] In this optional embodiment, regarding distance, the maximum outline radius of the vehicle refers to the radius of the largest circumscribed circle formed by the vehicle's projection onto the horizontal plane. Alternatively, it can be set as half the maximum width of the vehicle plus a preset safety margin (e.g., 0.3m). This parameter reflects the lateral space occupied by the vehicle during driving or U-turns. When the distance between the area and the obstacle is less than the maximum outline radius of the vehicle, collisions with obstacles are likely during U-turns, making obstacle avoidance safety difficult to guarantee. Conversely, when the distance between the area and the obstacle is not less than the maximum outline radius of the vehicle, the area meets the obstacle avoidance safety requirements and can be considered a candidate area for subsequent verification. By constraining both area and distance, it is ensured that the identified U-turn candidate areas meet the kinematic feasibility of vehicle U-turns, guarantee obstacle avoidance safety, and provide a reliable basis for the subsequent selection of effective U-turn points.

[0027] Step 20: Select the target U-turn point from the candidate U-turn point area.

[0028] In this embodiment, after identifying the candidate U-turn point area, it indicates the existence of potential U-turn locations that meet the basic space and obstacle avoidance conditions. These potential locations can support the initial needs of the vehicle to complete the U-turn operation. A target U-turn point can then be selected from this candidate area for use as a connecting node in subsequent hierarchical path search. This target U-turn point will serve as a key guiding node in the hierarchical path search process, facilitating segmented path planning between the parking space and the vehicle's initial position. This prevents the vehicle from getting stuck in a local optimum search within narrow passages, providing clear directional guidance for the planning of the subsequent two path segments and improving the overall feasibility and search efficiency. Methods for selecting target U-turn points include, but are not limited to, multi-dimensional feasibility verification and scoring methods as described later in this application, or quantitative evaluation algorithms such as the RRT* (Rapidly-exploring Random Tree Star) fast verification method and dynamic grid sampling method. The selection of the target U-turn point is sufficient to ensure that it meets vehicle kinematic constraints and has path reachability, enabling efficient connection of hierarchical path search and improving the overall parking path planning efficiency and execution robustness.

[0029] Optionally, in some embodiments, selecting a target U-turn point from the candidate U-turn point area includes: filtering valid U-turn points in the candidate U-turn point area based on preset kinematic constraint indicators and preset accessibility indicators; scoring each valid U-turn point based on U-turn space sufficiency indicators and proximity to parking spaces indicators, and selecting the valid U-turn point with the highest score as the target U-turn point.

[0030] In this optional embodiment, a method for selecting U-turn points is specifically described, which uses multi-dimensional indicators to screen effective U-turn points and quantifies and scores them to determine the target U-turn point. The screened points can be randomly selected from the candidate U-turn point area or selected according to specific rules, so as to select effective U-turn points that meet vehicle kinematic constraints and path accessibility requirements.

[0031] In this optional embodiment, the criteria for selecting effective U-turn points specifically include preset kinematic constraint indicators and preset accessibility indicators. The preset kinematic constraint indicators refer to the multi-dimensional kinematic constraint requirements that a vehicle must meet when entering the point from a narrow passage (such as turning radius adaptability, gear shifting operation feasibility, speed constraint compliance, and vehicle profile passability). The selection of these indicators aims to eliminate candidate areas that cannot safely turn around due to unmet kinematic constraints, ensuring the feasibility of subsequent paths. For example, a point where the minimum turning radius of the vehicle entering the point from the narrow passage matches the vehicle's kinematic parameters (such as wheelbase and minimum turning radius) can be determined as an effective U-turn point, ensuring that effective U-turn points are all points that the vehicle can reach by turning, while eliminating candidate areas that the vehicle cannot smoothly enter due to excessively large or small turning radii.

[0032] In this optional embodiment, the preset accessibility index refers to the collision-free connectivity requirement between the point and the parking space and the vehicle's initial position. The screening based on this index aims to ensure that the point can serve as a connecting node between two paths, avoiding planning failure due to path interruption. For example, this screening process can determine whether the point can connect the parking space and the vehicle's initial position via a collision-free path, ensuring that effective U-turn points are all points that the vehicle can reach via a safe path, thereby excluding points in areas inaccessible due to obstacles or narrow paths. Through the dual verification of these two indexes, it can be ensured that the selected effective U-turn points satisfy both the vehicle's kinematic constraints and path accessibility, laying a reliable foundation for the subsequent selection of target U-turn points.

[0033] In this optional embodiment, after selecting valid U-turn points, a scoring mechanism is used to quantitatively evaluate each valid U-turn point based on a U-turn space sufficiency index and a proximity index to parking spaces. The U-turn space sufficiency index is a core parameter measuring the amount of space available for vehicles to make U-turns in the area where the point is located. Incorporating this index into the scoring mechanism allows for the priority selection of locations with larger U-turn spaces. For example, the U-turn space radius r of the point can be used as a metric. k (i.e., the radius of the largest circular area in the region where the point is located that can accommodate a vehicle to complete a U-turn) is used as this indicator to avoid the number of times the vehicle needs to maneuver (i.e., the number of times the vehicle needs to move back and forth to adjust its position, such as r) due to insufficient space. k =6m requires 5 kneading cycles, r k =When the distance is 8m, only 3 turns are needed (too many turns lead to low path execution efficiency, so it is important to ensure the efficiency of path planning); while the proximity index to the parking space is a key parameter for measuring the straight-line distance between the point and the parking space. By incorporating this index into the scoring mechanism, priority can be given to U-turn points that are closer to the parking space. For example, the straight-line distance between the point and the parking space can be used as a metric. k This metric aims to prevent the path planning from becoming too long and complex due to excessive distance, and to avoid the reduction in robustness caused by the accumulation of perceived noise interference, thus ensuring the stability and real-time performance of path planning.

[0034] In this optional embodiment, after considering the above two indicators, it can be specifically determined according to the formula f(k) = a·r. k - b·dist k The score f(k) for the kth valid U-turn point is calculated, where r k The dist is an indicator of the sufficiency of U-turn space (specifically, the U-turn space radius of the kth effective U-turn point). kThe proximity index to parking spaces (specifically, the straight-line distance between the k-th valid U-turn point and the parking space) is used. 'a' and 'b' are dynamically adjusted weighting coefficients that can be set according to the characteristics of the scenario. For example, in a narrow passage scenario, 'a' can be set to 1.2-1.5 to prioritize ensuring U-turn space, while when the parking space and the narrow passage are tightly connected, 'b' can be set to 1.0-1.2 to prioritize shortening the path distance. This ensures that the scoring results balance the sufficiency of U-turn space and the proximity to parking spaces. After calculating the scores of all valid U-turn points, the valid U-turn point with the highest score can be selected as the target U-turn point, providing a key connecting node for subsequent hierarchical path search. The target U-turn point selected in this way can meet the vehicle's kinematic constraints and path accessibility, while also taking into account the sufficiency of U-turn space and proximity to parking spaces. This ensures that the vehicle has enough space to complete the U-turn operation, reducing the number of times it has to maneuver around the parking space and improving path execution efficiency. At the same time, it enables the vehicle to reach the parking space via a shorter path, reducing path complexity and perception noise interference, enhancing the robustness and real-time performance of path planning, and providing reliable and efficient guidance for subsequent hierarchical path search. This is conducive to improving the quality and feasibility of the overall parking path planning.

[0035] Step 30: Based on the narrow channel parameters, search for the first effective path from the parking spot to the target U-turn point.

[0036] Step 40: Based on the dead-end road parameters, search for a second valid path from the target U-turn point to the vehicle's starting point.

[0037] In this embodiment, after determining the target U-turn point, a hierarchical path search approach is used to decompose the complex narrow passage + dead-end road combination scenario into two independent paths for processing, adapting the path search to the parameter requirements of different scenarios. Specifically, the target U-turn point serves as the connecting node, processing the narrow passage scenario from the target parking space to the target U-turn point and the dead-end road scenario from the target U-turn point to the current vehicle's location. Different parameters (such as different grid resolutions and search logic parameter configurations) are used for the two-layer path search process to balance path planning accuracy and search efficiency. This decouples the complex combined scenario into path planning for two scenarios, reducing overall computational complexity. This method differs from the traditional one-size-fits-all approach of using a single parameter to cover the entire scenario or using mixed parameters to balance accuracy and efficiency. It can specifically adapt to the differentiated needs of different stages in the combined scenario, improving the feasibility and efficiency of path planning. Furthermore, the first effective path from the parking space to the target U-turn point (corresponding to...) is searched first. Figure 3 The reverse search path of path S3), then search for the second valid path from the target U-turn point to the vehicle's own point (corresponding to Figure 3The search order of the reverse search path of the middle path S1 can also guide the path to converge to the open turning area, realize the gradient decomposition of scene complexity and accurate parameter matching. Compared with the traditional forward search method, it is more conducive to quickly finding feasible paths in complex scenes and reducing invalid attempts and waste of computing resources in the search process.

[0038] In this embodiment, it is necessary to understand that Figure 3 The example illustrates a driving path for a scenario involving a combination of a residential driveway and a dead-end road in actual driving, which is the path required after processing in step 50. The first and second effective paths in steps 30 and 40 are paths obtained after path search. During path search, the path from the starting point to the target point is considered to end near the target point (the criteria for determining proximity can be set based on the continuity requirements of subsequent processing, scenario complexity, sensor accuracy, and robustness requirements of path execution; this application does not specifically limit this). Therefore, the first and second effective paths are usually not directly connected and lack smoothness, requiring path connection and optimization processing in subsequent steps to determine a path such as... Figure 3 The path for turning around in the middle path S2.

[0039] In this embodiment, narrow channel parameters are specifically used as the core configuration for the first effective path search. Narrow channel parameters refer to parameters adapted to the high-precision path planning requirements of narrow channel scenarios, emphasizing path planning accuracy more than the dead-end road scenarios described later. For example, a narrow channel parameter configuration with a grid size of 0.1m and a step size of 0.2m can be used to search for the path from the parking spot to the target U-turn point, ensuring full utilization of the limited space within the narrow channel. These narrow channel parameters can be specifically set according to the characteristics of the narrow channel scenario, spatial constraints, and obstacle density, to adapt to the high-precision obstacle avoidance requirements of narrow channels; this application does not impose specific limitations on this. Based on these parameters, the path from the parking spot to the target U-turn point can be searched using the A* search algorithm or other heuristically guided path search algorithms, thereby generating a first effective path that meets the requirements of the narrow channel scenario. This path search process, by using narrow channel parameters as the core configuration, ensures that the resolution of the path from the parking spot to the target U-turn point is sufficiently fine to fully utilize the limited space within the narrow channel and avoid the poor path feasibility problem caused by coarse grids.

[0040] In this embodiment, during the specific search process, a monotonic queue (denoted as pq1) can be maintained to record the nodes to be expanded and their status information. During the path search process, the starting point is the parking spot, and the target U-turn point is the target point. The nodes are expanded according to the rules adapted to narrow passage scenarios. When a node reaches the vicinity of the target U-turn point, the first effective path between the parking spot and the target U-turn point can be obtained. These nodes can be added to another monotonic queue (denoted as pq2) maintained in the second effective path search process to provide initial nodes for the second layer search, ensuring the continuity of path connection and avoiding redundancy of re-search. When the number of nodes in pq2 meets the preset threshold (such as 10, which can be set according to the real-time requirements of path planning), it means that the first layer search has obtained enough connecting nodes. pq1 can be cleared, the first layer search can be stopped, and the second layer search stage can be entered.

[0041] In this embodiment, the search for the second effective path is configured based on dead-end road parameters. These parameters are designed to meet the high-efficiency path search requirements of dead-end road scenarios, and their grid size and step size are larger than those of narrow channel parameters. For example, a dead-end road parameter configuration with a grid size of 0.3m and a step size of 0.5m can be used to search for the path from the target U-turn point to the vehicle's point, thereby improving the search speed. These dead-end road parameters can be specifically set according to factors such as the characteristics of the dead-end road scenario, obstacle distribution, and search efficiency requirements. They can also be dynamically adjusted based on obstacle density, spatial openness, or real-time requirements to meet the need for efficient U-turns on dead-end roads. This application does not impose specific limitations on these parameters. Based on these parameters, the path from the target U-turn point to the vehicle's starting point can also be searched using the A* search algorithm or other heuristic-guided path search algorithms, thereby generating a second effective path that meets the requirements of the dead-end road scenario. This path search process uses the dead-end road parameter as the core configuration of the path search, which can ensure efficient search of the path from the target U-turn point to the vehicle's starting point, reduce unnecessary fine-grained calculations, and improve the real-time performance of the overall planning.

[0042] In this embodiment, it should be understood that since the vehicle is usually in a relatively controllable motion state when near the target point, kinematic constraints are easier to satisfy. Therefore, the two endpoints of the first and second valid paths do not need to correspond exactly to the precise coordinates of the starting point and the target point. As long as the searched node is near the target point, it can be determined as a valid path. For example, if the distance between the searched node and the target point is less than 3m and the orientation error between the node and the target point is less than 20°, the node can be determined to be near the target point, and then the valid path is determined according to the corresponding monotonic queue. In addition, the path search determines multiple nodes that meet the scene adaptation parameter constraints. The trajectories between nodes can be generated by interpolation, smoothing, etc., to form a continuous and executable path trajectory, which serves as the corresponding valid path.

[0043] In this embodiment, the first and second effective paths found through two sets of adaptive scenario parameters can fully leverage the advantages of hierarchical search, balancing path accuracy in narrow passage scenarios with search efficiency in dead-end road scenarios. This effectively avoids the limitations of traditional single parameters, reduces redundant calculations in path planning, and improves the robustness of path planning. In complex and space-constrained parking scenarios, it can provide vehicles with a segmented path foundation that meets both accuracy requirements and high efficiency. Subsequently, based on these two effective paths, path fusion processing is performed on the two segments to obtain a complete parking path, realizing parking path planning. This can improve the success rate and real-time performance of path planning, ensuring that vehicles can complete parking operations safely and smoothly.

[0044] Optionally, in some embodiments, searching for a first effective path from the parking spot to the target U-turn point based on narrow channel parameters includes: searching for a number of nodes with the lowest obstacle avoidance priority cost from the parking spot to the target U-turn point based on narrow channel parameters to obtain the first effective path; wherein, the obstacle avoidance priority cost includes a first actual cost and a first heuristic cost, the first actual cost includes the steering angle cost from the parking spot to the corresponding node and the narrow channel boundary distance penalty cost, and the first heuristic cost includes the motion curve length penalty cost from the corresponding node to the target U-turn point and the number of gear shifts cost.

[0045] In this optional embodiment, the path search process in narrow passage scenarios is constrained by an obstacle avoidance priority cost to ensure obstacle avoidance safety and space utilization within a limited space. The obstacle avoidance priority cost is a cost metric that balances path feasibility and execution efficiency, with the high-precision obstacle avoidance requirement in narrow passage scenarios as the core design objective. This ensures that the first effective path found has the highest obstacle avoidance priority, making full use of the limited space within the narrow passage and reducing the number of path recursions due to insufficient space.

[0046] In this optional embodiment, the obstacle avoidance priority cost specifically includes a first actual cost representing the actual driving cost from the starting point (parking spot) to the current node, and a first heuristic cost representing the predicted guidance cost from the current node to the target point (target U-turn point). The first actual cost, calculated based on the actual driving trajectory, can comprehensively represent the steering angle cost (encouraging smooth small-angle steering and penalizing large-angle steering) and the narrow passage boundary distance penalty cost (encouraging paths to approach the narrow passage boundary within a safe range and penalizing paths that are too far from the narrow passage boundary resulting in wasted space or too close to it posing a collision risk), resulting in smoother path steering, more efficient space utilization, and higher obstacle avoidance safety.

[0047] In this optional embodiment, the first heuristic cost, since it is calculated based on the predicted trajectory according to kinematic constraints, can comprehensively represent the motion curve length penalty cost (encouraging the path to converge to the target U-turn point along the narrow channel direction with the shortest trajectory that satisfies the vehicle's kinematic constraints, and penalizing long paths that deviate from the narrow channel direction or do not satisfy the kinematic constraints. This motion curve length penalty cost can specifically adopt the length cost of the Reeds-Shepp curve or Dubins curve that incorporates the narrow channel direction constraint. These curves are the shortest trajectories that satisfy the vehicle's kinematic constraints. Incorporating the direction constraint can prevent the path from deviating from the narrow channel direction and improve the convergence accuracy) and the shift number cost (encouraging smooth paths with fewer shifts and penalizing frequent shifts. The weight value can be 1.5 times the normal value to prioritize reducing the number of shifts in the narrow channel and improve the smoothness and efficiency of path execution) to enable the path to converge to the target U-turn point quickly and improve the execution smoothness. This application does not impose specific restrictions on the calculation method of the specific cost factors in the first actual cost and the first heuristic cost, as long as these cost factors can guide the path search to prioritize obstacle avoidance safety and adapt to the spatial constraints of narrow passage scenarios.

[0048] In this optional embodiment, the obstacle avoidance priority cost can be specifically calculated using an obstacle avoidance priority cost function. For example, the obstacle avoidance priority cost function is Cost1(n) = g1(n) + h1(n), where Cost1(n) is the obstacle avoidance priority cost function at the current search node n during the path search from the parking spot to the target U-turn point, g1(n) is the first actual cost, including the steering angle cost and the narrow passage boundary distance penalty cost, and h1(n) is the first heuristic cost, including the motion curve length penalty cost and the gear shifting number cost. By using the obstacle avoidance priority cost as the core evaluation index for path search, the search system can be guided to prioritize expanding nodes that are safe in obstacle avoidance and make full use of space. This ensures that the first effective path in the narrow passage not only avoids collisions with obstacles but also maximizes the use of limited space, laying the foundation for efficient search in subsequent dead-end road scenarios.

[0049] Optionally, in some embodiments, searching for a second effective path from the target U-turn point to the vehicle's starting point based on the dead-end road parameters includes: searching for a number of nodes with the lowest efficiency-priority cost from the target U-turn point to the vehicle's starting point based on the dead-end road parameters to obtain the second effective path; wherein, the efficiency-priority cost includes a second actual cost and a second heuristic cost, the second actual cost includes the path length cost from the target U-turn point to the corresponding node and the steering smoothness penalty cost, and the second heuristic cost includes the motion curve length penalty cost from the corresponding node to the vehicle's starting point and the number of gear shifts cost.

[0050] In this optional embodiment, the path search process in the dead-end road scenario is constrained by an efficiency-first cost to ensure both search efficiency and execution smoothness. The efficiency-first cost is designed with the core objective of generating efficient paths in dead-end road scenarios, balancing search speed and execution smoothness. This ensures that the second effective path found is prioritized for efficiency, quickly generating a U-turn trajectory and reducing unnecessary fine-grained calculations.

[0051] In this optional embodiment, the efficiency-priority cost specifically includes a second actual cost representing the actual driving cost from the starting point (target U-turn point) to the current node, and a second heuristic cost representing the predicted guidance cost from the current node to the target point (vehicle point). The second actual cost, calculated based on the actual driving trajectory, can comprehensively represent the path length cost (encouraging short paths and penalizing long paths) and the steering smoothness penalty cost (encouraging smooth small-angle steering and penalizing large-angle steering), resulting in shorter path lengths and smoother steering, adapting to the steering needs of open spaces on dead-end roads.

[0052] In this optional embodiment, the second heuristic cost, since it is calculated based on the predicted trajectory according to kinematic constraints, can comprehensively represent the motion curve length penalty cost (encouraging the path to converge to the vehicle point along the dead-end direction with the shortest trajectory that satisfies the vehicle's kinematic constraints, and penalizing long paths that deviate from the dead-end direction or do not satisfy the kinematic constraints. This motion curve length penalty cost can also adopt the length cost of the Reeds-Shepp curve or Dubins curve that incorporates the dead-end direction constraint, which will not be elaborated here) and the shift number cost (encouraging smooth paths with fewer shifts, penalizing operations with frequent shifts, with the weight set to a normal value to balance the smoothness of path execution and search efficiency), so that the path converges to the vehicle point quickly, improving search efficiency and execution smoothness. This application does not impose specific restrictions on the calculation method of the specific cost factors in the second actual cost and the second heuristic cost, as long as these cost factors can guide the path search to prioritize meeting efficiency requirements and adapt to the open space constraints of the dead-end scenario.

[0053] In this optional embodiment, the efficiency-first cost can be specifically calculated using an efficiency-first cost function. For example, the efficiency-first cost function is Cost2(m) = g2(m) + h2(m), where Cost2(m) is the efficiency-first cost function at the current search node m during the path search from the target U-turn point to the vehicle's starting point, g2(m) is the second actual cost, including path length cost and steering smoothness penalty cost, and h2(m) is the second heuristic cost, including motion curve length penalty cost and gear shift count cost. By using the efficiency-first cost as the core evaluation index for path search, the search system can be guided to prioritize expanding nodes with short paths and smooth steering, ensuring that the second effective path within the dead-end road guarantees both search speed and steering smoothness requirements, providing an efficient segmentation basis for subsequent path fusion.

[0054] Step 50: Merge the first and second valid paths to obtain the parking path.

[0055] In this embodiment, after obtaining the first and second effective paths, a high-precision obstacle avoidance path from the parking spot to the target U-turn point and an efficient convergence path from the target U-turn point to the vehicle's point are determined. The two paths can be fused (e.g., smoothing the connecting area and the overall trajectory, trajectory optimization, etc.) to eliminate discontinuities such as turns and abrupt changes that may occur at the connection points of the two paths, ensuring the continuity and smoothness of the parking path and improving the executability of the path.

[0056] In this embodiment, path fusion processing can employ various methods. For example, a polynomial interpolation-based method can be used to smooth the connection area between two path segments. By selecting an appropriate interpolation polynomial, the trajectory transition in the connection area can be made natural, reducing the bumps and swaying of the vehicle during driving. Alternatively, an optimization algorithm-based method can be used, with path continuity, smoothness, and safety as optimization objectives, to optimize the overall trajectory of the two path segments and obtain a parking path that better suits the vehicle's driving characteristics.

[0057] In this embodiment, the parking path obtained after path fusion processing possesses both high-precision obstacle avoidance capabilities in narrow passage scenarios and efficient convergence characteristics in dead-end road scenarios. It can provide a safe, smooth, and efficient parking path for vehicles in complex and space-constrained parking scenarios, effectively improving the success rate and real-time performance of path planning and ensuring that vehicles can successfully complete parking operations. Furthermore, the parking path planning method in this embodiment can be flexibly adjusted and optimized according to the needs of actual application scenarios, including narrow passage parameters, dead-end road parameters, obstacle avoidance priority cost function, and efficiency priority cost function, to adapt to parking requirements in different scenarios. This improves the scenario adaptability and scalability of the corresponding parking solution, which is conducive to promoting the application of intelligent driving vehicles in scenarios combining narrow passages and dead-end roads.

[0058] Optionally, in some embodiments, the first effective path and the second effective path are fused to obtain a parking path, including: smoothing the first effective path, the second effective path and the path connection segment using B-spline curves to obtain the parking path; wherein, the smoothing process constrains the continuity index of the path connection segment to adapt to U-turns and constrains the maximum curvature of the parking path to be no greater than the reciprocal of the vehicle's minimum turning radius.

[0059] This optional embodiment specifically describes a smoothing method using B-spline curves and the constraints on path continuity and kinematic constraints. A B-spline curve is a parametric curve that achieves smooth transitions in local trajectories through piecewise polynomial interpolation. It possesses the characteristics of continuous curvature and the fact that local adjustments do not affect the global trajectory. Smoothing with B-spline curves ensures the continuity of the first effective path, the second effective path, and the connecting segments during U-turns, avoiding vehicle instability caused by abrupt trajectory changes. Specifically, during the smoothing process, cubic B-spline curves can be used to smooth the connecting segments at U-turn points. First, K nodes (e.g., K=5) at the end of the first effective path and M nodes (e.g., M=5) at the beginning of the second effective path are extracted as input nodes for the connecting segment. Then, control points for the B-spline curve are generated for the connecting segment to ensure the continuity of the heading angle and curvature at the connection point. Finally, cubic B-spline fitting is performed on the overall trajectory of the first effective path, the second effective path, and the connecting segment, preserving the high-precision obstacle avoidance characteristics of narrow passages and the efficient convergence characteristics of dead-end sections. This method allows for a natural transition between the two paths, avoiding bumps and swaying during actual parking and improving the feasibility of the path.

[0060] In this optional embodiment, the smoothing process ensures the smoothness and safety of the path through continuity index constraints and kinematic constraints. The continuity index for U-turns ensures the continuity of the heading angle and curvature during U-turns, avoiding the risk of vehicle instability or loss of control due to sudden steering changes. Specifically, this continuity index is quantified by calculating the difference in heading angle and curvature between the trajectories before and after the transition section. When these differences are less than a preset threshold, the continuity requirement is considered met. Simultaneously, kinematic constraints limit the maximum curvature of the parking path to no more than the reciprocal of the vehicle's minimum turning radius (e.g., when the vehicle's minimum turning radius is 4m, the maximum curvature ≤ 0.25m). -1 This constraint ensures that the path is planned while meeting the vehicle's steering capabilities, avoiding situations where the vehicle cannot follow the path due to excessive curvature, thus further improving the safety and feasibility of the path. In practical applications, by combining B-spline curve smoothing with continuity indices and kinematic constraints, the first and second effective paths can be effectively merged into a complete parking path that conforms to the vehicle's driving characteristics. In complex scenarios combining narrow passages and dead ends, this path can provide stable, safe, and efficient parking guidance for the vehicle, retaining the high-precision obstacle avoidance capability in narrow passage scenarios and possessing the efficient convergence characteristics in dead-end scenarios. The resulting parking path has continuous curvature, meets vehicle kinematic constraints, has strong scenario adaptability, and high robustness, effectively improving the success rate and real-time performance of parking path planning.

[0061] To perform the corresponding steps in the above method embodiments and various possible implementations, an implementation method for a parking path planning device is provided below. Please refer to... Figure 4 , Figure 4 This is a schematic diagram of a parking path planning device 60 according to an embodiment of the present invention. In this embodiment, the parking path planning device 60 includes: The U-turn point preprocessing module 61 is used to identify candidate regions for U-turn points and select target U-turn points from the candidate regions for U-turn points; The hierarchical path search module 62 is used to search for a first effective path from the parking spot to the target U-turn point based on the narrow channel parameter, and to search for a second effective path from the target U-turn point to the vehicle point based on the dead-end road parameter. The path fusion module 63 is used to fuse the first valid path and the second valid path to obtain a parking path.

[0062] Optionally, in some embodiments, the parking path planning device further includes an environmental perception module and a scene modeling module; the environmental perception module includes a lidar and a surround-view camera; the scene modeling module includes a grid map construction unit and a dynamic resolution adjustment unit; the environmental perception module is used to output obstacle point cloud data and semantic segmentation results to provide raw perception information to the scene modeling module, and the scene modeling module is used to construct a grid map based on the obstacle point cloud data and semantic segmentation results.

[0063] Optionally, in some embodiments, the U-turn point preprocessing module 61 is further configured to: identify areas within a preset drivable range whose area is not less than the sum of the vehicle's minimum U-turn space area and the safety redundancy area, and whose distance from each obstacle is not less than the vehicle's maximum outline radius, as candidate U-turn point areas.

[0064] Optionally, in some embodiments, the U-turn point preprocessing module 61 is further configured to: filter valid U-turn points in the candidate region based on preset kinematic constraint indicators and preset accessibility indicators; score each valid U-turn point based on U-turn space sufficiency indicators and proximity to parking spaces indicators, and select the valid U-turn point with the highest score as the target U-turn point.

[0065] Optionally, in some embodiments, the hierarchical path search module 62 is further configured to: search for a number of nodes with the lowest obstacle avoidance priority cost from the parking spot to the target U-turn point based on the narrow channel parameters, and obtain a first effective path; wherein the obstacle avoidance priority cost includes a first actual cost and a first heuristic cost, the first actual cost includes the steering angle cost from the parking spot to the corresponding node and the narrow channel boundary distance penalty cost, and the first heuristic cost includes the motion curve length penalty cost from the corresponding node to the target U-turn point and the number of gear shifts cost.

[0066] Optionally, in some embodiments, the hierarchical path search module 62 is further configured to: search for a number of nodes with the lowest efficiency-priority cost from the target U-turn point to the vehicle's starting point based on the dead-end road parameters, and obtain a second effective path; wherein the efficiency-priority cost includes a second actual cost and a second heuristic cost, the second actual cost includes the path length cost from the target U-turn point to the corresponding node and the steering smoothness penalty cost, and the second heuristic cost includes the motion curve length penalty cost from the corresponding node to the vehicle's starting point and the number of gear shifts cost.

[0067] Optionally, in some embodiments, the path fusion module 63 is further configured to: use B-spline curves to smooth the first effective path, the second effective path, and the path connection segments to obtain a parking path; wherein the smoothing process constrains the continuity index of the path connection segments to adapt to U-turns and constrains the maximum curvature of the parking path to be no greater than the reciprocal of the vehicle's minimum turning radius.

[0068] In this embodiment, the parking path planning device can be a path planning controller of an intelligent driving vehicle, or one of the hardware units with path planning and environmental perception fusion capabilities, such as an on-board computing unit or an autonomous driving domain controller. In the process of autonomous parking path planning in scenarios involving narrow passages and dead-end roads using this parking path planning device, high-precision environmental data can be collected first using an environmental perception module (e.g., using a LiDAR scanner at a scanning frequency of 10Hz to output obstacle point clouds with a ranging accuracy of ±2cm, and using a surround-view camera to output semantic segmentation images at a resolution of 1920×1080, extracting key environmental information such as obstacles, narrow passage boundaries, and parking spaces). Then, a scene modeling module can be used to construct a grid map adapted to the scene characteristics based on the perception data (e.g., using a fine grid resolution of 5cm×5cm for narrow passage areas with a width ≤2.5m to ensure high-precision obstacle avoidance, and using a dynamic grid resolution of 10cm×10cm for dead-end road areas with an area ≥6m² to balance efficiency and accuracy), providing a high-precision environmental spatial model for subsequent modules. Finally, a U-turn point preprocessing module can identify and filter target U-turn points (e.g., using clustering algorithms to extract dead-end road boundaries from the obstacle point cloud, and using dilation operations to...). The candidate area for U-turn points is defined by using half the maximum width of the vehicle plus a safety distance of 0.3m as the radius. Then, the kinematic constraints and accessibility of points within the candidate area are quickly verified by the RRT* algorithm to filter out effective U-turn points and select the target U-turn point with the highest score. Then, the hierarchical path search module performs segmented path search. The first layer searches for the first effective path from the parking point to the target U-turn point based on narrow channel parameters (such as fine grid resolution, obstacle avoidance priority cost function, and shift cost weight of 1.5 times the normal value). The second layer searches for the second effective path from the target U-turn point to the vehicle point based on dead-end road parameters (such as dynamic grid resolution, efficiency priority cost function, and narrow channel boundary distance penalty coefficient of 2 times the normal value). Finally, the path fusion module performs smooth optimization on the two paths. For example, cubic B-spline curves are used to process the path connection segment to ensure that the path curvature is continuous and the maximum curvature does not exceed the reciprocal of the vehicle's minimum turning radius, and the complete parking path that satisfies the vehicle's kinematic constraints is output. The parameter configurations of each module and unit in the above examples can be dynamically adjusted according to the characteristics of the scenario and the real-time requirements. This application does not impose specific restrictions on these parameter configurations, as long as they meet the functional requirements of scenario adaptability and path planning.

[0069] In this embodiment, since the embodiment of the device 60 corresponds to the embodiment of the method described above, the description of the parking path planning device 60 provided by the present invention can be referred to the above method embodiment. The present invention embodiment will not be described again here, and it has the same beneficial effects as the above parking path planning method.

[0070] See Figure 5 , Figure 5This is a schematic diagram of an embodiment of the storage medium provided in this application.

[0071] The storage medium 70 stores program data 71, which, when executed by the processor, implements, as follows: Figures 1 to 3 The parking path planning method described.

[0072] The program data 71 is stored in a storage medium 70 and includes several instructions for causing a network device (such as a router, personal computer, or server) or processor to execute all or part of the steps of the methods described in the various embodiments of this application.

[0073] Optionally, the storage medium 70 can be any medium capable of storing program data 71, such as a USB flash drive, portable hard drive, read-only memory (ROM), random access memory (RAM), disk, or optical disc.

[0074] See Figure 6 , Figure 6 This is a schematic diagram of the structure of an embodiment of the vehicle-mounted device provided in this application.

[0075] The on-board device 80 includes a processor 82 and a memory 81 connected to each other. The memory 81 stores a computer program, and when the processor 82 executes the computer program, it implements, for example, Figures 1 to 3 The parking path planning method described herein. The memory 81 may include the storage medium 70, or may be other separately developed memory.

[0076] See Figure 7 , Figure 7 This is a structural schematic diagram of an embodiment of the vehicle provided in this application.

[0077] The vehicle 90 includes, for example Figure 4 The parking path planning device 60 described herein, such as Figure 5 The described storage medium 70 and such Figure 6 At least one of the described vehicle-mounted devices 80, Figure 7 The example provided is a vehicle including the on-board equipment 80. The vehicle 90 can be a gasoline vehicle, an electric vehicle, or a hybrid vehicle, and can be a small car, a compact car, a mid-to-large car, or a large car. This application does not impose specific restrictions on the vehicle's power type, vehicle size, or specific model, as long as the vehicle 90 can autonomously complete parking path planning and perform parking operations in a combined scenario of narrow passages and dead-end roads using the parking path planning device 60, storage medium 70, and on-board equipment 80 as described above.

[0078] Unlike existing technologies, this application discloses a parking path planning method, device, storage medium, in-vehicle equipment, and vehicle. By identifying candidate U-turn point areas and filtering target U-turn points, clear guidance is provided for path search, preventing vehicles from falling into local optima traps due to heuristic irrationality in scenarios involving narrow passages and dead-end roads. Through a hierarchical path search strategy, the complex scenario of narrow passages and dead-end roads is decomposed into two independent paths: "from the parking spot to the target U-turn point" and "from the target U-turn point to the vehicle's starting point." These paths can be searched using search methods adapted to narrow passages and dead-end roads, respectively, balancing path planning accuracy and search speed, and ensuring a balance between feasibility and efficiency in different scenarios. By fusing the paths found in the two layers, the path curvature is ensured to be continuous and satisfy vehicle kinematic constraints, improving path executability and eliminating abrupt changes at path junctions to ensure the smoothness of the overall path. This application effectively solves the problems of poor path feasibility, low search efficiency, and easy getting trapped in local optima in traditional path planning methods in scenarios with narrow passages and dead ends. It achieves high-precision and high-efficiency path planning in complex and restricted scenarios, which can significantly improve the success rate, search efficiency and robustness of path planning. At the same time, it has strong scenario adaptability and scalability, which is conducive to promoting the application of intelligent driving vehicles in scenarios with narrow passages and dead ends.

[0079] In practical tests, this application demonstrates a 200% to 600% improvement in search efficiency compared to the traditional A* algorithm, significantly reducing path planning response time and minimizing resource consumption by the onboard computing unit. Simultaneously, the path planning success rate is increased to over 95%, far exceeding the typically less than 60% success rate of the traditional A* algorithm. Furthermore, in practical scenarios based on this hierarchical search method, such as a combination of a 2.8m wide dead-end road and a 5.5m wide, 15m long narrow passage, no collision risks or operational smoothness degradation due to abrupt path transitions were observed, and the total planning time was below 200ms. In contrast, the traditional A* algorithm typically takes over 800ms on average in the same scenario and frequently experiences search timeouts. Therefore, this application's solution effectively balances the high-precision obstacle avoidance requirements of narrow passage scenarios with the efficient convergence requirements of dead-end road scenarios, effectively addressing the core pain points of traditional algorithms in scenarios combining narrow passages and dead-end roads. It significantly improves the real-time performance and robustness of path planning and possesses high adaptability and executability in complex and constrained scenarios.

[0080] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on describing the differences from other embodiments. In particular, the device embodiments, storage medium embodiments, vehicle-mounted device embodiments, and vehicle embodiments are basically similar to the method embodiments, so the descriptions are relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0081] The above description is merely an embodiment of this application and does not limit the patent scope of this application. Any equivalent structural or procedural transformations made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

Claims

1. A parking path planning method, characterized in that, include: Identify candidate regions for U-turn points; Select a target U-turn point from the candidate U-turn point region; Based on the narrow channel parameters, search for the first effective path from the parking spot to the target U-turn point; Based on the dead-end road parameters, search for a second valid path from the target U-turn point to the vehicle's starting point; The first valid path and the second valid path are merged to obtain the parking path.

2. The parking path planning method according to claim 1, characterized in that, The candidate region for identifying U-turn points includes: The area within the preset drivable range whose area is not less than the sum of the vehicle's minimum U-turn space area and the safety redundancy area, and whose distance from each obstacle is not less than the vehicle's maximum outline radius, is identified as the candidate area for the U-turn point.

3. The parking path planning method according to claim 1, characterized in that, Selecting a target U-turn point from the candidate U-turn point region includes: Based on preset kinematic constraint indicators and preset reachability indicators, valid U-turn points in the candidate region of U-turn points are selected. Each valid U-turn point is scored based on the U-turn space sufficiency index and the proximity to parking spaces index, and the valid U-turn point with the highest score is selected as the target U-turn point.

4. The parking path planning method according to claim 1, characterized in that, The step of searching for the first effective path from the parking spot to the target U-turn point based on narrow channel parameters includes: Based on the narrow channel parameters, search for several nodes with the lowest obstacle avoidance priority cost from the parking spot to the target U-turn point to obtain the first effective path; The obstacle avoidance priority cost includes a first actual cost and a first heuristic cost. The first actual cost includes the steering angle cost from the parking spot to the corresponding node and the narrow passage boundary distance penalty cost. The first heuristic cost includes the motion curve length penalty cost from the corresponding node to the target U-turn point and the number of gear shifts cost.

5. The parking path planning method according to claim 4, characterized in that, The step of searching for a second valid path from the target U-turn point to the vehicle's starting point based on dead-end road parameters includes: Based on the dead-end road parameters, search for several nodes with the lowest efficiency priority and cost from the target U-turn point to the vehicle point to obtain the second effective path; The efficiency-priority cost includes a second actual cost and a second heuristic cost. The second actual cost includes the path length cost from the target U-turn point to the corresponding node and the steering smoothness penalty cost. The second heuristic cost includes the motion curve length penalty cost from the corresponding node to the vehicle point and the number of gear shifts cost.

6. The parking path planning method according to claim 1, characterized in that, The process of fusing the first valid path and the second valid path to obtain the parking path includes: The parking path is obtained by smoothing the first effective path, the second effective path, and the path connection segments using B-spline curves. The smoothing process constrains the continuity index of the path connection segment for adapting to U-turns and steering, and constrains the maximum curvature of the parking path to be no greater than the reciprocal of the vehicle's minimum turning radius.

7. A parking path planning device, characterized in that, include: The U-turn point preprocessing module is used to identify candidate regions for U-turn points and select target U-turn points from the candidate regions. The hierarchical path search module is used to search for a first effective path from the parking spot to the target U-turn point based on the narrow channel parameter, and to search for a second effective path from the target U-turn point to the vehicle point based on the dead-end road parameter. The path fusion module is used to fuse the first valid path and the second valid path to obtain the parking path.

8. A storage medium storing program data thereon, characterized in that, When the program data is executed by the processor, it implements the steps of the parking path planning method as described in any one of claims 1 to 6.

9. A vehicle-mounted device, characterized in that, It includes an interconnected processor and a memory, the memory storing a computer program, and the processor executing the computer program to implement the steps of the parking path planning method as described in any one of claims 1 to 6.

10. A vehicle, characterized in that, The vehicle includes at least one of the parking path planning device as claimed in claim 7, the storage medium as claimed in claim 8, and the on-board equipment as claimed in claim 9.

Images