Parking control method of automatic driving articulated vehicle and automatic driving vehicle

Abstract

The invention provides a parking control method of an automatic driving articulated vehicle and the automatic driving vehicle, and the method comprises the steps: constructing a grid map of a to-be-parked area, constructing a heuristic map based on the grid map, the heuristic map being a map data structure for providing path search guide information for each grid unit in the grid map, and the grid units being provided with path search guide information for the to-be-parked area in the to-be-parked area in the to-be-parked area in the to-be-parked area. Comprising a heuristic flow field structure used for calculating a parking end point, a heuristic distance field structure used for calculating obstacle avoidance and a gradient field structure used for calculating an obstacle avoidance direction. And performing parking path search based on the heuristic map and a vehicle model of the articulated vehicle to control the articulated vehicle to perform automatic parking according to the parking path. According to the method and the device, the technical problem that the real-time requirement of an actual parking task is difficult to meet due to low path searching efficiency during parking of an automatic driving articulated vehicle in the related technology is solved, and the technical effect of efficient parking is further achieved.

Description

Technical Field

[0001] This application relates to the field of autonomous driving, and in particular to parking control methods for autonomous articulated vehicles and autonomous vehicles. Background Technology

[0002] With the development of autonomous driving technology, the demand for automated parking of articulated vehicles (such as semi-trailer trucks) in scenarios such as ports, logistics parks, and long-haul freight is gradually increasing. During parking, autonomous semi-trailer trucks need to complete precise path planning and vehicle control within a limited working area to ensure safe and smooth entry into designated parking spaces despite complex obstacle layouts. Compared to passenger cars, semi-trailer trucks face more complex kinematic constraints during low-speed parking, such as greater vehicle length, an articulated structure between the tractor and trailer, and opposite movement directions and poor dynamic stability during reversing.

[0003] Furthermore, in unstructured environments such as port areas and logistics parks, parking areas typically lack clear lane line constraints and are open, continuous spaces. Vehicles are allowed to move within any feasible range of this space, and path searching must be performed in a multi-dimensional state space, considering factors such as position, attitude, articulation angle, and gear. Because semi-trailer vehicles have numerous state variables with continuous value ranges, high-dimensional state space searches are prone to dimensionality explosion, such as an exponential increase in the number of search nodes. This significantly increases the time required for path searching, making it difficult to meet the real-time requirements of actual parking tasks.

[0004] There is no effective solution yet to address the technical problem of low path search efficiency in related technologies. Summary of the Invention

[0005] This application provides a parking control method for an autonomous articulated vehicle and an autonomous vehicle to solve one or more of the aforementioned technical problems.

[0006] In a first aspect, embodiments of this application provide a parking control method for an autonomous articulated vehicle, comprising: constructing a grid map of a parking area; constructing a heuristic map based on the grid map, wherein the heuristic map is a map data structure that provides path search guidance information to each grid cell in the grid map, including a heuristic flow field structure for calculating the parking endpoint, a heuristic distance field structure for calculating obstacle avoidance, and a gradient field structure for calculating the obstacle avoidance direction; and performing a parking path search based on the heuristic map and a vehicle model of the articulated vehicle to control the articulated vehicle to automatically park according to the parking path.

[0007] Secondly, embodiments of this application provide an autonomous driving vehicle, including: a controller, the controller including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the method as described in any of the preceding claims.

[0008] Thirdly, embodiments of this application provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the method described in any of the above-mentioned embodiments.

[0009] Fourthly, embodiments of this application provide a computer program product, including a computer program / instructions, which, when executed by a processor, implement the steps of any of the methods described above.

[0010] In the automatic parking process of an autonomous articulated vehicle, this application first constructs a grid map of the parking area, and then builds a heuristic map based on the grid map. This heuristic map is a map data structure that provides path search guidance information to each grid cell in the grid map, including a heuristic flow field structure for calculating the parking endpoint, a heuristic distance field structure for calculating obstacle avoidance, and a gradient field structure for calculating the obstacle avoidance direction. Finally, based on the heuristic map and the vehicle model of the articulated vehicle, a parking path search is performed to control the articulated vehicle to automatically park according to the parking path. In other words, at least three layers of structure (flow field structure, distance field structure, and gradient field structure) are constructed on top of the grid map to obtain the heuristic map. Guided search based on this heuristic map can significantly accelerate and optimize path search, solving the technical problem in related technologies where low path search efficiency in autonomous articulated vehicle parking makes it difficult to meet the real-time requirements of actual parking tasks, thus achieving the technical effect of efficient parking.

[0011] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application, it can be implemented according to the contents of the specification. In order to make the above and other objects, features and advantages of this application more obvious and understandable, specific embodiments of this application are given below. Attached Figure Description

[0012] In the accompanying drawings, unless otherwise specified, the same reference numerals throughout the various drawings denote the same or similar parts or elements. These drawings are not necessarily drawn to scale. It should be understood that these drawings depict only some embodiments according to this application and should not be construed as limiting the scope of this application.

[0013] Figure 1A flowchart of a parking control method for an autonomous articulated vehicle provided in an embodiment of this application is shown;

[0014] Figure 2 This illustration shows a flow field structure diagram in a heuristic map construction method provided in an embodiment of this application.

[0015] Figure 3a This illustration shows a schematic diagram of an original raster map provided in an embodiment of this application;

[0016] Figure 3b This illustration shows a distance field structure in a heuristic map construction method provided in an embodiment of this application.

[0017] Figure 4 This illustration shows a schematic diagram of an N-circle coverage model provided in an embodiment of this application;

[0018] Figure 5 A flowchart of a path planning algorithm based on hybrid A* provided in an embodiment of this application is shown;

[0019] Figure 6a This illustration shows a multi-resolution structure diagram in a path search provided in an embodiment of this application;

[0020] Figure 6b This illustration shows a schematic diagram of a multi-resolution structure in another path search provided in an embodiment of this application;

[0021] Figure 7 This illustration shows a schematic diagram of the outline sampling of an obstacle avoidance vehicle provided in an embodiment of this application;

[0022] Figure 8a This illustration shows a hard constraint obstacle avoidance method provided in an embodiment of this application.

[0023] Figure 8b This illustration shows a schematic diagram of an obstacle avoidance method combining hard constraints and soft constraints provided in an embodiment of this application.

[0024] Figure 9a This illustration shows a search path without shifting and trajectory optimization without shifting provided in an embodiment of this application;

[0025] Figure 9b This illustration shows a search path without shifting gears, with trajectory optimization autonomously adding one shifting gears, provided in an embodiment of this application.

[0026] Figure 10a This illustration shows a schematic diagram of trajectory end smoothing processing provided in an embodiment of this application;

[0027] Figure 10b This application provides a corresponding trajectory dynamic information curve in an embodiment.

[0028] Figure 10c This application provides another corresponding trajectory dynamic information curve in an embodiment.

[0029] Figure 11 This paper shows a structural block diagram of a parking control device for an autonomous vehicle provided in an embodiment of this application;

[0030] Figure 12 This paper shows a block diagram of a parking control system for an autonomous vehicle provided in an embodiment of this application.

[0031] Figure 13 A block diagram of a controller used to implement embodiments of this application is shown;

[0032] Figure 14 A schematic diagram of the structure of an autonomous articulated vehicle provided in an embodiment of this application is shown. Detailed Implementation

[0033] In the following description, only certain exemplary embodiments are briefly described. As those skilled in the art will recognize, the described embodiments can be modified in various ways without departing from the concept or scope of this application. Therefore, the drawings and description are considered to be exemplary in nature and not restrictive.

[0034] To facilitate understanding of the technical solutions of the embodiments of this application, the relevant technologies of the embodiments of this application are described below. The following relevant technologies are optional solutions and can be combined with the technical solutions of the embodiments of this application in any way, and all of them fall within the protection scope of the embodiments of this application.

[0035] The parking navigation task for autonomous articulated vehicles (such as semi-trailer trucks) can be formalized as follows: in an open, continuous space, transforming perception data into obstacle information in a reasonable manner, and planning steering and gear shifting actions based on obstacle priors, while satisfying hard constraints such as obstacle avoidance and semi-trailer truck kinematics. However, related technologies generally suffer from low parking navigation efficiency and low accuracy.

[0036] In view of the above-mentioned technical problems, the technical solution of this application and how the technical solution of this application solves the aforementioned technical problems will be described in detail below with specific embodiments. The listed specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0037] like Figure 1The diagram shown is a flowchart of a parking control method for an autonomous articulated vehicle according to an embodiment of this application. The method can be executed by an electronic device (or controller). The electronic device can be a central control unit with data processing function on the vehicle, a background server or service cluster associated with the vehicle, or a cloud server that provides vehicle data processing. This embodiment does not specifically limit the method.

[0038] The parking control method for autonomous articulated vehicles provided in this application includes, but is not limited to, parking and docking tasks in open, continuous spaces that do not rely on regular road networks or closed grid maps. This includes autonomous parking navigation for articulated vehicles in environments such as large logistics parks, port yards, airport aprons, mining areas, and construction sites. In these scenarios, vehicles can autonomously plan feasible parking paths on a continuous plane within vast or semi-structured areas without fixed parking space markings, based on real-time perception information and kinematic constraints, and achieve precise parking or docking at designated locations.

[0039] like Figure 1 As shown, the parking control method for an automated articulated vehicle provided in this embodiment of the invention may include the following steps:

[0040] Step S102: Construct a grid map of the parking area.

[0041] Optionally, in this embodiment, a grid map can be constructed based on dynamic obstacle information and static environment information collected by sensors installed on the articulated vehicle. This allows for a unified representation and real-time updating of the spatial occupancy of open or semi-structured environments, providing an efficient basic data structure for parking navigation path planning and collision detection. For example, the dynamic obstacle information can be the position and speed information of moving vehicles, pedestrians, and transfer equipment, while the static environment information can be fixed objects such as walls, storage yard materials, loading and unloading platforms, and road boundaries.

[0042] Step S104: Construct a heuristic map based on the grid map. The heuristic map is a map data structure that provides path search guidance information to each grid cell in the grid map. It includes a heuristic flow field structure for calculating the parking endpoint, a heuristic distance field structure for calculating obstacle avoidance, and a gradient field structure for calculating the obstacle avoidance direction.

[0043] It should be noted that, in the embodiments of this application, the heuristic map mentioned above includes, but is not limited to, flow field structures, distance field structures, and gradient field structures, and may also include other structural types that can reflect the spatial relationship between location and target. In the parking task, the structural form adopted can be configured according to the required computational efficiency, path accuracy, and suitable environment type, and this application does not limit it in this regard.

[0044] S106, Based on the heuristic map and the vehicle model of the articulated vehicle, a parking path is searched to control the articulated vehicle to automatically park according to the parking path.

[0045] It is understood that the vehicle model in this application embodiment can be used to describe the motion capabilities and space occupancy of an autonomous articulated vehicle during parking. In one embodiment, the vehicle model may include: a vehicle kinematics model supporting gear shifting, used to simulate the vehicle's motion constraints, steering capabilities, and dynamic characteristics in forward, reverse, and gear-engaged states; and a geometric shape (e.g., circular) overlay model for collision detection, which approximates the vehicle's outline by covering several geometric shapes, quickly calculating the distance to obstacles when the vehicle's attitude changes, thereby achieving efficient collision judgment. In other embodiments, the composition of the vehicle model is not limited to this, and may also include vehicle dynamics models for different types of scenarios, polygonal overlay models for fine geometric detection, or dynamic shape models incorporating vehicle sensor information, etc.

[0046] Through steps S102-S106 above, in the automatic parking process of an autonomous articulated vehicle, a grid map of the parking area is first constructed. Then, a heuristic map is constructed based on this grid map. This heuristic map is a map data structure that provides path search guidance information to each grid cell in the grid map. It includes a heuristic flow field structure for calculating the parking endpoint, a heuristic distance field structure for calculating obstacle avoidance, and a gradient field structure for calculating the obstacle avoidance direction. Finally, a parking path search is performed based on this heuristic map and the vehicle model of the articulated vehicle to control the articulated vehicle to automatically park according to the parking path. In other words, at least three layers of structure (flow field structure, distance field structure, and gradient field structure) are constructed on top of the grid map to obtain the heuristic map. The guided search based on this heuristic map can significantly accelerate and optimize path search, solving the technical problem in related technologies where low path search efficiency in autonomous articulated vehicle parking makes it difficult to meet the real-time requirements of actual parking tasks, thus achieving the technical effect of efficient parking.

[0047] In one possible implementation, step S102 may include: S11, acquiring static environmental information and dynamic obstacle information of the parking area collected by the sensor; S12, after constructing a grid coordinate system based on the parking point and parking space, performing layered projection on the static environmental information and the dynamic obstacle information to obtain the grid map.

[0048] Optionally, embodiments of this application also provide an interface for downstream applications to perform custom resolution raster reconstruction to adapt to the needs of different planning modules.

[0049] Step S12 above may include: S121, for the static environment information, extracting the location and semantic information and projecting it into the corresponding grid cell; S122, for the dynamic obstacle information, extracting the obstacle outline corresponding to the current moment, and using the line segment rasterization method of integer iteration operation to convert the continuous geometric line segments into all grid cells covered by the line segments in the raster map, and using a recursive flooding fill algorithm to fill the dynamic obstacle polygon.

[0050] Optionally, in the embodiments of this application, the line segment rasterization method of the above-mentioned integer iteration operation can be the line segment rasterization method of the Bresenham line algorithm.

[0051] The grid map constructed through the above steps S11~S12 and S121~S122 provides efficient and reliable data support for real-time path planning and feasibility detection of autonomous vehicles, and is especially suitable for vehicle systems with limited computing resources.

[0052] The following section will elaborate on the construction of flow field structure, distance field structure, and gradient field structure.

[0053] For constructing the flow field structure, this application proposes the following: S21, using the parking endpoint as the cost source point; S22, using a breadth-first search algorithm to update the global cost layer by layer for all grid cells in the parking area, so as to allocate the minimum cost value to the parking endpoint for each grid cell.

[0054] This can be understood as using the target point of the navigation task (i.e., the parking endpoint) as the source point, and performing a global cost update based on a breadth-first search algorithm. Specifically, this can be achieved by repeatedly retrieving the cell with the lowest current cost from an open list (which is essentially a priority queue; a double-ended queue data structure can be used to pop candidate cells in ascending order of cost). traverse all of its non-obstacle effective neighbors Calculate from arrive New costs For specific calculation methods, please refer to Formula 1 below:

[0055] (Formula 1)

[0056] If the cost is lower Current cost Then update The cost is then added to the open table. For example... Figure 2 The diagram shows the flow field structure in heuristic map construction, where the position with a value of 0 is the target point of the navigation task.

[0057] For constructing the distance field structure, this application embodiment proposes: S31, based on the principle of parabolic lower envelope, and using... The complexity is to calculate the distance field in N-dimensional space, where the vertex of the parabola is related to the obstacle, and N is greater than 1.

[0058] Specifically, let For a series of obstacles If the parabola has vertex A, then The projection of the lower envelope formed by the series of parabolas constitutes the range field. The calculation of the lower envelope is related to the intersection points of the parabolas. Solving the equations of the two parabolas simultaneously, as shown in Formula 2:

[0059] (Formula 2)

[0060] We can obtain the following formula 3:

[0061] (Formula 3)

[0062] That is, any two parabolas have one and only one point of intersection. .set up Let it be the set of parabolas that actually form the lower envelope. Indicates the first The vertex of the parabola, Indicates the first Article and Section The intersection of the parabolas, the interval Indicates the first The lower envelope of a parabola. During the process of traversing and solving for the lower envelope, for a new parabola... , and The latest parabola in China intersection There are two possibilities:

[0063] like Then Add to And update , ;

[0064] like Then the first The parabola does not participate in forming the lower envelope; it should be considered from... Delete and recalculate With the new first The intersection of the parabolas until exist Right side.

[0065] like Figure 3a The image shown is the original raster map. Figure 3bThe diagram shows the distance field structure in heuristic map building, where the position with a value of 0 represents an obstacle.

[0066] To construct the gradient field structure, this application embodiment proposes: S41, using bilinear interpolation to obtain the obstacle avoidance gradient of the corresponding grid cell from the distance field.

[0067] Specifically, let The gradient represents the distance between a location and an obstacle, and is expressed as follows: Formula 4:

[0068] (Formula 4)

[0069] In one possible implementation, the above method further includes: S51, constructing a vehicle model of an articulated vehicle, wherein the vehicle model of the articulated vehicle includes a vehicle kinematic model supporting gear shifting and a multi-circle coverage model for collision detection.

[0070] Optionally, the above-mentioned vehicle kinematic model supporting gear shifting can adopt the following state equations:

[0071] (Formula 5)

[0072] in, This is the center position of the rear axle of the tractor. The heading angle of the tractor unit. This is the trailer heading angle. For the direction of motion, It is the distance of a single movement. It is the wheelbase of the tractor unit. That's the trailer wheelbase. The operating space is determined by the gear shift. and front wheel cornering composition.

[0073] like Figure 4 The multi-circle (N-circle) covering model shown can be calculated using the following method. Taking a tractor as an example, let the coordinates of the rear axle center be... According to geometric relationships, the coordinates of the centers of each covering circle are as shown in Formula 6:

[0074] (Formula 6)

[0075] in, It is the heading angle. It represents the number of covered circles.

[0076] In one possible implementation, step S106 may include: S61, selecting the first target grid cell with the lowest cost from the first set, and obtaining the neighborhood grid cell set of the first target cell based on the vehicle kinematics model; S62, for any second target grid cell in the neighborhood grid cell set, determining whether the second target grid cell is in the second set or determining whether a collision will occur when the articulated vehicle is located at the second target grid cell based on the multi-circle coverage model, wherein the grid cells in the second set are the grid cells visited during the path search process; S63, if the determination result is negative, calculating the cost of the second target grid cell based on the first target grid cell, adding the second target grid cell to the first set, and setting other grid cells in the neighborhood grid cell set as the second target grid cell to continue executing the determination and cost calculation steps until there are no expandable grid cells.

[0077] It is understandable that the parking path described above ultimately outputs a kinematic obstacle avoidance path, where each discrete waypoint contains the following state information:

[0078]

[0079] in, This is the center position of the rear axle of the tractor. The heading angle of the tractor unit. This is the trailer heading angle. This is the gear information (1 indicates forward gear, 0 indicates reverse gear). This refers to the turning angle of the front wheels of the tractor.

[0080] The following section uses a grid cell as the node, the first set as the open node table, and the second set as the closed node table as an example to describe the specific implementation process of step S106 above. Figure 5 As shown, it includes:

[0081] Step S501: Initialize the open node table and the closed node table;

[0082] Step S502: Determine whether the open node table is empty. If the result is yes, proceed to step S512. If the result is no, proceed to step S503.

[0083] Step S503: Remove the node n with the lowest cost from the open node list;

[0084] Step S504: Determine whether node n is the endpoint. If the result is yes, proceed to step S505. If the result is no, proceed to step S507.

[0085] Step S505: Backtrack from the endpoint to the starting point to form a path;

[0086] Step S506, search successful;

[0087] Step S507: Generate the set C of neighboring nodes of node n;

[0088] Step S508: Determine whether the neighborhood node set C is empty. If the result is yes, execute S502. If the result is no, execute step S509.

[0089] Step S509: Remove node c from the neighborhood node set C;

[0090] Step S510: Determine whether node c is in the closed node list or whether a collision has occurred. If the determination result is yes, proceed to step S508; if the determination result is no, proceed to step S511.

[0091] S511, calculate the cost of node c based on node n and add it to the open node table;

[0092] S512, search failed.

[0093] In the path search process described above, embodiments of this application also propose a discretization and caching structure to accelerate computational efficiency, a multi-resolution structure to balance refined obstacle avoidance and search efficiency, and a cost design to guide the robot to efficiently reach the target state. These will be described in detail below.

[0094] In one possible implementation, the method further includes: S71, determining whether a collision will occur when the articulated vehicle is located in the second target grid cell based on the grid map.

[0095] The process of obtaining the set of neighboring grid cells of the first target grid cell based on the vehicle kinematics model includes: S81, downsampling the grid map at a preset ratio to obtain a search map; S82, determining the set of neighboring grid cells based on the search map.

[0096] Understandably, in step S71 above, a high-resolution raster map built from perceived information is directly used for collision detection to achieve refined obstacle avoidance. In step S82 above, the high-resolution raster map is downsampled at a certain ratio to obtain a search map, which is used to perform the node expansion process of path planning, thereby accelerating the search process and avoiding interference from redundant information. Figure 6a The diagram shown illustrates a search resolution of 1.0m / pixel. Figure 6b The diagram shown is a collision detection resolution of 0.2m / pixel.

[0097] Optionally, in this embodiment of the application, before determining whether a collision will occur when the articulated vehicle is located in the second target grid cell, the above method further includes: S91, caching a collision table, wherein the collision table is a mapping relationship between the combined information of the tractor's heading angle and the trailer's heading angle and the collision information.

[0098] It is understandable that in step S91 above, the discretized tractor heading angles can be combined pairwise. Table and trailer heading angle The table, with the rear axle center of the tractor unit at the origin, iterates through each discrete heading angle pair. This can uniquely determine the information of the vehicle and N covering circles. (The last part is incomplete and likely refers to a different context.) A collision table cache is established with the key as the key and the collision information as the value. During the collision detection stage of node expansion, only the table needs to be looked up, without recalculating the covering circle model.

[0099] Before performing parking path search based on the heuristic map and the vehicle model of the articulated vehicle, the above method further includes: S101, discretizing the state space and action space of the articulated vehicle, wherein the state space includes at least one of the following: the center position of the rear axle of the tractor. The tractor's heading angle and the trailer's heading angle, the range of motion including at least one of the following: gear position Front wheel steering angle.

[0100] It is understandable that the aforementioned discretization to accelerate computational efficiency includes: state space discretization and action space discretization. Specifically, state space discretization involves dividing the state of the semi-trailer truck on the map according to... Discretize in four dimensions, including the center position of the rear axle of the tractor. Discretize into a two-dimensional discrete structure according to the search resolution; tractor heading angle and trailer heading angle Divide the 360° angle into several equal parts with a certain resolution to form a one-dimensional discrete structure; the motion space is discretized into gears within the motion space. Set as ,in Indicates moving forward. Indicates reversing; front wheel steering angle In the interval The state equation is discretized into several parts at equal intervals with a certain resolution. All parts of the state equation involving action space operations are cached during the discretization stage, so that the table can be looked up when the node is expanded.

[0101] To further accelerate the search process, the above method also includes: S111, mapping the discretized state space to the corresponding target index according to a preset encoding method; S112, storing the visited grid cells in the second set in the manner corresponding to the target index.

[0102] It is understood that the aforementioned second set is a closed-node table, and in the embodiments of this application, the discretized set can be represented according to certain encoding principles. State mapping to a unique index Information about visited nodes is stored in the closed node table as an index. When a node is expanded, the current state is checked to see if it is in the closed node table to avoid repeated visits.

[0103] To guide autonomous vehicles to reach the target state efficiently, embodiments of this application propose using a target cost model to calculate the aforementioned cost. The target cost model is determined by the existing cost of the grid cell, the path length cost from the parent grid cell to the current grid cell, the turning cost, the reversing cost, the gear shifting cost, and the heuristic cost obtained through the flow field structure in the heuristic map.

[0104] Optionally, the above target cost model can be as shown in Formula 7:

[0105] (Formula 7)

[0106] in, The node already has a cost, It is the path length cost from the parent node to the current node. It is the cost of turning around. It's the cost of reversing. It's the cost of shifting gears. Heuristic coefficients It is the heuristic cost obtained directly from the flow field structure in the heuristic map.

[0107] After obtaining the parking path (which can also be understood as the initial solution), this application also proposes a path optimization strategy based on spatiotemporal joint optimization modeling, which will be described in detail below.

[0108] Optionally, in this embodiment of the application, the above method further includes: S121, constructing a target cost, target constraints, and a target optimization stage number, wherein the target cost includes at least one of the following: curvature cost, first derivative of curvature cost, second derivative of curvature cost, acceleration cost, jerk cost, lateral acceleration cost, lateral jerk cost, and time-optimal cost; the target constraints include at least one of the following: curvature constraint, first derivative of curvature constraint, second derivative of curvature constraint, velocity constraint, acceleration constraint, second derivative of velocity constraint, tractor-trailer angle constraint, starting state constraint, and ending state constraint; the target optimization stage number is determined based on the total time and total number of steps of the trajectory estimated by the search path; S122, optimizing the parking path based on the target cost, the target constraints, and the target optimization stage number.

[0109] It is understood that, in the embodiments of this application, a nonlinear optimization problem based on the interior point method can be constructed, as shown in Equation 8.

[0110]

[0111] st f( , ,p)=0,i=1,…,N-1, (Formula 8)

[0112] g( p)≥0, i=1,…,N

[0113] in, This represents the total number of steps in the optimization problem. Indicates the first The total objective cost of the step is the sum of the objective costs of the following items; The variable to be optimized is designed as follows in the embodiments of this application: ; The parameters representing the optimization problem include weights, upper and lower bounds of constraints, vehicle kinematic parameters, etc. The kinematic ordinary differential equation constraints representing the optimization problem are designed in this invention as Equation 9:

[0114] (Formula 9)

[0115] in, Indicates the yaw angle of the tractor unit, Indicates the trailer yaw angle, Indicates trailer wheelbase, This represents the distance between the articulation point of the tractor and the center of the rear axle of the tractor, and the control variable is selected as the second derivative of curvature. and velocity second derivative As a control variable. This represents all constraints in the optimization problem, including all equality and inequality constraints.

[0116] The cost function described in the embodiments of this application includes the following basic cost:

[0117] Curvature cost Effort_Kappa:

[0118] (Formula 10)

[0119] The cost of the first derivative of curvature, Effort_Dkappa:

[0120] (Formula 11)

[0121] The cost of the second derivative of curvature: Effort_Ddkappa

[0122] (Formula 12)

[0123] Acceleration cost Effort_Acc:

[0124] (Formula 13)

[0125] The cost of accelerometer Effort_Jerk:

[0126] (Formula 14)

[0127] Lateral acceleration cost Effort_Lateral_Acc:

[0128] (Formula 15)

[0129] Lateral jerk cost Effort_Lateral_Jerk:

[0130] (Formula 16)

[0131] The optimal time cost Effort_Dt:

[0132] (Formula 17)

[0133] In one possible implementation, the above time-optimal trajectory planning can estimate the total time and total number of steps of the trajectory based on the search path as the number of optimization stages, thus avoiding the occurrence of invalid and redundant trajectory segments.

[0134] The constraints described above in the embodiments of this application may include the following basic constraints:

[0135] Curvature constraint Kappa_limit:

[0136] (Formula 18)

[0137] (Formula 19)

[0138] First derivative of curvature constraint Dkappa_limit:

[0139] (Formula 20)

[0140] (Formula 21)

[0141] Curvature second derivative constraint Ddkappa_limit:

[0142] (Formula 22)

[0143] (Formula 23)

[0144] Speed ​​limit:

[0145] (Formula 24)

[0146] (Formula 25)

[0147] Acceleration constraint Acc_limit:

[0148] (Formula 26)

[0149] (Formula 27)

[0150] Velocity second derivative constraint Jerk_limit:

[0151] (Formula 28)

[0152] (Formula 29)

[0153] Jacknife_angle_limit constraint between tractor and trailer:

[0154] (Formula 30)

[0155] (Formula 31)

[0156] The starting state constraint Start_state_limit is shown in Equation 32:

[0157]

[0158]

[0159]

[0160]

[0161]

[0162] (Formula 32)

[0163]

[0164]

[0165] The endpoint state constraint End_state_limit is shown in Equation 33:

[0166]

[0167]

[0168]

[0169] (Formula 33)

[0170]

[0171]

[0172] In this embodiment, the initial solution for the above-mentioned parking trajectory optimization is based on the search path, obtaining the tractor's pose and the trailer's heading angle, and processing the angles to make their range within a certain range. This does not affect the calculation of the yaw angle difference between the tractor and trailer. Since the number of search path points is less than the number of optimized trajectory points, the path points are set when setting the initial solution. The points between them are obtained by linear interpolation and passed to the optimizer as a warm start for parking trajectory optimization.

[0173] This application also proposes an obstacle avoidance optimization strategy, which will be described in detail below.

[0174] Optionally, in this embodiment of the application, the above method further includes: S131, setting an obstacle avoidance hard boundary and an obstacle avoidance soft boundary, wherein the obstacle avoidance hard boundary prohibits the trajectory from passing through the obstacle through inequality hard constraints, and the obstacle avoidance soft boundary is used to calculate the obstacle avoidance term in the target cost; S132, combining the obstacle avoidance hard boundary and the obstacle avoidance soft boundary to optimize the parking path.

[0175] It can be understood that, assuming Represents a point in Euclidean space The nearest distance to the obstacle is then subject to the following inequality constraint, Equation 34:

[0176] (Formula 34)

[0177] The gradient information is given by the following formulas 35-38:

[0178] (Formula 35)

[0179] (Formula 36)

[0180] (Formula 37)

[0181] (Formula 38)

[0182] This application embodiment samples several points uniformly along the geometric contour of the vehicle front and trailer, applying the above inequality constraints, such as... Figure 7 The diagram shows the outline of the obstacle avoidance vehicle.

[0183] In the obstacle avoidance function of this invention, two safety buffers are set: : Represents a hard boundary for obstacle avoidance, which directly prohibits the trajectory from passing through obstacles through hard constraints of inequalities, ensuring that physical collisions are absolutely avoided; : Represents the soft boundary for obstacle avoidance, used to calculate the obstacle avoidance term in the cost function, enabling the vehicle to maintain a reasonable and comfortable distance from obstacles during movement, thus improving the experience of the automatic parking system; where This allows us to keep the vehicle as far away from obstacles as possible while satisfying hard constraints. Specifically, let the slack variable be... Then the optimization problem becomes as shown in Equation 39:

[0184]

[0185] (Formula 39)

[0186]

[0187] like Figures 8a-8b The diagram shows a schematic of the obstacle avoidance algorithm for parking trajectory optimization. Figure 8a This is a schematic diagram of hard constraint obstacle avoidance. Figure 8b This is a schematic diagram of obstacle avoidance combining hard and soft constraints. It can be seen that the obstacle avoidance path combining hard and soft constraints can maintain a more reasonable distance from obstacles.

[0188] This application also proposes an autonomous gear shifting optimization in its embodiments, which will be described in detail below.

[0189] Optionally, in this embodiment of the application, the above method further includes: S141, obtaining the initial speed and initial acceleration of gear shifting using a uniform acceleration model based on the parking path and combined with the set maximum speed and acceleration constraints; S142, adding the attraction cost for the pose of the search path to the target cost based on the kinematic constraints of the initial speed and the initial acceleration, and reducing the upper and lower bounds of the curvature constraints, wherein the attraction cost includes at least one of the following: x-position attraction cost, y-position attraction cost, tractor heading angle attraction cost, and trailer heading angle attraction cost.

[0190] It is understood that in this embodiment, the gear shifting information of the search path can be used as a reference, rather than explicitly setting the upper and lower bound constraints on the velocity of each trajectory point based on the gear shifting information of the search path. This avoids the problem of poor trajectory optimization quality or even solution failure in some parking scenarios due to insufficient gear shifting times in the search path. First, based on the search path and the set maximum speed and acceleration, the initial solutions for speed and acceleration are obtained using a uniform acceleration model. At the same time, the attraction of the pose of the search path is added to the objective function, which can, to a certain extent, ensure the geometry of the optimized trajectory, and allow for setting smaller curvature constraint upper and lower bounds. With this design, the trajectory will autonomously shift gears reasonably to meet various constraints during the optimization process, and the trajectory shape is effectively guaranteed, greatly improving the flexibility and generalization of parking trajectory optimization.

[0191] The traction cost of the autonomous gear shifting function of this invention includes the following additional cost functions.

[0192] Attraction_X at position x:

[0193] (Formula 40)

[0194] Attraction_Y at position y:

[0195] (Formula 41)

[0196] Attraction_Tractor_Yaw:

[0197] (Formula 42)

[0198] Trailer yaw attraction:

[0199] (Formula 43)

[0200] like Figures 9a-9bThe image shows the same parking scenario. Figure 9a This is a diagram illustrating the search path without shifting gears and the trajectory optimization without shifting gears. Figure 9b This diagram illustrates the addition of an automatic gear shift during trajectory optimization, eliminating the need for gear shifts in the search path. A comparison before and after the implementation of the automatic gear shift function for parking trajectory optimization shows that the resulting trajectory is more reasonable, with less curvature and a smoother overall path.

[0201] This application also proposes a strategy for smoothing the end of the trajectory, which will be described in detail below.

[0202] Optionally, in this embodiment of the application, the above method further includes: S151, gradually reducing the upper and lower bounds of curvature and the soft constraint weight of curvature at several trajectory points at the end of the parking path.

[0203] Understandably, by applying additional constraints to several trajectory points at the end of the trajectory, the trajectory is made to return to its correct position in a timely manner and park in a straight line. The design approach involves gradually reducing the upper and lower bounds of curvature and the weight of the soft curvature constraint, with the following effect: Figures 10a-10c As shown, where Figure 10a This is a schematic diagram of the smoothing process at the end of the trajectory. Figures 10b-10c The corresponding trajectory dynamic information curve shows that as the trajectory is about to end, the curvature and its first and second derivatives are close to zero. The corresponding steering wheel angle and speed are also close to zero, which makes the steering wheel adjustment range minimal when finally parking in the parking space, while also ensuring safety in narrow parking spaces.

[0204] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, use and processing of the relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and corresponding operation entry points are provided for users to choose to authorize or refuse.

[0205] Corresponding to the application scenarios and methods provided in the embodiments of this application, the embodiments of this application also provide a parking control device for an autonomous articulated vehicle. For example... Figure 11 The diagram shown is a structural block diagram of a parking control device for an automated articulated vehicle according to an embodiment of this application, which may include:

[0206] The first construction module 1102 is used to construct a grid map of the parking area;

[0207] The second construction module 1104 is used to construct a heuristic map based on the grid map. The heuristic map is a map data structure that provides path search guidance information to each grid cell in the grid map, including a heuristic flow field structure for calculating the parking endpoint, a heuristic distance field structure for calculating obstacle avoidance, and a gradient field structure for calculating the obstacle avoidance direction.

[0208] The control module 1106 is used to search for parking paths based on the heuristic map and the vehicle model of the articulated vehicle, so as to control the articulated vehicle to automatically park according to the parking path.

[0209] pass Figure 11 The apparatus shown first constructs a grid map of the parking area during the automatic parking process of an autonomous articulated vehicle. Then, a heuristic map is constructed based on this grid map. This heuristic map is a map data structure that provides path search guidance information to each grid cell in the grid map. It includes a heuristic flow field structure for calculating the parking endpoint, a heuristic distance field structure for calculating obstacle avoidance, and a gradient field structure for calculating the obstacle avoidance direction. Finally, a parking path is searched based on this heuristic map and the vehicle model of the articulated vehicle to control the articulated vehicle to automatically park according to the parking path. In other words, at least three layers of structure (flow field structure, distance field structure, and gradient field structure) are constructed on top of the grid map to obtain the heuristic map. Guided search based on this heuristic map can significantly accelerate and optimize path search, solving the technical problem in related technologies where low path search efficiency in autonomous articulated vehicle parking makes it difficult to meet the real-time requirements of actual parking tasks, thus achieving efficient parking.

[0210] The articulated vehicle is equipped with sensors. In one possible implementation, the first construction module 1102 includes: a first acquisition unit, used to acquire static environmental information and dynamic obstacle information of the parking area collected by the sensors; and a second acquisition unit, used to perform layered projection of the static environmental information and the dynamic obstacle information after constructing a grid coordinate system based on the parking point and the parking space, so as to obtain the grid map.

[0211] Optionally, the second acquisition unit includes: a first acquisition subunit, used to extract the position and semantic information of the static environment information and project it into the corresponding grid cell; and a second acquisition subunit, used to extract the obstacle outline corresponding to the current moment for the dynamic obstacle information, and use an integer iterative line segment rasterization method to convert continuous geometric line segments into all grid cells covered by the line segments in the raster map, and use a recursive flooding fill algorithm to fill the dynamic obstacle polygon.

[0212] The second construction module 1104 mentioned above is also used to construct the heuristic flow field structure, including: taking the parking endpoint as the cost source point; and using a breadth-first search algorithm to update the global cost layer by layer for all grid cells in the parking area, so as to allocate the minimum cost value to the parking endpoint for each grid cell.

[0213] The second building module 1104 is also used to construct the heuristic distance field structure, including: based on the principle of parabolic lower envelope, and with The complexity is to calculate the distance field in N-dimensional space, where the vertex of the parabola is related to the obstacle, and N is greater than 1.

[0214] The second building module 1104 is also used to build the gradient field structure, including: obtaining the obstacle avoidance gradient of the corresponding grid cell from the distance field using bilinear interpolation.

[0215] The aforementioned apparatus also includes a module for constructing a vehicle model of the articulated vehicle, wherein the vehicle model of the articulated vehicle includes a vehicle kinematic model supporting gear shifting and a multi-circle coverage model for collision detection.

[0216] The aforementioned control module 1106 is further configured to include: a third acquisition unit, configured to select a first target grid cell with the lowest cost from the first set, and acquire a set of neighboring grid cells of the first target cell based on the vehicle kinematics model; a judgment unit, configured to, for any second target grid cell in the set of neighboring grid cells, determine whether the second target grid cell is in the second set or, based on the multi-circle coverage model, determine whether a collision will occur when the articulated vehicle is located at the second target grid cell, wherein the grid cells in the second set are grid cells visited during the path search process; and a control unit, configured to, if the judgment result is negative, calculate the cost of the second target grid cell based on the first target grid cell, add the second target grid cell to the first set, and set other grid cells in the set of neighboring grid cells as the second target grid cell to continue executing the judgment and cost calculation steps until there are no expandable grid cells.

[0217] The aforementioned judgment unit is also used to determine, based on the grid map, whether a collision will occur when the articulated vehicle is located in the second target grid cell.

[0218] The articulated vehicle includes a tractor and a trailer. The control module 1106 further includes a cache unit for caching a collision table before determining whether a collision will occur when the articulated vehicle is located in the second target grid unit. The collision table is a mapping relationship between the combined information of the tractor's heading angle and the trailer's heading angle and the collision information.

[0219] The third acquisition unit is also used to downsample the raster map at a preset ratio to obtain a search map; and to determine the neighborhood raster cell set based on the search map.

[0220] Before performing parking path search based on the heuristic map and the vehicle model of the articulated vehicle, the aforementioned device further includes: a first processing module for discretizing the state space and motion space of the articulated vehicle, wherein the state space includes at least one of the following: the center position of the rear axle of the tractor. The tractor's heading angle and the trailer's heading angle, the range of motion including at least one of the following: gear position Front wheel steering angle.

[0221] Optionally, the above device further includes: a second processing module, used to map the discretized state space to the corresponding target index according to a preset encoding method; and to store the accessed grid cells in the second set in the manner corresponding to the target index.

[0222] The aforementioned control unit is also used to calculate the cost using a target cost model, wherein the target cost model is determined by the existing cost of the grid cell, the path length cost from the parent grid cell to the current grid cell, the turning cost, the reversing cost, the shifting cost, and the heuristic cost obtained through the flow field structure in the heuristic map.

[0223] The aforementioned device further includes: a third construction module, used to construct a target cost, target constraints, and a target optimization stage number, wherein the target cost includes at least one of the following: curvature cost, first derivative of curvature cost, second derivative of curvature cost, acceleration cost, jerk cost, lateral acceleration cost, lateral jerk cost, and time-optimal cost; the target constraints include at least one of the following: curvature constraint, first derivative of curvature constraint, second derivative of curvature constraint, velocity constraint, acceleration constraint, second derivative of velocity constraint, tractor-trailer angle constraint, starting state constraint, and ending state constraint; the target optimization stage number is determined based on the total time and total number of steps of the trajectory estimated by the search path; and a first optimization module, used to optimize the parking path based on the target cost, the target constraints, and the target optimization stage number.

[0224] The aforementioned device further includes: a second optimization module, used to set obstacle avoidance hard boundaries and obstacle avoidance soft boundaries, wherein the obstacle avoidance hard boundaries use inequality hard constraints to prohibit the trajectory from passing through obstacles, and the obstacle avoidance soft boundaries are used to calculate the obstacle avoidance term in the target cost; and the parking path is optimized by combining the obstacle avoidance hard boundaries and the obstacle avoidance soft boundaries.

[0225] The aforementioned device further includes: a third optimization module, used to obtain the initial shift speed and initial acceleration using a uniform acceleration model based on the parking path and combined with the set maximum speed and acceleration constraints; based on the kinematic constraints of the initial speed and initial acceleration, to add an attraction cost for the pose of the search path to the target cost, and to reduce the upper and lower bounds of the curvature constraints, wherein the attraction cost includes at least one of the following: x-position attraction cost, y-position attraction cost, tractor heading angle attraction cost, and trailer heading angle attraction cost.

[0226] The aforementioned device also includes a third optimization module, used to gradually reduce the upper and lower bounds of curvature and the soft constraint weights of curvature at several trajectory points at the end of the parking path.

[0227] The functions of each module in each system of the embodiments of this application can be found in the corresponding description in the above method, and they have corresponding beneficial effects, which will not be repeated here.

[0228] Corresponding to the application scenarios and methods provided in the embodiments of this application, the embodiments of this application also provide a parking control system for an autonomous articulated vehicle. For example... Figure 12 The diagram shown is a block diagram of a parking control system for an automated articulated vehicle according to an embodiment of this application, which may include:

[0229] Grid map construction: Construct a closed grid map based on dynamic / static obstacle information from the sensor.

[0230] Heuristic map construction involves building a priori heuristic map, including a flow field for calculating the endpoint heuristic, a distance field for calculating the obstacle avoidance heuristic, and a gradient field for calculating the obstacle avoidance direction.

[0231] Hybrid A* path planning includes discretization and caching structures to accelerate computational efficiency, multi-resolution structures to balance fine-grained obstacle avoidance and search efficiency, and a cost design to guide the robot efficiently to the target state. The final output is a kinematically sound obstacle avoidance path, where each discrete waypoint contains the following state information:

[0232]

[0233] in This is the center position of the rear axle of the tractor. The heading angle of the tractor unit. This is the trailer heading angle. This is the gear information (1 indicates forward gear, 0 indicates reverse gear). The turning angle of the front wheels of the tractor;

[0234] Vehicle models, including a precise kinematic model that supports gear shift search, and a collision detection algorithm based on an N-circle coverage model;

[0235] The optimization module, based on a spatiotemporal joint trajectory optimization algorithm, includes the design of the objective function and constraints for trajectory optimization. It mainly covers obstacle avoidance optimization based on vehicle contour sampling and Euclidean Signed Distance Field map (ESDF), autonomous gear shifting optimization based on the initial solution, and optimization based on parking terminal smoothing. The final output is a time-optimal parking trajectory that satisfies kinematic constraints and possesses obstacle avoidance and autonomous gear shifting capabilities. Each discrete trajectory point contains the following state information:

[0236]

[0237] in Indicates the center position of the rear axle of the tractor. The heading angle of the tractor unit. This is the trailer heading angle. Indicates curvature. The first derivative of curvature is represented. Indicates speed, Indicates acceleration. The second derivative of curvature is represented. Indicates jerk. This indicates the time interval between adjacent trajectory points.

[0238] pass Figure 12 The system described above, during the automatic parking process of an autonomous articulated vehicle, first constructs a grid map of the parking area, and then builds a heuristic map based on this grid map. This heuristic map is a map data structure that provides path search guidance information to each grid cell in the grid map, including a heuristic flow field structure for calculating the parking endpoint, a heuristic distance field structure for calculating obstacle avoidance, and a gradient field structure for calculating the obstacle avoidance direction. Then, based on this heuristic map and the vehicle model of the articulated vehicle, a parking path is searched to control the articulated vehicle to automatically park according to the parking path. Finally, the parking path is optimized, which not only significantly accelerates and optimizes the path search, solving the technical problem of low path search efficiency in autonomous articulated vehicle parking that makes it difficult to meet the real-time requirements of actual parking tasks, but also improves the accuracy of the path search, thereby achieving efficient and high-precision parking.

[0239] This application also provides a controller. Figure 13 This is a block diagram of an electronic device used to implement embodiments of this application. For example... Figure 13As shown, the controller 130 includes a memory 1301 and a processor 1302. The memory 1301 stores a computer program that can run on the processor 1302. When the processor 1302 executes the computer program, it implements the method described in the above embodiments. The number of memories 1301 and processors 1302 can be one or more.

[0240] The electronic device also includes:

[0241] The communication interface 1303 is used to communicate with external devices and exchange and transmit data.

[0242] If the memory 1301, processor 1302, and communication interface 1303 are implemented independently, they can be interconnected via a bus to communicate with each other. This bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. This bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 13 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0243] Optionally, in a specific implementation, if the memory 1301, processor 1302, and communication interface 1303 are integrated on a single chip, then the memory 1301, processor 1302, and communication interface 1303 can communicate with each other through an internal interface.

[0244] Corresponding to the application scenarios and methods provided in the embodiments of this application, the embodiments of this application also provide an autonomous driving vehicle. For example... Figure 14 The diagram shown is a schematic diagram of an autonomous vehicle structure according to an embodiment of this application, which may include the controller 130 described above.

[0245] This application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method provided in this application.

[0246] This application also provides a chip including a processor for calling and executing instructions stored in a memory, causing a communication device with the chip installed to perform the method provided in this application.

[0247] This application also provides a chip, including: an input interface, an output interface, a processor, and a memory. The input interface, output interface, processor, and memory are connected through an internal connection path. The processor is used to execute code in the memory. When the code is executed, the processor is used to execute the method provided in the application embodiment.

[0248] It should be understood that the aforementioned processor can be a Central Processing Unit (CPU), or other general-purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field-Programmable Gate Arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. General-purpose processors can be microprocessors or any conventional processor. It is worth noting that the processor can be a processor supporting Advanced Reduced Instruction Set Machines (ARM) architecture.

[0249] Further, optionally, the aforementioned memory may include read-only memory and random access memory. The memory may be volatile memory or non-volatile memory, or may include both. Non-volatile memory may include read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. Volatile memory may include random access memory (RAM), which serves as an external cache. By way of example, but not limitation, many forms of RAM are available. Examples include Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDR SDRAM), Enhanced Synchronous DRAM (ESDRAM), Sync Link DRAM (SLDRAM), and Direct Rambus RAM (DR RAM).

[0250] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions according to this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another.

[0251] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of those different embodiments or examples.

[0252] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "a plurality of" means two or more, unless otherwise explicitly specified.

[0253] Any process or method described in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing a particular logical function or process. Furthermore, the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functionality involved.

[0254] The logic and / or steps described in the flowchart or otherwise herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus or device (such as a computer-based system, a processor-included system or other system that can fetch and execute instructions from, an instruction execution system, apparatus or device).

[0255] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. All or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware, the program being stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiments.

[0256] Furthermore, the functional units in the various embodiments of this application can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. This storage medium can be a read-only memory, a disk, or an optical disk, etc.

[0257] The above description is merely an exemplary embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various variations or substitutions within the technical scope described in this application, and these should all 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 parking control method of an automated driving articulated vehicle, characterized by, include: Construct a grid map of the parking waiting area; A heuristic map is constructed based on the grid map, wherein the heuristic map is a map data structure that provides path search guidance information to each grid cell in the grid map, including a heuristic flow field structure for calculating the parking endpoint, a heuristic distance field structure for calculating obstacle avoidance, and a gradient field structure for calculating the obstacle avoidance direction. Parking path search is performed based on the heuristic map and the vehicle model of the articulated vehicle to control the articulated vehicle to automatically park according to the parking path.

2. The method of claim 1, wherein, The articulated vehicle is equipped with sensors, and the construction of a grid map of the parking area includes: Acquire static environmental information and dynamic obstacle information of the parking area collected by the sensor; After constructing a grid coordinate system based on parking points and parking spaces, the static environment information and the dynamic obstacle information are projected in layers to obtain the grid map.

3. The method according to claim 2, characterized in that, The static environment information and the dynamic obstacle information are layered and projected to obtain the grid map, including: For the static environment information, the location and semantic information are extracted and projected onto the corresponding grid cell; For the dynamic obstacle information, the obstacle outline corresponding to the current moment is extracted, and the continuous geometric line segments are converted into all grid cells covered by the line segments in the grid map using the line segment rasterization method of integer iteration operation. Then, the dynamic obstacle polygon is filled using a recursive flooding fill algorithm.

4. The method according to claim 1, characterized in that, Constructing the heuristic flow field structure includes: The parking endpoint is used as the cost source point; A breadth-first search algorithm is used to update the global cost of all grid cells within the parking area layer by layer, so as to allocate the minimum cost value to the parking endpoint for each grid cell.

5. The method according to claim 1, characterized in that, Constructing the heuristic distance field structure includes: Based on the principle of parabolic envelope, and using The complexity is calculated for the distance field in N-dimensional space, where the vertex of the parabola is related to the obstacle, and N is greater than 1.

6. The method according to claim 5, characterized in that, Constructing the gradient field structure includes: The obstacle avoidance gradient of the corresponding grid cell is obtained from the distance field using bilinear interpolation.

7. The method according to claim 1, characterized in that, The method further includes: constructing a vehicle model of the articulated vehicle, wherein the vehicle model of the articulated vehicle includes a vehicle kinematic model supporting gear shifting and a multi-circle coverage model for collision detection.

8. The method according to claim 7, characterized in that, Parking path search is performed based on the heuristic map and the vehicle model of the articulated vehicle, including: Select the first target grid cell with the lowest cost from the first set, and obtain the set of neighboring grid cells of the first target cell based on the vehicle kinematics model; For any second target grid cell in the neighborhood grid cell set, determine whether the second target grid cell is in the second set or determine whether a collision will occur when the articulated vehicle is located in the second target grid cell based on the multi-circle coverage model, wherein the grid cells in the second set are the grid cells that have been visited during the path search process; If the judgment result is negative, the cost of the second target grid cell is calculated based on the first target grid cell, and the second target grid cell is added to the first set. Other grid cells in the neighborhood grid cell set are set as the second target grid cell, and the judgment and cost calculation steps are continued until there are no more expandable grid cells.

9. The method according to claim 8, characterized in that, The method further includes: Based on the grid map, it is determined whether a collision will occur when the articulated vehicle is located in the second target grid cell.

10. The method according to claim 8, characterized in that, The articulated vehicle includes a tractor and a trailer. Before determining whether a collision will occur when the articulated vehicle is located in the second target grid cell, the method further includes: A cached collision table, wherein the collision table is a mapping relationship between the combined information of the tractor's heading angle and the trailer's heading angle and the collision information.

11. The method according to claim 8, characterized in that, The set of neighboring grid cells of the first target grid cell is obtained based on the vehicle kinematics model, including: The search map is obtained by downsampling the grid map at a preset ratio; The neighborhood grid cell set is determined based on the search map.

12. The method according to claim 8, characterized in that, The articulated vehicle includes a tractor and a trailer. Before performing parking path search based on the heuristic map and the vehicle model of the articulated vehicle, the method further includes: The state space and motion space of the articulated vehicle are discretized, wherein the state space includes at least one of the following: the center position of the rear axle of the tractor, the heading angle of the tractor, and the heading angle of the trailer, and the motion space includes at least one of the following: gear position and front wheel angle.

13. The method according to claim 12, characterized in that, The method further includes: The discretized state space is mapped to the corresponding target index according to the preset encoding method; The visited raster cells are stored in the second set according to their corresponding target indices.

14. The method according to claim 8, characterized in that, The cost is calculated using a target cost model, which is determined by the existing cost of the grid cell, the path length cost from the parent grid cell to the current grid cell, the turning cost, the reversing cost, the shifting cost, and the heuristic cost obtained from the flow field structure in the heuristic map.

15. The method according to claim 1, characterized in that, The articulated vehicle includes a tractor and a trailer, and the method further includes: The objective cost, objective constraints, and number of objective optimization stages are constructed. The objective cost includes at least one of the following: curvature cost, first derivative of curvature cost, second derivative of curvature cost, acceleration cost, jerk cost, lateral acceleration cost, lateral jerk cost, and time-optimal cost. The objective constraints include at least one of the following: curvature constraint, first derivative of curvature constraint, second derivative of curvature constraint, velocity constraint, acceleration constraint, second derivative of velocity constraint, tractor-trailer angle constraint, starting state constraint, and ending state constraint. The number of objective optimization stages is determined based on the total time and total number of steps of the estimated trajectory in the search path. The parking path is optimized based on the target cost, the target constraints, and the target number of optimization stages.

16. The method according to claim 15, characterized in that, The method further includes: Set obstacle avoidance hard boundaries and obstacle avoidance soft boundaries, wherein the obstacle avoidance hard boundaries prevent the trajectory from passing through obstacles through inequality hard constraints, and the obstacle avoidance soft boundaries are used to calculate the obstacle avoidance term in the target cost; The parking path is optimized by combining the obstacle avoidance hard boundary and the obstacle avoidance soft boundary.

17. The method according to claim 15, characterized in that, The method further includes: Based on the parking path, and combined with the set maximum speed and acceleration constraints, the initial speed and initial acceleration for gear shifting are obtained using a uniform acceleration model. Based on the kinematic constraints of the initial velocity and the initial acceleration, an attraction cost for the pose of the search path is added to the target cost, and the upper and lower bounds of the curvature constraint are reduced. The attraction cost includes at least one of the following: x-position attraction cost, y-position attraction cost, tractor heading angle attraction cost, and trailer heading angle attraction cost.

18. The method according to claim 15, characterized in that, The method further includes: The upper and lower bounds of curvature and the soft constraint weight of curvature are gradually reduced for several trajectory points at the end of the parking path.

19. An autonomous vehicle, characterized in that, include: A controller, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the steps of the method as claimed in any one of claims 1 to 18.

20. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the method described in claims 1-18.

21. A computer program product, characterized in that, Includes a computer program / instructions that, when executed by a processor, implement the steps of the method as described in any one of claims 1-18.

Images