Vehicle and parking control method thereof

Abstract

Embodiments of the present application provide a vehicle and a parking control method thereof. The method is applied to a vehicle control unit arranged in the vehicle, and includes: receiving a global reference path and an environment grid map issued by a cloud scheduling platform, wherein the global reference path is used to represent a reference path for guiding the vehicle to travel from a current position to a target parking space, and the environment grid map is used to represent a grid map representing the spatial distribution of obstacles in a parking lot; fusing the environment grid map and obstacle data perceived by the vehicle to generate a target grid map; generating a local control instruction of the vehicle based on the global reference path, the target grid map, and vehicle state data of the vehicle; and controlling the vehicle to perform a parking operation based on the local control instruction. The present application solves the technical problem of low accuracy of parking control of the vehicle in the related art.

Description

Technical Field

[0001] This application relates to the field of vehicles, and more specifically, to a vehicle and a parking control method thereof. Background Technology

[0002] With the continuous growth of urban vehicle ownership and the increasing scarcity of parking resources, parking has become a frequent and difficult task in the driving process. Especially in narrow and dense parking environments such as shopping malls and old residential areas, problems such as blind spots, dynamic obstacles, and limited parking space significantly increase the difficulty of parking.

[0003] Currently, parking path planning methods are mainly divided into two categories: one is the traditional algorithm based on geometry or graph search. This type of method generates parking trajectories by pre-setting path templates or spatial search. It has reliability in structured environments, but it is difficult to improve the smoothness of the planned path. The other type is the end-to-end method based on deep reinforcement learning. Although this type of method can generate smooth trajectories, it is difficult to adapt to changing parking environments, resulting in low accuracy of vehicle parking control in related technologies.

[0004] There is currently no good solution to the above problems. Summary of the Invention

[0005] This application provides a vehicle and its parking control method to at least solve the technical problem of low parking control accuracy of vehicles in related technologies.

[0006] According to one aspect of the embodiments of this application, a parking control method for a vehicle is provided, applied to an on-board control unit, the on-board control unit being deployed in the vehicle, comprising: receiving a global reference path and an environmental grid map issued by a cloud scheduling platform, wherein the global reference path is used to represent a reference path guiding the vehicle from its current position to a target parking space, and the environmental grid map is used to represent a grid map characterizing the spatial distribution of obstacles in the parking lot; fusing the environmental grid map and obstacle data perceived by the vehicle to generate a target grid map; generating local control commands for the vehicle based on the global reference path, the target grid map, and the vehicle's vehicle state data; and controlling the vehicle to perform a parking operation based on the local control commands.

[0007] Furthermore, the environmental grid map and the obstacle data perceived by the vehicle are fused to generate a target grid map, including: determining the obstacle types corresponding to multiple grids in the environmental grid map; updating the environmental grid map based on the obstacle types and obstacle data corresponding to multiple grids to generate the target grid map; preferably, updating the environmental grid map based on the obstacle types and obstacle data corresponding to multiple grids to generate the target grid map includes: when the obstacle type in any grid is a static obstacle, determining the grid value corresponding to the grid as a preset grid value; when the obstacle type in any grid is a dynamic obstacle, updating the grid value corresponding to the grid based on the obstacle data; and generating the target grid map based on the grid values ​​corresponding to multiple grids.

[0008] Furthermore, based on the global reference path, the target grid map, and the vehicle's state data, local control commands for the vehicle are generated, including: smoothing the global reference path based on the vehicle state data to obtain a global reference trajectory; constructing a multi-dimensional evaluation model based on the global reference trajectory and vehicle state data, wherein the multi-dimensional evaluation model is used to evaluate the tracking deviation and steering wheel control deviation of the vehicle when traveling along the global reference trajectory within a preset control cycle; constructing target constraints based on the target grid map and the vehicle's preset safety conditions; and solving the multi-dimensional evaluation model based on the target constraints to obtain local control commands.

[0009] Furthermore, the global reference path is smoothed based on vehicle state data to obtain a global reference trajectory, including: determining the position of the preview point on the global reference path based on the vehicle position data and vehicle attitude data in the vehicle state data; determining the target curvature based on the preview point position and the position of the rear axle center point of the vehicle; and smoothing the global reference path based on the target curvature to obtain the global reference trajectory.

[0010] Furthermore, based on the target grid map and the vehicle's preset safety conditions, target constraints are constructed, including: constructing dynamic constraints based on the vehicle's kinematic equations; constructing obstacle avoidance constraints based on the target grid map and preset safety distance thresholds in the preset safety conditions; constructing control constraints based on preset steering wheel control thresholds in the preset safety conditions; constructing boundary constraints based on preset boundary thresholds in the preset safety conditions; and determining target constraints based on dynamic constraints, obstacle avoidance constraints, control constraints, and / or boundary constraints. Preferably, the obstacle avoidance constraints are constructed based on the target grid map and preset safety distance thresholds in the preset safety conditions, including: determining the passable area at the current moment based on the target grid map; determining the interval distance based on the location data of the passable area and the location data of the obstacle grid; and determining an interval distance greater than the preset safety distance threshold as an obstacle avoidance constraint.

[0011] Furthermore, the multi-dimensional evaluation model is solved based on the target constraints to obtain local control commands, including: solving the multi-dimensional evaluation model to obtain a control parameter sequence, wherein the control parameter sequence satisfies the target constraints; and determining the control parameter of the first control cycle in the control parameter sequence as the local control command.

[0012] Furthermore, after controlling the vehicle to perform a parking operation based on local control commands, the method also includes: acquiring the rear axle center point position data of the vehicle; determining the deviation data based on the rear axle center point position data and preset vehicle center position data; determining the vehicle status as successfully parked when the deviation data is less than the preset deviation and the vehicle speed is zero, and acquiring the target vehicle pose data; and sending the target vehicle pose data and the parking success signal to the cloud scheduling platform.

[0013] According to another aspect of the embodiments of this application, a parking control method for a vehicle is also provided, applied to a cloud-based scheduling platform, comprising: when a vehicle enters the target communication range in a parking lot, receiving vehicle status data uploaded by an on-board control unit and obtaining environmental data in the parking lot; determining a global reference path based on the vehicle status data and the environmental data; generating an environmental grid map corresponding to the global reference path; and sending the global reference path and the environmental grid map to the on-board control unit.

[0014] Furthermore, based on vehicle status data and environmental data, a global reference path is determined, including: constructing a global accessible area map of the parking lot based on vehicle status data and environmental data; performing accessibility and parking feasibility analysis on all vacant parking spaces based on the global accessible area map and vehicle status data to obtain a priority list of available parking spaces; determining the first parking space in the priority list as the target parking space; and generating a global reference path based on the current location data in the vehicle status data, the location data corresponding to the target parking space, and the principle of the shortest global path.

[0015] According to another aspect of the embodiments of this application, a parking control device for a vehicle is also provided, applied to an on-board control unit. The on-board control unit is deployed in the vehicle and includes: a first receiving module, used to receive a global reference path and an environmental grid map sent by a cloud scheduling platform, wherein the global reference path is used to represent a reference path guiding the vehicle from its current position to a target parking space, and the environmental grid map is used to represent a grid map characterizing the spatial distribution of obstacles in the parking lot; a processing module, used to perform fusion processing on the environmental grid map and obstacle data perceived by the vehicle to generate a target grid map; a first generating module, used to generate local control commands for the vehicle based on the global reference path, the target grid map, and the vehicle's vehicle status data; and a control module, used to control the vehicle to perform parking operations based on the local control commands.

[0016] According to another aspect of the embodiments of this application, a parking control device for a vehicle is also provided, applied to a cloud-based scheduling platform, comprising: a second receiving module, configured to receive vehicle status data uploaded by an on-board control unit and acquire environmental data in the parking lot when the vehicle enters the target communication range in the parking lot; a determining module, configured to determine a global reference path based on the vehicle status data and the environmental data; a second generating module, configured to generate an environmental grid map corresponding to the global reference path; and a sending module, configured to send the global reference path and the environmental grid map to the on-board control unit.

[0017] According to another aspect of the embodiments of this application, a vehicle is also provided, including: a memory storing an executable program; and a processor for running the program, wherein the program executes the methods in various embodiments of this application when it runs.

[0018] According to another aspect of the embodiments of this application, a computer-readable storage medium is also provided, the computer-readable storage medium including a stored executable program, wherein, when the executable program is running, it controls the device where the computer-readable storage medium is located to perform the methods of various embodiments of this application.

[0019] According to another aspect of the embodiments of this application, a computer program product is also provided, including a computer program that, when executed by a processor, implements the methods of various embodiments of this application.

[0020] According to another aspect of the embodiments of this application, a computer program product is also provided, including a non-volatile computer-readable storage medium storing a computer program that, when executed by a processor, implements the methods in various embodiments of this application.

[0021] According to another aspect of the embodiments of this application, a computer program is also provided, which, when executed by a processor, implements the methods of the various embodiments of this application.

[0022] In this embodiment, a parking control method for a vehicle is applied to an on-board control unit deployed in the vehicle. The method first receives a global reference path and an environmental grid map from a cloud-based scheduling platform; then, it fuses the environmental grid map with obstacle data perceived by the vehicle to generate a target grid map; next, based on the global reference path, the target grid map, and the vehicle's state data, it generates local control commands for the vehicle; finally, based on the local control commands, it controls the vehicle to perform a parking operation. This application adopts a cloud-based scheduling platform that provides global prior guidance, vehicle-side real-time perception fusion, and local dynamic response. By fusing the reference path and grid map planned in the cloud that covers the entire static environment with the dynamic obstacle information perceived in real time by the vehicle's own sensors, a target grid map containing both global structure and real-time changes is formed. This achieves the goal of accurate cognition and path constraint update of complex and dynamic parking environments. Thus, based on the feasibility of the global path, local control behavior is dynamically corrected according to real-time environmental changes, avoiding collisions or path failures caused by the disconnect between static planning and dynamic environment. This solves the technical problem of low accuracy of vehicle parking control in related technologies. Attached Figure Description

[0023] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0024] Figure 1 This is a flowchart of a vehicle parking control method according to an embodiment of this application;

[0025] Figure 2 This is a flowchart of another vehicle parking control method according to an embodiment of this application;

[0026] Figure 3 This is a flowchart of an optional vehicle parking control method according to an embodiment of this application;

[0027] Figure 4 This is a schematic diagram of a grid map according to an embodiment of this application;

[0028] Figure 5 This is a schematic diagram of dynamic path planning according to an embodiment of this application;

[0029] Figure 6 This is a schematic diagram of the structure of a vehicle parking control system according to an embodiment of this application;

[0030] Figure 7This is a schematic diagram of a vehicle parking control device according to an embodiment of this application;

[0031] Figure 8 This is a schematic diagram of a vehicle parking control device according to an embodiment of this application. Detailed Implementation

[0032] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0033] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. 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 apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0034] According to an embodiment of this application, an embodiment of a parking control method for a vehicle is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0035] This embodiment provides a parking control method for a vehicle, which is applied to an on-board control unit deployed in the vehicle. Figure 1 This is a flowchart of a vehicle parking control method according to an embodiment of this application, such as... Figure 1 As shown, the process includes the following steps:

[0036] Step S102: Receive the global reference path and environmental grid map sent by the cloud scheduling platform. The global reference path is used to represent the reference path that guides the vehicle from the current location to the target parking space, and the environmental grid map is used to represent the grid map that characterizes the spatial distribution of obstacles in the parking lot.

[0037] The aforementioned vehicle control unit refers to a computing and control module deployed inside the vehicle. The type of vehicle control unit may include, but is not limited to, vehicle domain controllers, autonomous driving domain control modules, etc., and the specific vehicle control unit needs to be determined based on the vehicle type. The vehicle control unit integrates vehicle sensor data processing, communication interfaces, path planning algorithms, and actuator control logic, and is the core decision-making and execution unit for realizing the automatic parking function.

[0038] The aforementioned cloud-based scheduling platform can refer to an intelligent system deployed on a remote server cluster. This platform can be used to perceive the parking space occupancy status, dynamic obstacle trajectories, lane congestion, and performance parameters of each vehicle in the entire parking lot. It then allocates target parking spaces to each vehicle requesting parking, generates a globally optimal driving path, and distributes structured environmental information (such as a grid map) to the corresponding vehicles, achieving collaborative decision-making between the vehicle, the parking lot, and the cloud.

[0039] The aforementioned global reference path refers to the recommended driving trajectory from the vehicle's current position to the target parking space, calculated by the cloud-based scheduling platform based on the overall parking lot topology and the vehicle's own dynamic constraints. The types of global reference paths can include, but are not limited to, spline curves, polynomial trajectories, polygonal paths, and discrete path point sequences; the specific global reference path needs to be determined according to actual needs. Global reference paths can be used to provide macro-level navigation guidance to the vehicle control unit and serve as a benchmark reference for local path planning.

[0040] The aforementioned environmental grid map refers to two-dimensional map data in which the parking lot ground area is divided into discrete grid cells of fixed size, and the spatial attributes of each cell are represented by numerical encoding. The environmental grid map may include, but is not limited to, static obstacle grids, dynamic risk grids, and passable grids; the specific environmental grid map needs to be determined based on the actual conditions of the parking lot. The environmental grid map can be used to provide structured and computable environmental perception information for the vehicle control unit, supporting the application of obstacle avoidance constraints during the local path planning stage and compensating for the limited sensing range or latency of vehicle sensors.

[0041] The aforementioned vehicles can refer to motor vehicles equipped with an Automatic Parking Assist System (APA) or an Autonomous Valet Parking System (AVP). These vehicles must possess a drive-by-wire chassis (drive-by-wire steering, drive-by-wire braking), a multi-sensor fusion perception system (ultrasonic, camera, radar), and an onboard communication module. These vehicles can serve as the execution platform for parking tasks, automatically completing the entire process of entering from the external area, path tracking, dynamic obstacle avoidance, precise parking, and parking lock, driven by the onboard control unit and based on path and map information distributed through cloud-field collaboration.

[0042] The aforementioned target parking space refers to a designated parking location selected and allocated by the cloud-based scheduling platform from all available parking spaces in the parking lot based on vehicle size, parking preferences, parking space availability, and accessibility, for the purpose of completing the parking operation. The target parking space can serve as the endpoint of the global path planning, and the onboard control unit performs final attitude correction based on the geometric boundaries (length, width, orientation, and distance from the edge) of the target parking space to achieve high-precision parking.

[0043] The aforementioned parking lots can refer to enclosed or semi-enclosed parking areas equipped with intelligent sensing devices, communication networks, and management systems. These parking lots can provide automated parking services for multiple vehicles. Parking lots can include, but are not limited to, underground parking garages, multi-level parking garages, shopping mall parking lots, transportation hub parking areas, and unmanned parking areas in industrial parks; the specific type of parking lot needs to be determined based on the actual situation. Parking lots can serve as the physical environment for a "cloud-vehicle-parking lot" collaborative architecture. The IoT sensing devices deployed within them continuously collect information on parking space status and dynamic obstacles, uploading this data to a cloud-based scheduling platform to form a foundation for global environmental awareness.

[0044] In one optional embodiment, the vehicle control unit receives a global reference path and an environmental grid map from a cloud-based scheduling platform. The global reference path serves as the sole reference for guiding the vehicle from its current location to the target parking space, while the environmental grid map represents the spatial distribution of all obstacles within the parking lot. This receiving action enables the vehicle control unit to obtain pre-planned macroscopic driving guidance and discretized spatial representations of static and dynamic obstacles from the cloud, providing the input data foundation for subsequent path tracking and local obstacle avoidance. This ensures that the vehicle's decision-making process always proceeds based on the global information provided by the cloud.

[0045] Step S104: The environmental grid map and the obstacle data perceived by the vehicle are fused to generate the target grid map.

[0046] The aforementioned fusion processing refers to the computational process by which the vehicle control unit integrates the environmental grid map sent from the cloud with obstacle data perceived in real time by the vehicle's own sensors, according to spatial coordinate alignment, time synchronization, and priority logic, to ensure consistency of multi-source data. The core of fusion processing lies in prioritizing the insurmountability of static obstacles and dynamically correcting the existence state of moving targets based on the vehicle's perception results, ensuring that the final output map reflects both the global environmental structure and responds to instantaneous changes.

[0047] The obstacle data mentioned above refers to the real-time detection and output of obstacle location, size, and movement trend information around the vehicle by its onboard sensing devices such as ultrasonic sensors, surround-view cameras, and millimeter-wave radar. Obstacle data can be expressed in coordinate form, including the obstacle's lateral and longitudinal distances relative to the vehicle, its speed, and confidence level, with its timestamp synchronized with the vehicle's positioning information. Specific obstacle data needs to be determined based on the actual situation. Obstacle data can be used to reflect the dynamic or partially static obstacles actually present around the vehicle at any given moment, such as pedestrians, cyclists, temporarily parked vehicles, and construction equipment.

[0048] The aforementioned target grid map refers to a two-dimensional discretized environmental occupancy map generated after fusion processing, used to guide local vehicle path planning. The target grid map divides the parking lot surface into several uniformly sized grid cells, each corresponding to a fixed physical area (e.g., 0.01m × 0.01m), and encodes the traffic status of that area with integer values. The target grid map can be used as the sole environmental constraint input to the model's predictive control module, ensuring that the path planning results are physically feasible and safety-verifiable. The above values ​​are for illustrative purposes only; specific values ​​need to be determined based on actual requirements.

[0049] In one optional embodiment, the onboard control unit fuses the environmental grid map sent by the cloud-based scheduling platform with obstacle data collected in real time by the vehicle's own sensors to generate a target grid map. This fusion process spatially aligns and logically overlays the prior information of static and dynamic obstacles in the environmental grid map with the obstacle position and motion status information output by the onboard ultrasonic sensors, surround-view cameras, and millimeter-wave radar, all within a unified global coordinate system and time reference. This process uses the cloud map as its framework and onboard perception as its dynamic update source, prioritizing the retention of static obstacles and covering dynamic risk areas with real-time perception results, thereby eliminating uncertainties caused by perception delays or blind spots. The generated target grid map is a structured, discrete representation of environmental occupancy. Each grid cell takes only a value of 0, 1, or 2, corresponding to passable, static obstacles, and dynamic risk areas, respectively. Its output is deterministic binary logic, without probability or intermediate states, and is directly used for obstacle avoidance constraint construction in subsequent model predictive control, ensuring that local paths are safe, reliable, and executable in dynamic environments.

[0050] Step S106: Based on the global reference path, the target grid map, and the vehicle's status data, generate local control commands for the vehicle.

[0051] The aforementioned vehicle status data refers to a set of discrete parameters characterizing the vehicle's current motion and attitude, collected and output in real time by the vehicle's own sensors and onboard control system. Vehicle status data may include, but is not limited to, the vehicle's lateral and longitudinal position coordinates, heading angle, longitudinal velocity, lateral velocity, front wheel steering angle, vehicle attitude angle, and wheel speed information. Specific vehicle status data needs to be determined based on actual requirements. The vehicle status data originates from the vehicle's built-in inertial measurement unit, wheel speed sensors, steering wheel angle sensor, and high-precision positioning module. It does not contain predicted values, estimated values, or external fusion results, but only reflects the actual measurable physical state of the system within the current control cycle. As the initial condition for local path planning, the vehicle status data is used to determine the vehicle's relative position and motion trend on the global reference path. It serves as the real-time basis for trajectory tracking and dynamic obstacle avoidance. Its update frequency is consistent with the control system's sampling cycle, ensuring the real-time and synchronous generation of control commands.

[0052] The aforementioned local control command can refer to the first output value in a short-term control sequence directly driving the vehicle's underlying actuators, generated online by the onboard control unit based on the global reference path, target grid map, and vehicle state data. Local control commands may include, but are not limited to, target front wheel steering angle and target longitudinal acceleration; the specific local control command needs to be determined based on actual control requirements and calculation results. Local control commands enable the vehicle to conform to the geometry of the global reference path while satisfying obstacle avoidance constraints defined by the target grid map, and simultaneously limiting abrupt changes in steering action, achieving smooth, safe, and low-vibration parking trajectory tracking.

[0053] In one optional embodiment, local control commands for the vehicle are generated based on a global reference path, a target grid map, and the vehicle's state data. The global reference path provides macroscopic driving guidance from the vehicle's current position to the target parking space; the target grid map clearly indicates traversable areas and obstacle distribution in the environment; and the vehicle's state data reflects its current pose, velocity, and motion parameters. These three elements serve as input information to calculate real-time control actions that meet kinematic constraints and spatial obstacle avoidance requirements. This ensures that the vehicle dynamically adjusts steering and acceleration / deceleration commands based on real-time environmental conditions while adhering to the global path, thereby generating local control commands for the direct-drive actuators.

[0054] Step S108: Based on local control commands, control the vehicle to perform a parking operation.

[0055] In one optional embodiment, controlling the vehicle to perform a parking operation based on local control commands means that the generated local control commands are directly transmitted to the vehicle's underlying actuators, driving the vehicle to move from its current position to the target parking space. This command, without human intervention or additional decision-making layers, serves as the sole control input, driving the vehicle's power, steering, and braking systems to work in coordination, causing the vehicle to gradually approach and stop at the target parking space along an adjusted trajectory. This process continuously responds to real-time environmental changes, ensuring that the vehicle always follows the movement path defined by the local control commands until the parking operation is complete.

[0056] In this embodiment, a parking control method for a vehicle is applied to an on-board control unit deployed in the vehicle. The method first receives a global reference path and an environmental grid map from a cloud-based scheduling platform; then, it fuses the environmental grid map with obstacle data perceived by the vehicle to generate a target grid map; next, based on the global reference path, the target grid map, and the vehicle's state data, it generates local control commands for the vehicle; finally, based on the local control commands, it controls the vehicle to perform a parking operation. This application adopts a cloud-based scheduling platform that provides global prior guidance, vehicle-side real-time perception fusion, and local dynamic response. By fusing the reference path and grid map planned in the cloud that covers the entire static environment with the dynamic obstacle information perceived in real time by the vehicle's own sensors, a target grid map containing both global structure and real-time changes is formed. This achieves the goal of accurate cognition and path constraint update of complex and dynamic parking environments. Thus, based on the feasibility of the global path, local control behavior is dynamically corrected according to real-time environmental changes, avoiding collisions or path failures caused by the disconnect between static planning and dynamic environment. This solves the technical problem of low accuracy of vehicle parking control in related technologies.

[0057] Optionally, the environmental grid map and obstacle data perceived by the vehicle are fused to generate a target grid map, including: determining the obstacle types corresponding to multiple grids in the environmental grid map; updating the environmental grid map based on the obstacle types and obstacle data corresponding to the multiple grids to generate the target grid map; preferably, updating the environmental grid map based on the obstacle types and obstacle data corresponding to the multiple grids to generate the target grid map includes: when the obstacle type in any grid is a static obstacle, determining the grid value corresponding to the grid as a preset grid value; when the obstacle type in any grid is a dynamic obstacle, updating the grid value corresponding to the grid based on the obstacle data; and generating the target grid map based on the grid values ​​corresponding to the multiple grids.

[0058] The aforementioned grids refer to a set of continuous two-dimensional discrete units in an environmental raster map, divided by fixed spatial dimensions, used to represent the parking lot ground area. Each grid corresponds to a specific physical spatial region (e.g., 0.01m × 0.01m), and its location is uniquely identified by its row and column index in the global coordinate system. Multiple grids together constitute a structured spatial discrete representation of the parking lot environment, serving as the basic unit for information fusion and obstacle state mapping. Multiple grids can provide a unified spatial reference framework for prior environmental information from the cloud and vehicle-mounted perception data, enabling fusion processing to compare and update data from different sources unit by unit under the same coordinates, ensuring spatial consistency in environmental modeling. The above values ​​are for illustrative purposes only; specific values ​​need to be determined based on actual needs and are not limited here.

[0059] The obstacle types mentioned above can refer to category labels classified according to the physical characteristics and motion state of obstacles. Obstacle types can include, but are not limited to, static obstacles and dynamic obstacles; the specific obstacle type needs to be determined based on the actual obstacle. Obstacle types can serve as the logical basis for the fusion process, determining the subsequent grid value update rules.

[0060] The aforementioned obstacle data refers to real-time spatial information about surrounding environmental entities directly collected and output by vehicle-mounted sensing devices such as ultrasonic sensors, surround-view cameras, and millimeter-wave radar within the current control cycle. Obstacle data may include, but is not limited to, the lateral and longitudinal distances, speed, and presence confidence levels of obstacles relative to the vehicle body; the specific obstacle data needs to be determined based on actual requirements. Obstacle data can be used to provide real-time perception input to the onboard control unit, dynamically correcting the status of obstacles not covered or changed in the cloud map, and is a key basis for achieving closed-loop updates of environmental perception.

[0061] The aforementioned static obstacles refer to fixed physical structures that remain in place during parking, show no tendency to move, and are difficult for vehicles to traverse. Static obstacles can include, but are not limited to, walls, pillars, bollards, fixed curbs, and parking lot structural beams; the specific static obstacles need to be determined based on the actual situation. In the fusion processing, static obstacles have a high priority. Once their corresponding grid state is determined, it must not be overwritten or cleared by onboard perception data. This ensures that vehicles can identify and avoid such insurmountable physical obstacles under any circumstances, serving as the underlying constraint for ensuring parking safety.

[0062] The aforementioned preset grid value can refer to a numerical code pre-set in the system to uniquely identify a specific obstacle type. Its value is an integer and remains fixed during system operation. In this application, the preset grid value is 1, specifically used to represent the grid cell occupied by a static obstacle. This value is not dependent on real-time perception but is fixed during system initialization based on parking lot structure information. Its function is to provide a deterministic judgment benchmark for fusion processing, ensuring that all static obstacles are expressed in a uniform and unchanging manner in the target grid map, avoiding the failure of safety boundaries due to perception errors or communication delays.

[0063] The aforementioned dynamic obstacles refer to variable entities that move during parking and exhibit a tendency to move. Dynamic obstacles can include, but are not limited to, pedestrians, cyclists, temporarily moving vehicles, service robots, or carts; the specific dynamic obstacles must be determined based on the actual situation. The boundaries and positions of dynamic obstacles are obtained in real-time by onboard sensors. Their status is time-sensitive, and they are only considered a threat within the effective detection window of the sensors (e.g., 200 milliseconds). If they are not re-detected outside this window, their impact will automatically become invalid. Dynamic obstacles can be used to provide real-time disturbance information in the environment to dynamically correct passable areas in the cloud map, enabling vehicles to respond quickly to sudden obstacles. However, their priority is lower than that of static obstacles; they are only allowed to be updated to specific values ​​in the corresponding grid and must not be cleared or overwritten. The values ​​above are for illustrative purposes only; specific values ​​need to be determined based on actual needs and are not limited here.

[0064] The aforementioned raster value refers to the discrete integer code assigned to each raster cell. The raster value characterizes the accessibility of the area at the current moment and is the unique data carrier of the target raster map. Raster values ​​are only allowed to be 0, 1, or 2, representing accessible areas, static obstacle areas, and dynamic risk areas, respectively. The updating of raster values ​​is determined by the fusion processing logic: if a static obstacle exists in the corresponding area, it is fixed at 1; if a dynamic obstacle exists and is confirmed by vehicle perception, it is updated to 2; if there are no obstacles, it remains at 0. Raster values ​​can be used to provide a deterministic, computable, and constrained representation of environmental occupancy for subsequent model predictive control, ensuring that path planning results are physically safe and computationally efficient.

[0065] In one optional embodiment, the obstacle types corresponding to multiple grids in the environmental grid map are first determined. Then, the environmental grid map is updated based on the obstacle types and obstacle data corresponding to the multiple grids to generate a target grid map. Specifically, in any grid, if the corresponding obstacle type is a static obstacle, its preset grid value is directly retained to ensure the stability of the prior environmental information and prevent interference or mis-coverage by temporary sensing noise. If the corresponding obstacle type is a dynamic obstacle, the corresponding grid value is dynamically adjusted according to the obstacle data to achieve accurate response and trajectory tracking of moving targets. This effectively preserves the fixed structure information of the parking lot during the fusion process while updating the real-time changing obstacle state with high precision. Finally, a target grid map with both prior environmental reliability and real-time perception accuracy is generated, providing a solid and safe environmental perception foundation for the subsequent generation of local control commands based on the global path and vehicle status. This significantly improves the path planning accuracy and operational safety of the automatic parking system in complex and dynamic parking scenarios.

[0066] Optionally, based on the global reference path, the target grid map, and the vehicle's state data, local control commands for the vehicle are generated, including: smoothing the global reference path based on the vehicle state data to obtain a global reference trajectory; constructing a multi-dimensional evaluation model based on the global reference trajectory and the vehicle state data, wherein the multi-dimensional evaluation model is used to evaluate the tracking deviation and steering wheel control deviation of the vehicle when traveling along the global reference trajectory within a preset control cycle; constructing target constraints based on the target grid map and the vehicle's preset safety conditions; and solving the multi-dimensional evaluation model based on the target constraints to obtain local control commands.

[0067] The aforementioned smoothing process refers to the process by which the vehicle control unit, after receiving the global reference path from the cloud, continuously adjusts the geometric transitions between adjacent path points based on vehicle status data. This eliminates corner points, abrupt curvature changes, or discontinuous curvature variations caused by discrete sampling or path generation algorithms. Smoothing methods can include, but are not limited to, interpolation and low-pass filtering; the specific method must be determined based on actual needs. Smoothing can provide a physically realizable reference path with smoothly changing control inputs for model predictive control, avoiding severe steering wheel vibrations or system oscillations caused by path jumps, thus improving the stability and comfort of trajectory tracking.

[0068] The aforementioned global reference trajectory refers to the vehicle's desired driving path, generated after smoothing and consisting of a continuous sequence of spatial coordinate points. The global reference trajectory includes the lateral and longitudinal positions of each point, as well as the corresponding heading angle, in meters and radians, with a sampling interval consistent with the vehicle control cycle. The global reference trajectory can be used as a benchmark for local path planning, allowing the model predictive control module to calculate tracking errors and ensure that the vehicle continues to move towards the target parking space as a whole during local dynamic obstacle avoidance, maintaining directional consistency in the parking task.

[0069] The aforementioned multi-dimensional evaluation model can be defined as a mathematical expression in the form of a cost function, constructed to evaluate the comprehensive control performance of a vehicle traveling along a global reference trajectory within a preset control cycle, with quantitative indicators as the objective. The multi-dimensional evaluation model consists of two core dimensions: first, the spatial deviation between the vehicle's actual state and the global reference trajectory, termed tracking deviation; and second, the variation in steering wheel control commands between adjacent control cycles, termed steering wheel control deviation. This model combines these two dimensions into a single optimization objective through a weighted summation method. Its function is to guide the model's predictive control to minimize trajectory deviation while suppressing drastic fluctuations in control quantities, achieving a coordinated adjustment between trajectory tracking accuracy and handling smoothness, and avoiding frequent steering system movements, driving discomfort, or actuator overload caused by excessive pursuit of tracking accuracy.

[0070] The aforementioned preset control cycle refers to the discrete sampling and calculation cycle with a fixed time interval set in the vehicle control unit for the multi-dimensional evaluation model. The preset control cycle can be used to provide a unified time reference for trajectory prediction, cost calculation, and control command generation, ensuring that the system makes rolling predictions of the state within a finite future time period based only on the current vehicle state and environmental information in each cycle, and outputs only the control command for the current moment, thereby achieving low-latency, high-real-time closed-loop control.

[0071] The aforementioned tracking deviation refers to the spatial difference between the vehicle's actual driving state (including lateral position and heading angle) and the expected state at the corresponding predicted time on the global reference trajectory within a preset control period. Tracking deviation only reflects the geometric deviation between the vehicle's current moment and the desired path, and is used to quantify the vehicle's ability to follow the global path, serving as a direct control basis for achieving precise parking positioning.

[0072] The aforementioned steering wheel control deviation refers to the difference between the target front wheel steering angle output at the current moment and the previous moment within a preset control cycle. Steering wheel control deviation can reflect the degree of adjustment of steering action by the control system in continuous control cycles. It can serve as a constraint indicator for control smoothness in a multi-dimensional evaluation model, used to suppress high-frequency steering wheel vibration, avoid excessive wear of actuators or "jerk" (abrupt acceleration) perceived by the driver and passengers, thereby improving the comfort and system reliability of the parking process while ensuring safety.

[0073] The aforementioned preset safety conditions refer to rigid physical constraints pre-set to ensure that vehicles do not collide or cross boundaries during parking. Preset safety conditions may include, but are not limited to, preset safety distance thresholds, preset steering wheel control thresholds, and preset boundary thresholds; the specific preset safety conditions need to be determined based on real-time requirements. Preset safety conditions provide mandatory boundaries for model predictive control, ensuring that all predicted trajectories are physically legal and safety-verifiable, and are the core guarantee mechanism for achieving zero-collision parking.

[0074] The aforementioned target constraints refer to the complete set of hard constraints used to limit the solution space of model predictive control, which consists of preset safety conditions and the vehicle kinematic model. Target constraints may include, but are not limited to, dynamic constraints, obstacle avoidance constraints, control constraints, and boundary constraints; the specific target constraints need to be determined based on actual needs. Target constraints are used to ensure that the local control commands output by the Model Predictive Control (MPC) algorithm meet both the trajectory tracking objective and safety and physical feasibility requirements.

[0075] In one optional embodiment, the global reference path is first smoothed based on the vehicle's real-time status data to form a global reference trajectory that matches the vehicle's dynamic characteristics, thereby improving the path's adaptability to the vehicle's execution capabilities. Based on this, a multi-dimensional evaluation model is constructed. This model comprehensively quantifies the trajectory tracking deviation and steering wheel control fluctuations during the vehicle's journey along the reference trajectory, directly linking the target to trajectory following accuracy and control smoothness. Simultaneously, target constraints are established based on the target grid map and preset safety conditions, including dynamic constraints, obstacle avoidance constraints, control constraints, and boundary constraints, ensuring that obstacle avoidance safety is incorporated into the decision-making loop in real time. Finally, by jointly solving the multi-dimensional evaluation model under the target constraints, local control commands that balance trajectory tracking accuracy, control smoothness, and obstacle avoidance safety are output, achieving a synergistic improvement in global guidance and local response. This effectively solves the problem of simultaneously failing to consider path guidance, control stability, and environmental safety in dynamic parking scenarios, achieving a comprehensive improvement in the efficiency, smoothness, and safety of the parking process.

[0076] Optionally, the global reference path is smoothed based on vehicle state data to obtain a global reference trajectory, including: determining the position of the preview point on the global reference path based on the vehicle position data and vehicle attitude data in the vehicle state data; determining the target curvature based on the preview point position and the position of the rear axle center point of the vehicle; and smoothing the global reference path based on the target curvature to obtain a global reference trajectory.

[0077] The vehicle position data mentioned above can refer to the vehicle's two-dimensional coordinates in the global coordinate system. This vehicle position data can be used as a reference to determine the vehicle's relative position on the global reference path, for subsequent geometric calculations of preview points, and is the initial input for path tracking.

[0078] The aforementioned vehicle attitude data can refer to the heading angle parameter, which characterizes the vehicle's orientation in the global coordinate system and is calculated by the fusion of an onboard inertial measurement unit (IMU) or a steering angle sensor and wheel speed sensor. Vehicle attitude data can be used to represent the angle between the vehicle's longitudinal axis and a global coordinate system reference direction (such as true north or the lane centerline). Vehicle attitude data, together with vehicle position data, constitutes the vehicle's current pose state, used to calculate the relative angular relationship between the vehicle and the global reference path, and is a key input for determining the aiming direction and target curvature.

[0079] The aforementioned preview point position can refer to a path point located along the global reference path, extending a preset distance forward from the center point of the vehicle's rear axle along the tangent direction of the path. The coordinates of the preview point position are determined by geometric sequence interpolation of the global reference path. The preview point position can be used to provide target direction for pure tracking algorithms, enabling the vehicle to perform steering control based on the direction of the path ahead rather than the current position, thereby achieving natural tracking of curved paths and avoiding path deviation due to hysteresis response.

[0080] The aforementioned rear axle center point position refers to the precise spatial coordinates of the intersection of the left and right rear wheel axles in the global coordinate system within the vehicle's chassis geometry. It serves as the reference centroid in the vehicle's kinematic model. The rear axle center point position can be used as a benchmark reference point for calculating the geometric relationship between the vehicle and the global reference path. It determines the relative spatial relationship between the aiming point and the vehicle body, and is the sole geometric starting point for calculating target curvature and steering commands, ensuring the physical consistency of path tracking calculations.

[0081] The aforementioned target curvature refers to the ideal steering curvature value required for the vehicle to travel along the global reference path, calculated using a pure tracking algorithm based on the geometric relationship between the pre-aiming point and the vehicle's rear axle center point. The target curvature represents the degree of path curvature at that point. It serves as a bridge between the global reference path and local control, providing a basic curvature input for subsequent smoothing processes, enabling the vehicle to achieve stable, low-oscillation tracking of complex parking paths without relying on complex models.

[0082] In one optional embodiment, firstly, based on vehicle position and attitude data from the vehicle state data, the position of a pre-aiming point on the global reference path is accurately located. Then, the pre-aiming point position is combined with the position of the vehicle's rear axle center point to calculate a target curvature that conforms to the vehicle's kinematic characteristics. Using this target, the original global reference path is continuously smoothed to generate a physically feasible and abrupt global reference trajectory. This effectively avoids trajectory tracking errors and control system oscillations caused by discontinuous or physically unrealizable path curvature, providing a high-precision and highly stable input foundation for the subsequent construction of multi-dimensional evaluation models and target constraints. This, in turn, improves the quality of local control command generation and the safety, smoothness, and environmental adaptability of the parking process.

[0083] Optionally, based on the target grid map and the vehicle's preset safety conditions, target constraints are constructed, including: constructing dynamic constraints based on the vehicle's kinematic equations; constructing obstacle avoidance constraints based on the target grid map and preset safety distance thresholds in the preset safety conditions; constructing control constraints based on preset steering wheel control thresholds in the preset safety conditions; constructing boundary constraints based on preset boundary thresholds in the preset safety conditions; and determining target constraints based on dynamic constraints, obstacle avoidance constraints, control constraints, and / or boundary constraints. Preferably, based on the target grid map and preset safety distance thresholds in the preset safety conditions, obstacle avoidance constraints are constructed, including: determining the passable area at the current moment based on the target grid map; determining the interval distance based on the location data of the passable area and the location data of the obstacle grid; and determining an interval distance greater than the preset safety distance threshold as an obstacle avoidance constraint.

[0084] The aforementioned vehicle kinematic equations can be considered nonlinear mathematical models describing the deterministic relationship between the motion state (lateral position, longitudinal position, and heading angle) of the vehicle's rear axle center point in a two-dimensional plane and the control inputs (front wheel steering angle and longitudinal velocity). These equations provide a predictive basis for model predictive control (MPC) of the vehicle's future state, ensuring that the generated control commands are physically executable and avoiding unrealistic trajectories (such as sideslip or rollover). Their function is to establish a calculable mapping between the control quantities (steering wheel angle) and the vehicle's future trajectory, serving as the mathematical basis for constructing dynamic constraints.

[0085] The aforementioned dynamic constraints refer to the physically feasible boundary conditions derived from the vehicle's kinematic equations, limiting the vehicle's performance in all future predicted time domains. Dynamic constraints may include, but are not limited to, vehicle speed constraints, front wheel steering angle constraints, and steering rate constraints; specific dynamic constraints need to be determined based on actual requirements. Dynamic constraints ensure that the control sequences generated by Model Predictive Control (MPC) are physically realizable, preventing outputs from exceeding the vehicle's actuator capabilities (e.g., the steering wheel cannot be turned instantaneously to its limit) or leading to unrealistic trajectories (e.g., zero-radius turns). Dynamic constraints embed the vehicle's inherent motion capabilities as hard constraints into the optimization process, ensuring the executability of control commands.

[0086] The aforementioned preset safety distance threshold refers to a fixed distance value pre-set to ensure that the vehicle maintains a minimum physical distance from any obstacle (static or dynamic) during parking. The preset safety distance threshold may include, but is not limited to, 0.25 meters, 0.3 meters, 0.35 meters, etc., and the specific preset safety distance threshold needs to be determined based on the actual parking lot type and safety requirements. The preset safety distance threshold can serve as a benchmark for judging obstacle avoidance constraints, ensuring that the vehicle's outline maintains a safe margin with the obstacle at every moment of the predicted trajectory, avoiding collisions caused by perception errors, vehicle body sway, or control delays.

[0087] The aforementioned obstacle avoidance constraints can refer to a set of mathematical inequalities constructed for model predictive control based on the target grid map and a preset safety distance threshold, requiring that the Euclidean distance between all predicted points on the vehicle's future trajectory and the center of the obstacle grid must not be less than this threshold. Obstacle avoidance constraints can be used to transform environmental perception results into computable geometric constraints, enabling trajectory generation to possess environmental perception response capabilities.

[0088] The aforementioned preset steering wheel control threshold refers to the maximum permissible change in steering angle set to limit the variation of steering wheel control commands within adjacent control cycles. The preset steering wheel control threshold can be used to suppress drastic jumps in control inputs, prevent high-frequency steering wheel vibration caused by path disturbances or perceived noise, improve handling smoothness, reduce actuator wear, and meet driving comfort requirements.

[0089] The aforementioned control constraints refer to a set of mathematical constraints, consisting of preset steering wheel control thresholds and other control input limitations, that restrict the variation of MPC variables (steering wheel angle and its rate of change) within the feasible region. These control constraints ensure that the control sequence output by the MPC remains within the capabilities of the vehicle's actuators, preventing system instability or hardware overload caused by excessive control.

[0090] The aforementioned preset boundary threshold refers to the minimum permissible distance between the vehicle's outer contour and the boundary, set to prevent vehicles from leaving the drivable area of ​​a parking lot (such as the road edge, guardrail, or parking space boundary). The preset boundary threshold ensures that vehicles remain within the legal driving area during parking, preventing collisions, jams, or public safety risks caused by exceeding the boundary.

[0091] The aforementioned boundary constraints can refer to a set of geometric constraints based on a preset boundary threshold, requiring that the shortest distance between the vehicle's outer contour point and the boundary of the drivable area of ​​the parking lot at all times within the prediction time domain must not be less than that threshold. By incorporating the physical boundaries of the parking lot (such as curbs, barriers, and walls) as static obstacles into the MPC optimization framework, it ensures that vehicles do not cross the boundary when parking in narrow spaces, thereby extending obstacle avoidance capabilities to the environmental structural boundaries and achieving "all-scenario safety".

[0092] The aforementioned passable area refers to the connected region formed by all grid cells marked "0" in the target grid map, representing the ground space through which the rear axle center of the vehicle can safely pass. The passable area serves as the legal spatial basis for MPC path planning and is the environmental input source for constructing obstacle avoidance and boundary constraints. Its function is to abstract the complex 3D parking environment into a 2D discrete map, achieving efficient environment modeling and constraint generation.

[0093] The aforementioned interval distance can refer to the Euclidean distance between any point on the vehicle's predicted trajectory (such as the rear axle center) in the target grid map and the center point of the nearest obstacle grid. The interval distance can be used as a quantized input for obstacle avoidance constraints to determine whether the current trajectory point meets the minimum safe distance requirement.

[0094] In one optional embodiment, during the optimization process of Model Predictive Control (MPC), the onboard control unit establishes dynamic constraints based on the vehicle's kinematic equations to ensure that the vehicle's state changes in the future prediction time domain conform to its physical motion characteristics and avoid unrealistic trajectories. Simultaneously, by combining the obstacle locations marked in the target grid map with a preset safety distance threshold, obstacle avoidance constraints are calculated to ensure that the vehicle does not collide in the dynamic environment. Furthermore, based on a preset steering wheel control threshold, the variation in steering wheel angle within adjacent control cycles is limited to form control constraints, suppressing steering jitter and improving comfort. Furthermore, based on preset boundary thresholds, structural boundaries such as parking lot fences and road edges are considered insurmountable areas, constructing boundary constraints to prevent vehicles from crossing boundaries and scraping. Finally, the aforementioned dynamic constraints, obstacle avoidance constraints, control constraints, and boundary constraints are integrated into target constraints, serving as a set of hard constraints for the MPC online optimization problem, ensuring that the generation of each control command is completed within a joint constraint space that is physically feasible, safe and compliant, and smooth in operation.

[0095] The above process significantly improves the reliability and safety of parking in complex, dynamic, and confined environments by unifying environmental perception, vehicle dynamics, control execution, and safety boundaries into a deterministic, non-learning, and analytical multi-dimensional hard constraint system within the automated parking system. It achieves a "zero-collision" design through mandatory constraints rather than probabilistic avoidance, drastically reducing the rate of manual intervention. Furthermore, all constraints are constructed based on discrete grid maps and fixed thresholds, requiring no high-performance computing power and possessing strong engineering feasibility and cross-vehicle compatibility. This provides a standardized and replicable control paradigm for highly robust automated parking under a cloud-vehicle-park collaborative architecture.

[0096] In one optional embodiment, the passable area at the current moment is determined based on the target grid map, i.e., all cells with a grid value of 0 are extracted, representing the unobstructed area through which the center of the vehicle's rear axle can safely pass. Subsequently, the Euclidean distance from each discrete point on the vehicle's predicted trajectory to the center point of the nearest obstacle grid (with a value of 1 or 2) is calculated in real time to obtain the interval distance between each trajectory point and the obstacle. Finally, all interval distances are required to be greater than a preset safety distance threshold as a hard constraint condition, which is embedded in the optimization framework of model predictive control (MPC) to ensure that the vehicle's trajectory maintains a physical safety margin with respect to obstacles at any predicted moment. The above process achieves direct mapping between environmental perception and control constraints through a discrete grid map, avoiding reliance on complex sensor fusion or probabilistic models, and has the advantages of deterministic calculation, real-time response, and simple implementation.

[0097] Optionally, the multi-dimensional evaluation model is solved based on the target constraints to obtain local control commands, including: solving the multi-dimensional evaluation model to obtain a sequence of control parameters, wherein the sequence of control parameters satisfies the target constraints; and determining the control parameters of the first control cycle in the sequence of control parameters as local control commands.

[0098] The aforementioned control parameter sequence refers to the set of several consecutive control actions calculated by the algorithm in each control cycle within the prediction time domain of Model Predictive Control (MPC) to achieve target trajectory tracking and dynamic obstacle avoidance. The control parameter sequence may include, but is not limited to, the steering wheel angle increment sequence; the specific control parameter sequence needs to be determined based on the actual control objective. The control parameter sequence can be used to transform the vehicle behavior in the future prediction time domain into a set of executable control commands, enabling the system to possess forward-looking obstacle avoidance and trajectory smoothing capabilities, rather than merely responding to the current state.

[0099] The first control cycle mentioned above can refer to the time period during which the first control action, generated earliest in the control parameter sequence and corresponding to the current moment, is applied. The first control cycle can serve as the execution entry point for the MPC "rolling optimization" mechanism. The control parameters of the first control cycle are the only instructions actually issued to the vehicle's underlying actuators; the remaining sequence elements are only used for prediction and optimization. The role of the first control cycle is to implement a closed-loop control strategy of "execute only one step, then replan the next," ensuring predictive foresight while responding to environmental changes in real time, thus avoiding control failure due to model mismatch or perception delay.

[0100] In one optional embodiment, the multi-dimensional evaluation model is solved online under target constraints to generate a sequence of control parameters that meets both safety and control continuity objectives. Finally, the control parameters of the first control cycle in the control parameter sequence are output as local control commands in real time, thereby achieving coordinated adjustment of trajectory tracking accuracy and control smoothness in a dynamic parking environment. This effectively suppresses control jitter and sudden steering caused by environmental changes or perception delays, and significantly improves stability and safety during the parking process.

[0101] Optionally, after controlling the vehicle to perform a parking operation based on local control commands, the method further includes: acquiring the rear axle center point position data of the vehicle; determining deviation data based on the rear axle center point position data and preset vehicle center position data; determining the vehicle status as successfully parked when the deviation data is less than the preset deviation and the vehicle speed is zero, and acquiring the target vehicle pose data; and sending the target vehicle pose data and the parking success signal to the cloud scheduling platform.

[0102] The aforementioned rear axle center point position data refers to the real-time position coordinates of the vehicle's rear axle center (i.e., the midpoint of the left and right rear wheel axles) in a two-dimensional plane coordinate system. This rear axle center point position data can be used as a benchmark reference point for evaluating parking accuracy, and is used to calculate the relative positional deviation between the vehicle and the target parking space.

[0103] The aforementioned preset vehicle center position data can refer to the coordinates of the ideal target position where the rear axle center of the vehicle should be parked, predefined in the cloud-based scheduling platform or parking lot map. This preset vehicle center position data can serve as a reference benchmark for successful parking determination, comparing it with the actual rear axle center position to quantify the "target-actual" deviation. It transforms the abstract task of "parking in a space" into a quantifiable geometric constraint problem, ensuring that parking acceptance has objective, repeatable, and auditable criteria.

[0104] The aforementioned deviation data refers to the spatial difference between the rear axle center point position data and the preset vehicle center position data. Deviation data may include, but is not limited to, lateral deviation and longitudinal deviation; the specific deviation data needs to be determined based on actual requirements. Deviation data is used to quantify the degree of deviation between the vehicle's final parking position and the target ideal position, and is a core input for determining whether parking is successful. Its function is to achieve "quantifiable position accuracy and programmable judgment criteria," avoiding subjective or visual judgment and ensuring the system's automated closed-loop operation.

[0105] The aforementioned preset deviations refer to the maximum permissible lateral and longitudinal positional deviation thresholds pre-set to determine successful parking. Preset deviations may include, but are not limited to, lateral and longitudinal preset deviations; the specific preset deviations need to be determined based on actual needs. Preset deviations are used to ensure that the vehicle's final parking position is within the tolerance and user-acceptable range.

[0106] The aforementioned target vehicle position data refers to the complete spatial state information of the vehicle after successful parking. This target vehicle position data can be used to report the final parking status of the vehicle to the cloud-based dispatch platform, enabling real-time updates of parking lot resources and providing a basis for accurate subsequent vehicle dispatching.

[0107] The aforementioned parking success signal can refer to a status confirmation command automatically generated and sent by the vehicle control unit after meeting the dual conditions of "deviation data < preset deviation" and "vehicle speed is zero". The parking success signal can be used as the end trigger of the system closed loop to notify the cloud to release resources, update parking space status, push user notifications, and trigger extended services such as charging / cleaning.

[0108] In one optional embodiment, after performing a parking operation, the vehicle control unit acquires the rear axle center point position data of the vehicle and compares it with preset vehicle center position data to calculate the deviation data. Then, if the deviation data is less than the preset deviation and the vehicle speed is zero, it determines that the vehicle has parked in the target parking space, thereby avoiding misidentification caused by sensor misjudgment or temporary vibration. After confirming successful parking, it actively acquires and uploads the target vehicle pose data and parking success signal to the cloud scheduling platform, realizing closed-loop feedback of parking results. This allows the cloud to monitor the vehicle's parking status in real time and use it for subsequent scheduling adjustments. This effectively solves the problems of global scheduling disconnect and resource waste caused by the lack of reliable judgment criteria for successful parking and the inability to transmit accurate pose information back to the cloud in the prior art, thus improving the coordination, safety, and intelligence level of the automatic parking system in dynamic environments.

[0109] According to an embodiment of this application, a vehicle parking control method is also provided, applied to a cloud-based dispatching platform. Figure 2 This is a flowchart of another vehicle parking control method according to an embodiment of this application, such as... Figure 2 As shown, the device includes the following steps:

[0110] Step S202: When the vehicle enters the target communication range in the parking lot, the vehicle status data uploaded by the vehicle control unit is received, and environmental data in the parking lot is obtained.

[0111] Step S204: Determine the global reference path based on vehicle status data and environmental data; generate an environmental raster map corresponding to the global reference path.

[0112] Step S206: The global reference path and environmental grid map are sent to the vehicle control unit.

[0113] The aforementioned target communication range refers to the physical space area within the parking lot covered by wireless communication infrastructure, ensuring low-latency, high-reliability, two-way data interaction between vehicles and the cloud-based dispatch platform. This target communication range can serve as the trigger boundary for cloud-based dispatch services, ensuring that the system only initiates global path planning and resource allocation processes after a vehicle enters the effective parking service area, thus avoiding invalid communication, resource waste, and computational redundancy. Its function is to achieve "on-demand service and precise triggering," ensuring system response efficiency and energy efficiency.

[0114] The aforementioned vehicle status data refers to a structured data set describing the vehicle's characteristics and real-time motion status, actively uploaded by the onboard control unit to the cloud-based scheduling platform after entering the target communication range. Vehicle status data may include, but is not limited to, the vehicle's unique identification number, vehicle dimensions (length, width, wheelbase, track width), maximum turning angle, minimum turning radius, current positioning coordinates (x, y), current heading angle, current speed, and electronic stability system status. Specific vehicle status parameters need to be determined based on actual requirements. Vehicle status data can be used as input parameters for cloud-based planning algorithms to evaluate the vehicle's accessibility and parking feasibility in a parking environment.

[0115] The aforementioned environmental data refers to multi-source heterogeneous sensing information describing the spatial structure and dynamic elements of the parking lot, collected and uploaded in real time by IoT sensing devices deployed within the parking lot (such as high-definition cameras, millimeter-wave radar, ultrasonic arrays, geomagnetic sensors, lidar, and parking space occupancy detectors). Environmental data may include, but is not limited to, coordinates and status of vacant parking spaces, locations of static obstacles (pillars, walls, barriers), predicted trajectories of dynamic obstacles (pedestrians, moving vehicles), lane congestion status, ground marking information, lighting and weather conditions, etc. Specific environmental data needs to be determined based on actual needs. Environmental data can be used as environmental semantic input for cloud-based global path planning, for constructing a accessibility map and providing a basis for obstacle avoidance decisions.

[0116] In one optional embodiment, when a vehicle enters the target communication range within the parking lot, the onboard control unit automatically triggers a communication handshake protocol and uploads vehicle status data to the cloud dispatch platform. This data includes static vehicle parameters (such as wheelbase and maximum turning angle), dynamic pose (current position, heading, and speed), and system self-check status. Simultaneously, the cloud dispatch platform receives environmental data from the parking lot's IoT sensing network, covering static obstacle distribution, vacant parking space coordinates, dynamic obstacle trajectory prediction, and lane access status. Next, based on the vehicle status data and environmental data, the cloud dispatch platform generates a global reference path from the vehicle's current position to the target parking space and simultaneously generates a corresponding environmental grid map, where each grid cell is labeled as 0 (passable), 1 (static obstacle), or 2 (dynamic risk). Finally, the cloud encapsulates the global reference path and high-precision grid map into a structured data packet and sends it to the onboard control unit via the 5G network with low latency, serving as an environmental prior and reference benchmark for its subsequent local path planning. The above process, through precise triggering within the target communication range, structured uploading of vehicle status data, and multi-source environmental data fusion modeling, constructs the core perception-decision closed loop of the cloud-vehicle-park collaborative parking system.

[0117] Optionally, a global reference path is determined based on vehicle status data and environmental data, including: constructing a global accessible area map of the parking lot based on vehicle status data and environmental data; performing accessibility analysis and parking feasibility analysis on all vacant parking spaces based on the global accessible area map and vehicle status data to obtain a priority list of available parking spaces; determining the first parking space in the priority list of available parking spaces as the target parking space; and generating a global reference path based on the current location data in the vehicle status data, the location data corresponding to the target parking space, and the principle of the shortest global path.

[0118] The aforementioned globally accessible area map refers to a global environmental semantic map constructed by a cloud-based dispatch platform based on parking lot environmental data, represented in the form of a rasterized two-dimensional matrix. In this map, 0 represents accessible, 1 represents static obstacles (impassable), and 2 represents dynamic risk zones, i.e., the predicted trajectory envelope area of ​​moving obstacles (such as pedestrians and shuttle vehicles). The globally accessible area map can serve as the sole environmental constraint basis for global path planning, eliminating areas that vehicles cannot pass through and ensuring the physical feasibility of the generated paths. It overcomes the shortcomings of onboard sensors, such as short line-of-sight, low resolution, and susceptibility to interference, enabling a comprehensive, "one-stop" macro-level path decision-making system.

[0119] The aforementioned vacant parking spaces refer to standardized parking spaces that are marked as available for parking and are currently unoccupied on the global accessible area map. Vacant parking spaces serve as a set of candidate destinations for route planning, forming the basic resource pool for the system to assign parking targets to vehicles.

[0120] The aforementioned accessibility analysis can refer to determining whether a vehicle can reach the entrance or parking start point of a certain vacant parking space from its current location based on vehicle status data (minimum turning radius, wheelbase, maximum turning angle) and a global passable area map, using a geometric path search algorithm, and ensuring that all points on the path satisfy vehicle dynamics constraints and environmental obstacle constraints.

[0121] The above-mentioned parking feasibility analysis can refer to, on the premise that the parking space is accessible, further determining whether the vehicle can complete the complete parking action from the starting point to the target position without exceeding the parking space boundary and without colliding with surrounding obstacles, based on the vehicle size (length, width, wheelbase) and the geometric parameters of the target parking space, through parking kinematics simulation.

[0122] The aforementioned parking space priority list refers to an ordered candidate list generated after accessibility and parking feasibility analyses are completed, based on a comprehensive ranking of all qualified vacant parking spaces according to preset scoring rules. This priority list can be used as input for cloud-based route planning, determining which target parking space a vehicle should be assigned to. Its purpose is to prioritize "better parking spaces over the nearest ones," improving the overall parking experience for users and the utilization rate of parking lot resources.

[0123] The aforementioned principle of global path shortestness refers to determining the target parking space, starting from the vehicle's current position and ending at the target parking space entrance or parking start point, and using a path planning algorithm to find the continuous collision-free trajectory with the minimum path cost within the globally accessible area map. This principle can be used as the core objective of global path generation, ensuring that the system prioritizes the shortest path while satisfying all constraints, thereby reducing vehicle travel time, lowering energy consumption, and improving the overall throughput of the parking lot.

[0124] In one optional embodiment, a global accessible area map is first constructed based on vehicle size and motion performance parameters (such as minimum turning radius and maximum turning angle) from vehicle status data and environmental data. Next, accessibility analysis is used to determine whether each available parking space can be reached from the vehicle's current location. Then, a parking feasibility analysis is performed on the accessible parking spaces to verify whether the vehicle can complete the parking maneuver without crossing boundaries or colliding. After double screening, a priority list of available parking spaces is generated for all eligible spaces according to preset scoring rules (including path length, parking space type, environmental interference, and distance to the exit). Subsequently, the first parking space in the list is selected as the target parking space, and based on the principle of the shortest global path, a path from the vehicle's current location to the starting point of the target parking space is searched in the global accessible area map to generate a global reference path. This path is a coordinate sequence containing several discrete path points, which will subsequently be compared with the corresponding grid points. Figure 1The data is simultaneously sent to the vehicle-mounted terminal for trajectory tracking and local optimization. This overcomes the shortcomings of traditional systems, such as random target parking space selection and lack of joint decision-making based on vehicle adaptability and environmental constraints. It enables intelligent selection of a unique target from a large number of available parking spaces and matching the most reliable path, significantly improving the success rate and safety of automatic parking in complex dynamic environments.

[0125] In one alternative embodiment, Figure 3 This is a flowchart of an optional vehicle parking control method according to an embodiment of this application, such as... Figure 3 As shown, the method includes the following steps: Step 1, the vehicle enters the monitoring range, the cloud allocates parking spaces and issues a global path; Step 2, the vehicle sensors and cloud maps are fused to generate a real-time occupancy grid map; Step 3, a pure tracking algorithm is used to track the cloud-recommended path; Step 4, an optimization problem is constructed based on model predictive control, and control commands are solved in real-time; Step 5, the vehicle executes the control commands, dynamically avoids obstacles and drives into the target parking space; Step 6, the status is reported after parking is completed, and if there is an anomaly, minimum risk parking is triggered.

[0126] Specifically, in step one, once the vehicle enters the monitoring range, the cloud assigns a parking space and issues a global path. That is, when the vehicle enters the parking lot's communication coverage area, the onboard control unit sends vehicle status data, including vehicle size, performance parameters (minimum turning radius, maximum turning angle), real-time positioning, and attitude information, to the cloud dispatch platform via 5G communication.

[0127] After receiving the vehicle status, the cloud platform integrates the environmental information uploaded by the parking lot IoT devices (including the distribution of available parking spaces, the location of dynamic obstacles, lane congestion, etc.) to build a global passable map.

[0128] Based on vehicle parameters and the principle of the shortest global path, the cloud-based scheduling platform generates a priority list of available parking spaces, selects a target parking space, and calculates the globally recommended path from the vehicle's current location to the target parking space. Subsequently, the cloud sends the parking space coordinates, the global path, and the corresponding grid map to the vehicle control unit.

[0129] Figure 4 This is a schematic diagram of a grid map according to an embodiment of this application, such as... Figure 4 As shown, the raster map here refers to a parking lot map in grid format, where each grid has a fixed size, representing a real ground unit at a unit size (usually 0.01m²). Figure 4As shown in the legend, diagonal shading represents static obstacles, horizontal shading represents dynamic obstacles, white unshaded areas represent passable areas, and the leftmost black frame represents the target parking space. Grid cell values ​​represent different meanings: 0 represents a passable area (no obstacles and curvature radius > 5m), 1 represents static obstacles (static objects such as walls / pillars), and 2 represents a dynamic risk zone (predicted trajectory envelope of moving objects, with a valid timestamp of 200ms).

[0130] At this point, the vehicle's infotainment system and the mobile application (APP) receive the parking list from the cloud. Users can then select whether to park on either the infotainment system or the mobile app. If the driver clicks to enter parking mode, the vehicle's Automatic Parking Assist System (APA) performs a self-check. If the self-check passes, the system enters parking space selection mode and then route planning. If the driver clicks to exit or the self-check fails, the system exits parking mode, and vehicle control is returned to the driver.

[0131] Step two involves fusing onboard sensor data with cloud-based maps to generate a real-time occupancy grid map. This means combining real-time obstacle information detected by the vehicle's sensors, a bird's-eye view (BEV), with a grid map downloaded from the cloud. Merge the data to generate a real-time occupied grid map. The fusion rules are as follows: If a grid cell in the cloud map is a static obstacle (value = 1), it is directly retained; for dynamic obstacles, the grid value is updated based on the vehicle's perception results; finally, a real-time map for path planning is generated. ,in, This indicates that the area is safe and feasible.

[0132] Step 3: Use a pure tracking algorithm to track the cloud-recommended path. That is, after receiving the recommended path from the cloud, the vehicle uses the Pure Pursuit algorithm for global path tracking. Based on the vehicle's current position and attitude, the algorithm calculates the positions of preview points on the reference path and their preview distances. Defined as:

[0133] ;

[0134] in, Based on the vehicle speed, the system calculates the target curvature according to the geometric relationship between the vehicle's rear axle center and the pre-aiming point. :

[0135] ;

[0136] Where α is the angle between the current heading and the tangent of the target point. The target curvature output by the Pure Pursuit algorithm is used to generate a smooth reference trajectory for subsequent use by the local planner.

[0137] Step four: Construct an optimization problem based on model predictive control and solve control commands in real time. That is, under the guidance of the global reference trajectory, the vehicle system performs local path planning and dynamic obstacle avoidance based on the model predictive control (MPC) method.

[0138] (a) Problem Formulation for Optimization:

[0139] The system is within a finite prediction time domain ( Predict the future state of the vehicle in 20 steps (0.05s each), and construct a cost function (corresponding to the multi-dimensional evaluation model above) with the objective of minimizing trajectory deviation and control variation:

[0140] ;

[0141] Where k is the time step index (step size is 0.05s). =20 represents the prediction time domain, indicating that the controller has predicted forward by 20 control cycles (20 The state of 0.05s = 1s. This indicates the predicted lateral position of the vehicle at time step k. This indicates the lateral position on the reference trajectory at time step k. This represents the predicted heading angle of the vehicle at time step k. This represents the heading angle of the reference trajectory at time step k. This represents the time domain of the control action that needs to be optimized. The control increment, i.e., the change in steering wheel angle between two adjacent cycles, Q, R, and S are positive definite weight matrices, which are used to adjust the optimization weights for lateral deviation, heading deviation, and control change rate, respectively.

[0142] (ii) Constraints:

[0143] Dynamic constraints: The predicted trajectory must conform to the vehicle's kinematics model to ensure that the generated path is physically feasible;

[0144] Control constraints: Steering wheel angle and its rate of change are limited;

[0145] Obstacle avoidance constraints: Every point on the predicted trajectory must be located within a passable area (grid value 0) in the real-time grid map, and its distance from any obstacle grid (value ≠ 0) must be greater than a set safety threshold (e.g., 0.3m). This constraint is achieved by calculating the Euclidean distance between the trajectory point and the nearest obstacle and applying an inequality constraint.

[0146] Boundary constraints: The vehicle profile must not exceed the boundary of the drivable area (such as road edges, guardrails, etc.) throughout the entire prediction time domain.

[0147] (III) Online Solving and Scrolling Optimization:

[0148] In each control cycle (0.05s), the current vehicle state is used as the initial condition, and the above problem is solved online based on the latest environmental perception information (real-time grid map). A better control sequence is output for the future finite time domain. Only the first element of the control sequence is taken. The actual control command (such as the target front wheel steering angle) at the current moment is sent to the vehicle's underlying actuator, and the rest is re-optimized and updated in the next control cycle (50ms).

[0149] Step five: The vehicle executes control commands, dynamically avoids obstacles, and drives into the target parking space. Specifically, the vehicle executes the front wheel steering angle and longitudinal acceleration control commands output by the MPC module. The controller works in conjunction with the underlying actuators to dynamically avoid obstacles while maintaining trajectory stability, and continuously updates the local path based on real-time environmental changes. As the vehicle gradually enters the target parking space, the control system automatically decelerates to zero speed.

[0150] Step six: After parking, report the status. If an anomaly is detected, trigger minimum risk parking. That is, when the system determines the lateral deviation between the vehicle's rear axle center and the target parking space center... Longitudinal deviation and heading angle deviation All are less than the set tolerance ( , , When the vehicle speed is zero, parking is considered successful. The vehicle automatically shifts into P gear (parking gear), engages the electronic parking brake (EPB), and turns off the "parking" indicator light.

[0151] The vehicle system sends a "parking successful" signal and the final vehicle position to the cloud platform via 5G communication. After receiving the signal, the cloud platform updates the status of the parking space to "occupied" on the global map and releases the computing and communication resources allocated for this task. The mobile APP receives the success notification and displays the "parking completed" interface.

[0152] If an unrecoverable malfunction occurs at any stage (e.g., multiple consecutive planning failures by the planner, loss of key sensor signals, vehicle drive-by-wire system failure, or communication link interruption), the APA (Automatic Parking Assist) relinquishes vehicle control. The Integrated Power Brake System (IPB) immediately applies maximum deceleration until the vehicle speed reaches zero, the Electronic Parking Brake (EPB) engages, the Body Domain Controller (BDM) illuminates warning lights, and the driver is alerted to take over the vehicle via the instrument cluster, center console text display, and voice announcement. The values ​​in the above steps are for illustrative purposes only; specific values ​​need to be determined based on actual needs and are not limited here.

[0153] In one alternative embodiment, Figure 5 This is a schematic diagram of dynamic path planning according to an embodiment of this application, such as... Figure 5 As shown, the dashed line represents the initial path, the solid line represents the replanned path, and the vehicle outlines on the top and bottom sides of the figure represent the statically parked vehicles that have been parked on both sides of the target parking space. Their positions are fixed and constitute the boundary constraints in the parking environment. "P" is the target parking space assigned by the cloud scheduling platform and is marked at the center of the parking space.

[0154] The initial path is a theoretical path generated by the vehicle system based on a global grid map (containing only static obstacles and parking space information) sent from the cloud. This path directly connects the vehicle's current position to the target parking space P, without considering real-time dynamic environmental changes. As shown in the figure, this path crosses the predicted trajectory envelope of a moving pedestrian (labeled "Dynamic Obstacle: Pedestrian"), indicating a collision risk if this path is followed, classifying it as a high-risk trajectory.

[0155] The replanning path is the trajectory generated online in the MPC module based on a real-time fused grid map using the parking control method proposed in this application. For example... Figure 5 As shown by the solid line, the path actively shifts to the right within the prediction time domain, detouring to a safe distance outside the dynamic pedestrian trajectory envelope while maintaining a continuous heading and smooth curvature, ultimately accurately guiding the vehicle to the target parking space P. This comparison clearly demonstrates that traditional path planning methods relying solely on static maps struggle to handle sudden dynamic obstacles. In contrast, this application, through a fusion mechanism of MPC and real-time multi-source perception, achieves closed-loop control of "perception-prediction-replanning-execution," significantly improving parking safety and reliability in complex urban parking environments.

[0156] Figure 6 This is a schematic diagram of the structure of a vehicle parking control system according to an embodiment of this application, as shown below. Figure 6As shown, the system mainly includes: a cloud computing center, a vehicle-side computing center, field sensors, and vehicle-side sensors. Among them, the field sensors include a top-mounted camera and a geomagnetic parking space sensor, and the vehicle-side sensors include a chassis wheel speed / IMU / GPS module, a forward / surround view camera, and ultrasonic radar.

[0157] The global map and real-time parking space information collected by the top-mounted camera and geomagnetic parking space sensor, as well as the real-time vehicle pose collected by the chassis wheel speed / IMU / Global Positioning System (GPS) module, are uploaded to the cloud computing center. The cloud computing center calculates the global map / available parking spaces and the recommended optimal global path for the parking spaces, and sends the global map / available parking spaces and the recommended optimal global path for the parking spaces to the vehicle-side computing center. The vehicle-side computing center combines the real-time vehicle pose collected by the chassis wheel speed / IMU / GPS module with the BEV bird's-eye view and obstacle information collected by the front / surround view camera and ultrasonic radar to perform fusion calculations to obtain fused mapping and vehicle positioning, and performs MPC path optimization.

[0158] This system architecture realizes a three-level collaborative mechanism of cloud-based global scheduling, site-side full-domain perception, and vehicle-side real-time decision-making and execution. It breaks through the limitations of traditional single-vehicle intelligent systems in terms of perception coverage, computing power, and environmental cognition, and significantly improves the success rate, safety, and robustness of automatic parking in complex, dynamic, and unstructured parking environments.

[0159] 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.

[0160] According to an embodiment of this application, a parking control device for a vehicle is provided. It should be noted that this device can be used to execute the aforementioned parking control method for a vehicle. The specific implementation method is the same as the aforementioned parking control method for a vehicle, and will not be repeated here.

[0161] Figure 7 This is a schematic diagram of a vehicle parking control device according to an embodiment of this application, applied to an on-board control unit, which is deployed in the vehicle, such as... Figure 7 As shown, the device includes: a first receiving module 702, a processing module 704, a first generating module 706, and a control module 708.

[0162] The first receiving module 702 is used to receive a global reference path and an environmental grid map sent by the cloud scheduling platform. The global reference path is used to represent a reference path guiding the vehicle from its current position to the target parking space, and the environmental grid map is used to represent a grid map characterizing the spatial distribution of obstacles in the parking lot. The processing module 704 is used to fuse the environmental grid map and the obstacle data perceived by the vehicle to generate a target grid map. The first generating module 706 is used to generate local control commands for the vehicle based on the global reference path, the target grid map, and the vehicle's vehicle status data. The control module 708 is used to control the vehicle to perform parking operations based on the local control commands.

[0163] Optionally, the processing module is further configured to determine the obstacle types corresponding to multiple grids in the environmental grid map; update the environmental grid map based on the obstacle types and obstacle data corresponding to the multiple grids to generate a target grid map; preferably, the processing module is further configured to determine the grid value corresponding to the grid as a preset grid value when the obstacle type in any grid is a static obstacle; update the grid value corresponding to the grid based on the obstacle data when the obstacle type in any grid is a dynamic obstacle; and generate a target grid map based on the grid values ​​corresponding to multiple grids.

[0164] Optionally, the first generation module is further configured to smooth the global reference path based on vehicle state data to obtain a global reference trajectory; construct a multi-dimensional evaluation model based on the global reference trajectory and vehicle state data, wherein the multi-dimensional evaluation model is used to evaluate the tracking deviation and steering wheel control deviation of the vehicle when it travels along the global reference trajectory within a preset control cycle; construct target constraints based on the target grid map and the vehicle's preset safety conditions; and solve the multi-dimensional evaluation model based on the target constraints to obtain local control commands.

[0165] Optionally, the first generation module is further configured to: determine the position of the preview point on the global reference path based on the vehicle position data and vehicle attitude data in the vehicle state data; determine the target curvature based on the preview point position and the position of the rear axle center point of the vehicle; and smooth the global reference path based on the target curvature to obtain the global reference trajectory.

[0166] Optionally, the first generation module is further configured to construct dynamic constraints based on the vehicle's kinematic equations; construct obstacle avoidance constraints based on the target grid map and a preset safety distance threshold in the preset safety conditions; construct control constraints based on a preset steering wheel control threshold in the preset safety conditions; construct boundary constraints based on a preset boundary threshold in the preset safety conditions; and determine target constraints based on the dynamic constraints, obstacle avoidance constraints, control constraints, and / or boundary constraints. Preferably, the first generation module is further configured to determine the passable area at the current moment based on the target grid map; determine the interval distance based on the location data of the passable area and the location data of the obstacle grid; and determine the interval distance greater than the preset safety distance threshold as an obstacle avoidance constraint.

[0167] Optionally, the first generation module is also used to solve the multi-dimensional evaluation model to obtain a sequence of control parameters, wherein the sequence of control parameters satisfies the target constraint conditions; and to determine the control parameters of the first control cycle in the sequence of control parameters as local control commands.

[0168] Optionally, after controlling the vehicle to perform a parking operation based on local control commands, the device is also used to acquire the rear axle center point position data of the vehicle; determine the deviation data based on the rear axle center point position data and the preset vehicle center position data; determine the vehicle status as parking successful when the deviation data is less than the preset deviation and the vehicle speed is zero, and acquire the target vehicle pose data; and send the target vehicle pose data and parking success signal to the cloud dispatch platform.

[0169] According to an embodiment of this application, a parking control device for a vehicle is also provided. It should be noted that this device can be used to execute the aforementioned parking control method for a vehicle. The specific implementation method is the same as the aforementioned parking control method for a vehicle, and will not be repeated here.

[0170] Figure 8 This is a schematic diagram of a vehicle parking control device according to an embodiment of this application, applied to a cloud-based dispatching platform, such as... Figure 8 As shown, the device includes: a second receiving module 802, a determining module 804, and a second generating module 806.

[0171] The second receiving module 802 is used to receive vehicle status data uploaded by the vehicle control unit and obtain environmental data in the parking lot when the vehicle enters the target communication range in the parking lot; the determining module 804 is used to determine the global reference path based on the vehicle status data and environmental data; the second generating module 806 is used to generate an environmental grid map corresponding to the global reference path; and the sending module is used to send the global reference path and environmental grid map to the vehicle control unit.

[0172] Optionally, the determination module is also used to construct a global accessible area map of the parking lot based on vehicle status data and environmental data; perform accessibility analysis and parking feasibility analysis on all vacant parking spaces based on the global accessible area map and vehicle status data to obtain a priority list of available parking spaces; determine the first parking space in the priority list of available parking spaces as the target parking space; and generate a global reference path based on the current location data in the vehicle status data, the location data corresponding to the target parking space, and the principle of the shortest global path.

[0173] Embodiments of this application also provide a vehicle, including: a memory storing an executable program; and a processor for running the program, wherein the program executes the methods described in various embodiments of this application when it runs.

[0174] Embodiments of this application also provide a computer-readable storage medium including a stored executable program, wherein, when the executable program is running, it controls the device where the computer-readable storage medium is located to perform the methods of various embodiments of this application.

[0175] Embodiments of this application also provide a computer program product, including a computer program that, when executed by a processor, implements the methods of various embodiments of this application.

[0176] Embodiments of this application also provide a computer program product, including a non-volatile computer-readable storage medium for storing a computer program that, when executed by a processor, implements the methods in various embodiments of this application.

[0177] Embodiments of this application also provide a computer program that, when executed by a processor, implements the methods described in the various embodiments of this application.

[0178] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0179] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0180] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0181] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0182] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard drive, magnetic disk, or optical disk.

[0183] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A parking control method for a vehicle, characterized in that, An application to an in-vehicle control unit, wherein the in-vehicle control unit is deployed in a vehicle, including: Receive a global reference path and an environmental grid map sent by a cloud-based scheduling platform. The global reference path is used to represent a reference path that guides the vehicle from its current location to the target parking space, and the environmental grid map is used to represent a grid map that characterizes the spatial distribution of obstacles in the parking lot. The environmental grid map and the obstacle data perceived by the vehicle are fused together to generate a target grid map; Based on the global reference path, the target grid map, and the vehicle's status data, local control commands for the vehicle are generated. Based on the local control command, the vehicle is controlled to perform a parking operation.

2. The method according to claim 1, characterized in that, The environmental grid map and the obstacle data perceived by the vehicle are fused to generate a target grid map, including: Determine the obstacle types corresponding to multiple grids in the environmental grid map; The environmental grid map is updated based on the obstacle types corresponding to the multiple grids and the obstacle data to generate the target grid map; Preferably, updating the environmental grid map based on the obstacle types corresponding to the plurality of grids and the obstacle data to generate the target grid map includes: If the obstacle type in any grid is a static obstacle, the grid value corresponding to that grid is determined to be a preset grid value; If the obstacle type in any grid is a dynamic obstacle, the grid value corresponding to the grid is updated based on the obstacle data; The target raster map is generated based on the raster values ​​corresponding to the multiple raster cells.

3. The method according to claim 1, characterized in that, Based on the global reference path, the target grid map, and the vehicle's status data, local control commands for the vehicle are generated, including: The global reference path is smoothed based on the vehicle state data to obtain the global reference trajectory. Based on the global reference trajectory and the vehicle state data, a multi-dimensional evaluation model is constructed. The multi-dimensional evaluation model is used to evaluate the tracking deviation and steering wheel control deviation of the vehicle when it travels along the global reference trajectory within a preset control cycle. Based on the target grid map and the vehicle's preset safety conditions, target constraints are constructed; The multi-dimensional evaluation model is solved based on the target constraints to obtain the local control command.

4. The method according to claim 3, characterized in that, The global reference path is smoothed based on the vehicle state data to obtain the global reference trajectory, including: Based on the vehicle position data and vehicle attitude data in the vehicle status data, the position of the pre-aiming point on the global reference path is determined; The target curvature is determined based on the position of the pre-aiming point and the position of the rear axle center point of the vehicle; The global reference path is smoothed based on the target curvature to obtain the global reference trajectory.

5. The method according to claim 3, characterized in that, Based on the target grid map and the vehicle's preset safety conditions, target constraints are constructed, including: Construct dynamic constraints based on vehicle kinematic equations; Based on the target grid map and the preset safety distance threshold in the preset safety conditions, obstacle avoidance constraints are constructed. Based on the preset steering wheel control threshold in the preset safety conditions, control constraints are constructed. Based on the preset boundary thresholds in the preset safety conditions, boundary constraint conditions are constructed; Based on the dynamic constraints, the obstacle avoidance constraints, the control constraints, and / or the boundary constraints, the target constraints are determined. Preferably, obstacle avoidance constraints are constructed based on the target grid map and the preset safety distance threshold in the preset safety conditions, including: The passable area at the current moment is determined based on the target grid map; Based on the location data of the passable area and the location data of the obstacle grid, the interval distance is determined; The interval distance being greater than the preset safe distance threshold is determined as the obstacle avoidance constraint condition.

6. The method according to claim 3, characterized in that, The multi-dimensional evaluation model is solved based on the target constraints to obtain the local control commands, including: The multi-dimensional evaluation model is solved to obtain a sequence of control parameters, wherein the sequence of control parameters satisfies the target constraint conditions; The control parameters of the first control cycle in the control parameter sequence are determined as the local control command.

7. The method according to claim 1, characterized in that, After controlling the vehicle to perform a parking operation based on the local control command, the method further includes: Obtain the rear axle center point position data of the vehicle; Based on the rear axle center point position data and the preset vehicle center position data, the deviation data is determined; If the deviation data is less than the preset deviation and the vehicle speed is zero, the vehicle status is determined to be successful parking, and the target vehicle pose data is obtained. The target vehicle's position and pose data, along with a parking success signal, are sent to the cloud-based dispatch platform.

8. A parking control method for a vehicle, characterized in that, Applications in cloud-based scheduling platforms include: When a vehicle enters the target communication range in a parking lot, it receives vehicle status data uploaded by the vehicle control unit and obtains environmental data in the parking lot. Based on the vehicle status data and the environmental data, a global reference path is determined; Generate an environment raster map corresponding to the global reference path; The global reference path and the environmental grid map are sent to the vehicle control unit.

9. The method according to claim 8, characterized in that, Based on the vehicle status data and the environmental data, a global reference path is determined, including: Based on the vehicle status data and the environmental data, a global passable area map of the parking lot is constructed; Based on the global accessible area map and the vehicle status data, accessibility analysis and parking feasibility analysis are performed on all available parking spaces to obtain a priority list of available parking spaces. The first parking space in the list of available parking spaces is selected as the target parking space. Based on the current location data in the vehicle status data, the location data corresponding to the target parking space, and the principle of the shortest global path, the global reference path is generated.

10. A vehicle, characterized in that, include: Memory, which stores executable programs; A processor for running the program, wherein the program, when running, performs the method according to any one of claims 1 to 9.

Images