Offline automatic lightering tracking method and device

Abstract

The invention provides an off-line automatic lightering tracking method and device, and relates to the technical field of vehicle control, and the method comprises the steps: firstly, reading a vehicle identification code and a starting point number in a vehicle off-line starting region, associating the information with an actual position, and storing the information in a background database; and then, obtaining and storing the real-time pose, surrounding obstacle information, road environment information and a parking space state of a buffer area of a target vehicle, carrying out multi-target tracking based on a space-time constraint model and a cross mutual verification mechanism, and binding a vehicle identification code with a track fragment, thereby ensuring that each vehicle can be stably tracked even in a complex environment. And then, the distributed parking space number and the driving track of each vehicle are dynamically planned, and the vehicle is controlled to drive according to the planned path. And finally, verifying the accuracy of the operation by verifying the consistency of the vehicle identification code of the vehicle parked in the parking space and the number of the distributed parking space. According to the method, the matching precision and the operation efficiency in the off-line automatic lightering process of the vehicle are effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of vehicle control technology, and in particular to a method and device for tracking off-line automatic transfer. Background Art

[0002] The technology for autonomous vehicle transfer after the vehicle is offline is designed to address the high costs and low management efficiency of manual transfers, but existing technologies are deficient in terms of target vehicle code matching and tracking continuity, positioning accuracy, and real-time performance. Traditional methods rely primarily on sensor data matching and lack an authentication mechanism for active vehicle feedback, resulting in a high mismatch rate. In addition, in occlusion scenarios, when roadside sensors lose their target, they are unable to maintain continuous positioning through vehicle-side data, resulting in an increase in the target vehicle's code switching rate. At the same time, due to delays in roadside positioning data transmission, vehicle control lags, affecting trajectory tracking accuracy. Existing solutions fail to effectively combine vehicle-side and roadside data for real-time verification to ensure the continuity of target vehicle codes and high-precision positioning throughout the entire process. In particular, it is difficult to ensure accurate matching and precise control of driving trajectories when multiple similar vehicles enter the perception range at the same time. These problems limit the functionality and reliability of the automated transfer system. Summary of the Invention

[0003] The purpose of the present invention is to provide a method and device for tracking offline automatic transfer, so as to solve one or more technical problems existing in the prior art and at least provide a beneficial option or create conditions.

[0004] The present invention solves the technical problem as follows: On the one hand, the present invention provides an offline automatic transfer tracking method, comprising the following steps: In the vehicle off-line starting area, the vehicle identification code and starting point number of the target vehicle are read, a mapping relationship between the vehicle identification code and the actual position is constructed, and stored in the background database; Obtaining the real-time position of the target vehicle, surrounding obstacle information, road environment information and buffer parking space status, and storing them in the backend database; According to the backend database, based on the spatiotemporal constraint model and the cross-verification mechanism, multi-target tracking is performed on the target vehicle to obtain the binding relationship between the vehicle identification code and the trajectory segment of the target vehicle, and store it in the backend database; the trajectory segment is used to represent the vehicle trajectory point of the target vehicle in a continuous time period; the vehicle trajectory point is used to represent the real-time position and posture of the target vehicle at a specific time point; Perform dynamic planning based on the background database to generate the assigned parking space number and planned trajectory of the target vehicle, and control the target vehicle to travel along the planned trajectory; Verify the consistency between the vehicle identification code of the target vehicle parked in the parking space and the allocated parking space number; if the verification is consistent, update the background database to store the mapping relationship between the vehicle identification code and the allocated parking space number.

[0005] Furthermore, the method of reading the vehicle identification code and starting point number of the target vehicle in the vehicle off-line starting area, constructing a mapping relationship between the vehicle identification code and the actual position, and storing the mapping relationship in the background database includes the following steps: At the vehicle off-line starting area, reading the vehicle identification code and the starting point number; Based on the coordinate information of the starting point number in the electronic map, the actual position of the target vehicle is associated, and a mapping relationship between the vehicle identification code and the actual position is established and stored in the background database.

[0006] Furthermore, the acquisition of the real-time position of the target vehicle, surrounding obstacle information, road environment information, and buffer parking space status, and storage in the background database includes the following steps: Obtaining the real-time position and the surrounding obstacle information; Extracting the road environment information and the parking space status of the buffer zone according to the electronic map; wherein the road environment information includes the road topology and traffic rules; the parking space status of the buffer zone includes the vacant / occupied status and the parking space number; The real-time position, the surrounding obstacle information, the road environment information and the parking space status in the buffer zone are subjected to data fusion processing to obtain fused perception data, which is stored in a background database.

[0007] Furthermore, the method of performing multi-target tracking on the target vehicle based on the background database, a spatiotemporal constraint model and a cross-verification mechanism, obtaining a binding relationship between the vehicle identification code and the trajectory segment of the target vehicle, and storing the binding relationship in the background database includes the following steps: Performing a first match based on the backend database to construct a four-dimensional spatiotemporal association model, filtering candidate data that meets vehicle identification code binding constraints, spatial constraints, temporal constraints, and motion constraints from the backend database, and constructing a spatiotemporal association graph; in the spatiotemporal association graph, nodes represent the trajectory segments, and edge weights represent the degree of difference between the trajectory segments; A second global matching is performed based on the spatiotemporal correlation graph, and an improved Kuhn-Monkress algorithm is used to solve the minimum weight matching of the spatiotemporal correlation graph to obtain a binding relationship between the vehicle identification code and the trajectory segment, which is stored in the background database.

[0008] Furthermore, when the real-time position and posture of the target vehicle is lost, a third cross-verification is performed, including the following steps: Acquire the speed, steering wheel angle, and heading angle of the target vehicle as chassis bus data and store them in the backend database; Based on the historical pose of the target vehicle in the backend database and the chassis bus data, a cloud-side predicted pose of the target vehicle is calculated by a cloud platform, and a vehicle-side dead reckoning unit is used to calculate a vehicle-side predicted pose of the target vehicle; The cloud-side predicted posture and the vehicle-side predicted posture are longitudinally verified. If the verification passes, the trajectory segment is updated according to the vehicle-side predicted posture, and the binding relationship between the vehicle identification code and the trajectory segment is obtained and stored in the background database.

[0009] Furthermore, the vehicle identification code binding constraint is used to constrain the trajectory segment and the vehicle identification code to be uniquely matched; The real-time position and posture of the target vehicle are obtained through the laser radar and camera, including the laser radar positioning, camera positioning and heading angle; The spatial constraint is used to constrain the error between the current lidar positioning and the lidar positioning of the previous frame to not exceed a first preset error threshold, and the error between the current camera positioning and the camera positioning of the previous frame to not exceed a second preset error threshold; The time constraint is used to constrain the clock time difference between the cameras to not exceed a third preset error threshold; The motion constraint is used to constrain the error between the current heading angle and the heading angle of the previous frame to not exceed a fourth preset error threshold, and the relative speed deviation between the current vehicle speed and the vehicle speed of the previous frame to not exceed a fifth preset error threshold.

[0010] Furthermore, the verification of the consistency between the vehicle identification code of the target vehicle parked in the parking space and the assigned parking space number comprises the following steps: Identify the actual parking space number where the target vehicle is parked; Retrieving the assigned parking space number and vehicle identification code of the target vehicle from the backend database; Verify whether the actual parking space number is consistent with the allocated parking space number; If the verification is consistent, the backend database is updated, the buffer parking space status is marked as occupied, the transfer task status of the target vehicle is marked as completed, and the mapping relationship between the vehicle identification code and the allocated parking space number is stored; If the verification is inconsistent, the mapping relationship between the vehicle identification code and the allocated parking space number is corrected and a parking space is reallocated for the target vehicle.

[0011] In another aspect, the present invention provides an offline automatic transfer tracking device, comprising a vehicle-side module, a road-side module, and a cloud platform module; The vehicle-end module includes: An on-board diagnostic system unit, used to read the vehicle identification code and starting point number of the target vehicle at the vehicle off-line starting area; A vehicle-side trajectory following control unit is used to receive the planned trajectory sent by the cloud platform module and control the target vehicle to travel along the planned trajectory; a chassis sensing unit, configured to obtain the speed, steering wheel angle, and heading angle of the target vehicle as chassis bus data; The roadside module is used to collect the real-time position and surrounding obstacle information of the target vehicle and identify the actual parking space number where the target vehicle is parked; The cloud platform module includes: A cloud-based electronic map unit stores an electronic map and is used to construct a mapping relationship between the vehicle identification code and the actual location of the target vehicle based on the vehicle identification code and the starting point number, and store the mapping relationship in a backend database; A cloud-based multi-target tracking unit is configured to perform multi-target tracking on the target vehicle based on the backend database, a spatiotemporal constraint model, and a cross-verification mechanism, obtain a binding relationship between the target vehicle's vehicle identification code and the trajectory segment, and store the binding relationship in the backend database; The cloud-based scheduling and planning unit is configured to perform dynamic planning based on the backend database to generate the assigned parking space number and planned trajectory of the target vehicle; verify the consistency between the vehicle identification code of the target vehicle parked in the parking space and the assigned parking space number; if the verification is consistent, update the backend database to store the mapping relationship between the vehicle identification code and the assigned parking space number; The cloud-based backend database unit is used to manage the backend database.

[0012] Furthermore, the road end module includes: A laser radar, used to collect three-dimensional point cloud data and spatial position of the target vehicle; A camera is used to capture an image of the target vehicle's appearance, a road environment image, and an actual parking space image; The edge computing unit is used to detect the real-time position and surrounding obstacle information of the target vehicle and identify the actual parking space number in which the target vehicle is parked.

[0013] Furthermore, the vehicle-side module also includes a vehicle-side trajectory calculation unit, which is used to predict the vehicle-side predicted posture of the target vehicle based on the historical posture of the target vehicle in the background database and the chassis bus data.

[0014] The beneficial effects of the present invention are as follows: the present invention proposes a method for tracking automatic transfer of vehicles offline. First, the vehicle identification code and starting point number are read in the starting area of ​​the vehicle offline, and this information is associated with the actual position and stored in the background database; then, the real-time position, surrounding obstacle information, road environment information and buffer parking space status of the target vehicle are obtained and stored. Based on the spatiotemporal constraint model and the cross-verification mechanism, this method can perform multi-target tracking, bind the vehicle identification code with the trajectory segment, and ensure that each vehicle can be stably tracked even in a complex environment. Subsequently, the assigned parking space number and driving trajectory of each vehicle are dynamically planned based on the collected information, and the vehicle is controlled to drive according to the planned path. Finally, the accuracy of the operation is verified by checking the consistency between the vehicle identification code of the vehicle parked in the parking space and the assigned parking space number. This method effectively improves the matching accuracy and operating efficiency in the process of automatic transfer of vehicles offline. The present application also provides a corresponding device. The beneficial effects of the device are similar to those of the method and will not be repeated here.

[0015] Other features and advantages of the present application will be described in the following description, and in part will become apparent from the description, or will be understood by practicing the present application. The purposes and other advantages of the present application can be achieved and obtained through the structures particularly pointed out in the description, claims and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The accompanying drawings are used to provide a further understanding of the technical solution of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the technical solution of the present invention and do not constitute a limitation to the technical solution of the present invention.

[0017] Figure 1 It is a flow chart of the offline automatic lightering tracking method provided by this application; Figure 2 It is a structural diagram of the offline automatic transfer tracking device provided by this application; Figure 3 This is a schematic diagram of the application scenario of automatic off-line transfer tracking of vehicles provided by this application. DETAILED DESCRIPTION

[0018] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0019] The present application is further described below in conjunction with the accompanying drawings and specific embodiments. The described embodiments should not be considered as limiting the present application. All other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0020] In the following description, reference is made to “some embodiments”, which describes a subset of all possible embodiments, but it will be understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0021] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this application pertains. The terms used herein are for the purpose of describing the embodiments of this application only and are not intended to limit this application.

[0022] Autonomous vehicle transfer technology for off-line vehicles aims to address the high costs and low management efficiency of existing manual transfers. This technology requires extremely high standards for matching and tracking target vehicle codes, as well as positioning accuracy and real-time performance, to ensure that each off-line vehicle can accurately and autonomously reach its designated location.

[0023] Traditional tracking relies on matching sensor data to identify and track target vehicles, but this approach lacks an authentication mechanism based on active vehicle feedback. This means that when multiple similar vehicles enter the sensing range simultaneously, the system struggles to distinguish them, resulting in a high rate of false matches. For example, if multiple vehicles are simultaneously offline, sensors may mistakenly assign the track of one vehicle to another due to similar appearance or environmental interference, compromising overall tracking accuracy.

[0024] In real-world environments, roadside sensors may temporarily lose the target vehicle's location information due to obstruction by obstacles. Current technical solutions cannot effectively utilize vehicle-side data to maintain continuous positioning, resulting in a significant increase in the encoding switching rate in this situation. For example, when a vehicle passes through a tunnel or is obscured by other large vehicles, the roadside sensor may lose track of the target vehicle, resulting in tracking interruption or restart, reducing the stability and reliability of the system. Existing technologies typically rely on roadside positioning data for vehicle control, but due to data transmission delays, this can cause vehicle control to lag, affecting the accuracy of trajectory following. Especially at high speeds or in dense traffic environments, this delay may cause the vehicle to deviate from the intended trajectory, increasing the risk of collision or reducing the efficiency of path planning.

[0025] Furthermore, current technology has yet to address the issue of ensuring consistent target vehicle identification throughout the entire process and achieving real-time, high-precision positioning through active vehicle-side feedback and real-time verification of roadside data. This is particularly true when multiple target vehicles of the same model and color simultaneously enter the roadside sensing range. The cloud must accurately match and track each vehicle's VIN, simultaneously transmitting the trajectory and position of multiple vehicles to ensure that each target vehicle follows its own trajectory. Furthermore, maintaining stable tracking and high-precision positioning in obstructed or complex environments is another challenge facing existing technologies.

[0026] Therefore, this application proposes a method and device for tracking automatic off-line transportation, aiming to effectively improve the functions and performance indicators of the vehicle off-line automatic transportation system through cross-verification tracking between the vehicle side and the cloud side, and overcome the above-mentioned problems.

[0027] Specifically, the system reads the target vehicle's vehicle identification number and starting point number at the vehicle's off-line starting zone. Based on the coordinates of the starting point number on the electronic map, the system associates the target vehicle's actual location with the target vehicle's VIN. A mapping relationship is established and stored in the backend database, ensuring each vehicle's unique identification and accurate initial location in the system. Subsequently, the system acquires and integrates real-time data such as the target vehicle's position, surrounding obstacles, road environment information, and buffer parking status, providing comprehensive data support for precise tracking and path planning. Based on a spatiotemporal constraint model and a cross-validation mechanism, a four-dimensional spatiotemporal correlation model is used to select data that meets the constraints. An improved KM algorithm is used to solve the minimum weight matching, achieving high-precision multi-target vehicle tracking. Even in occluded scenarios, vehicle-based data can be used to compensate for targets lost by roadside sensors, maintaining tracking continuity. The system further performs dynamic planning to generate the target vehicle's assigned parking space number and planned trajectory, controlling the vehicle along the planned path. In the event of target loss by the roadside sensors, vehicle-based data is used to predict and compensate, ensuring accurate and continuous tracking. Finally, after the target vehicle parks in the designated space, the system verifies the consistency of the vehicle identification code with the assigned parking space number. If they match, the mapping relationship in the backend database is updated. Otherwise, the mapping relationship is corrected and the parking space is reassigned, ensuring the accuracy of all operations. This method overcomes the high mismatch rate caused by traditional tracking that relies on sensor data matching, improves tracking stability in occluded scenarios, and reduces the impact of positioning delays, significantly improving the efficiency and reliability of the vehicle offline automatic transfer system.

[0028] First, the offline automatic transfer tracking method provided by the embodiment of the present application will be described in detail with reference to the accompanying drawings.

[0029] Reference Figure 1 The implementation process of the offline automatic transfer tracking method provided in the embodiment of the present application includes but is not limited to the following steps.

[0030] Step S110: In the vehicle off-line starting area, the vehicle identification code and the starting point number of the target vehicle are read, a mapping relationship between the vehicle identification code and the actual position is constructed, and stored in the background database.

[0031] In step S110, the target vehicle's vehicle identification number and starting point number are read from the vehicle's off-line starting area. Based on the coordinates of the starting point number on the electronic map, the target vehicle's actual location is correlated. A mapping between the VIN and actual location is established and stored in the backend database. This process ensures the unique identification of each vehicle in the system and its accurate record of its initial location, providing a solid foundation for subsequent tracking, route planning, and management. In this way, the system can accurately identify and track each vehicle off the line, avoiding mismatches caused by similar appearances or environmental interference.

[0032] Step S120: obtaining the real-time position of the target vehicle, information about surrounding obstacles, road environment information, and parking space status in the buffer zone, and storing them in a backend database.

[0033] In step S120, data such as the target vehicle's real-time position, surrounding obstacle information, road environment information, and buffer parking space status are acquired. This data is then integrated and stored in a backend database. This step provides comprehensive and real-time data support for the entire automated transfer system. Through detailed perception of the vehicle's surroundings, the system can better understand current traffic conditions and road layout, providing a basis for subsequent route planning and dynamic adjustments. Furthermore, understanding the status of buffer parking spaces facilitates the rational allocation of parking resources, improving the system's overall efficiency and responsiveness.

[0034] In step S130 , according to the background database, based on the spatiotemporal constraint model and the cross-verification mechanism, multi-target tracking is performed on the target vehicle, and the binding relationship between the vehicle identification code and the trajectory segment of the target vehicle is obtained and stored in the background database.

[0035] Among them, the trajectory segment is used to represent the vehicle trajectory point of the target vehicle in a continuous time period. The vehicle trajectory point is used to represent the real-time position of the target vehicle at a specific time point.

[0036] In step S130, based on the information in the background database, the target vehicle is tracked in multiple targets based on the spatiotemporal constraint model and the cross-verification mechanism, and the binding relationship between the target vehicle's vehicle identification code and the trajectory segment is obtained, and the result is stored in the background database. The trajectory segment is used to represent the vehicle trajectory point of the target vehicle in a continuous time period, and the vehicle trajectory point represents the real-time position of the target vehicle at a specific time point. The data that meets the constraint conditions is filtered through the four-dimensional spatiotemporal correlation model, and the improved KM algorithm is used to solve the minimum weight matching to achieve high-precision vehicle tracking. This method not only improves the tracking stability in occlusion scenarios, but also effectively reduces the problem of tracking interruption caused by the sensor losing the target.

[0037] Step S140 , performing dynamic planning based on the background database, generating the assigned parking space number and planned trajectory of the target vehicle, and controlling the target vehicle to travel along the planned trajectory.

[0038] In step S140, dynamic planning is performed based on the information in the backend database to generate the target vehicle's assigned parking space number and planned trajectory, and the vehicle is controlled to travel along the planned path. This step takes into account the vehicle's current position and speed, as well as its historical trajectory and changes in the surrounding environment, to achieve more accurate path planning. If the roadside sensor loses the target, the system can use vehicle-side data to predict and compensate, ensuring continuous and accurate tracking. By dynamically adjusting the path planning, the system can flexibly respond to various changes in complex traffic environments, ensuring that each vehicle reaches its designated location safely and efficiently.

[0039] Step S150: Verify the consistency between the vehicle identification code of the target vehicle parked in the parking space and the assigned parking space number. If the verification is consistent, update the backend database to store the mapping relationship between the vehicle identification code and the assigned parking space number.

[0040] In step S150, after verifying that the target vehicle is parked in a parking space, the system checks the consistency between the vehicle identification code and the assigned parking space number. If they are consistent, the backend database is updated to store the mapping between the vehicle identification code and the assigned parking space number. If they are inconsistent, the mapping is corrected and the parking space is reassigned. This step ensures the accuracy of all operations and avoids confusion caused by incorrect parking space allocation or identification issues. In this way, the system not only promptly detects and corrects any potential errors, but also optimizes the use of parking resources, further improving the reliability and efficiency of the entire process.

[0041] In some embodiments of the present application, in step S110, in the vehicle offline starting area, the vehicle identification code and starting point number of the target vehicle are read, a mapping relationship between the vehicle identification code and the actual location is constructed, and the implementation process of storing it in the background database includes but is not limited to the following steps.

[0042] Step S210: Read the vehicle identification code and the starting point number at the vehicle off-line starting area.

[0043] In step S210, an automatic identification device (such as an OBD reader, RFID reader / writer, or onboard communication module) is installed at the vehicle's off-line starting area. By communicating with the target vehicle's onboard system, it automatically reads the vehicle identification number (VIN) and the current starting point number. The VIN is a globally unique 17-character vehicle identifier used to accurately identify each vehicle; the starting point number indicates the specific physical location where the vehicle enters the automated transfer process. This step provides accurate identification and initial location information for subsequent tracking and route planning, ensuring the system can associate the correct vehicle with the correct task and avoiding scheduling errors caused by identity confusion.

[0044] Step S220 , based on the coordinate information of the starting point number in the electronic map, the actual position of the target vehicle is associated, a mapping relationship between the vehicle identification code and the actual position is established, and stored in the background database.

[0045] In step S220, based on the geographic coordinate information corresponding to the preset starting point number in the electronic map, the system associates the read VIN of the target vehicle with its actual geographic location, constructs a "VIN-actual location" mapping relationship, and stores the mapping data in the background database. The significance of this process is that it not only achieves the precise binding of vehicle identity and spatial location, but also provides a reliable data foundation for subsequent multi-vehicle collaborative tracking, trajectory prediction and path planning. Through this mapping relationship, the system can call the initial position information of each vehicle at any time to assist in determining the changing trend of its motion trajectory, thereby improving the robustness and tracking accuracy of the system in complex environments. At the same time, it also provides traceable original data support for subsequent verification links (such as parking space consistency verification).

[0046] In some embodiments of the present application, in step S120, the real-time position and posture of the target vehicle, surrounding obstacle information, road environment information and buffer parking space status are obtained and stored in the background database, and the implementation process includes but is not limited to the following steps.

[0047] Step S310: Acquire real-time position and surrounding obstacle information.

[0048] In step S310, real-time information is collected about the target vehicle's position (including location, speed, and heading angle) and surrounding obstacles (such as other vehicles, pedestrians, and static obstacles). This step provides the system with data on the vehicle's current dynamic state and surrounding environment, forming the foundation for precise tracking, path planning, and obstacle avoidance. By continuously updating this information, the system maintains real-time visibility into the vehicle's operational status, ensuring safe and efficient completion of transfer tasks in complex traffic environments.

[0049] Step S320: extracting road environment information and buffer zone parking space status based on the electronic map.

[0050] The road environment information includes the road topology and traffic rules, and the buffer zone parking space status includes the free / occupied status and parking space number.

[0051] In step S320, the electronic map system is used to extract road environment information and buffer zone parking space status related to the target vehicle's travel path. This road environment information includes road topology (e.g., lane connectivity, intersection layout), traffic rules (e.g., speed limits, priority rules), etc.; buffer zone parking space status includes the availability / occupancy of each parking space and its corresponding parking space number. This process provides the system with global environmental awareness, enabling the automated transfer system to not only perceive local dynamic information but also make more efficient scheduling and routing decisions based on pre-defined road structures and parking resources, thereby improving the overall system's intelligence and operational efficiency.

[0052] Step S330: Perform data fusion processing on the real-time posture, surrounding obstacle information, road environment information, and buffer zone parking space status to obtain fused perception data, which is stored in the background database.

[0053] In step S330, the real-time position and obstacle information obtained in step S310 is subjected to multi-source data fusion processing with the road environment and parking status information extracted in step S320 to generate fused perception data in a unified format and store it in the background database. The key role of this step is to break the limitations of a single sensor or information source and improve the integrity, accuracy and robustness of the perception data. For example, by fusing dynamic obstacle information with static map data, traffic conditions can be predicted more accurately. Ultimately, the fused perception data provides high-quality data support for subsequent multi-target tracking, trajectory prediction and path planning, ensuring the stability and safety of the entire automatic transfer system.

[0054] In some embodiments of the present application, in step S130, according to the background database, based on the spatiotemporal constraint model and the cross-verification mechanism, multi-target tracking is performed on the target vehicle to obtain the binding relationship between the vehicle identification code and the trajectory segment of the target vehicle, and the implementation process of storing it in the background database includes but is not limited to the following steps.

[0055] In step S410 , a first matching is performed based on the background database to construct a four-dimensional spatiotemporal association model, candidate data that meets the vehicle identification code binding constraint, space constraint, time constraint and motion constraint is screened out from the background database, and a spatiotemporal association graph is constructed.

[0056] In the spatiotemporal correlation graph, nodes represent trajectory segments, and edge weights represent the degree of difference between trajectory segments.

[0057] In step S410, a first match is performed based on chassis data, historical trajectory segments, and other sensory data in the backend database to construct a four-dimensional spatiotemporal correlation model. This model, based on the four dimensions of space, time, motion, and identity, selects a candidate set of data that meets binding constraints and further constructs a spatiotemporal correlation graph with trajectory segments as nodes and degree of difference as edge weights. The core significance of this step lies in narrowing the matching search space and improving matching efficiency and accuracy through multi-dimensional modeling. In particular, the introduction of the VIN binding constraint ensures that the identity of each vehicle remains unique throughout the tracking process, avoiding the code switching problem common in traditional methods and enhancing the stability and robustness of the system.

[0058] In step S420 , a second global matching is performed based on the spatiotemporal correlation graph. The improved Kuhn-Moncrease algorithm is used to solve the minimum weight matching of the spatiotemporal correlation graph, and the binding relationship between the vehicle identification code and the trajectory segment is obtained and stored in the background database.

[0059] In step S420, based on the constructed spatiotemporal correlation graph, an improved Kuhn-Munkres (KM) algorithm is used to solve the minimum weight matching problem, thereby achieving optimal matching for all candidate trajectory segments. Ultimately, the binding relationship between each vehicle identification code and the corresponding trajectory segment is obtained, and the results are stored in the backend database. This step is crucial in completing the transition from local perception to global cognition, ensuring that tracking continuity for each vehicle is accurately maintained even in occlusion, intersecting paths, or high-density traffic environments. By optimizing the matching algorithm, the system significantly improves matching accuracy while maintaining computational efficiency, effectively addressing code drift and mismatching issues in multi-target tracking, and providing reliable data support for subsequent path planning and parking control.

[0060] Specifically, the improved KM algorithm ensures precise binding between vehicle identification numbers (VINs) and trajectory segments during multi-target tracking by optimizing the global optimal matching strategy. While the traditional KM algorithm is primarily used to solve the maximum weight matching problem in bipartite graphs, the improved version can better handle complex matching tasks in dynamic environments, such as maintaining high-precision tracking continuity in the presence of occlusion or sensor data loss. By introducing an optimization mechanism for spatiotemporal feature constraints, the improved KM algorithm can significantly improve matching accuracy while ensuring computational efficiency, effectively addressing issues such as code drift and mismatching in multi-target tracking. This provides reliable data support for path planning, obstacle avoidance control, and parking operations, enhancing the stability and accuracy of the entire automated transfer system.

[0061] In some embodiments of the present application, in step S130, when the real-time position and posture of the target vehicle is lost, the implementation process of performing the third cross-verification includes but is not limited to the following steps.

[0062] Step S510: Obtain the target vehicle's speed, steering wheel angle, and heading angle as chassis bus data and store them in the backend database. In step S510, the chassis sensing unit acquires key dynamic parameters of the target vehicle, including speed, steering wheel angle, and heading angle. This data is then stored as chassis bus data in a backend database. This process provides accurate vehicle motion state information for subsequent trajectory prediction and multi-target tracking. This chassis bus data not only provides information on the vehicle's current dynamic behavior but also serves as input for prediction models based on historical data, such as LSTM combined with Kalman filtering. By storing this data in the backend database, the system can analyze vehicle behavior patterns from a global perspective, improving tracking accuracy and robustness. Furthermore, this data provides an important reference for the subsequent cross-validation mechanism, ensuring high tracking continuity and accuracy even in the event of sensor data loss.

[0063] In step S520, based on the historical posture and chassis bus data of the target vehicle in the background database, the cloud platform uses a long short-term memory network combined with a Kalman filter to calculate the cloud-based predicted posture of the target vehicle.

[0064] In step S520, if the target vehicle's real-time pose is lost due to occlusion, communication interruption, or other reasons, the system uses a long short-term memory (LSTM) network combined with a Kalman filter algorithm to predict the vehicle's position and pose for a period of time in the future by calling upon the target vehicle's historical pose and chassis data (such as speed and steering wheel angle) stored in the backend database. This prediction results in a "cloud-based predicted pose." The core function of this step is that, in the absence of external sensory information, the system can generate a high-confidence position prediction based on historical motion patterns and vehicle dynamics models, thereby maintaining tracking continuity. LSTM excels at capturing long-term dependencies in time series, while Kalman filtering effectively integrates predictions with observed data. The combination of these two improves prediction accuracy and stability.

[0065] Step S530: Based on the historical pose and chassis data of the target vehicle in the background database, the vehicle-side dead reckoning unit calculates the vehicle-side predicted pose of the target vehicle.

[0066] In step S530, based on the target vehicle's historical pose data stored in the backend database and the latest chassis parameters, a local prediction model (deployed on the vehicle or edge node) is used to predict the vehicle's pose at the current moment. This is known as the vehicle-side predicted pose. For example, the vehicle-side predicted position can support a range of 60 meters. The significance of this process lies in providing a cloud-independent pose estimation method, avoiding the risk of bias caused by single-source predictions. Vehicle-side predictions do not rely on the data processing capabilities of remote servers, resulting in lower latency and higher response speeds, making them particularly suitable for scenarios with short-term pose loss. By retaining the vehicle's autonomous prediction capabilities, the system enhances the redundancy and reliability of the overall architecture, ensuring stable tracking even in complex environments.

[0067] In step S540, a longitudinal verification is performed on the cloud-side predicted posture and the vehicle-side predicted posture. If the verification passes, the trajectory segment is updated according to the vehicle-side predicted posture, and the binding relationship between the vehicle identification code and the trajectory segment is obtained and stored in the background database.

[0068] In step S540, the predicted posture on the cloud side is compared with the predicted posture on the vehicle side, and a longitudinal cross-check is performed. If the difference between the two is within the preset threshold range, for example, 50 centimeters, the prediction result is considered credible, and the current trajectory segment is updated based on the predicted posture on the vehicle side, the binding relationship between the VIN and the trajectory segment is continued to be maintained, and the updated data is written to the background database; otherwise, the system may trigger an exception handling mechanism, such as re-acquiring perception data or requesting manual intervention. The key role of this step is to improve the credibility of the posture prediction by verifying the consistency of multi-source prediction results, and to prevent coding drift or trajectory breakage problems caused by a single prediction error. Through this cross-verification mechanism, the system can maintain high-precision and high-stability multi-target tracking capabilities even when perception information is incomplete, providing reliable data support for subsequent path planning and parking control.

[0069] In some embodiments of the present application, a vehicle identification code binding constraint is used to constrain the unique matching of trajectory segments and vehicle identification codes. The vehicle identification code binding constraint ensures that a trajectory segment is uniquely matched to a specific vehicle. The purpose of this constraint is to avoid confusion between the tracks of multiple vehicles and ensure that each track can be accurately mapped to a unique vehicle. By strictly binding each trajectory segment to a specific vehicle identification code, the system can effectively reduce the problem of mismatching caused by inaccurate sensor data or environmental interference, thereby improving the reliability and accuracy of the entire tracking system.

[0070] In some embodiments of this application, the real-time position and orientation of the target vehicle is acquired using LiDAR and cameras, including LiDAR positioning, camera positioning, and heading angle. This data provides precise position and orientation information of the vehicle in three-dimensional space and is key to achieving high-precision tracking. Leveraging the complementary advantages of LiDAR and cameras can provide rich environmental perception information, not only improving positioning accuracy but also enhancing the ability to understand complex scenes. This is crucial for ensuring safe and efficient vehicle operation in dynamic environments.

[0071] In some embodiments of the present application, spatial constraints are used to constrain the error between the current lidar positioning and the lidar positioning of the previous frame to not exceed a first preset error threshold, and the error between the current camera positioning and the camera positioning of the previous frame to not exceed a second preset error threshold.

[0072] Spatial constraints limit the error between the current LiDAR position and the previous frame's LiDAR position to a first preset error threshold (e.g., 15 cm), and the error between the current camera position and the previous frame's camera position to a second preset error threshold (e.g., 50 cm). These constraints ensure the consistency and reliability of the position information provided by different sensors, preventing positioning deviations due to sensor errors or environmental factors. By setting reasonable error ranges, the system can maintain accuracy while improving its adaptability to external changes, enabling stable and reliable vehicle tracking even in complex urban environments. This spatial consistency check helps improve the robustness and stability of the system.

[0073] In some embodiments of the present application, the time constraint is used to constrain the clock time difference between the cameras to not exceed a third preset error threshold.

[0074] The timing constraint stipulates that the clock time difference between each camera must not exceed a third preset error threshold. For example, when the frame rate of each camera is 30fps, the clock time difference between each camera must not exceed 0.3 seconds. This constraint aims to solve the problem of multi-sensor time synchronization and ensure that data from different sensors can be fused and processed under the same time reference. Accurate time synchronization is crucial for correctly understanding the vehicle's motion state and changes in the surrounding environment, especially in high-speed or rapidly changing scenes. Any slight time deviation can lead to serious tracking errors. Therefore, by strictly controlling time errors, the overall performance and stability of the system can be significantly improved, ensuring the consistency and reliability of all sensor data.

[0075] In some embodiments of the present application, motion constraints are used to constrain the error between the current heading angle and the heading angle of the previous frame to not exceed a fourth preset error threshold, and the relative speed deviation between the current vehicle speed and the vehicle speed of the previous frame to not exceed a fifth preset error threshold.

[0076] Motion constraints limit the error between the current heading angle and the heading angle of the previous frame to no more than a fourth preset error threshold (e.g., 15 degrees). They also require the relative speed deviation between the current and previous frame speeds to no more than a fifth preset error threshold (e.g., 10%). This constraint helps verify that the vehicle's actual motion conforms to the expected dynamic model, further enhancing the credibility of the tracking process. By comparing actual measurements with theoretical predictions, not only can anomalies be detected, but tracking strategies can also be adjusted promptly to ensure the vehicle always follows the correct path. This provides strong support for precise path following and obstacle avoidance control, improving the overall operational efficiency and safety of the system.

[0077] In some embodiments of the present application, in step S150, the process of verifying the consistency between the vehicle identification code of the target vehicle parked in the parking space and the assigned parking space number includes but is not limited to the following steps.

[0078] Step S610: Identify the actual parking space number where the target vehicle is parked.

[0079] In step S610, the target vehicle, under cloud-based scheduling and trajectory planning, parks in the target parking space in the buffer zone. It then shuts down, locks the vehicle, closes the windows, and uploads a transfer completion flag to the cloud platform. Roadside sensors (such as cameras and lidar) identify the parking space number actually occupied by the target vehicle. The key to this process is accurately capturing the vehicle's entry into the space and determining its specific location. This step provides foundational data for subsequent verification, ensuring the system accurately determines the vehicle's final parking location. High-precision parking space recognition technology effectively avoids mission failures and resource waste due to misidentification.

[0080] Step S620: retrieve the assigned parking space number and vehicle identification code of the target vehicle from the backend database.

[0081] In step S620, the target vehicle's pre-assigned parking space number and vehicle identification code are retrieved from the backend database. This step serves to obtain the parking space information planned before the task is executed, which serves as a verification basis. By comparing the actual parking space with the planned parking space, the effectiveness of the automated transfer system can be evaluated. This step also ensures the integrity and consistency of the system's internal data, helping to improve overall management efficiency and service quality.

[0082] Step S630: Check whether the actual parking space number is consistent with the allocated parking space number.

[0083] In step S630, the actual parking space number identified in step S610 is compared with the assigned parking space number retrieved from the backend database in step S620 to check for consistency. The core of this step is to confirm that the vehicle has parked accurately in the designated parking space according to the planned schedule. Through a rigorous verification mechanism, any potential deviations can be promptly identified and corrected, ensuring the accuracy and reliability of the entire automated transfer process. This also provides a basis for subsequent decision-making, ensuring the smooth progress of subsequent steps.

[0084] Step S640: If the verification is consistent, the backend database is updated, the buffer parking space status is marked as occupied, the target vehicle's transfer task status is marked as completed, and the mapping relationship between the vehicle identification code and the allocated parking space number is stored.

[0085] In step S640, if the verification result indicates that the actual parking space number matches the assigned parking space number, the system updates the backend database, marking the buffer space status as "occupied" and the target vehicle's transfer task status as "completed." The system also stores the mapping between the vehicle identification code and the assigned parking space number. This not only records the successful completion of the current task but also provides a detailed historical record for future query and management. This real-time update mechanism helps maintain the latest system data, supporting dynamic management and resource scheduling optimization.

[0086] Step S650: If the verification is inconsistent, the mapping relationship between the vehicle identification code and the allocated parking space number is corrected and a parking space is reallocated for the target vehicle.

[0087] In step S650, if the verification process finds a discrepancy between the actual parking space number and the assigned space number, the system will correct the mapping between the vehicle identification code and the assigned space number and reassign an appropriate parking space to the target vehicle. This step aims to address parking space allocation errors caused by various reasons (such as sensor failure and path planning errors). By flexibly adjusting the parking space allocation plan, errors can be quickly corrected while minimizing the impact on other vehicles and maintaining the efficient operation of the entire automated transfer system. Furthermore, after the parking space is reassigned, the system will update the relevant records to ensure transparency and traceability of all operations.

[0088] Secondly, refer to Figure 2 The present application provides an offline automatic transfer tracking device, including a vehicle-side module 700, a road-side module 800 and a cloud platform module 900.

[0089] In some embodiments of the present application, the vehicle-side module 700 includes an on-board diagnostic system unit 701, a vehicle-side trajectory following control unit 702, a chassis sensing unit 703 and a vehicle-side track calculation unit 704.

[0090] The on-board diagnostic system unit 701 reads the target vehicle's vehicle identification code and starting point number at the vehicle's off-line starting area. This unit automatically obtains the vehicle's identity information and initial location by communicating with the vehicle's onboard OBD interface or CAN bus. It serves as the starting point for the entire automated transfer process. Its core function is to ensure that each vehicle entering the system is uniquely identified and tied to its physical location, providing an accurate data foundation for subsequent tracking, route planning, and task management.

[0091] The vehicle-side trajectory following control unit 702 receives the planned trajectory from the cloud platform module 900 and controls the target vehicle along the planned trajectory. Typically integrated into the vehicle's autonomous driving controller, this unit executes the path-tracking algorithm and coordinates chassis control systems such as steering, braking, and drive to ensure the vehicle operates stably and safely along the planned path. It is a key component for autonomous vehicle motion control and directly determines the driving accuracy and stability of the automated transfer system.

[0092] The chassis sensing unit 703 is used to acquire the target vehicle's speed, steering wheel angle, and heading angle as chassis bus data. This data reflects the vehicle's current motion state and dynamic characteristics and is an important input for pose prediction, trajectory generation, and multi-target tracking. By collecting chassis data at a high frequency, the system can more accurately understand vehicle behavior patterns, improving tracking accuracy and control response speed, which is particularly important in scenarios where sensor data is missing or obscured.

[0093] The vehicle-side dead reckoning unit 704 is used to predict the target vehicle's predicted position based on the target vehicle's historical position and chassis bus data in the backend database. This unit plays a key role when the target vehicle's real-time position is lost, maintaining an estimate of the vehicle's position without relying on external sensors. By integrating historical motion trends with the vehicle dynamics model, the vehicle-side dead reckoning unit 704 improves the system's robustness and fault tolerance, provides an independent prediction source for the cross-validation mechanism, and enhances the continuity and reliability of overall tracking.

[0094] In some embodiments of the present application, the chassis sensing unit 703 includes a wheel speed sensor, a steering wheel angle sensor, and an IMU inertial navigation sensor.

[0095] Wheel speed sensors monitor the rotational speed of each wheel in real time and transmit this data to the control system. By analyzing this wheel speed information, the system can calculate the vehicle's speed and distance traveled, which is crucial for precise vehicle motion control. Wheel speed data can also help identify tire slippage or locking, further enhancing vehicle safety and handling. The high-frequency speed feedback provided by wheel speed sensors is one of the foundations for precise trajectory following.

[0096] The steering wheel angle sensor detects the steering wheel's operating angle and converts it into an electrical signal for the control system. This information is crucial for understanding the vehicle's current steering intent and predicting future driving direction. Combined with data from other sensors, such as wheel speed sensors, the system can accurately calculate the vehicle's actual heading change, optimizing path planning and tracking control to ensure the vehicle adheres to its intended trajectory.

[0097] An IMU (Inertial Navigation Unit) integrates an accelerometer and gyroscope to sense changes in a vehicle's acceleration and angular velocity in three-dimensional space. Based on this data, the IMU can infer the vehicle's attitude (pitch, roll, and yaw) and short-term position changes, providing continuous position estimates even when GPS signal is poor or lost. The high-update attitude information provided by the IMU compensates for potential delays or errors associated with other sensors, enhancing the robustness and reliability of the entire system, particularly in complex urban environments or variable weather conditions.

[0098] In some embodiments of the present application, the vehicle-side module 700 further includes a wire-controlled chassis actuator 705 and a vehicle-side communication interface unit 706 .

[0099] The drive-by-wire chassis actuator 705 is a key component responsible for executing control system commands. It directly controls vehicle operations such as steering, acceleration, and braking. Unlike traditional mechanical or hydraulic control systems, the drive-by-wire system transmits commands via electronic signals, enabling more precise and responsive operations. With the support of the drive-by-wire chassis, the autonomous driving system can more flexibly adjust the vehicle's speed and direction, ensuring that the vehicle can accurately follow the planned path. In addition, the drive-by-wire technology improves the redundancy and safety of the system because it can quickly switch to a backup mode when a fault is detected, ensuring driving safety.

[0100] The vehicle-side communication interface unit 706 supports 5G / V2X (Vehicle-to-Everything) communications, providing high-speed, low-latency communication capabilities for vehicles, enabling information exchange with other vehicles (V2V), infrastructure (V2I), pedestrians (V2P), and cloud platforms (V2N). This advanced communication technology greatly enhances the real-time and collaborative nature of the automated transfer system.

[0101] Specifically: V2V communication allows vehicles to share information such as location, speed, and driving intentions, which helps prevent collisions and optimize traffic flow; V2I communication enables vehicles to receive information from roadside infrastructure, such as traffic light status, speed limit instructions, etc., to make smarter driving decisions; V2P communication enhances the vehicle's perception of pedestrians and improves pedestrian safety; V2N communication exchanges data with the cloud platform through the 5G network, supports remote monitoring, dynamic scheduling, big data analysis and other functions, and improves the system's intelligence level and service quality.

[0102] Leveraging 5G / V2X communication, the automated transfer system enables real-time data exchange and collaborative operations in complex traffic environments, significantly improving overall system reliability and efficiency. For example, if multiple vehicles enter the sensing range simultaneously, the system can quickly identify each vehicle through efficient communication mechanisms and assign it the appropriate route and parking space, avoiding potential conflicts and delays.

[0103] In some embodiments of the present application, the roadside module 800 includes a lidar 801, a camera 802 and an edge computing unit 803, which are used to collect the real-time position and surrounding obstacle information of the target vehicle and identify the actual parking space number where the target vehicle is parked.

[0104] The LiDAR 801 collects 3D point cloud data and spatial positions of target vehicles, offering high precision and all-weather operation. This device provides precise geometric structure of the surrounding environment, helping the system identify obstacles, vehicle outlines, and spatial distances. It is one of the core sensors for high-precision positioning and environmental perception. During the automated transfer process, the LiDAR 801 provides rich spatial data support for vehicle identification, trajectory tracking, and obstacle avoidance.

[0105] Camera 802 captures images of the target vehicle's exterior, the road environment, and the actual parking space, serving as the primary source of visual perception. Using image processing technology, the system can identify information such as license plates, vehicle body color, lane markings, traffic signs, and parking space occupancy status. The camera plays an irreplaceable role in identifying parking space numbers and verifying parking accuracy. It also assists LiDAR 801 in achieving multimodal perception fusion, enhancing the system's recognition capabilities and adaptability.

[0106] The edge computing unit 803 detects the target vehicle's real-time position and surrounding obstacle information, and identifies the actual parking space number in which the target vehicle is parked. This unit typically possesses strong local computing capabilities, enabling initial data processing and decision-making close to the data source, reducing the burden on the cloud and communication latency. The edge computing unit 803 is a key component for achieving low-latency response, efficient data preprocessing, and field-level intelligent control, helping to enhance the overall real-time performance and intelligence of the system.

[0107] In some embodiments of the present application, the camera 802 includes a box camera and a fisheye camera.

[0108] A box camera is a fixed-focal-length camera with high resolution and a long viewing distance, primarily used to capture detailed image information of the road environment. It provides clear image data of road signs, traffic lights, lane markings, and surrounding obstacles. Due to its high resolution and long focal length, a box camera is ideal for long-range object recognition and tracking, such as detecting other vehicles or pedestrians within hundreds of meters ahead. Furthermore, a box camera can help the system recognize and read text on roadside signs, which is crucial for dynamically adjusting path planning.

[0109] Fisheye cameras, on the other hand, utilize a wide-angle lens design, typically with a viewing angle of 180 degrees or even greater, providing nearly all-around coverage around the vehicle. This type of camera is particularly well-suited for close-range environmental perception and blind spot monitoring, such as detecting obstacles to the side or rear of the vehicle and identifying parking space occupancy. The wide viewing angle of a fisheye camera enables it to provide more comprehensive environmental information in confined spaces, which is particularly important for accurate parking and obstacle avoidance. Fisheye cameras can also assist in 360-degree panoramic image stitching, providing the driver or autonomous driving system with a complete view of the vehicle's surroundings, thereby improving driving safety and operational precision.

[0110] In some embodiments of the present application, the roadside module 800 also includes a roadside communication interface unit 804, which ensures that data collected by roadside sensors (such as the LiDAR 801 and the camera 802) can be transmitted to the cloud platform in a timely and accurate manner, and supports real-time information exchange between the vehicle-side module 700 and the roadside module 800. Its main functions include: The roadside communication interface unit 804 also supports 5G / V2X communication. Using high-speed network connections (such as 5G or dedicated V2X communication links), it ensures that data can be transmitted from roadside devices to the cloud or directly to nearby vehicles in record time. This is particularly important for scenarios requiring rapid response, such as emergency obstacle avoidance or dynamic path adjustment. This unit is responsible for integrating data from various roadside sensors (such as 3D point cloud data from the LiDAR 801 and image data from cameras), packaging this data into a unified format, and transmitting it to the cloud or other required terminals. It also receives instructions or updated information from the cloud platform module 900 and distributes it to the corresponding vehicle-side module 700 and roadside module 800.

[0111] Through efficient communication interfaces, the RCE can seamlessly collaborate with other traffic participants, such as other vehicles, pedestrians, and infrastructure. For example, at complex intersections, the RCE 804 can help coordinate the safe passage of vehicles from multiple directions, reducing traffic accidents and optimizing traffic flow.

[0112] In some embodiments of the present application, the cloud platform module 900 includes a cloud electronic map unit 901 , a cloud multi-target tracking unit 902 , a cloud scheduling and planning unit 903 , a cloud backend database unit 904 and a cloud communication interface unit 905 .

[0113] The cloud-based electronic map unit 901 stores an electronic map. Based on the target vehicle's VIN and starting point number, it constructs a mapping between the VIN and the actual location and stores it in a backend database. This unit not only provides a global geographic coordinate reference but also contains static information such as road topology, traffic regulations, and parking space distribution. It serves as a fundamental support module for route planning, task scheduling, and multi-vehicle collaboration. Through this electronic map, the system can achieve a transition from local perception to global cognition, enhancing the intelligence of the automated barge system.

[0114] The cloud-based multi-target tracking unit 902 is used to track multiple target vehicles based on a backend database, a spatiotemporal constraint model, and a cross-verification mechanism. The unit then obtains the binding relationship between the target vehicle's vehicle identification code and trajectory segment, and stores it in the backend database. This unit's core function is to maintain continuous identification of each target vehicle in complex environments, preventing issues such as target vehicle code switching and mismatching. By introducing a four-dimensional spatiotemporal correlation model and an improved KM algorithm, this unit significantly improves tracking accuracy and system stability in multi-vehicle scenarios.

[0115] The cloud-based scheduling and planning unit 903 is used to perform dynamic planning based on the backend database to generate the assigned parking space number and planned trajectory of the target vehicle. The consistency between the vehicle identification code of the target vehicle parked in the parking space and the assigned parking space number is verified. If the verification is consistent, the backend database is updated and the mapping relationship between the vehicle identification code and the assigned parking space number is stored. This unit comprehensively considers factors such as vehicle status, road environment, and parking space resources to formulate the optimal scheduling strategy to ensure that all vehicles complete the transfer task efficiently and orderly. In addition, this unit is also responsible for verifying whether the vehicle is accurately parked in the designated parking space. If it is inconsistent, the mapping relationship is corrected and reallocated. It is the core module for realizing closed-loop control of the entire process.

[0116] The cloud-based backend database unit 904 manages the backend database. This unit centrally manages and maintains the entire system's backend database, including the storage, query, and update of various data, including vehicle identification numbers, position information, trajectory segments, parking status, and task progress. This unit serves as the information hub for the entire system, ensuring data consistency, integrity, and real-time availability across all modules. Through its efficient database architecture, the system can rapidly respond to concurrent multi-vehicle operations, supporting efficient operation and intelligent management in large-scale automated transport scenarios.

[0117] The Cloud Communication Interface Unit 905 also supports 5G / V2X communication, enabling high-speed, low-latency data exchange within the automated transfer system. This not only supports real-time information exchange between vehicles and the cloud platform, but also enhances direct communication with other vehicles (V2V), infrastructure (V2I), and pedestrians (V2P), improving the system's interoperability and scalability. This design ensures rapid response for key functions such as dynamic path planning and emergency obstacle avoidance, while providing redundant data transmission paths, enhancing system reliability and fault tolerance, and laying a solid foundation for autonomous driving. Overall, the Cloud Communication Interface Unit 905 makes the entire automated transfer system more intelligent, efficient, and secure.

[0118] In some embodiments of the present application, reference is made to Figure 3 , showing the scene layout of the vehicle offline automatic transfer tracking method and device in actual application. Figure 3 As can be seen in the figure, the entire process begins at the end of the factory production line, at the vehicle off-line starting area. In this area, each vehicle is assigned a corresponding starting point number (such as S1, S2, S3, etc.). In this area, newly produced vehicles are lined up in order to enter the subsequent automatic transfer process.

[0119] After departing from the off-line starting area, vehicles travel along the factory campus roads, passing through several key nodes along the way. These nodes may include sensor monitoring points, communication base stations, and other facilities. These facilities collect real-time information such as the vehicle's location, speed, and direction, and use an improved KM algorithm to accurately match and track the vehicle. As the vehicles continue to move forward, they gradually approach the logistics buffer zone, where they complete the final parking operation.

[0120] The logistics buffer zone is the endpoint of the entire process and the location of the target parking space. This area is divided into a number of orderly arranged parking spaces, each with a clear number (such as A01, A02, A03, etc.) to facilitate system management and scheduling. When a vehicle arrives at its designated target parking space, the system rechecks the consistency of the vehicle identification code with the assigned parking space number to ensure that each vehicle is parked correctly in the corresponding space.

[0121] Throughout the entire process, the coordinated efforts of onboard diagnostic systems, multiple sensors, edge computing units, and cloud platforms enable comprehensive awareness and efficient management of vehicle status and environment. This not only enhances the intelligence level of automated off-line transfer, but also significantly improves system reliability and safety, providing strong support for factory production logistics.

[0122] In summary, the offline automatic transfer tracking method and device provided in the embodiments of the present application have the following technical effects.

[0123] The method and device for tracking automatic off-line transfers provided in the embodiments of the present application first read the vehicle identification number (VIN) and starting point number of the target vehicle through the on-board diagnostic system in the vehicle off-line starting area, and construct a "VIN-actual location" mapping relationship based on the electronic map to ensure the unique identification of each vehicle in the system and the accurate recording of its initial position. In combination with real-time posture, surrounding obstacle information, and road environment data obtained by multiple sensors (such as lidar, cameras, etc.), the system can fully perceive the vehicle status and its surrounding environment, providing a solid data foundation for subsequent precise tracking and path planning. Through the spatiotemporal constraint model and cross-verification mechanism, and using the improved KM algorithm for global optimal matching, the system achieves high-precision multi-target tracking and maintains stable tracking continuity even in complex environments.

[0124] Furthermore, this method and device boast efficient data fusion processing capabilities, ensuring low-latency, high-reliability communication through close collaboration between the vehicle and cloud. The vehicle-side module, comprised of wheel speed sensors, steering wheel angle sensors, IMU inertial navigation sensors, drive-by-wire chassis actuators, and a 5G / V2X communication unit, provides comprehensive vehicle dynamic parameters and rapid response capabilities. The roadside module utilizes lidar, cameras, and edge computing units for real-time monitoring and identification of target vehicles. The cloud platform module is responsible for global scheduling, multi-target tracking, and database management, ensuring the accuracy and efficiency of all operations. Finally, after the target vehicle parks in a parking space, the system verifies the consistency of the vehicle identification code with the assigned parking space number and updates the backend database, ensuring efficient resource management and smooth task completion. These combined designs not only address the high mismatch rate, tracking failure in occluded scenarios, and positioning delays inherent in traditional tracking technologies, but also effectively enhance the intelligence and reliability of the entire automated transfer process.

[0125] In some optional embodiments, the functions / operations mentioned in the block diagram may not occur in the order mentioned in the operation schematic diagram. For example, depending on the functions / operations involved, the two boxes shown in succession may actually be executed substantially simultaneously or the boxes can sometimes be executed in reverse order. In addition, the embodiments presented and described in the flow chart of the present application are provided in an exemplary manner for the purpose of providing a more comprehensive understanding of the technology. The disclosed method is not limited to the operations and logical flows presented herein. Optional embodiments are contemplated in which the order of the various operations is changed and the sub-operations described as a part of a larger operation are performed independently.

[0126] In addition, although the present application is described in the context of functional modules, it should be understood that, unless otherwise stated, one or more of the functions and / or features may be integrated into a single physical device and / or software module, or one or more functions and / or features may be implemented in separate physical devices or software modules. It is also understood that a detailed discussion of the actual implementation of each module is not necessary for understanding the present application. More specifically, given the properties, functions, and internal relationships of the various functional modules in the devices disclosed herein, the actual implementation of the module will be understood within the routine skills of an engineer. Therefore, a person skilled in the art will be able to implement the present application as set forth in the claims using ordinary techniques. It is also understood that the specific concepts disclosed are merely illustrative and are not intended to limit the scope of the present application, which is determined by the full scope of the appended claims and their equivalents.

[0127] If a function 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 the present invention, or the portion that contributes to the prior art, or the portion 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 programs that enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the various embodiments of the method of the present invention. The aforementioned storage medium includes various media that can store program code, such as a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0128] The logic and / or steps represented in a flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable programs for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, a program execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can retrieve and execute a program from a program execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, a program execution system, apparatus, or device.

[0129] More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic devices), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disc read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting, or processing it in a suitable manner as necessary, and then storing it in a computer memory.

[0130] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable program execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used: a discrete logic circuit having logic gate circuits for implementing logic functions on data signals, an application-specific integrated circuit having suitable combinational logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0131] In the above description of this specification, reference to the terms "one embodiment / implementation," "another embodiment / implementation," or "certain embodiments / implementations" means that the specific features, structures, materials, or characteristics described in conjunction with the embodiment or example are included in the embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0132] Although the embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the claims and their equivalents.

[0133] The above is a specific description of the preferred implementation of the present invention, but the present invention is not limited to the embodiments. Those skilled in the art can make various equivalent modifications or substitutions without violating the spirit of the present invention. These equivalent modifications or substitutions are all included in the scope defined by the claims of the present invention.

Claims

1. A method for tracking automatic off-line transfer, characterized in that: The following steps are involved: In the vehicle off-line starting area, the vehicle identification code and starting point number of the target vehicle are read, a mapping relationship between the vehicle identification code and the actual position is constructed, and stored in the background database; Obtaining the real-time position of the target vehicle, surrounding obstacle information, road environment information and buffer parking space status, and storing them in the backend database; According to the backend database, based on the spatiotemporal constraint model and the cross-verification mechanism, multi-target tracking is performed on the target vehicle to obtain the binding relationship between the vehicle identification code and the trajectory segment of the target vehicle, and store it in the backend database; the trajectory segment is used to represent the vehicle trajectory point of the target vehicle in a continuous time period; the vehicle trajectory point is used to represent the real-time position and posture of the target vehicle at a specific time point; Perform dynamic planning based on the background database to generate the assigned parking space number and planned trajectory of the target vehicle, and control the target vehicle to travel along the planned trajectory; Verify the consistency between the vehicle identification code of the target vehicle parked in the parking space and the allocated parking space number; if the verification is consistent, update the background database to store the mapping relationship between the vehicle identification code and the allocated parking space number.

2. The offline automatic transfer tracking method according to claim 1, characterized in that: The method comprises the following steps: reading the vehicle identification code and the starting point number of the target vehicle in the vehicle off-line starting area, constructing a mapping relationship between the vehicle identification code and the actual position, and storing the mapping relationship in the background database. At the vehicle off-line starting area, reading the vehicle identification code and the starting point number; Based on the coordinate information of the starting point number in the electronic map, the actual position of the target vehicle is associated, and a mapping relationship between the vehicle identification code and the actual position is established and stored in the background database.

3. The offline automatic transfer tracking method according to claim 1, characterized in that: The step of acquiring the real-time position of the target vehicle, surrounding obstacle information, road environment information, and buffer parking space status, and storing the information in the backend database comprises the following steps: Obtaining the real-time position and the surrounding obstacle information; Extracting the road environment information and the parking space status of the buffer zone according to the electronic map; wherein the road environment information includes the road topology and traffic rules; the parking space status of the buffer zone includes the vacant / occupied status and the parking space number; The real-time position, the surrounding obstacle information, the road environment information and the parking space status in the buffer zone are subjected to data fusion processing to obtain fused perception data, which is stored in a background database.

4. The offline automatic transfer tracking method according to claim 1, characterized in that: The method includes the following steps: performing multi-target tracking on the target vehicle based on the background database, a spatiotemporal constraint model and a cross-verification mechanism, obtaining a binding relationship between the target vehicle's vehicle identification code and the trajectory segment, and storing the binding relationship in the background database. Performing a first match based on the backend database to construct a four-dimensional spatiotemporal association model, filtering candidate data that meets vehicle identification code binding constraints, spatial constraints, temporal constraints, and motion constraints from the backend database, and constructing a spatiotemporal association graph; in the spatiotemporal association graph, nodes represent the trajectory segments, and edge weights represent the degree of difference between the trajectory segments; A second global matching is performed based on the spatiotemporal correlation graph, and an improved Kuhn-Monkress algorithm is used to solve the minimum weight matching of the spatiotemporal correlation graph to obtain a binding relationship between the vehicle identification code and the trajectory segment, which is stored in the background database.

5. The offline automatic transfer tracking method according to claim 4, characterized in that: When the real-time position of the target vehicle is lost, a third cross-verification is performed, including the following steps: Acquire the speed, steering wheel angle, and heading angle of the target vehicle as chassis bus data and store them in the backend database; Based on the historical pose of the target vehicle in the backend database and the chassis bus data, a cloud-side predicted pose of the target vehicle is calculated by a cloud platform, and a vehicle-side dead reckoning unit is used to calculate a vehicle-side predicted pose of the target vehicle; The cloud-side predicted posture and the vehicle-side predicted posture are longitudinally verified. If the verification passes, the trajectory segment is updated according to the vehicle-side predicted posture, and the binding relationship between the vehicle identification code and the trajectory segment is obtained and stored in the background database.

6. The offline automatic transfer tracking method according to claim 4, characterized in that: The vehicle identification code binding constraint is used to constrain the trajectory segment and the vehicle identification code to be uniquely matched; The real-time position and posture of the target vehicle are obtained through the laser radar and camera, including the laser radar positioning, camera positioning and heading angle; The spatial constraint is used to constrain the error between the current lidar positioning and the lidar positioning of the previous frame to not exceed a first preset error threshold, and the error between the current camera positioning and the camera positioning of the previous frame to not exceed a second preset error threshold; The time constraint is used to constrain the clock time difference between the cameras to not exceed a third preset error threshold; The motion constraint is used to constrain the error between the current heading angle and the heading angle of the previous frame to not exceed a fourth preset error threshold, and the relative speed deviation between the current vehicle speed and the vehicle speed of the previous frame to not exceed a fifth preset error threshold.

7. The offline automatic transfer tracking method according to claim 1, characterized in that: Verifying the consistency between the vehicle identification code of the target vehicle parked in the parking space and the assigned parking space number includes the following steps: Identify the actual parking space number where the target vehicle is parked; Retrieving the assigned parking space number and vehicle identification code of the target vehicle from the backend database; Verify whether the actual parking space number is consistent with the allocated parking space number; If the verification is consistent, the backend database is updated, the buffer parking space status is marked as occupied, the transfer task status of the target vehicle is marked as completed, and the mapping relationship between the vehicle identification code and the allocated parking space number is stored; If the verification is inconsistent, the mapping relationship between the vehicle identification code and the allocated parking space number is corrected and a parking space is reallocated for the target vehicle.

8. An offline automatic transfer tracking device, characterized in that: Including vehicle-side module, road-side module and cloud platform module; The vehicle-end module includes: An on-board diagnostic system unit, used to read the vehicle identification code and starting point number of the target vehicle at the vehicle off-line starting area; A vehicle-side trajectory following control unit is used to receive the planned trajectory sent by the cloud platform module and control the target vehicle to travel along the planned trajectory; A chassis sensing unit, configured to obtain the speed, steering wheel angle, and heading angle of the target vehicle as chassis bus data; The roadside module is used to collect the real-time position and surrounding obstacle information of the target vehicle and identify the actual parking space number where the target vehicle is parked; The cloud platform module includes: A cloud-based electronic map unit stores an electronic map and is used to construct a mapping relationship between the vehicle identification code and the actual location of the target vehicle based on the vehicle identification code and the starting point number, and store the mapping relationship in a backend database; A cloud-based multi-target tracking unit is configured to perform multi-target tracking on the target vehicle based on the backend database, a spatiotemporal constraint model, and a cross-verification mechanism, obtain a binding relationship between the target vehicle's vehicle identification code and the trajectory segment, and store the binding relationship in the backend database; The cloud-based scheduling and planning unit is configured to perform dynamic planning based on the backend database to generate the assigned parking space number and planned trajectory of the target vehicle; verify the consistency between the vehicle identification code of the target vehicle parked in the parking space and the assigned parking space number; if the verification is consistent, update the backend database to store the mapping relationship between the vehicle identification code and the assigned parking space number; The cloud-based backend database unit is used to manage the backend database.

9. The offline automatic transfer tracking device according to claim 8, characterized in that: The road end module includes: A laser radar, used to collect three-dimensional point cloud data and spatial position of the target vehicle; A camera is used to capture an image of the target vehicle's appearance, a road environment image, and an actual parking space image; The edge computing unit is used to detect the real-time position and surrounding obstacle information of the target vehicle and identify the actual parking space number in which the target vehicle is parked.

10. The offline automatic transfer tracking device according to claim 8, characterized in that: The vehicle-side module also includes a vehicle-side dead reckoning unit, which is used to predict the vehicle-side predicted posture of the target vehicle based on the historical posture of the target vehicle in the background database and the chassis bus data.

Images