Unmanned tractor control method with automatic hook supporting function

Abstract

The invention relates to a control method of an unmanned towing vehicle with an automatic hooking and hooking function. The towing vehicle comprises a vehicle body mechanism and an automatic hooking and hooking mechanism, a rear laser radar is arranged on the rear portion, navigation laser and front blind compensation laser are arranged on a vehicle roof, and left and right blind compensation laser are arranged on the two sides. The control method comprises the following steps that the tractor advances to a preset position point and then backs up along a planning curve to approach the skip car. And when the skip car distance is smaller than the recognition distance threshold value, the laser radar is started to recognize the accurate poses of the hooks. And the system continuously detects the distance, if identification is overtime, identification is stopped and whether the identification distance is adjusted is judged, if necessary, parameters are updated for retry, and otherwise, the system returns to the initial state. After the identification is successful, judging transverse deviation: if the deviation is too large, advancing, adjusting the vehicle body and reversing again; and if the deviation is within the allowable range, changing the end point of the path, closing the obstacle avoidance function, and continuing to reverse until the docking is completed. According to the method, full-automatic docking is realized, the fault-tolerant capability is realized, the collision risk is avoided, and the docking success rate is improved.

Description

Technical Field

[0001] This invention relates to the field of automatic traction hook technology, and in particular to a control method for an unmanned tractor with automatic hook-up function. Background Technology

[0002] With the rapid development of intelligent manufacturing and smart logistics, automated guided vehicles (AGVs) are widely used in factories, warehouses, airports, and other scenarios, undertaking the automatic transfer tasks of non-powered vehicles such as material carts, baggage carts, and trailers. Among them, the automatic hook docking between the AGV and the material cart is a key link to achieve fully unmanned operation.

[0003] Currently, existing automated docking solutions can be mainly categorized as follows: Firstly, there are docking systems based on manual assistance or semi-automatic methods. Operators control the tractor to reverse via remote control or onboard human-machine interface and rely on visual observation to complete the hook-up. While this method reduces labor intensity to some extent, it still requires human intervention and cannot achieve truly unmanned operation. Furthermore, docking efficiency is greatly affected by the operator's skill level, posing safety hazards.

[0004] Secondly, there are docking solutions based on a single sensor (such as a regular camera or ultrasonic radar). These solutions typically use visual recognition of the material cart hook or ultrasonic ranging to achieve a backward approach. However, cameras are susceptible to environmental factors such as lighting, dust, and obstructions, resulting in insufficient recognition stability; ultrasonic radar can only provide distance information and cannot obtain the precise position of the hook, making it difficult to guide the tractor to achieve high-precision docking. When there is a significant lateral or angular deviation in the material cart's parking position, these solutions often cannot automatically adjust, easily leading to docking failure or even collisions.

[0005] Thirdly, there is the LiDAR-based identification and docking solution. Some existing technologies use LiDAR to scan the front face of the material cart or the hook, and then use point cloud processing to estimate the pose. However, existing methods generally have the following shortcomings: They lack a systematic area filtering and multi-dimensional target screening mechanism during point cloud processing, making them susceptible to environmental interference (such as adjacent parked carts, pillars, walls, etc.), leading to misidentification; they lack mechanisms for handling abnormal situations such as identification timeouts and excessive lateral deviations, resulting in poor system fault tolerance; when there is a large lateral deviation between the material cart and the center of the storage location, existing methods often require replanning the global path or directly terminating the task, lacking adaptability; furthermore, existing solutions still have room for optimization in terms of hook pose calculation accuracy, multi-frame filtering smoothing, and dynamic adjustment of reference lines, making it difficult to balance docking success rate and operational efficiency.

[0006] In summary, existing unmanned tractor automatic hook-up technologies generally suffer from problems in practical applications, such as poor environmental adaptability, insufficient recognition robustness, low docking success rate, and imperfect anomaly handling mechanisms. Summary of the Invention

[0007] To address the aforementioned technical problems, the present invention aims to provide a control method for an unmanned tractor with an automatic hook-and-drop function, which can achieve efficient, stable, and safe automatic docking between the tractor and the material car through multi-sensor fusion, systematic point cloud processing and pose calculation, and flexible path adjustment strategies.

[0008] To achieve the objectives of the invention described above, the present invention adopts the following technical solution: A control method for an unmanned tractor with an automatic hook-up function is disclosed. The tractor includes a vehicle body structure and an automatic hook-up mechanism fixed to the rear of the vehicle body structure. A rear lidar for detecting the hook position is also provided on the upper part of the automatic hook-up mechanism on the vehicle body structure. A navigation laser and a front blind spot laser are also provided on the top of the vehicle body structure, and left and right blind spot lasers are provided on both sides of the vehicle body structure. The specific control steps are as follows: The tractor first moves forward to the predetermined position, then reverses from there, following a planned curved trajectory to approach the material car parked in the material car parking area C. When the distance to the material car is less than the recognition distance threshold, a recognition signal is issued. After activation, the LiDAR scanner scans and detects the precise pose of the material car's hook. The system in the industrial control computer on the tractor continuously monitors whether the distance to the material car is still within the recognition range. If it is, it waits to obtain the recognition result. If no result is obtained after the recognition timeout, a stop recognition signal is issued and it is determined whether the recognition distance needs to be changed. If it needs to be changed, the recognition distance parameter is updated and the recognition signal is reissued to continue trying. If it does not need to be changed, a stop recognition signal is issued to return to the initial state. After successful identification, the system first checks whether the lateral deviation between the tractor and the material car is less than the threshold. If the deviation is too large, the system moves forward to adjust the vehicle body to reduce the lateral deviation and then returns to the reversing state to approach again. If the lateral deviation is within the allowable range, the system directly changes the destination of the path, disables the obstacle avoidance function, and continues to reverse until the goods are detected and the task is completed.

[0009] As a preferred solution, the specific process of using LiDAR scanning to detect and identify the hooks of the material cart is as follows: Step S1: Parking Space Information Initialization: The tractor receives parking space information from the dispatch system and generates a search area for the trailer's front end. The starting point A(xa, ya) and the ending point B(xb, yb) of the parking space are connected. The direction of the line connecting the starting point A(xa, ya) and the ending point B(xb, yb) is the reference centerline for the trailer's expected orientation, which will be used later to determine whether the trailer's front end attitude exceeds the allowable deviation. Quadrilateral A0A1B1B0 is defined as the valid trailer search area. Step S2, Point Cloud Filtering: Includes the following steps: Vehicle filtering: Filter out point clouds hitting the tractor itself; Parking area filtering: Based on the quadrilateral search area generated in step S1, only retain the point cloud inside quadrilateral A0A1B1B0, and filter out irrelevant point clouds outside the area. Step S3, Point Cloud Clustering: Cluster the filtered point clouds based on spatial proximity, group adjacent point clouds according to a distance threshold, and group points with a distance less than the threshold into the same category, ultimately forming several independent point cloud clusters, and calculate the convex hull, length and width of the outer contour of each cluster. Step S4, End Face Recognition: The geometric feature of the trailer's front face is an approximately straight line segment with a known width. This feature is used to filter out the trailer's front face from the clustering results. Step S5, Target Tracking: Perform inter-frame tracking on the identified front face target. If the candidate target in the current frame matches the tracked target in the previous frame in terms of position and orientation, update the tracking list and perform temporal filtering on the position and orientation of the front face based on the extended Kalman filter model to suppress single-frame measurement noise. If no match is found, add the target as a new target to the tracking list. Step S6, Pose Calculation: This includes two stages: coarse positioning and fine calculation. Step S61, coarse positioning stage: Based on the RANSAC algorithm, straight line fitting is performed on the candidate point cloud of the front face to preliminarily determine whether the pose of the front face meets the docking requirements. Step S62, Fine Calculation Stage: The pose of the front face is accurately calculated using the least squares method for the interior point cloud that has passed the coarse positioning stage. Step S7: Hook pose calculation. Based on the front face pose calculation results, calculate the precise position of the hook. Step S8: The tractor obtains the identified hook position and the dispatch system issues the driving reference line L. The tractor plans the driving path from the current location of the vehicle along the reference line L to the hook position and performs automatic towing and docking operation.

[0010] As a preferred solution, when there is a lateral deviation between the trailer's parking position and the center of the storage space, causing the hook to shift laterally relative to the reference line L, a reference line offset adjustment step is also included. Let the vertical distance from the hook position (Phook) to the reference line L be the horizontal offset δ, and define two thresholds: Docking threshold δdock: If the lateral offset is within this range, the tractor can directly complete the docking along the original reference line without adjustment; Safety threshold δsafe: The upper limit of the lateral offset. If this value is exceeded, the deviation is considered too large and safe docking cannot be achieved. The system will alarm and terminate the task. When δdock < δ ≤ δsafe, the specific method for adjusting the reference line offset is as follows: The original reference line L is shifted a distance δ along its normal vector direction to generate a new offset reference line L'. The offset direction is from the reference line towards the hook position, i.e.: ,in The unit normal vector of the reference line L is directed towards the hook side; the offset reference line L' passes through the lateral position of the projection point of the hook position on the original reference line, so that when the tractor travels along the new reference line L', the center line of the vehicle body is laterally aligned with the hook, thus enabling successful docking.

[0011] As a preferred option, the reference line offset adjustment step is followed by a re-interlocking process after the offset: After the reference line offset is complete, the tractor unit needs to perform the following sequence of actions to enter the new reference line L' and restart automatic docking: 1. Moving forward and out: The tractor unit moves forward in the current direction of travel, leaving the docking area in front of the trailer and moving to a safe distance; 2. Lateral switching: The tractor moves laterally from the original reference line L to the offset reference line L', so that the center line of the vehicle body is aligned with L'; 3. Reversing into the area: The tractor unit reverses along the new reference line L' and re-enters the trailer docking area with its rear facing the direction of the hook-up. 4. Restart docking: After reversing into position, restart the process of scanning and identifying the hook of the material car with the laser radar (6) after restarting, and complete the precise docking based on the new reference line L'.

[0012] As a preferred embodiment, in step 1, the unit vector from the berth end point B to the berth start point A is: (xv, yv), The two normal vectors are respectively (-yv, xv) and (yv, -xv), berth width is w; based on the above information, generate a search quadrilateral A0A1B1B0, with the coordinates of its four vertices as follows: Wherein: the starting point A of the berth is on the side where the tractor approaches the material car, and the ending point B of the berth is on the side where the material car body is in the longitudinal direction; the width w of the berth is greater than the width of the material car body; the line connecting the starting point A and the ending point B of the berth is the centerline. Central axis This serves as a benchmark for the expected orientation of the trailer, used for subsequent positional deviation determination.

[0013] As a preferred embodiment, step S2 further includes discrete point filtering: taking any point P as the center and r as the radius, count the number N of neighboring points within the radius. If N is less than a set threshold, then point P is considered a discrete noise point and is filtered out.

[0014] As a preferred embodiment, the rules for end face recognition in step S4 are as follows: 1. Distance sorting: Project the center points of all clustered targets onto the centerline of the berth. The targets are sorted from closest to furthest from the starting point A of the parking space, with priority given to targets closer to the towing vehicle. 2. Size screening: Select targets whose length matches the width of the trailer's front face, and eliminate interfering objects whose dimensions do not match; 3. Direction filtering: Based on the centerline of the berth. The direction information is used to determine whether the angle between the principal direction angle of the candidate target and the direction of the normal to the central axis exceeds a threshold; if it exceeds the threshold, the target is considered to deviate from the expected orientation and is excluded. 4. Straightness test: Perform straight line fitting on the point cloud of the candidate target and calculate the fitting residual. The front face of the trailer is a planar structure, and its point cloud projection should have good straightness. Targets with residuals exceeding the threshold are excluded.

[0015] As a preferred embodiment, the specific process of step S61, the coarse positioning stage, is as follows: 1. Select the projection point pt that is closest to the starting point A of the berth; 2. Select point clouds from all clustered targets whose distance from pt is less than maxDist, where maxDist = w ×sin(θ_max × π / 180), and θ_max is the maximum allowable deviation angle for trailer parking; 3. Perform RANSAC line fitting on the point cloud that meets the requirements to obtain the direction vector of the fitted line and a point on the line; 4. Calculate the angle between the fitted straight line and the normal direction of the centerline of the berth. If the angle exceeds the maximum allowable deviation angle θmax, the current frame detection is considered to have failed, and the next detection cycle begins.

[0016] As a preferred embodiment, the specific process of step S62, the fine calculation stage, is as follows: 1. Line Fitting: Perform least-squares line fitting on the set of interior points to obtain the direction vector of the front face. and the fitted linear equation; 2. Midpoint calculation: Project all interior points onto the fitted line, and take the midpoint of the projected points as the center point Pf(xf, yf) of the front face; 3. Orientation calculation: The direction of the front face normal vector is the trailer orientation θ = atan2(v_fit_x, -v_fit_y). The normal vector is perpendicular to the fitted line and points inward to the vehicle body. 4. Multi-frame accumulation: The weighted average of N frames is calculated by accumulating the results of multiple frames and taking the average value.

[0017] As a preferred embodiment, the precise position of the hook in step S7 is calculated as follows: xhook = xf + d×cosθ, yhook = yf + d×sinθ, Where d is the known fixed offset of the hook from the front face, and θ is the trailer orientation angle; the final output is the hook pose (xhook, yhook, θ), which is used by the tractor path planning module to perform automatic docking operation.

[0018] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention's method defines a complete process: "advancing to a predetermined position, reversing along a planned trajectory to approach, identifying the hook's pose, adjusting deviations, and completing the docking." This achieves fully automated operation from the initial position to the final docking, requiring no manual intervention. Furthermore, it introduces two judgment nodes: an "identification distance threshold" and a "lateral deviation threshold." When identification times out, the system can dynamically adjust the identification distance parameter and retry, demonstrating the system's fault tolerance and adaptability. When the lateral deviation is too large, the system automatically performs a "forward adjustment of the vehicle body" operation instead of forcibly reversing, avoiding collision risks and improving the docking success rate. Attached Figure Description

[0019] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments of this application and their descriptions are used to explain this application and do not constitute a limitation thereof.

[0020] Figure 1 This is a schematic diagram of the overall structure of the unmanned tractor vehicle of the present invention; Figure 2 and Figure 3 These are schematic diagrams of the automatic unhooking mechanism of the present invention from two different angles; Figure 4 This is a schematic diagram of the overall structure of the trailer of the present invention; Figure 5 This is a schematic diagram of the docking operation between the unmanned tractor and trailer of the present invention; Figure 6 This is a schematic diagram of the process of docking the unmanned tractor and trailer of the present invention; Figure 7 This is a flowchart illustrating the automatic hook-up posture recognition process of the unmanned tractor of the present invention. Figure 8 This is a schematic diagram of a typical operation process of the unmanned tractor vehicle of the present invention in a dual-cargo-area warehouse environment.

[0021] The labels in the attached diagram are as follows: 100, vehicle body mechanism; 1, navigation laser; 2, front blind spot laser; 3, left and right blind spot lasers; 5, industrial control computer; 6, rear lidar; 7, automatic hook release mechanism; 71, mounting bracket; 72, pin positioning bar; 73, support plate; 74, pin; 75, locking pin; 76, spring; 77, connecting socket; 78, electric push rod; 79, upper proximity switch; 710, lower proximity switch; 711, support plate. Detailed Implementation

[0022] It should be noted that the following detailed descriptions are illustrative and intended to provide further explanation of this application. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0023] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments according to this application. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0024] Furthermore, in the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "clockwise," and "counterclockwise," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0025] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, unless otherwise stated, "a plurality of" means two or more, unless explicitly defined otherwise.

[0026] In this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0027] In this invention, unless otherwise explicitly specified and limited, "above" or "below" the second feature can include direct contact between the first and second features, or contact between the first and second features through another feature between them. Furthermore, "above," "over," and "on top" of the second feature includes the first feature directly above or diagonally above the second feature, or simply indicates that the first feature is at a higher horizontal level than the second feature. "Below," "below," and "under" the second feature includes the first feature directly below or diagonally below the second feature, or simply indicates that the first feature is at a lower horizontal level than the second feature.

[0028] The present invention will be further described below with reference to the accompanying drawings and embodiments: like Figures 1 to 3 As shown, an unmanned tractor with an automatic hook-and-hook mechanism includes a body structure 100 and an automatic hook-and-hook mechanism 7 fixed to the rear of the body structure 100. The automatic hook-and-hook mechanism 7 includes a mounting frame 71, a pin 74, a connecting socket 77, and an electric push rod 78. The connecting socket 77 is fixed to the mounting frame 71. The pin 74 is inserted into the top plate of the mounting frame 71 and the connecting socket 77. The top of the pin 74 is also provided with a support plate 73. The electric push rod 78 is fixed to the mounting frame 71. The connecting socket 77 is provided with at least one proximity switch for detecting whether the hook is in place. The body structure 100 is also provided with an industrial control computer 5. The industrial control computer 5 controls the electric push rod 78 to extend or retract through the signal of the proximity switch, thereby driving the pin 74 to disengage from or insert into the connecting socket 77 through the support plate 73, and to pull it out or insert it into the hook, so that the hook is detached from or connected to the body structure.

[0029] The pin 74 is also equipped with a locking pin 75, and a spring 76 is fitted onto the pin 74. The two ends of the spring 76 abut against the top plate of the mounting bracket 71 and the locking pin 75, respectively. Under the action of the spring 76, the pin 74 is always inserted into the connecting socket 77. The spring provides a continuous preload, keeping the pin inserted under normal conditions, thus providing a mechanical anti-disengagement function. Even if the electric actuator fails or the power is cut off, the pin will not accidentally disengage, improving safety. Simultaneously, during disengagement, the electric actuator must overcome the spring force to lift the pin, ensuring the certainty of the disengagement action and avoiding accidental disengagement due to vibration or impact.

[0030] The top plate of the mounting bracket 71 is also equipped with a pin positioning bar 72, and the support plate 73 is slidably sleeved on the pin positioning bar 72. The pin positioning bar guides and limits the lifting and lowering movement of the support plate, ensuring that the pin remains vertical and does not deviate during the up and down movement, thus improving the alignment accuracy of the pin with the connecting socket and hook hole. At the same time, this structure reduces the risk of friction and jamming between moving parts, improving the reliability of long-term operation.

[0031] The connecting slot 77 is a flared guide groove. The flared structure provides a certain margin of error when the hook is inserted, allowing it to be smoothly guided into the connecting slot even with slight horizontal deviations. This design reduces the positioning accuracy requirements for unmanned tractor docking during reversing, and improves the success rate and adaptability of automatic docking.

[0032] Two electric actuators 78 are provided, each contacting one end of the support plate 73. Using two actuators to synchronously drive the support plate ensures more even force distribution compared to single-point drive, preventing pin tilting or jamming due to uneven load. Furthermore, the dual actuators are redundant; even if one fails, the other can still perform emergency disengagement or maintain the system's state, improving fault tolerance.

[0033] The mounting bracket 71 has side wing plates extending to both sides from its lower part, and two electric actuators 78 are fixed to the side wing plates by support plates 711. The side wing plate structure provides a stable mounting base for the electric actuators, allowing the thrust line of the electric actuators to act perpendicularly to both ends of the support plate, reducing additional bending moment. This layout makes full use of the space under the mounting bracket, has a compact structure, and facilitates the disassembly and maintenance of the electric actuators.

[0034] The upper and lower ends of the connection port 77 are respectively equipped with an upper proximity switch 79 and a lower proximity switch 710. The industrial control computer 5 controls the extension or retraction of the electric actuator 78 through signals from the upper proximity switch 79 and / or the lower proximity switch 710. Through the two proximity switches, the system can simultaneously detect whether the hook is fully inserted and whether it is at a critical position about to disengage. This dual detection mechanism can effectively determine different states of the hook, providing a more accurate basis for the timing of the electric actuator's action, preventing accidental locking when the hook is not fully inserted, or accidental movement when it is not fully disengaged.

[0035] The mounting bracket 71 is also equipped with a limit switch 712 located below the connection socket 77 to determine whether the pin 74 is fully inserted. This limit switch directly detects the pin's descent into place, serving as a final confirmation signal that the pin is fully inserted into the hook hole. Together with the proximity switch, it forms a closed-loop feedback to ensure that the pin is indeed locking the hook in the connected state, preventing disengagement during operation due to the pin not being fully inserted, thus further improving connection safety.

[0036] The vehicle body structure 100 is also equipped with a rear lidar 6 above the automatic unhooking mechanism 7 to detect the hook position. The rear lidar can detect the three-dimensional position of the hook in real time during the unmanned tractor's reversing docking process, providing the vehicle with accurate navigation and positioning information. Through data fusion with the industrial control computer, precise alignment and insertion can be achieved, significantly improving the success rate of automatic docking, and can identify whether there are any abnormalities or obstacles in the hook before docking.

[0037] The top of the vehicle body structure 100 is equipped with a navigation laser 1 and a front blind spot laser 2, while the sides of the vehicle body structure 100 are equipped with left and right blind spot lasers 3. This multi-laser radar layout constructs an omnidirectional perception system. The top navigation laser is used for global positioning and path planning; the front blind spot laser eliminates the near-field blind spot in front of the vehicle, facilitating precise docking; the side blind spot lasers cover the sides of the vehicle body, ensuring safe operation in narrow passages or complex environments. Overall, this improves the environmental perception capability of the unmanned tractor, providing a reliable positioning and safety foundation for automatic uncoupling operations.

[0038] Figure 4The demonstration showcased a standardized trailer structure designed for use with a tractor unit: the trailer features a dual-wheel chassis design with a triangular drawbar at the front. A passive hook (female head, U-shaped or annular opening) is mounted at the top of the drawbar for docking with the active hook (male head) at the rear of the tractor unit. The trailer body is surrounded by guardrails for easy loading of materials, and the floor can accommodate multiple cargo boxes or standardized cargo units such as battery packs. Multiple trailers of this type can be grouped together in a "train-like" configuration, significantly improving the efficiency of a single transport operation. This is particularly suitable for large-volume material delivery, inter-workshop transfers, and warehouse inbound / outbound scenarios. Compared to a single-trailer configuration, multi-trailer grouping can increase transport efficiency by 3-5 times, while reducing the number of trips by the tractor unit, thus lowering energy consumption and equipment wear.

[0039] Figure 5 The diagram illustrates the complete path and process of automatic towing operation for a tractor-trailer: the tractor-trailer first moves forward to the predetermined position A, then reverses from A along a planned curved trajectory (an arc from A to B) to approach the target trailer parked in the trailer parking area C. When the tractor-trailer reaches point B, the rear lidar begins scanning and detecting the precise position of the trailer hook. Based on the recognition results, the system determines whether lateral deviations require path adjustment, ultimately achieving precise hook-up. This diagram clearly illustrates the core operation process of the invention: moving forward to position (point A), planning the reversing path, approaching backward (the curve from A to B), the lidar starting identification and detection (point B), and precise hook-up, demonstrating the complete process of automatic hook-up technology from macroscopic path planning to microscopic precise control.

[0040] Figure 6 This demonstrates an adaptive automated cargo retrieval process based on LiDAR recognition: The tractor begins reversing to approach the trailer. When the trailer distance is less than the recognition distance threshold, a recognition signal is issued to initiate radar scanning. The system continuously monitors whether the trailer distance is still within the recognition range. If it is, it waits to obtain the recognition result. Upon successful recognition, it first checks whether the lateral deviation is less than the threshold. If the deviation is too large, it performs a forward adjustment to reduce the lateral deviation and then returns to the reversing state to approach again. If the lateral deviation is within the allowable range, it directly changes the path endpoint, disables obstacle avoidance, and continues reversing until the collision sensor confirms that the cargo has been retrieved and the task is completed. If the recognition timeout occurs and no result is obtained, a stop recognition signal is issued, and it is determined whether the recognition distance needs to be changed (shortened from 1.5m to 1.0m for a second retry). If a change is needed, the recognition distance parameters are updated, and a new recognition signal is issued to continue trying. If no change is needed, a stop recognition signal is issued to return to the initial state. This flowchart fully embodies the core technical features of this invention: a dual recognition retry mechanism, adaptive judgment and adjustment of lateral deviation, dynamic path correction, and closed-loop feedback control, ensuring that high-precision automated cargo retrieval operations can still be successfully completed even when there is a deviation in the trailer position.

[0041] like Figure 7 As shown, the specific process of rear lidar scanning detection and identification of the hook of the material cart is as follows: Step S1: Initialize parking space information: Based on the parking space information received from the scheduling system (including the starting point, ending point, and width of the parking space), generate the search area for the front face of the trailer.

[0042] Assume the starting point of the berth is A(xa, ya) and the ending point is B(xb, yb). The unit vector from the ending point to the starting point is... (xv, yv), The two normal vectors are respectively (-yv, xv) and (yv, -xv), berth width is w.

[0043] Based on the above information, a search quadrilateral A0A1B1B0 is generated, with the coordinates of its four vertices as follows: Wherein: the starting point A is the side where the tractor approaches the trailer, and the ending point B is the side where the trailer body extends longitudinally. The parking space width w should be greater than the trailer body width, leaving a certain margin to allow for lateral deviations in trailer parking. The line connecting the starting and ending points is the reference centerline for the trailer's expected orientation, which is subsequently used to determine whether the trailer's front face attitude exceeds the allowable deviation. This quadrilateral defines the effective trailer search area. Point clouds outside the area will be filtered out in subsequent filtering steps, thus significantly reducing the processing range. Meanwhile, the centerline... It serves as a benchmark for the expected orientation of the trailer and is used for subsequent positional deviation determination.

[0044] Step S2, Point Cloud Filtering Module: The purpose of point cloud filtering is to remove invalid point clouds, thereby accelerating the processing efficiency of subsequent steps. It includes the following steps: Self-filtering: Filters out point clouds hitting the tractor itself.

[0045] Parking area filtering: Based on the quadrilateral search area generated in step 3.2, only the point cloud inside quadrilateral A0A1B1B0 is retained, and irrelevant point clouds outside the area are filtered out.

[0046] Discrete point filtering: With any point P as the center and r as the radius, count the number N of neighboring points within the radius. If N is less than a set threshold, point P is considered a discrete noise point and is filtered out.

[0047] Step S3, Point Cloud Clustering: The filtered point clouds are clustered based on spatial proximity. Adjacent point clouds are grouped according to a distance threshold. Points with a distance less than the threshold are grouped into the same category, ultimately forming several independent point cloud clusters. The geometric features such as the convex hull, length, and width of the outer contour of each cluster are calculated.

[0048] Step S4, End Face Identification: The trailer's front face has obvious geometric features: an approximately straight line segment with a known width. Using this prior feature, the trailer's front face is selected from the clustering results. Distance sorting: Project the center points of all clustered targets onto the centerline of the berth. The targets are sorted from closest to furthest from the starting point A of the parking space, with priority given to targets that are closer to the towing vehicle.

[0049] Size filtering: Select targets whose length matches the width of the trailer's front face (within the set tolerance range) and exclude interfering objects whose size is obviously inconsistent.

[0050] Direction filtering: based on the centerline of the berth The system uses directional information to determine whether the angle between the principal orientation angle of a candidate target and the normal direction of the central axis exceeds a threshold. If it exceeds the threshold, the target's orientation is considered to deviate significantly from the expected orientation and is therefore excluded.

[0051] Straightness check: Perform straight line fitting on the point cloud of candidate targets and calculate the fitting residual. The front face of the trailer is a planar structure, and its point cloud projection should have good straightness. Targets with residuals exceeding the threshold are excluded.

[0052] Step S5, Target Tracking: Perform inter-frame tracking on the identified front-end target. If the candidate target in the current frame matches the tracked target in the previous frame in terms of position and orientation, update the tracking list and perform temporal filtering on the position and orientation of the front-end face based on the Extended Kalman Filter (EKF) model to suppress single-frame measurement noise. If no match is found, add the target as a new target to the tracking list.

[0053] Step S6, pose calculation, includes two stages: Step S61, coarse localization stage: Based on the RANSAC algorithm, straight line fitting is performed on the candidate point cloud of the front face to preliminarily determine whether the pose of the front face meets the docking requirements. Select the projection point pt that is closest to the starting point A of the berth.

[0054] Select point clouds from all clustered targets whose distance from pt is less than maxDist, where maxDist = w ×sin(θ_max × π / 180), and θmax is the maximum allowable deviation angle for trailer parking.

[0055] Perform RANSAC line fitting on the point cloud that meets the requirements to obtain the direction vector of the fitted line and a point on the line.

[0056] Calculate the angle between the fitted straight line and the normal direction of the centerline of the berth. If the angle exceeds the maximum allowable deviation angle θmax, the current frame detection is considered to have failed, and the next detection cycle begins.

[0057] Step S62, Fine Calculation Stage: For the inliers point cloud obtained from the coarse positioning stage, the least squares method is used to accurately calculate the pose of the front face. Line Fitting: Perform least-squares line fitting on the set of interior points to obtain the direction vector of the front face. And fit the linear equation.

[0058] Midpoint calculation: Project all interior points onto the fitted line, and take the midpoint of the projected points as the center point Pf(xf, yf) of the front face.

[0059] Orientation calculation: The direction of the front face normal vector is the trailer orientation θ = atan2(v_fit_x, -v_fit_y) (the normal vector is perpendicular to the fitted line and points inward to the vehicle body).

[0060] Multi-frame accumulation: Due to the limited number of point clouds in a single frame, the calculation result of a single frame has a certain error. A multi-frame accumulation and averaging method is adopted to continuously calculate the weighted average of the results of N frames, thereby improving pose accuracy.

[0061] Step S7, Hook Position Calculation: Based on the front face pose calculation results, calculate the precise position of the hook: xhook = xf + d × cosθ, yhook = yf + d × sinθ, where d is the known fixed offset of the hook from the front face, and θ is the trailer orientation angle. The final output is the hook pose (xhook, yhook, θ), which is used by the tractor path planning module to perform automatic docking operations.

[0062] Step S8, Reference Line Offset Adjustment: The driving reference line L issued by the dispatch system is located on the center line of the parking space (i.e., the line connecting the starting point A and the ending point B of the parking space). Ideally, the trailer is parked exactly in the center of the parking space, the hook is on the reference line, and the tractor can complete the docking by driving straight along the reference line. However, in actual working conditions, there is a lateral deviation in the parking of the trailer, which causes the hook to shift laterally relative to the reference line.

[0063] Let δ be the vertical distance (lateral offset) from the hook position Phook to the reference line L, and define two thresholds: δdock (docking threshold): If the lateral offset is less than this value, the tractor can directly complete the docking along the original reference line without adjustment.

[0064] δsafe (safety threshold): The upper limit of the lateral offset. If this value is exceeded, the deviation is considered too large, and safe docking is not possible. The system will then alarm and terminate the task.

[0065] When δdock < δ ≤ δsafe, the specific method for adjusting the reference line offset is as follows: The original reference line L is shifted a distance δ along its normal vector direction to generate a new offset reference line L'. The offset direction is from the reference line towards the hook position, i.e.: .

[0066] in Let L' be the unit normal vector of the reference line L, pointing towards the hook side. The offset reference line L' passes through the lateral position of the projection point of the hook position on the original reference line, so that when the tractor travels along L', the center line of the vehicle body is laterally aligned with the hook, thus enabling a smooth docking.

[0067] Step S9, Re-docking process after offset: After the reference line offset is completed, the tractor needs to perform the following sequence of actions to enter the new reference line L' and restart automatic docking: Forward movement: The tractor moves forward in the current direction of travel, moves away from the docking area in front of the trailer, and moves to a safe distance (ensuring sufficient adjustment space for subsequent reversing).

[0068] Lateral switching: The tractor moves laterally from the original reference line L to the offset reference line L', so that the center line of the vehicle body is aligned with L'.

[0069] Reverse entry: The tractor unit reverses along the new reference line L' and re-enters the trailer docking area with its rear facing the hook-up direction.

[0070] Restart docking: After reversing into position, the system re-executes the complete hook identification and automatic docking process (steps S1 to S8), and completes precise docking based on the new reference line L'.

[0071] The entire process can be simplified to: moving forward, changing lanes, reversing, and re-docking. This process ensures that after the lateral deviation is eliminated, the tractor can approach the hook again with the correct attitude and path, ultimately completing the trailer docking.

[0072] Figure 8The demonstration showcased a typical operational process for a tractor unit in a dual-cargo-area warehousing environment: After receiving a task from the dispatch system, the tractor unit departs from the standby area and heads towards the source cargo area (e.g., area A). Navigation lasers guide it to the vicinity of the designated parking space. As the vehicle reverses and approaches the target trailer, the rear laser radar continuously scans and identifies the hook position. Based on the identification results, the system determines whether lateral deviation needs adjustment. If the deviation is significant, the system automatically advances a distance to correct the path and approaches again. Once the deviation is acceptable, the tractor unit precisely reverses and docks until the collision sensor confirms successful hook-up. Subsequently, the tractor unit pulls the trailer along the central aisle to the target cargo area (e.g., area B). During this process, navigation lasers, front blind spot lasers, and left and right blind spot lasers provide real-time obstacle avoidance and coordinate with other tractor units for obstacle avoidance. Upon reaching the destination, the tractor unit reverses to the unloading position for precise docking, unlocks the hook, and advances to detach from the trailer to complete unloading. Finally, the tractor unit returns to the standby area or directly executes the next handling task. The entire process enables automated and unmanned cross-area handling operations between different cargo areas.

[0073] The dispatching system intelligently allocates tasks based on task priority and tractor location, avoiding conflicts caused by multiple tractors entering adjacent parking spaces in the same cargo area at the same time. When two tractors are traveling towards each other in the central aisle, navigation lasers, front blind spot lasers, and left and right blind spot lasers detect each other's positions in real time and maintain a safe lateral distance to ensure safe passing. If a tractor is performing hook-up operations in area A, the dispatching system will prioritize assigning other tasks in area A to another tractor or wait for the current tractor to complete its task before entering, thereby avoiding operational interference.

[0074] Through this multi-vehicle collaborative mechanism, the system can realize bidirectional parallel handling from area A to area B and from area B to area A, increasing the overall warehousing throughput several times. At the same time, the real-time perception capability of LiDAR ensures safety in the multi-vehicle environment. Even if dynamic situations such as personnel entering or temporary obstacles occur, the tractor can automatically detect and take measures such as slowing down, stopping, or detouring, ensuring the efficient, safe, and reliable operation of the entire logistics system.

[0075] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0076] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention without departing from the principles and spirit of the present invention. Any simple modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention shall still fall within the scope of the technical solution of the present invention.

Claims

1. A control method for an unmanned tractor with an automatic hook-and-lift function, characterized in that: The tractor unit includes a body structure (100) and an automatic unhooking mechanism (7) fixed to the rear of the body structure (100). The body structure (100) is also equipped with a rear lidar (6) for detecting the hook position on the upper part of the automatic unhooking mechanism (7). The top of the body structure (100) is also equipped with a navigation laser (1) and a front blind spot laser (2). The sides of the body structure (100) are also equipped with left and right blind spot lasers (3). The specific control steps are as follows: The tractor first moves forward to the predetermined position, and then reverses from the predetermined position, reversing along the planned curved trajectory to approach the material car parked in the material car parking area C. When the distance to the material car is less than the recognition distance threshold, a recognition signal is issued. After activation, the lidar (6) scans and detects the precise pose of the hook of the material car. The system in the industrial control computer (5) on the tractor continuously detects whether the distance to the material car is still within the recognition range. If it is within the range, it waits to obtain the recognition result. If no result is obtained after the recognition timeout, a stop recognition signal is issued and it is determined whether the recognition distance needs to be changed. If it needs to be changed, the recognition distance parameter is updated and the recognition signal is reissued to continue trying. If it does not need to be changed, a stop recognition signal is issued to return to the initial state. After successful identification, the system first checks whether the lateral deviation between the tractor and the material car is less than the threshold. If the deviation is too large, the system moves forward to adjust the vehicle body to reduce the lateral deviation and then returns to the reversing state to approach again. If the lateral deviation is within the allowable range, the system directly changes the destination of the path, disables the obstacle avoidance function, and continues to reverse until the goods are detected and the task is completed.

2. The control method for an unmanned tractor with automatic hook-lifting function according to claim 1, characterized in that, The specific process of the rear lidar (6) scanning and detecting the hook of the material cart is as follows: Step S1: Parking Space Information Initialization: The tractor receives parking space information from the dispatch system and generates a search area for the trailer's front end. The starting point A(xa, ya) and the ending point B(xb, yb) of the parking space are connected. The direction of the line connecting the starting point A(xa, ya) and the ending point B(xb, yb) is the reference centerline for the trailer's expected orientation, which will be used later to determine whether the trailer's front end attitude exceeds the allowable deviation. Quadrilateral A0A1B1B0 is defined as the valid trailer search area. Step S2, Point Cloud Filtering: Includes the following steps: Vehicle filtering: Filter out point clouds hitting the tractor itself; Parking area filtering: Based on the quadrilateral search area generated in step S1, only retain the point cloud inside quadrilateral A0A1B1B0, and filter out irrelevant point clouds outside the area. Step S3, Point Cloud Clustering: Cluster the filtered point clouds based on spatial proximity, group adjacent point clouds according to a distance threshold, and group points with a distance less than the threshold into the same category, ultimately forming several independent point cloud clusters, and calculate the convex hull, length and width of the outer contour of each cluster. Step S4, End Face Recognition: The geometric feature of the trailer's front face is an approximately straight line segment with a known width. This feature is used to filter out the trailer's front face from the clustering results. Step S5, Target Tracking: Perform inter-frame tracking on the identified front face target. If the candidate target in the current frame matches the tracked target in the previous frame in terms of position and orientation, update the tracking list and perform temporal filtering on the position and orientation of the front face based on the extended Kalman filter model to suppress single-frame measurement noise. If no match is found, add the target as a new target to the tracking list. Step S6, Pose Calculation: This includes two stages: coarse positioning and fine calculation. Step S61, coarse positioning stage: Based on the RANSAC algorithm, straight line fitting is performed on the candidate point cloud of the front face to preliminarily determine whether the pose of the front face meets the docking requirements. Step S62, Fine Calculation Stage: The pose of the front face is accurately calculated using the least squares method for the interior point cloud that has passed the coarse positioning stage. Step S7: Hook pose calculation. Based on the front face pose calculation results, calculate the precise position of the hook. Step S8: The tractor obtains the identified hook position and the dispatch system issues the driving reference line L. The tractor plans the driving path from the current location of the vehicle along the reference line L to the hook position and performs automatic towing and docking operation.

3. The control method for an unmanned tractor with automatic hook-up function according to claim 2, characterized in that, When there is a lateral deviation between the trailer's parking position and the center of the storage space, the hook will shift laterally relative to the reference line L. This also includes a reference line offset adjustment step. Let the vertical distance from the hook position (Phook) to the reference line L be the horizontal offset δ, and define two thresholds: Docking threshold δdock: If the lateral offset is within this range, the tractor can directly complete the docking along the original reference line without adjustment; Safety threshold δsafe: The upper limit of the lateral offset. If this value is exceeded, the deviation is considered too large and safe docking cannot be achieved. The system will alarm and terminate the task. When δdock < δ ≤ δsafe, the specific method for adjusting the reference line offset is as follows: The original reference line L is shifted a distance δ along its normal vector direction to generate a new offset reference line L'. The offset direction is from the reference line towards the hook position, i.e.: ,in The unit normal vector of the reference line L is directed towards the hook side; the offset reference line L' passes through the lateral position of the projection point of the hook position on the original reference line, so that when the tractor travels along the new reference line L', the center line of the vehicle body is laterally aligned with the hook, thus enabling successful docking.

4. The unmanned tractor control method with automatic hook-lift function according to claim 3, characterized in that, Following the reference line offset adjustment steps is the re-interlocking process after the offset: After the reference line offset is complete, the tractor unit needs to perform the following sequence of actions to enter the new reference line L' and restart automatic docking:

1. Moving forward and out: The tractor unit moves forward in the current direction of travel, leaving the docking area in front of the trailer and moving to a safe distance; 2. Lateral switching: The tractor moves laterally from the original reference line L to the offset reference line L', so that the center line of the vehicle body is aligned with L'; 3. Reversing into the area: The tractor unit reverses along the new reference line L' and re-enters the trailer docking area with its rear facing the direction of the hook-up.

4. Restart docking: After reversing into position, restart the process of scanning and identifying the hook of the material car with the laser radar (6) after restarting, and complete the precise docking based on the new reference line L'.

5. A control method for an unmanned tractor with an automatic hook-and-lift function according to claim 2, characterized in that, In step 1, the unit vector from the end point B of the berth to the starting point A of the berth is: (xv, yv), The two normal vectors are respectively (-yv, xv) and (yv, -xv), berth width is w; based on the above information, generate a search quadrilateral A0A1B1B0, with the coordinates of its four vertices as follows: Wherein: the starting point A of the parking space is on the side where the tractor approaches the material car, and the ending point B of the parking space is on the side where the material car body is in the longitudinal direction; the width w of the parking space is greater than the width of the material car body; the line connecting the starting point A and the ending point B of the parking space is the centerline. Central axis The expected orientation of the trailer is used as a reference for subsequent positional deviation determination.

6. A control method for an unmanned tractor with an automatic hook-and-lift function according to claim 2, characterized in that, Step S2 further includes discrete point filtering: taking any point P as the center and r as the radius, count the number N of neighboring points within the radius. If N is less than a set threshold, then point P is considered a discrete noise point and is filtered out.

7. A control method for an unmanned tractor with an automatic hook-and-lift function according to claim 2, characterized in that, The rules for end face recognition in step S4 are as follows:

1. Distance sorting: Project the center points of all clustered targets onto the centerline of the berth. The targets are sorted from closest to furthest from the starting point A of the parking space, with priority given to targets closer to the towing vehicle.

2. Size screening: Select targets whose length matches the width of the trailer's front face, and eliminate interfering objects whose dimensions do not match; 3. Direction filtering: Based on the centerline of the berth. The direction information is used to determine whether the angle between the principal direction angle of the candidate target and the normal direction of the central axis exceeds the threshold. If the value exceeds the threshold, the target is considered to be deviating from the expected orientation and is excluded.

4. Straightness test: Perform straight line fitting on the point cloud of the candidate target and calculate the fitting residual. The front face of the trailer is a planar structure, and its point cloud projection should have good straightness. Targets with residuals exceeding the threshold are excluded.

8. A control method for an unmanned tractor with an automatic hook-and-lift function according to claim 1, characterized in that, The specific process of step S61, the coarse positioning stage, is as follows:

1. Select the projection point pt that is closest to the starting point A of the berth; 2. Select point clouds from all clustered targets whose distance from pt is less than maxDist, where maxDist = w ×sin(θ_max × π / 180), and θ_max is the maximum allowable deviation angle for trailer parking; 3. Perform RANSAC line fitting on the point cloud that meets the requirements to obtain the direction vector of the fitted line and a point on the line; 4. Calculate the angle between the fitted straight line and the normal direction of the centerline of the berth. If the angle exceeds the maximum allowable deviation angle θmax, the current frame detection is considered to have failed, and the next detection cycle begins.

9. A control method for an unmanned tractor with an automatic hook-and-lift function according to claim 1, characterized in that, The specific process of step S62, the fine calculation stage, is as follows:

1. Line Fitting: Perform least-squares line fitting on the set of interior points to obtain the direction vector of the front face. and the fitted linear equation; 2. Midpoint calculation: Project all interior points onto the fitted line, and take the midpoint of the projected points as the center point Pf(xf, yf) of the front face; 3. Orientation calculation: The direction of the front face normal vector is the trailer orientation θ = atan2(v_fit_x, -v_fit_y). The normal vector is perpendicular to the fitted line and points towards the inside of the vehicle body.

4. Multi-frame accumulation: The weighted average of N frames is calculated by accumulating the results of multiple frames and taking the average value.

10. A control method for an unmanned tractor with an automatic hook-and-lift function according to claim 9, characterized in that, The precise position of the hook in step S7 is calculated as follows: xhook = xf + d×cosθ, yhook = yf + d×sinθ, Where d is the known fixed offset of the hook from the front face, and θ is the trailer orientation angle; the final output is the hook pose (xhook, yhook, θ), which is used by the tractor path planning module to perform automatic docking operation.

Images