Anti-collision control method and system for AGV lifting platform

Abstract

The application provides a kind of AGV lifting platform anti-collision control method and system, including the height information H_cmd of target library position in the docking task issued by AGV receiving dispatching system;AGV travels to the front of target library position, starts 3D vision acquisition module installed on AGV;Three-dimensional image data containing target library position and its adjacent library position is collected by 3D vision acquisition module;Three-dimensional image data is processed, and the actual available height H_actual of target library position is calculated by recognizing the jack feature or bearing surface feature of target library position;The control instruction of allowing lifting, prohibiting lifting or alarm is generated according to the comparison result by comparing height information H_cmd with actual available height H_actual.The application can verify the height instruction issued by dispatching system twice before executing lifting action, and fundamentally eliminate the toppling accident caused by instruction error.

Description

Technical Field

[0001] This invention relates to the field of automated guided vehicle (AGV) control technology, specifically to an anti-collision control method and system for an AGV lifting platform, which is particularly suitable for the automatic storage and retrieval of goods on multi-layer shelves. Background Technology

[0002] With the development of automated warehousing, more and more AGVs are equipped with lifting platforms or forks for storing and retrieving goods on high-level racks. The scheduling system issues height instructions for the target storage location to the AGVs based on the records of the warehouse management system (WMS).

[0003] However, in practical applications, AGVs are prone to serious accidents such as tipping over other shelves due to the following reasons: Information asynchrony: The storage location height information in the scheduling system database is not synchronized with the actual storage location situation in a timely manner (e.g., the shelf height has been manually adjusted, or the goods are not placed in a standardized manner).

[0004] Recognition error: The cumulative positioning error of the AGV during its movement causes it to be aligned with the target column, but the lifting height is aimed at the upper or lower layer.

[0005] Protruding Goods: Goods may be protruding from the shelves below or above the target storage location, occupying the vertical space of the target storage location.

[0006] Existing technical solutions, such as Chinese invention patent CN110040670A, typically focus on alignment via visual recognition of pallet slots, but they do not solve the problem of physical collisions caused by incorrect height commands issued by the scheduling system. Once an incorrect command is issued, the AGV will blindly execute it, tipping over the shelves and causing significant losses. Summary of the Invention

[0007] In view of the technical problems existing in the AGV when storing and retrieving goods, such as information asynchrony, large recognition error and prominent goods position, the purpose of this invention is to provide an anti-collision control method and system for AGV lifting platforms. It can perform secondary verification of the height command issued by the scheduling system before executing the lifting action, thereby fundamentally eliminating the overturning accident caused by command error.

[0008] The present invention achieves the above objectives through the following technical solutions: A collision avoidance control method for an AGV lifting platform, comprising: Step S1: The AGV receives the docking task issued by the scheduling system. The docking task includes the height information H_cmd of the target storage location. Step S2: The AGV travels to the front of the target storage location and starts the 3D vision acquisition module installed on the AGV; Step S3: Acquire three-dimensional image data containing the target storage location and its adjacent storage locations using the 3D vision acquisition module; Step S4: Process the three-dimensional image data to identify the socket features or bearing surface features of the target storage location, and calculate the actual usable height H_actual of the target storage location; Step S5: Compare the height information H_cmd with the actual available height H_actual; if H_cmd ≤ H_actual, the AGV lifting platform is allowed to perform the lifting action; if H_cmd > H_actual, the lifting action is prohibited, and an alarm is sent to the upper-level system.

[0009] According to the anti-collision control method for an AGV lifting platform provided by the present invention, step S4, processing the three-dimensional image data includes: Statistical filtering or radius filtering is used to filter and denoise the raw point cloud data acquired by the 3D vision acquisition module to remove outliers; Based on the current position of the AGV and the estimated position of the target storage location, a region of interest is set, and point cloud data within the region of interest is extracted to remove ground point cloud and irrelevant far-field data. The retained point cloud data is segmented into planes using a random sampling consensus algorithm or a region growing algorithm to extract each plane component; Euclidean distance clustering was used to separate the structure of the shelf columns, beams and other components into independent point cloud clusters.

[0010] According to the anti-collision control method for an AGV lifting platform provided by the present invention, the identification method of the socket feature includes: For point cloud data after planar segmentation, extract all candidate regions of horizontal planes with downward normal directions; For each candidate region on the horizontal plane, four boundary lines distributed in a rectangle are searched around it. The boundary lines are obtained by fitting straight lines to the edge points of the point cloud. Determine whether the point cloud density inside the rectangular region is lower than a preset density threshold. If so, determine that the region is a rectangular concave region with missing internal point cloud. Linear or planar fitting is performed on the lower edge point cloud of the region identified as a rectangular depression, and the height value of the fitting result is taken as the bearing reference.

[0011] According to the anti-collision control method for an AGV lifting platform provided by the present invention, the identification method of the bearing surface feature includes: For point cloud data after planar segmentation, calculate the normal direction of each planar region, and filter horizontal candidate regions with the normal direction pointing upward and the angle between the normal direction and the vertical direction being less than a preset angle threshold. For each candidate region on the horizontal plane, its boundary point cloud is extracted. The set of edge points with the highest height in the boundary point cloud is fitted with a straight line or a plane, and the height value of the fitting result is taken as the bearing benchmark.

[0012] According to the anti-collision control method of the AGV lifting platform provided by the present invention, after determining the bearing reference, the height of the bearing reference is used as the starting surface, and a ray projection or columnar region scan is performed in the vertical upward direction. The position where the point cloud density first appears to exceed the preset continuity threshold along the scanning path is detected, and the height value of the position is used as the lower edge of the obstacle above.

[0013] According to the anti-collision control method for an AGV lifting platform provided by the present invention, the actual usable height H_actual is calculated as follows: Starting from the bearing reference height determined in step S4, scan vertically upwards to the position where the continuous point cloud is first detected, and record the height of this position as Z_obs; Then H_actual = Z_obs - Z_base - Δ, Where Z_base is the height value of the bearing reference, Z_obs is the height value of the lower edge of the obstacle above, and Δ is the preset safety margin.

[0014] According to the anti-collision control method for an AGV lifting platform provided by the present invention, step S4 of calculating the actual usable height H_actual of the target storage location includes: The lower edge plane of the target storage location and the lower plane of the first obstacle immediately above the lower edge plane were identified using point cloud data. Calculate the vertical distance between the lower edge plane and the lower plane of the obstacle, and use it as the actual usable height H_actual; Where H_actual = Z2- Z1- Δ, Z1 is the height value of the lower edge plane of the target storage location, Z2 is the height value of the lower plane of the upper obstacle, and Δ is the preset safety margin.

[0015] According to the anti-collision control method for an AGV lifting platform provided by the present invention, step S4 further includes: After identifying the socket features or bearing surface features, the validity of the identification results is verified. Validity verification includes any of the following conditions: If the point cloud density of the identified bearing reference plane is lower than the preset density threshold, or the fitting error after fitting the plane exceeds the preset error threshold, it is determined that the clear target storage location features have not been identified. If, after multiple scans, no point cloud structure matching the characteristics of a socket or a bearing surface is detected in the region of interest, it is determined that a clear target storage location feature has not been identified. When it is determined that the target storage location cannot be clearly identified, the AGV lifting platform is prohibited from performing lifting actions, and an environmental anomaly alarm is sent to the upper-level system.

[0016] According to the anti-collision control method of the AGV lifting platform provided by the present invention, the scheduling system calculates the target lifting height required for docking based on the target storage location parameters, and sends it to the AGV as height information H_cmd. The target storage location parameters include at least one of the following: shelf height, shelf spacing, current inventory height, and preset safety gap.

[0017] An anti-collision control system for an AGV lifting platform includes: The scheduling interface module is used to receive the docking task and the height information H_cmd of the target storage location issued by the scheduling system; The 3D vision acquisition module, installed on the AGV, is used to acquire three-dimensional image data of the target storage location; The vision processing and pose calculation module is used to process 3D image data, identify storage location features, and calculate the actual usable height H_actual. The safety interlock control module is used to compare the height information H_cmd with the actual available height H_actual, and generate control commands to allow lifting, prohibit lifting, or trigger an alarm based on the comparison result; The output signal of the safety interlock control module is connected in series to the lifting control circuit of the AGV lifting platform to form a hardware-level or software-level interlock.

[0018] Therefore, compared with the prior art, the anti-collision control method and system for the AGV lifting platform proposed in this invention have the following beneficial effects: 1. This invention introduces a 3D vision-based actual usable height verification mechanism, H_actual, before the AGV lifting platform performs a lifting action. This mechanism compares the theoretical height command H_cmd issued by the scheduling system with the actual usable height obtained in real-time from data collected and calculated by the 3D camera. Lifting is only permitted when H_cmd ≤ H_actual. This mechanism acts as the last line of defense for verifying the scheduling system's "theoretical height" command at the execution level. It effectively prevents collisions or even overturning accidents caused by scheduling system data errors, human error, or changes in the storage environment. This fundamentally eliminates the safety hazards inherent in existing technologies that rely solely on upper-level system commands without real-time verification at the execution end.

[0019] 2. This invention does not rely solely on pre-stored storage location parameter data in the scheduling system or warehouse management system. Instead, it uses a 3D camera installed on the AGV to perform real-time three-dimensional perception of the target storage location and its adjacent storage locations, dynamically calculating the actual usable height. Even if the storage location information is not updated in time, goods are displaced, or the storage location structure is temporarily changed, the AGV can still make accurate judgments based on the current environmental conditions, thereby ensuring operational safety and significantly improving the system's robustness and environmental adaptability in complex dynamic environments.

[0020] 3. This invention deeply integrates 3D visual perception and height comparison logic into the control process of the AGV lifting platform, enabling the AGV to have the ability to autonomously perceive and intelligently confirm the target storage location environment before performing lifting actions. It is an important technological manifestation of the development of logistics automation equipment towards a higher level of intelligence and safety, realizing the upgrade of the AGV's operation mode from blind execution to intelligent confirmation, and has significant engineering application value and promotion prospects.

[0021] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments. Attached Figure Description

[0022] Figure 1 This is a flowchart of an embodiment of an anti-collision control method for an AGV lifting platform according to the present invention.

[0023] Figure 2 This is a flowchart illustrating the processing of three-dimensional image data in an embodiment of the anti-collision control method for an AGV lifting platform according to the present invention.

[0024] Figure 3 This is a flowchart illustrating the identification of socket features in an embodiment of the anti-collision control method for an AGV lifting platform according to the present invention.

[0025] Figure 4 This is a flowchart illustrating the identification of bearing surface features in an embodiment of the anti-collision control method for an AGV lifting platform according to the present invention.

[0026] Figure 5 This is a schematic diagram illustrating the validity verification of the identification results in an embodiment of the anti-collision control method for an AGV lifting platform according to the present invention.

[0027] Figure 6 This is a system schematic diagram of an embodiment of the anti-collision control system for an AGV lifting platform according to the present invention. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0029] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0030] An embodiment of an anti-collision control method for an AGV lifting platform See Figure 1 This embodiment provides an anti-collision control method for an AGV lifting platform, including: Step S1: The AGV receives the docking task issued by the scheduling system. The docking task includes the height information H_cmd of the target storage location. Step S2: The AGV travels to the front of the target storage location and starts the 3D vision acquisition module installed on the AGV; Step S3: Acquire three-dimensional image data containing the target storage location and its adjacent storage locations using the 3D vision acquisition module; Step S4: Process the three-dimensional image data to identify the socket features or bearing surface features of the target storage location, and calculate the actual usable height H_actual of the target storage location; Step S5: Compare the height information H_cmd with the actual available height H_actual; if H_cmd ≤ H_actual, the AGV lifting platform is allowed to perform the lifting action; if H_cmd > H_actual, the lifting action is prohibited, and an alarm is sent to the upper-level system.

[0031] In step S4 above, as Figure 2 As shown, processing of 3D image data includes: Statistical filtering or radius filtering is used to filter and denoise the raw point cloud data acquired by the 3D vision acquisition module to remove outliers; Based on the current position of the AGV and the estimated position of the target storage location, a region of interest is set, and point cloud data within the region of interest is extracted to remove ground point cloud and irrelevant far-field data. The retained point cloud data is segmented into planes using a random sampling consensus algorithm or a region growing algorithm to extract each plane component; Euclidean distance clustering was used to separate the structure of the shelf columns, beams and other components into independent point cloud clusters.

[0032] Specifically, the Random Sample Consensus (RANSAC) algorithm or the region growing algorithm is used to perform planar segmentation on the preserved point cloud data: If the RANSAC algorithm is used, the following steps are performed: Randomly select the minimum number of point pairs (no less than 3 points) from the retained point cloud data, and fit an initial planar model based on the point pairs; Calculate the distance from each point in the retained point cloud data to the initial plane model, and count the number of interior points whose distance is less than the preset plane fitting error threshold; Repeat the above steps of random selection, fitting, and statistics until the preset number of iterations is reached or the number of interior points exceeds the preset interior point threshold. Select the plane model with the most interior points as the target plane. Remove the interior points of the target plane from the retained point cloud data, and repeat the above steps on the remaining point cloud data until all planar components are extracted.

[0033] If the region growing algorithm is used, the following steps are performed: Select a preset number of seed points from the retained point cloud data and calculate the normal direction of each seed point; Using each seed point as the center, search for neighboring points within a preset neighborhood radius. If the angle between the normal direction of the neighboring point and the normal direction of the seed point is less than a preset angle threshold, then add the neighboring point to the current growth region. Repeat the neighborhood search and addition steps above until no new points are added to the current growth region, forming a planar component; Remove the point cloud corresponding to the extracted planar components from the retained point cloud data, and repeat the above steps on the remaining point cloud data until all planar components are extracted.

[0034] In this embodiment, point cloud data after planar segmentation is obtained, which includes point clouds of non-planar structures such as shelf columns and beams; Initialize cluster centers: Randomly select a preset number of points from the point cloud data as initial cluster centers. The preset number is set according to the complexity of the shelf structure. Perform Euclidean distance clustering iteration: Calculate the Euclidean distance from each point in the point cloud data to each initial cluster center, and assign each point to the cluster center that is closest to it and whose distance is less than a preset distance threshold to form an initial point cloud cluster; Calculate the average coordinates of all points within each initial point cloud cluster, and use these as the new cluster centers for that cluster. Repeat the above steps of distance calculation, point assignment, and cluster center update until the change in cluster center coordinates is less than the preset convergence threshold or the preset maximum number of iterations is reached. Filter invalid clusters: Count the number of points in each point cloud cluster, remove point cloud clusters with fewer points than the preset minimum cluster size, and retain point cloud clusters with more or less points than the preset minimum cluster size as independent point cloud clusters for structures such as shelf columns and beams.

[0035] In this embodiment, as Figure 3 As shown, the methods for identifying socket features include: For point cloud data after planar segmentation, extract all candidate regions of horizontal planes with downward normal directions; For each candidate region on the horizontal plane, four boundary lines distributed in a rectangle are searched around it. The boundary lines are obtained by fitting straight lines to the edge points of the point cloud. Determine whether the point cloud density inside the rectangular region is lower than a preset density threshold. If so, determine that the region is a rectangular concave region with missing internal point cloud. Linear or planar fitting is performed on the lower edge point cloud of the region identified as a rectangular depression, and the height value of the fitting result is taken as the bearing reference.

[0036] In this embodiment, as Figure 4 As shown, the identification methods for bearing surface features include: For point cloud data after planar segmentation, calculate the normal direction of each planar region, and filter horizontal candidate regions with the normal direction pointing upward and the angle between the normal direction and the vertical direction being less than a preset angle threshold. For each candidate region on the horizontal plane, its boundary point cloud is extracted. The set of edge points with the highest height in the boundary point cloud is fitted with a straight line or a plane, and the height value of the fitting result is taken as the bearing benchmark.

[0037] In this embodiment, after determining the bearing reference, the height of the bearing reference is used as the starting surface, and a ray projection or columnar region scan is performed in the vertical upward direction. The position where the point cloud density first exceeds the preset continuity threshold is detected along the scanning path, and the height value of the position is used as the lower edge of the obstacle above.

[0038] In this embodiment, the actual usable height H_actual is calculated as follows: Starting from the bearing reference height determined in step S4, scan vertically upwards to the position where the continuous point cloud is first detected, and record the height of this position as Z_obs; Then H_actual = Z_obs - Z_base - Δ, Where Z_base is the height value of the bearing reference, Z_obs is the height value of the lower edge of the obstacle above, and Δ is the preset safety margin.

[0039] In this embodiment, calculating the actual usable height H_actual of the target storage location in step S4 above includes: The lower edge plane of the target storage location and the lower plane of the first obstacle immediately above the lower edge plane were identified using point cloud data. Calculate the vertical distance between the lower edge plane and the lower plane of the obstacle, and use it as the actual usable height H_actual; Where H_actual = Z2- Z1- Δ, Z1 is the height value of the lower edge plane of the target storage location, Z2 is the height value of the lower plane of the upper obstacle, and Δ is the preset safety margin.

[0040] In this embodiment, as Figure 5 As shown, step S4 further includes: After identifying the socket features or bearing surface features, the validity of the identification results is verified. Validity verification includes any of the following conditions: If the point cloud density of the identified bearing reference plane is lower than the preset density threshold, or the fitting error after fitting the plane exceeds the preset error threshold, it is determined that the clear target storage location features have not been identified. If, after multiple scans, no point cloud structure matching the characteristics of a socket or a bearing surface is detected in the region of interest, it is determined that a clear target storage location feature has not been identified. When it is determined that the target storage location cannot be clearly identified, the AGV lifting platform is prohibited from performing lifting actions, and an environmental anomaly alarm is sent to the upper-level system.

[0041] In step S1 above, the scheduling system calculates the target lifting height required for docking based on the target storage location parameters and sends it to the AGV as height information H_cmd. The target storage location parameters include at least one of the following: shelf height, shelf spacing, current inventory height, and preset safety gap.

[0042] In practical applications, this embodiment is applicable to a stacking AGV with lifting forks, used for storing and retrieving palletized goods on multi-layer shelves.

[0043] Under the command of the scheduling system, the AGV arrives at the target shelf. The scheduling system instructs the AGV to place the goods into the third storage location, and issues a height instruction H_cmd of 1500mm, meaning the forks need to be raised to a height of 1500mm off the ground.

[0044] The 3D camera installed on the AGV (a TOF camera is used in this embodiment, but a binocular stereo camera can also be used) is activated to scan the target storage area and generate three-dimensional point cloud data. The vision processing module first defines the region of interest (ROI) in the point cloud data based on the current positioning information of the AGV. This region covers the socket structure of the target storage location (3rd layer) and the lower edge area of ​​the shelf of the upper layer (4th layer).

[0045] Subsequently, the visual processing module filters and denoises the point cloud data within the ROI and extracts the region of interest, removing ground point clouds and irrelevant far-field data; then, it performs planar segmentation and Euclidean clustering on the remaining point clouds, separating the shelving columns, beams, and other structures into independent point cloud clusters.

[0046] Based on this, the module identifies the lower support surface of the socket in the target storage location (3rd layer) and obtains plane A through plane fitting; at the same time, it identifies the lower surface of the shelf in the upper layer (4th layer) and obtains plane B through plane fitting. The vertical distance between plane A and plane B is calculated to obtain the actual usable height H_actual.

[0047] In this embodiment, since the fourth layer of goods is stacked at a low height, the actual distance between the lower edge of the shelf and the support surface of the third layer's insertion hole is only 1400mm, i.e., H_actual=1400mm.

[0048] The safety interlock control module receives H_cmd = 1500mm and H_actual = 1400mm, and compares them to determine the result. Since H_cmd (1500mm) > H_actual (1400mm), the judgment command height exceeds the actual available space. If the lifting action is performed, the forks will inevitably hit the 4th shelf or the goods.

[0049] The safety interlock control module immediately cuts off the lifting control loop, prohibiting the AGV from performing lifting actions. It also sends an alarm message, "Target storage location height abnormal, lifting prohibited," to the upper-level system via the scheduling interface module, prompting the administrator to check the storage location. The AGV remains in place, awaiting further instructions.

[0050] When the target storage location uses a bearing surface feature instead of a socket feature: The dispatch system issues an H_cmd = 1200mm instruction, requiring the goods to be placed on the second layer. After scanning by the 3D camera, the vision processing module identifies the upper edge height Z1 = 1180mm of the second layer's bearing surface and the lower plane height Z2 = 1350mm of the obstacle on the third layer above it, with a preset safety margin Δ = 50mm.

[0051] The actual usable height is: H_actual=Z2-Z1-Δ=1350-1180-50=120mm Because H_cmd (1200mm) > H_actual (120mm), the safety interlock control module determines that the height is insufficient, prohibits lifting, and triggers an alarm.

[0052] Exception handling scenarios for target location feature recognition failure: After the AGV arrives at the target storage location, the 3D camera completes the scan, but the vision processing module still fails to detect a point cloud structure that matches the characteristics of the socket or the bearing surface within the ROI after multiple scans, and determines that it cannot identify clear target storage location features.

[0053] Based on this, the safety interlock control module determines that the environment does not meet the docking conditions, prohibits the AGV from performing lifting actions, and sends an environmental anomaly alarm to the upper-level system. The AGV remains in place, not performing any lifting operations, waiting for manual intervention or a new task from the scheduling system.

[0054] An embodiment of an anti-collision control system for an AGV lifting platform like Figure 6 As shown, this embodiment provides an anti-collision control system for an AGV lifting platform, comprising: The scheduling interface module is used to receive the docking task and the height information H_cmd of the target storage location issued by the scheduling system; The 3D vision acquisition module, installed on the AGV, is used to acquire three-dimensional image data of the target storage location; The vision processing and pose calculation module is used to process 3D image data, identify storage location features, and calculate the actual usable height H_actual. The safety interlock control module compares the height information H_cmd with the actual available height H_actual, and generates control commands to allow lifting, prohibit lifting, or trigger an alarm based on the comparison result.

[0055] Furthermore, the scheduling interface module includes: The communication interface unit is used to establish a communication connection with the upper-level scheduling system, receive docking tasks and altitude instructions H_cmd issued by the scheduling system, and send alarm information to the upper-level scheduling system. The data parsing unit is used to parse the height instruction H_cmd in the received docking task and extract the height value and target storage location identification information. The alarm generation unit is used to generate alarm information, including the target storage location identifier, the type of abnormality, and suggested handling measures, based on the judgment result of the safety interlock control module. The data transmission unit is used to transmit the parsed height command H_cmd to the safety interlock control module, and to send the alarm information generated by the alarm generation unit to the upper-level scheduling system through the communication interface unit.

[0056] Furthermore, the 3D vision acquisition module includes: A 3D camera unit is installed at the front or top of the AGV to perform three-dimensional scanning of the target storage area and generate raw point cloud data. The 3D camera unit includes at least one of a binocular stereo camera, a time-of-flight (TOF) camera, or a structured light camera, and its field of view covers the shelf area within a preset range in front of the AGV. The data acquisition and control unit is used to receive the scanning trigger command from the AGV main control unit, control the 3D camera unit to start scanning, and set the scanning parameters (including at least one of resolution, scanning frequency, and exposure time). At the same time, it transmits the raw point cloud data generated by the scanning to the vision processing module. The preprocessing unit is used to perform preliminary processing on the raw point cloud data, including: Statistical filtering or radius filtering algorithms are used to remove isolated noise points; Voxel filtering algorithm is used to downsample point cloud data to reduce data density; Transform the point cloud data from the camera coordinate system to the AGV coordinate system or the world coordinate system to ensure alignment with the AGV positioning data. The preprocessed point cloud data is transmitted to the vision processing and pose calculation module.

[0057] Furthermore, the vision processing and pose calculation module is specifically used for: filtering and denoising the raw point cloud data acquired by the 3D vision acquisition module and extracting the region of interest to remove ground point cloud and far-field irrelevant data; performing planar segmentation and Euclidean clustering on the retained point cloud to separate the shelving columns, beams and other structures into independent point cloud clusters; identifying the features of the sockets or the bearing surface and calculating the actual usable height H_actual.

[0058] Furthermore, the output signal of the safety interlock control module is connected in series with the lifting control circuit of the AGV lifting platform in the form of a hardware relay, forming a hardware-level or software-level interlock.

[0059] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0060] The above embodiments are merely preferred embodiments of the present invention and should not be construed as limiting the scope of protection of the present invention. Any non-substantial changes and substitutions made by those skilled in the art based on the present invention shall fall within the scope of protection claimed by the present invention.

Claims

1. A collision avoidance control method for an AGV lifting platform, characterized in that, include: Step S1: The AGV receives the docking task issued by the scheduling system. The docking task includes the height information H_cmd of the target storage location. Step S2: The AGV travels to the front of the target storage location and starts the 3D vision acquisition module installed on the AGV; Step S3: Acquire three-dimensional image data containing the target storage location and its adjacent storage locations using the 3D vision acquisition module; Step S4: Process the three-dimensional image data to identify the socket features or bearing surface features of the target storage location, and calculate the actual usable height H_actual of the target storage location; Step S5: Compare the height information H_cmd with the actual available height H_actual; if H_cmd ≤ H_actual, the AGV lifting platform is allowed to perform the lifting action; if H_cmd > H_actual, the lifting action is prohibited, and an alarm is sent to the upper-level system.

2. The method according to claim 1, characterized in that, In step S4, the processing of the three-dimensional image data includes: Statistical filtering or radius filtering is used to filter and denoise the raw point cloud data acquired by the 3D vision acquisition module to remove outliers; Based on the current position of the AGV and the estimated position of the target storage location, a region of interest is set, and point cloud data within the region of interest is extracted to remove ground point cloud and irrelevant far-field data. The retained point cloud data is segmented into planes using a random sampling consensus algorithm or a region growing algorithm to extract each plane component; Euclidean distance clustering was used to separate the structure of the shelf columns, beams and other components into independent point cloud clusters.

3. The method according to claim 2, characterized in that, Methods for identifying socket features include: For point cloud data after planar segmentation, extract all candidate regions of horizontal planes with downward normal directions; For each candidate region on the horizontal plane, four boundary lines distributed in a rectangle are searched around it. The boundary lines are obtained by fitting straight lines to the edge points of the point cloud. Determine whether the point cloud density inside the rectangular region is lower than a preset density threshold. If so, determine that the region is a rectangular concave region with missing internal point cloud. Linear or planar fitting is performed on the lower edge point cloud of the region identified as a rectangular depression, and the height value of the fitting result is taken as the bearing reference.

4. The method according to claim 2, characterized in that, Methods for identifying the features of the bearing surface include: For point cloud data after planar segmentation, calculate the normal direction of each planar region, and filter horizontal candidate regions with the normal direction pointing upward and the angle between the normal direction and the vertical direction being less than a preset angle threshold. For each candidate region on the horizontal plane, its boundary point cloud is extracted. The set of edge points with the highest height in the boundary point cloud is fitted with a straight line or a plane, and the height value of the fitting result is taken as the bearing benchmark.

5. The method according to claim 3 or 4, characterized in that: After determining the bearing reference, the height of the bearing reference is used as the starting surface, and ray projection or columnar region scanning is performed in the vertical upward direction. The position where the point cloud density first exceeds the preset continuity threshold is detected along the scanning path, and the height value of the position is used as the lower edge of the obstacle above.

6. The method according to claim 1, characterized in that, The actual usable height H_actual can be calculated in the following ways: Starting from the bearing reference height determined in step S4, scan vertically upwards until the position where the continuous point cloud is first detected, and record the height of this position as Z_obs; Then H_actual = Z_obs - Z_base - Δ, Where Z_base is the height value of the bearing reference, Z_obs is the height value of the lower edge of the obstacle above, and Δ is the preset safety margin.

7. The method according to claim 1, characterized in that, Step S4, calculating the actual usable height H_actual of the target storage location, includes: The lower edge plane of the target storage location and the lower plane of the first obstacle immediately above the lower edge plane were identified using point cloud data. Calculate the vertical distance between the lower edge plane and the lower plane of the obstacle, and use it as the actual usable height H_actual; Where H_actual = Z2 - Z1 - Δ, Z1 is the height value of the lower edge plane of the target storage location, Z2 is the height value of the lower plane of the upper obstacle, and Δ is the preset safety margin.

8. The method according to any one of claims 1 to 7, characterized in that, Step S4 also includes: After identifying the socket features or bearing surface features, the validity of the identification results is verified. Validity verification includes any of the following conditions: If the point cloud density of the identified bearing reference plane is lower than the preset density threshold, or the fitting error after fitting the plane exceeds the preset error threshold, it is determined that the clear target storage location features have not been identified. If, after multiple scans, no point cloud structure matching the characteristics of a socket or a bearing surface is detected in the region of interest, it is determined that a clear target storage location feature has not been identified. When it is determined that the target storage location cannot be clearly identified, the AGV lifting platform is prohibited from performing lifting actions, and an environmental anomaly alarm is sent to the upper-level system.

9. The method according to any one of claims 1 to 7, characterized in that: The scheduling system calculates the target lifting height required for docking based on the target storage location parameters and sends it to the AGV as height information H_cmd. The target storage location parameters include at least one of the following: shelf height, shelf spacing, current inventory height, and preset safety clearance.

10. A collision avoidance control system for an AGV lifting platform, characterized in that, include: The scheduling interface module is used to receive the docking task and the height information H_cmd of the target storage location issued by the scheduling system; The 3D vision acquisition module, installed on the AGV, is used to acquire three-dimensional image data of the target storage location; The vision processing and pose calculation module is used to process 3D image data, identify storage location features, and calculate the actual usable height H_actual. The safety interlock control module is used to compare the height information H_cmd with the actual available height H_actual, and generate control commands to allow lifting, prohibit lifting, or trigger an alarm based on the comparison result; The output signal of the safety interlock control module is connected in series to the lifting control circuit of the AGV lifting platform to form a hardware-level or software-level interlock.

Images