A three-dimensional mapping adaptive scheduling method and system for a smart factory

Abstract

This invention provides a three-dimensional mapping adaptive scheduling method and system for smart factories, belonging to the field of digital twin technology. The key technical points include: first, forming a digital twin for each floor based on panoramic images of each floor in an automated warehouse; then, for each automated device operating across floors, merging the floors with cross-floor operations into the same floor to form a merged digital twin; and finally, based on each merged digital twin and a determined first scheduling strategy, adaptively scheduling the automated devices corresponding to each digital twin sub-body within the merged digital twin. This invention solves the problem of reduced efficiency caused by the inability to communicate data across floors, leading to corridor occupancy, by setting up digital twins.

Description

Technical Field

[0001] This invention belongs to the field of digital twins and relates to a three-dimensional mapping adaptive scheduling method and system for smart factories. Background Technology

[0002] In the construction of smart factories, automated warehouses serve as core logistics hubs connecting upstream and downstream operations. Their multi-layered, high-density structure places extremely high demands on the intelligent management of equipment. Existing automated warehouse management systems are prone to problems such as discontinuous digital twin representations, asynchronous real-time positioning data, and difficulties in verifying scheduling strategies when equipment operates across floors. This leads to ineffective control over equipment operating across floors, resulting in collisions between automated equipment and low utilization rates of public facilities (such as elevators), exhibiting numerous shortcomings. Summary of the Invention

[0003] To address the shortcomings of existing technologies, the present invention aims to provide a three-dimensional mapping adaptive scheduling method and system for smart factories. This system solves the problems that existing automated warehouse management systems often encounter when equipment operates across floors, such as discontinuous digital twin representations, asynchronous real-time positioning data, difficulties in verifying scheduling strategies, and ineffective control over equipment operating across floors. These problems can easily lead to collisions between automated equipment and low utilization of public equipment.

[0004] To achieve the above objectives, the present invention provides the following technical solution:

[0005] A three-dimensional mapping adaptive scheduling method for smart factories includes the following steps:

[0006] Step 1) Obtain panoramic images of each floor in the smart factory to be scheduled. Each panoramic image includes patterns of various automated equipment and background patterns of the automated warehouse environment.

[0007] Step 2) Based on the panoramic image obtained in Step 1), generate a digital twin for each floor, wherein the digital twin includes digital twin sub-units of each automated device;

[0008] Step 3) For each automated device that needs to operate across floors, merge the two digital twins that operate across floors into the same floor based on their distance relationship to form a merged digital twin;

[0009] Step 4) For each merged digital twin generated in step 3), based on the determined first scheduling strategy, adaptively schedule the automatic devices corresponding to each digital twin sub-sub ...

[0010] In an optional embodiment, step 1) of "obtaining panoramic images of each floor in the smart factory to be scheduled" specifically includes the following sub-steps:

[0011] 2a) Determine the overall data collection area for the automated warehouse on each floor of the smart factory, and divide it into several sub-collection areas;

[0012] 2b) Control the preset panoramic cameras on each floor to capture images sequentially according to the acquisition sub-area, and set the acquisition resolution to the preset acquisition resolution and the acquisition angle to the preset acquisition angle;

[0013] 2c) The images of each acquired sub-area are stitched and corrected to remove image distortion and form a panoramic image covering the entire area of ​​the corresponding floor. The panoramic image also includes patterns of cross-floor corridors, and marks the corridor width, entrance location and curve features.

[0014] In an optional embodiment, step 2) "generating a digital twin of each floor based on the panoramic image" specifically includes the following sub-steps:

[0015] 3a) Preprocess the panoramic image obtained in step 1), including denoising, enhancing image contrast, and highlighting the differences between the patterns of automated equipment, the background patterns of the automated warehouse environment, and the corridor patterns.

[0016] 3b) Extract feature data of automated equipment, background features of the automated warehouse environment, and corridor features from the preprocessed panoramic image;

[0017] 3c) The extracted feature data is mapped to a preset three-dimensional spatial coordinate system to construct a digital twin with a preset ratio to the physical scene of the corresponding floor. The digital twin is also associated with a task attribute unit to store the task urgency and estimated task duration data of the automated equipment.

[0018] In an optional embodiment, step 3) "merging two digital twins that have cross-floor operations into the same floor based on their distance relationship to form a merged digital twin" specifically includes the following sub-steps:

[0019] 4a) Determine the digital twin of the current floor where the cross-floor device is located, denoted as the current twin; and determine the digital twin of the target floor, denoted as the target twin;

[0020] 4b) Calculate the vertical distance between the current twin and the target twin at the corresponding floor level. Using the standard floor height of the smart factory as the unit, multiply the difference between the target floor number and the current floor number by the standard floor height to obtain the distance value.

[0021] 4c) Determine if the distance value is less than or equal to the preset merging threshold. If it is less than or equal to the threshold, perform the merging operation. If it is greater than the threshold, first split the target twin into several sub-twins, and then select the sub-twin closest to the current twin to perform the merging.

[0022] 4d) Perform data fusion on two twins that meet the merging conditions. During the fusion process, a new corridor twin unit is added. The corridor twin unit is used to collect corridor occupancy status, movement speed data and corridor congestion coefficient data to form a merged digital twin.

[0023] In an optional embodiment, step 4d) "the corridor twin unit is used to collect corridor occupancy status, movement speed data, and corridor congestion coefficient data" specifically includes the following sub-steps:

[0024] 5a) Associate a preset number of lidar sensors at the entrance, exit and bend of the corridor, and collect the location data and moving speed data of the equipment in the corridor at a preset collection frequency;

[0025] 5b) Determine whether the corridor is occupied based on the collected location data. If there are devices and the distance between devices is less than the preset safety distance, it is determined to be occupied.

[0026] 5c) Calculate the corridor congestion coefficient, which is the ratio of the actual passage time of the equipment to the preset theoretical passage time. The occupancy status, movement speed data and corridor congestion coefficient data are synchronized to the merged digital twin in real time.

[0027] In an optional embodiment, step 4) "adaptive scheduling includes equipment queuing across floors and corridors" specifically includes the following sub-steps:

[0028] 6a) Extract cross-floor task attribute data that requires the use of corridors from the task attribute units of the merged digital twin, including task urgency score and estimated task duration;

[0029] 6b) Calculate task priority, which is determined based on task urgency and estimated task duration;

[0030] 6c) Generate an initial waiting queue in descending order of priority and store it in the waiting queue management subunit of the merged digital twin;

[0031] 6d) At each preset update cycle, based on the real-time status and task attribute changes of the corridor twin units of the merged digital twin, perform insertion, removal or rearrangement operations on the waiting queue to update the queue order;

[0032] 6e) Based on the updated waiting list, send status commands such as waiting, about to wait, or waiting cancelled to the digital twin of the corresponding automated device to guide the device to prepare for passage through the corridor.

[0033] In an optional embodiment, step 6d) of “performing insertion, removal, or rearrangement operations on the waiting queue” specifically includes the following sub-steps:

[0034] 7a) Determine whether the priority of the newly added task is not lower than the sum of the differences between the priority of the current queue’s first task and the preset priority. If yes, insert the new task at the head of the queue, and the number of insertions per hour shall not exceed the preset insertion limit. If no, insert the task into the queue at the position with the closest and lower priority.

[0035] 7b) If a device in the waiting queue malfunctions or its task is transferred, the device is removed from the queue, the priority of the remaining tasks is recalculated, and the queue order is adjusted.

[0036] 7c) If the waiting time of a device in the queue is not less than the product of its estimated task execution time and the preset waiting multiple, then the priority of the device will be temporarily increased by the preset increase score, and the queue will be reordered.

[0037] In an optional embodiment, step 4) "adaptive scheduling includes task transfer across floors and corridors" specifically includes the following sub-steps:

[0038] 8a) Based on real-time data from the corridor twin unit, determine whether the corridor is occupied. If occupied, calculate the estimated waiting time for the current waiting device.

[0039] 8b) If the estimated waiting time is greater than the product of the estimated task execution time and the preset trigger multiple, and there are similar devices within the task execution range that meet the preset conditions, then the task transfer process will begin.

[0040] 8c) Filter similar devices. First, filter out devices that do not match the device type or whose load rate is not lower than the preset load threshold to obtain a list of candidate devices;

[0041] 8d) The candidate devices are weighted and scored according to their distance from the task start point, remaining battery power, and historical task matching degree, and the device with the highest score is selected as the task receiving device.

[0042] 8e) Calculate the total cost of task transfer, which is the sum of the transfer instruction issuance time, the time for the receiving device to travel to the task start point, and the task handover time;

[0043] 8f) If the total transfer cost is less than the estimated waiting time, a task transfer instruction is generated and sent to the digital twin of the original device and the receiving device, and the task management unit and waiting management subunit of the merged digital twin are updated synchronously; if the total transfer cost is not less than the estimated waiting time, the task transfer is terminated and the original waiting queue is retained.

[0044] In an optional embodiment, step 8f) of "generating a task transfer instruction and sending it to the digital twin of the original device and the receiving device" specifically includes the following sub-steps:

[0045] 9a) The task transfer instruction includes basic task information, handover requirements and subsequent task allocation information of the original equipment. The basic task information includes task ID, task type, start and end coordinates, and handover requirements include handover time window and handover location.

[0046] 9b) The task transfer command is sent to the digital twin of the original equipment and the receiving equipment via industrial Ethernet, and the two complete the task information synchronization through near-field communication;

[0047] 9c) After the original equipment completes the task handover at the handover location, it sends a handover confirmation signal to the merged digital twin and then proceeds to the starting point of the subsequent task; after receiving the task information, the receiving equipment updates the task status of its own digital twin sub-entity and prepares to pass through the corridor according to the waiting queue order.

[0048] 9d) If the receiving equipment malfunctions en route to the handover location, repeat steps 8c)-8f) to screen for new candidate equipment; if no equipment meets the criteria in the second screening, terminate the transfer, add the original equipment back to the waiting queue, and temporarily increase its priority.

[0049] In an optional embodiment, step 4) "adaptively scheduling the automated devices corresponding to each digital twin sub-sub ...

[0050] 10a) Adjust the movement speed of the device in the corridor according to the task type data of the task attribute unit and the occupancy status data of the corridor twin unit; if the task is of high urgency and the corridor is not occupied, set it to the preset maximum movement speed; if the task is of low urgency or there are multiple devices intersecting in the corridor, set it to the preset safe movement speed.

[0051] 10b) Based on the queue length data of the waiting management subunit, switch the corridor access mode; if the queue length is not lower than the preset switching threshold, switch to one-way access mode; if the queue length is less than the preset switching threshold, switch to two-way access mode.

[0052] 10c) Based on the estimated task duration data, allocate the corridor time slice length, wherein the time slice length is the product of the estimated corridor passage time and the preset buffer multiple, to ensure that the equipment completes corridor passage within the time slice.

[0053] In an optional embodiment, step 4) is followed by the following steps:

[0054] 11a) The first scheduling strategy is parsed into control instructions corresponding to each digital twin sub-entity and sent to the corresponding physical devices, including automatic devices and common devices;

[0055] 11b) Collect execution data of physical equipment in real time, including actual movement path, operation completion time, equipment fault information and corridor access data, and feed the execution data back to the merged digital twin;

[0056] 11c) The digital twin is merged to update the status data of the corridor twin unit, waiting management sub-unit and task attribute unit according to the execution data. If the device position deviation exceeds the preset deviation threshold or the corridor congestion coefficient exceeds the preset congestion threshold, the adaptive scheduling in step 4) is re-executed to generate a second scheduling strategy, replace the first scheduling strategy and send it to the corresponding device.

[0057] In an optional embodiment, step 11b) of "collecting execution data of physical devices in real time, including actual movement path, job completion time, equipment fault information and corridor access data" specifically includes the following sub-steps:

[0058] 12a) Obtain the real-time position coordinates, running speed and motor speed data of the automated equipment through the equipment controller, and use them as the basis for calculating the actual movement path and operation completion time;

[0059] 12b) Collect equipment fault signals, including motor faults and low battery alarms, through equipment sensors and mark them as equipment fault information;

[0060] 12c) The actual moving speed and passage time of the device in the corridor are collected by the lidar sensor associated with the corridor twin unit as corridor passage data;

[0061] 12d) According to the preset feedback frequency, the actual movement path, operation completion time, equipment failure information and corridor access data are integrated into execution data and uploaded to the merged digital twin.

[0062] In an optional embodiment, the following steps are also included:

[0063] 13a) Periodically integrate the execution data collected in step 11b) with the corresponding first scheduling strategy, waiting queue adjustment record, and task transfer record into new training samples;

[0064] 13b) Add new training samples to the sample library of the AI ​​scheduling model and update the sample library according to the preset optimization cycle;

[0065] 13c) The gradient descent algorithm is used to iteratively train the AI ​​scheduling model, and the priority calculation weight, transfer cost evaluation parameters and corridor parameter adaptation threshold are adjusted.

[0066] 13d) Verify the accuracy of the trained model. If the accuracy improvement meets the preset accuracy threshold, replace the original model; if it does not improve, retain the original model and continue to accumulate new training samples.

[0067] In an optional embodiment, step 13c) "adjusting the priority calculation weight, transfer cost assessment parameters, and corridor parameter adaptation threshold" specifically includes the following sub-steps:

[0068] 14a) Based on the task completion time data in the new training samples, adjust the urgency weight and duration weight of priority calculation to improve the completion time of high-urgency tasks.

[0069] 14b) Based on the task transfer success rate data, adjust the time coefficient in the transfer cost assessment parameters to reduce the probability of transfer failure;

[0070] 14c) Based on the data of corridor congestion rate and passage efficiency, adjust the switching threshold and buffer multiple in the corridor parameter adaptation threshold to balance corridor passage efficiency and safety.

[0071] 14d) Synchronize the adjusted parameters to the corresponding sub-units of the merged digital twin to ensure that subsequent scheduling matches the optimized model parameters.

[0072] A three-dimensional mapping adaptive scheduling system for a smart factory includes:

[0073] Panoramic Image Acquisition Module: Acquires panoramic images of each floor in the smart factory to be scheduled. Each panoramic image includes patterns of various automated equipment and the background pattern of the automated warehouse environment.

[0074] Digital twin generation module: Based on the acquired panoramic images, it generates a digital twin for each floor, wherein the digital twin includes digital twin sub-units of each automated device;

[0075] Twin merging module: For each automated device that needs to operate across floors, the two digital twins that operate across floors are merged into the same floor based on their distance relationship to form a merged digital twin;

[0076] Adaptive scheduling module: For each generated merged digital twin, based on a determined first scheduling strategy, adaptively schedules the automatic devices corresponding to each digital twin sub-sub ...

[0077] The beneficial effects of the present invention are as follows: The present invention provides a three-dimensional mapping adaptive scheduling method and system for smart factories. First, a digital twin for each floor is formed based on the panoramic images of each floor in the automated warehouse. Then, for each automated device that operates across floors, the floors with cross-floor operations are merged into the same floor to form a merged digital twin, so as to solve the problem of reduced efficiency caused by the inability of cross-floor data to communicate. Attached Figure Description

[0078] Figure 1This is a schematic diagram of a three-dimensional mapping adaptive scheduling method for a smart factory.

[0079] Figure 2 This is a structural diagram of a three-dimensional mapping adaptive scheduling system for a smart factory. Detailed Implementation

[0080] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solution of the present invention, rather than limitations thereof. In the absence of conflict, the embodiments of the present invention and the technical features in the embodiments can be combined with each other.

[0081] This application provides a three-dimensional mapping adaptive scheduling method for smart factories, such as... Figure 1 As shown, it includes the following steps:

[0082] Step 1) Obtain panoramic images of each floor in the smart factory to be scheduled. Each panoramic image includes patterns of various automated equipment and background patterns of the automated warehouse environment.

[0083] Step 2) Based on the panoramic image obtained in Step 1), generate a digital twin for each floor, wherein the digital twin includes digital twin sub-units of each automated device;

[0084] Step 3) For each automated device that needs to operate across floors, merge the two digital twins that operate across floors into the same floor based on their distance relationship to form a merged digital twin;

[0085] Step 4) For each merged digital twin generated in step 3), based on the determined first scheduling strategy, adaptively schedule the automatic devices corresponding to each digital twin sub-sub ...

[0086] In this embodiment, the core of the 3D mapping adaptive scheduling method for smart factories is to construct a digital scene through panoramic images, and then optimize the twin and scheduling logic based on cross-layer operation requirements, ultimately achieving efficient cross-layer operation of equipment. It should be noted that this method is not limited to a specific type of smart factory; it is applicable to factory scenarios that include automated warehouses and require equipment to operate across floors. For example, it can be applied to smart electronics factories, smart food processing factories, etc., but this application is not limited to these.

[0087] First, panoramic images of each floor in the smart factory to be scheduled need to be acquired. Specifically, a panoramic image is an image that can cover the entire area of ​​the corresponding floor, including images of various automated equipment and background images of the automated warehouse environment. These images are obtained by taking pictures with image acquisition devices (such as panoramic cameras) deployed on each floor. During the acquisition process, it is necessary to ensure that the images can fully present the position and shape of the automated equipment and background elements such as shelves and aisles in the warehouse, so as to provide a real physical scene basis for the subsequent construction of a digital twin.

[0088] Next, a digital twin of each floor is generated based on the aforementioned panoramic images. Specifically, the digital twin is a three-dimensional digital representation of the physical floor, containing digital twin sub-units of each automated device. This twin is generated through feature extraction and three-dimensional modeling of the panoramic images. Specifically, features of the automated devices and the warehouse background are first separated from the panoramic images, and then these feature data are mapped onto a preset three-dimensional spatial coordinate system, thereby constructing a digital twin that matches the proportions of the physical floors. This transforms the physical scene into a calculable and controllable digital scene, providing a digital carrier for subsequent cross-floor twin merging and adaptive scheduling.

[0089] Then, for each automated device requiring cross-floor operation, the two digital twins involved in the cross-floor operation are merged into the same floor based on their distance relationship, forming a merged digital twin. Automated devices requiring cross-floor operation are specifically those whose start and end points are located on different floors. This identification is based on floor information obtained from the device's task instructions. The distance relationship is specifically the spatial distance between the corresponding physical floors of the two twins to be merged, derived by calculating the vertical distance between the two floors (e.g., using the standard floor height as the unit, combined with the difference in floor numbers). The merged digital twin is specifically a digital scene resulting from the fusion of the digital twin data from the two floors. This fusion is achieved by eliminating data redundancy and establishing a spatial coordinate mapping relationship, addressing the issue that a single-floor twin cannot cover cross-floor operation scenarios, and providing unified digital scene support for cross-floor device scheduling.

[0090] Finally, for each merged digital twin, adaptive scheduling is performed on each digital twin sub-unit within the merged digital twin, outputting a first scheduling strategy. Adaptive scheduling is specifically a process of dynamically adjusting the scheduling scheme based on device status and scene constraints within the merged digital twin, including device waiting and task transfer across floors. The first scheduling strategy is specifically the device operation plan generated after scheduling, which can be calculated by an AI scheduling model. The model input includes data such as device status and scene constraints, and the output includes information such as device movement paths and operation time nodes, used to provide specific operation guidance for cross-floor devices, ensuring that devices complete cross-floor operations in an orderly and efficient manner.

[0091] In an optional embodiment, step 1) of "obtaining panoramic images of each floor in the smart factory to be scheduled" specifically includes the following sub-steps:

[0092] 2a) Determine the overall data collection area for the automated warehouse on each floor of the smart factory, and divide it into several sub-collection areas;

[0093] 2b) Control the preset panoramic cameras on each floor to capture images sequentially according to the acquisition sub-area, and set the acquisition resolution to the preset acquisition resolution and the acquisition angle to the preset acquisition angle;

[0094] 2c) The images of each acquired sub-area are stitched and corrected to remove image distortion and form a panoramic image covering the entire area of ​​the corresponding floor. The panoramic image also includes patterns of cross-floor corridors, and marks the corridor width, entrance location and curve features.

[0095] In this embodiment, the process of acquiring panoramic images is divided into three steps to ensure the integrity and accuracy of the images. It should be noted that the acquisition equipment and processing logic involved in this process can be adjusted according to the actual scenario, and this application is not limited to these.

[0096] The first step is to determine the overall data acquisition area for each floor of the automated warehouse in the smart factory, and then divide it into several sub-areas. The overall data acquisition area is the complete range within the corresponding floor that needs to be imaged. This can be determined based on the actual physical boundaries of the floor (such as wall locations and warehouse area boundaries), ensuring coverage of all automated equipment and the warehouse background. Sub-areas are smaller sections of the overall data acquisition area, used to avoid image distortion or loss of detail due to excessively large single acquisition areas. The division can be based on layout features such as shelf arrangement and aisle distribution within the floor, for example, dividing sub-areas according to the number of shelf columns. Both sub-areas provide a range basis for subsequent precise regional data acquisition, ensuring that the acquired images are completely covered and clear in detail.

[0097] The second step involves controlling the pre-set panoramic cameras on each floor to sequentially acquire images according to the sub-regions. The acquisition resolution and angle are set to preset values. Specifically, the pre-set panoramic cameras are cameras pre-deployed near the acquisition sub-regions on each floor. Their deployment positions must ensure coverage of the corresponding sub-regions, determined based on the range and viewing angle requirements of the acquisition sub-regions. The preset acquisition resolution is a pre-set image sharpness parameter, determined based on the required detail precision for subsequent digital twin construction. If the twin needs to depict the fine structure of the device, the resolution needs to be set higher; otherwise, it can be appropriately lowered. The preset acquisition angle is the camera's field of view, determined based on the size of the acquisition sub-region, ensuring that the image of a single sub-region completely covers that area. The purpose of these three parameters is to ensure that the acquired sub-region images meet the requirements of subsequent stitching and twin construction, avoiding data loss due to insufficient resolution or incomplete viewing angles.

[0098] The third step involves stitching and correcting the images from each acquired sub-region to remove image distortion and create a panoramic image covering the entire corresponding floor. This panoramic image also includes patterns of stairwells across floors, and labels indicating stairwell width, entrance location, and curve features. Image stitching specifically involves merging multiple sub-region images into a global image. This can be achieved by using image registration algorithms (such as feature point-based registration) to find overlapping parts between sub-region images and then fusing and stitching them together. Image correction specifically involves removing distortions caused by camera lens or shooting angle (such as barrel distortion and perspective distortion), which can be achieved using distortion correction algorithms (such as camera intrinsic parameter-based correction). The stairwell patterns and labeling information specifically refer to the stairwell images across floors presented in the panoramic image and the parameters describing the stairwell structure. The labeling information is extracted from the stairwell patterns using image measurement algorithms (such as measuring stairwell width and locating entrance location). The purpose of this entire step is to generate a complete, distortion-free panoramic image containing key stairwell information, laying the foundation for the subsequent generation of a digital twin containing stairwell features.

[0099] In an optional embodiment, step 2) "generating a digital twin of each floor based on the panoramic image" specifically includes the following sub-steps:

[0100] 3a) Preprocess the panoramic image obtained in step 1), including denoising, enhancing image contrast, and highlighting the differences between the patterns of automated equipment, the background patterns of the automated warehouse environment, and the corridor patterns.

[0101] 3b) Extract feature data of automated equipment, background features of the automated warehouse environment, and corridor features from the preprocessed panoramic image;

[0102] 3c) The extracted feature data is mapped to a preset three-dimensional spatial coordinate system to construct a digital twin with a preset ratio to the physical scene of the corresponding floor. The digital twin is also associated with a task attribute unit to store the task urgency and estimated task duration data of the automated equipment.

[0103] In this embodiment, generating a digital twin based on panoramic images requires three steps, the core of which is converting image data into a three-dimensional digital scene. It should be noted that the algorithms and coordinate system settings involved in this process can be flexibly adjusted, and this application is not limited to them.

[0104] The first step is to preprocess the panoramic image, including denoising and enhancing image contrast to highlight the differences between the automated equipment patterns, the background patterns of the automated warehouse environment, and the corridor patterns. Image preprocessing specifically involves optimizing the original panoramic image. Since the original image may contain noise, such as light interference during shooting or low contrast (e.g., dim lighting making it difficult to distinguish between equipment and background), processing is necessary for accurate feature extraction. Denoising refers to removing noise points from the image using filtering algorithms (such as Gaussian filtering). Enhancing contrast involves improving the brightness differences between the equipment, background, and corridor patterns using grayscale stretching or histogram equalization algorithms. The purpose of this step is to improve image quality, enabling subsequent feature extraction to accurately separate the features of the automated equipment, warehouse background, and corridor, avoiding feature extraction deviations caused by image quality issues.

[0105] The second step involves extracting feature data from the preprocessed panoramic image, including feature data of automated equipment, background features of the automated warehouse environment, and corridor features. Automated equipment feature data specifically describes the attributes of the automated equipment, such as equipment size, outline shape, and unique identifier. This can be achieved by separating the automated equipment pattern from the preprocessed image using image segmentation algorithms (such as semantic segmentation), followed by feature extraction algorithms (such as edge detection and contour extraction). Automated warehouse background feature data specifically describes the warehouse background attributes, such as shelf location, aisle width, and shelf height. This is obtained similarly to the automated equipment feature data, by separating the background pattern from the image and then extracting it. Corridor feature data specifically describes the corridor attributes, such as corridor length, width, and bend location. This can be extracted from the corridor pattern after separating it from the image. The purpose of these three data points is to provide specific feature parameters for constructing a digital twin, ensuring that the twin accurately reproduces the equipment, background, and corridor structure in the physical scene.

[0106] The third step is to map the extracted feature data to a preset three-dimensional spatial coordinate system to construct a digital twin that is proportional to the physical scene of the corresponding floor. This digital twin is also associated with task attribute units to store the task urgency and estimated task duration data of the automated equipment. The preset three-dimensional spatial coordinate system is a pre-established coordinate system used to map physical space. It can be set according to the actual physical space size of the smart factory. For example, with a corner of the factory as the origin, the X, Y, and Z axes are set to correspond to the horizontal, vertical, and height directions, respectively. The units of the coordinate system are consistent with the physical space. The preset ratio is the size ratio between the digital twin and the physical scene. It is usually set to 1:1 to ensure that the twin and the physical scene are completely matched, but this embodiment does not limit this. The task attribute unit is a unit attached to the digital twin to store task information. It is used to associate equipment tasks with the digital twin. The source of task information is synchronously obtained from the smart factory's production management system. For example, the task urgency is determined according to the production process priority, and the estimated task duration is calculated based on the task distance and equipment speed. The purpose of the whole step is to generate a digital twin containing equipment, background, corridor features and associated task information, providing complete digital scene and task data for subsequent cross-layer twin merging and adaptive scheduling.

[0107] In an optional embodiment, step 3) "merging two digital twins that have cross-floor operations into the same floor based on their distance relationship to form a merged digital twin" specifically includes the following sub-steps:

[0108] 4a) Determine the digital twin of the current floor where the cross-floor device is located, denoted as the current twin; and determine the digital twin of the target floor, denoted as the target twin;

[0109] 4b) Calculate the vertical distance between the current twin and the target twin at the corresponding floor level. Using the standard floor height of the smart factory as the unit, multiply the difference between the target floor number and the current floor number by the standard floor height to obtain the distance value.

[0110] 4c) Determine if the distance value is less than or equal to the preset merging threshold. If it is less than or equal to the threshold, perform the merging operation. If it is greater than the threshold, first split the target twin into several sub-twins, and then select the sub-twin closest to the current twin to perform the merging.

[0111] 4d) Perform data fusion on two twins that meet the merging conditions. During the fusion process, a new corridor twin unit is added. The corridor twin unit is used to collect corridor occupancy status, movement speed data and corridor congestion coefficient data to form a merged digital twin.

[0112] In this embodiment, merging cross-floor digital twins requires four steps, with the core being unified digital management of cross-floor scenarios based on distance relationships. It should be noted that the merging thresholds and splitting logic involved in this process can be adjusted according to the actual factory layout, and this application is not limited to this.

[0113] The first step is to determine the digital twin of the cross-floor device on the current floor, denoted as the current twin, and the digital twin of the target floor, denoted as the target twin. A cross-floor device is an automated device that needs to perform operations between different floors. Its identification is based on the start and end floor information obtained from the device's task instructions. If the start and end floors are different, it is determined to be a cross-floor device. The current twin is the digital twin already generated on the physical floor where the cross-floor device is currently located. The target twin is the digital twin already generated on the target floor in the cross-floor device's task instructions. The purpose of both is to clearly identify the two twin objects that need to be merged, ensuring that the merging operation covers the complete scenario of the device's cross-floor operations.

[0114] The second step is to calculate the vertical distance between the current twin and the target twin on the corresponding floors. Using the standard floor height of the smart factory as the unit, the distance is obtained by multiplying the difference between the target floor number and the current floor number by the standard floor height. The standard floor height is specifically the standard vertical distance between floors in the smart factory, determined by the factory's architectural design parameters and is a fixed value. The distance value is determined based on the vertical distribution characteristics of the physical floors; the difference in floor numbers reflects the number of floors crossed, and multiplying this by the standard height yields the actual vertical distance. The purpose of this step is to obtain the spatial distance between the two twins in their corresponding physical scenes, providing a distance basis for subsequent decisions on whether and how to merge them.

[0115] The third step is to determine whether the distance value is less than or equal to a preset merging threshold. If so, the merging operation is performed; otherwise, the target twin is first split into several sub-twins, and then the sub-twin closest to the current twin is selected for merging. The preset merging threshold is a pre-set distance standard for determining whether to directly merge two twins. It is determined based on the floor spacing and data processing capabilities of the smart factory. If the distance is too far, direct merging will result in an excessive amount of twin data, increasing the computational load. Therefore, a threshold needs to be set. Splitting the target twin into sub-twins specifically involves dividing the target twin into multiple smaller twins according to the functional partitions of the physical floors, such as by shelf areas or aisle areas. The splitting logic in this embodiment is to reduce the amount of merged data, merging only the sub-areas closest to the current twin and related to cross-floor operations. The purpose of this step is to reasonably select the merging method based on the distance, ensuring coverage of cross-floor operation scenarios while avoiding excessive data processing load.

[0116] The fourth step involves data fusion of two twins that meet the merging criteria. During the fusion process, a new corridor twin unit is added. This corridor twin unit is used to collect corridor occupancy status, movement speed data, and corridor congestion coefficient data, ultimately forming a merged digital twin. Specifically, data fusion integrates the three-dimensional spatial data, equipment twin data, and background feature data of the two twins. The integration logic is based on establishing a spatial coordinate mapping relationship, such as adjusting the coordinates of the target twin using the current twin's coordinate system as a reference, eliminating duplicate data, and achieving seamless data connection for identical public equipment data in the two twins. The corridor twin unit is a dedicated digital unit for managing cross-floor corridors. The purpose of setting up the corridor twin unit is that cross-floor operations need to be carried out through corridors, thus requiring separate collection of corridor status data. This unit's data collection is achieved through sensors (such as LiDAR) deployed in the corridors. The overall goal of this step is to generate a merged digital twin containing cross-floor scenes, equipment data, and corridor status, providing unified digital scene and corridor data support for subsequent adaptive scheduling of cross-floor equipment.

[0117] In an optional embodiment, step 4d) "the corridor twin unit is used to collect corridor occupancy status, movement speed data, and corridor congestion coefficient data" specifically includes the following sub-steps:

[0118] 5a) Associate a preset number of lidar sensors at the entrance, exit and bend of the corridor, and collect the location data and moving speed data of the equipment in the corridor at a preset collection frequency;

[0119] 5b) Determine whether the corridor is occupied based on the collected location data. If there are devices and the distance between devices is less than the preset safety distance, it is determined to be occupied.

[0120] 5c) Calculate the corridor congestion coefficient, which is the ratio of the actual passage time of the equipment to the preset theoretical passage time. The occupancy status, movement speed data and corridor congestion coefficient data are synchronized to the merged digital twin in real time.

[0121] In this embodiment, the data collection by the corridor twin unit needs to be performed in three steps. The core is to obtain the real-time operating status of the corridor to provide a basis for scheduling. It should be noted that the types of sensors and the collection frequency involved in this process can be adjusted according to the corridor layout, and this application is not limited to this.

[0122] The first step involves attaching a pre-set number of LiDAR sensors at the entrances, exits, and bends of the stairwell. These sensors collect location and speed data of the equipment within the stairwell at a pre-set frequency. The equipment in the stairwell is essentially an automated system operating across floors within a merged digital twin, requiring it to traverse different floors. The pre-set number of LiDAR sensors refers to the number of sensors pre-deployed at key locations in the stairwell. These locations are determined based on the stairwell's structural characteristics; for example, entrances, exits, and bends are critical nodes for equipment movement and require focused data collection. The number of sensors is determined by the stairwell length and the sensor's detection range to ensure full coverage. The pre-set frequency is the time interval between sensor data collection, determined by the scheduling requirements for real-time data. If the scheduling needs to respond quickly to changes in the stairwell's status, the frequency needs to be set higher. The purpose of this step is to acquire real-time location and speed data of the equipment within the stairwell, providing fundamental data for determining the stairwell's occupancy status and congestion level.

[0123] The second step involves determining whether the corridor is occupied based on the collected location data. If equipment is present and the distance between devices is less than a preset safety distance, the corridor is considered occupied. The location data specifically refers to the coordinates of the equipment within the corridor's coordinate system, collected by sensors. The preset safety distance is a pre-set threshold for determining whether a collision risk exists, determined based on the size of the automated equipment and safety operation requirements. If the distance between devices is less than this threshold, a collision risk may exist, thus the corridor is considered occupied. This step clarifies whether the corridor is currently occupied, providing a basis for subsequent equipment deployment. The equipment movement data changes in real time, collected in real-time by calculating the rate of change of the equipment's position within the corridor's coordinate system using sensors.

[0124] The third step is to calculate the corridor congestion coefficient, which is the ratio of the actual travel time of the equipment to the preset theoretical travel time. The occupancy status, movement speed data, and corridor congestion coefficient data are synchronized to the merged digital twin in real time. The actual travel time is specifically the actual time it takes for the equipment to travel from the corridor entrance to the exit, calculated by the time difference between the equipment entering and leaving the corridor, collected by sensors. The preset theoretical travel time is specifically the theoretical time it takes for the equipment to travel in the corridor at a preset speed, calculated based on the corridor length and the equipment's preset movement speed data. The corridor congestion coefficient reflects the degree of corridor congestion through the ratio of actual to theoretical time; the larger the ratio, the more severe the congestion. The purpose of this step is to quantify the corridor congestion status and synchronize corridor occupancy, speed, and congestion data to the merged digital twin, providing real-time data support for subsequent adaptive scheduling adjustments to corridor parameters and equipment waiting lists.

[0125] In an optional embodiment, step 4) "adaptive scheduling includes equipment queuing across floors and corridors" specifically includes the following sub-steps:

[0126] 6a) Extract cross-floor task attribute data that requires the use of corridors from the task attribute units of the merged digital twin, including task urgency score and estimated task duration;

[0127] 6b) Calculate task priority, which is determined based on task urgency and estimated task duration;

[0128] 6c) Generate an initial waiting queue in descending order of priority and store it in the waiting queue management subunit of the merged digital twin;

[0129] 6d) At each preset update cycle, based on the real-time status and task attribute changes of the corridor twin units of the merged digital twin, perform insertion, removal or rearrangement operations on the waiting queue to update the queue order;

[0130] 6e) Based on the updated waiting list, send status commands such as waiting, about to wait, or waiting cancelled to the digital twin of the corresponding automated device to guide the device to prepare for passage through the corridor.

[0131] In this embodiment, the queuing of cross-floor corridor equipment is a crucial step in adaptive scheduling, requiring five steps. The core principle is to achieve orderly queuing of cross-floor equipment based on task attributes. It should be noted that the weights and update cycles involved in this process can be adjusted according to factory production needs, and this application is not limited to this.

[0132] The first step is to extract cross-floor task attribute data requiring the use of corridors from the task attribute units of the merged digital twin. This includes the task urgency score and the estimated task duration. Cross-floor tasks requiring corridor access are specifically those where equipment must pass through corridors to perform cross-floor operations. The identification criteria are whether the task instructions include corridor access. The task attribute data consists of task-related information stored in the task attribute units. The task urgency score is determined based on the priority of the production process; tasks affecting critical production nodes have higher scores. The estimated task duration is calculated based on the task distance and the equipment's preset speed. This step aims to obtain the task characteristic data needed for queueing and ranking, providing a basis for subsequent priority calculations.

[0133] The second step is to calculate the task priority, which is calculated as: Task Urgency Score × Preset Urgency Weight + (1 / Estimated Task Duration) × Preset Duration Weight. The preset urgency weight and preset duration weight are pre-set weighting parameters used to balance urgency and efficiency. These are determined based on the smart factory's emphasis on task urgency and operational efficiency. If production continuity is prioritized, the urgency weight can be set higher; if corridor utilization is prioritized, the duration weight can be set higher. (1 / Estimated Task Duration) indicates that the shorter the task duration, the higher this value, and the higher the priority, preventing long-duration tasks from occupying corridors for extended periods. The purpose of this step is to obtain task priority through quantitative calculation, providing a unified standard for sorting the waiting list.

[0134] The third step involves generating an initial waiting list in descending order of priority, which is then stored in the waiting list management subunit of the merged digital twin. Specifically, the initial waiting list is a queue of devices arranged from highest to lowest task priority. This sorting logic ensures that devices corresponding to high-priority tasks can use the corridors first. The waiting list management subunit is a dedicated digital unit for storing and managing the waiting list, used for centralized control of the waiting process and to prevent queue chaos. The purpose of this step is to generate an ordered initial waiting list, providing a foundation for subsequent devices to use the corridors in sequence.

[0135] The fourth step involves setting a preset update cycle at each interval. Based on the real-time status and task attribute changes of the corridor twin unit, insertion, removal, or rearrangement operations are performed on the waiting queue to update the queue order. The preset update cycle is the time interval for queue adjustments, determined based on the frequency of task changes and corridor status changes, ensuring that the queue can reflect the latest situation in a timely manner. The real-time status of the corridor twin unit specifically refers to the occupancy and congestion of the corridor. If the congestion eases, the queue order can be adjusted appropriately. Task attribute changes specifically refer to changes in the task urgency score or estimated duration in the cross-floor task attribute data, such as a task suddenly becoming a production-critical task with increased urgency. Insertion, removal, or rearrangement operations are performed to address dynamic changes in the queue: new tasks need to be inserted, equipment failures need to be removed, and changes in task attributes need to be rearranged. The purpose of this step is to ensure that the waiting queue can adapt to dynamic changes in real time, avoiding scheduling inefficiencies due to information lag.

[0136] The fifth step involves sending status commands—waiting, about to enter, or cancelled—to the digital twin of the corresponding automated device based on the updated waiting queue, guiding the device to prepare for passage through the stairwell. Specifically, the status commands inform the device of its current waiting stage. The command type is determined by the device's position in the queue: if the device is at the back of the queue, a waiting command is sent; if the device is about to enter the stairwell, an about to enter command is sent; if the task is cancelled or transferred, a cancelled command is sent. Guiding the device to prepare involves the device completing relevant preparations in advance (such as adjusting its speed or checking its battery level). The purpose of this step is to allow the device to know its waiting status in advance, reducing preparation time before entering the stairwell and improving passage efficiency.

[0137] Furthermore, in addition to the method described above for determining priority based on task urgency and estimated task duration, this embodiment provides another step for determining priority. Specifically, a multi-dimensional target matrix is ​​first constructed. Each row in the matrix corresponds to one task, and there are a total of There are [number] tasks, with each column corresponding to a task feature dimension, totaling [number]. Each task has a characteristic dimension, including a task urgency score, a task value coefficient, and a device execution cost. Each element represents the weight coefficient of the task in that dimension. The task value coefficient is determined based on the benefits brought about by completing the task, and the device execution cost is determined based on the resource costs required for the device to complete the task, such as device energy consumption, wear and tear costs, and time occupation costs. This embodiment does not impose any restrictions on this. The weight coefficient of each task in this dimension can be determined by those skilled in the art based on the actual situation corresponding to each task. This embodiment does not impose any restrictions on this.

[0138] Then, through nonnegative matrix factorization, the multidimensional target matrix is ​​split into two nonnegative submatrices, called the task feature matrix. and weight matrix ,in This represents the number of hidden priority components. Each element in the task feature matrix represents the score of the task under that hidden priority component, and each element in the weight matrix represents the weight of that hidden priority component under that task feature dimension. Then, the global weight corresponding to each hidden priority component is calculated. For example, for the _ ... Each implicit priority component has a corresponding global weight. , Represents the weight matrix of the first element. Line number The corresponding elements are listed, and then the overall score for each task is calculated. , Represents the first element in the task feature matrix. Line number The corresponding elements in the column are then sorted in descending order of the comprehensive score to obtain the initial sequence. The Pareto optimal correction sequence is then used to obtain the final priority sequence. For example, for any two tasks in the initial sequence, the weight coefficients of the two tasks in each task feature dimension are compared. If one task satisfies that the weight coefficients of all task feature dimensions are not less than the weight coefficients of the other task in all task feature dimensions, and at least one task feature dimension has a higher weight coefficient than the other task, then regardless of whether the comprehensive score of the task is lower than that of the other task, the priority of the task is strictly set to be higher than that of the other task.

[0139] In this embodiment, the implicit priority component is a potential, comprehensive priority dimension extracted from the explicit task feature dimensions through non-negative matrix factorization. It is not a directly defined feature, but rather a combination of priority features hidden behind multiple explicit dimensions. For example, the original task feature dimensions include task urgency score, task value coefficient, and equipment execution cost. However, the actual task priority is often a combination of these dimensions. For instance, some tasks have high task urgency and low equipment execution cost, while others have high task value coefficient and low task urgency. The implicit priority component extracted through matrix factorization can transform these combined features into explicit priority dimensions, such as setting... In this case, the two implicit priority components can be efficiency-type priority components and value-type priority components. Efficiency-type priority components correspond to combinations with high task urgency and low equipment execution cost, while value-type priority components correspond to components with high task value coefficient and low task urgency. This example provides... The above examples are merely illustrative and are not intended to limit the scope of this embodiment.

[0140] This embodiment sets implicit priority components as a key intermediate layer connecting multiple scattered explicit dimensions with the final task priority. This not only integrates the relationships between explicit dimensions but also simplifies the priority evaluation logic. Compared to the methods mentioned above that rely on task urgency and estimated duration to determine priority, this embodiment determines priority through matrix factorization and Pareto optimality. By using matrix factorization to extract implicit priority components, it integrates the relationships between various explicit dimensions, avoiding subjective biases from manual judgment. Furthermore, by incorporating Pareto optimality correction steps, it can also compensate for the potential defects of comprehensive score ranking. Ultimately, the priority determination comprehensively integrates the multi-dimensional attributes of the task and accurately adapts to the actual needs of complex cross-layer task scheduling.

[0141] In an optional embodiment, step 6d) of “performing insertion, removal, or rearrangement operations on the waiting queue” specifically includes the following sub-steps:

[0142] 7a) Determine whether the priority of the newly added task is not lower than the sum of the differences between the priority of the current queue’s first task and the preset priority. If yes, insert the new task at the head of the queue, and the number of insertions per hour shall not exceed the preset insertion limit. If no, insert the task into the queue at the position with the closest and lower priority.

[0143] 7b) If a device in the waiting queue malfunctions or its task is transferred, the device is removed from the queue, the priority of the remaining tasks is recalculated, and the queue order is adjusted.

[0144] 7c) If the waiting time of a device in the queue is not less than the product of its estimated task execution time and the preset waiting multiple, then the priority of the device will be temporarily increased by the preset increase score, and the queue will be reordered.

[0145] In this embodiment, the dynamic adjustment of the waiting queue needs to be executed in three steps, the core of which is to balance queue stability and task flexibility. It should be noted that the thresholds and scores involved in this process can be adjusted according to scheduling fairness requirements, and this application is not limited to this.

[0146] The first step involves inserting a new task at the head of the queue if its priority is equal to or greater than the priority of the current queue's first task plus a preset priority difference. The number of insertions per hour cannot exceed a preset insertion limit. If the priority difference is less than the preset value, the task is inserted into the queue at the position with the closest but lower priority. Newly added tasks are cross-floor tasks requiring the use of a corridor added within the queue update cycle. Their priority calculation method is the same as the initial queue. The preset priority difference is a pre-set threshold for determining whether a new task should be inserted at the head of the queue, used to avoid frequent insertions of high-priority tasks that could cause queue chaos. Insertion at the head is only allowed when the new task's priority far exceeds that of the first task. The preset insertion limit is the maximum number of insertions allowed at the head of the queue per hour, used to further restrict disordered insertions and ensure overall queue stability. This step aims to rationally handle the insertion of new tasks, ensuring high-priority tasks are prioritized while preventing queue chaos.

[0147] The second step involves removing a device from the queue if it malfunctions or its task is transferred, then recalculating the priority of the remaining tasks and adjusting the queue order. Device malfunctions are determined from the status data of the device's digital twin, such as a fault signal or insufficient battery power to continue operation. Task transfer specifically refers to the transfer of a device's cross-layer task to another device, which can be determined based on transfer records obtained from the task management unit. Priority recalculation is necessary because the relative priorities of the remaining tasks may change after removing the device, ensuring accurate sorting. This step aims to promptly remove invalid devices from the queue, ensuring all devices in the queue have the ability to execute tasks and avoiding resource waste.

[0148] The third step involves temporarily increasing the priority of a device in the queue by a predetermined score if the device's waiting time is greater than or equal to its estimated task execution time multiplied by a preset waiting multiple. The waiting time is calculated from the device's entry time recorded by the queue management subunit. The preset waiting multiple is a pre-set threshold for determining whether a device has been waiting too long, determined based on the device's waiting tolerance and scheduling fairness. If the waiting time exceeds this multiple, the device is considered to have been waiting too long and its priority needs to be increased to ensure fairness. The preset increase score is the temporary priority increase value, determined based on the priority differences among tasks in the queue. This ensures that the device can be queued earlier after the increase, but without affecting excessively high-priority tasks. This step aims to prevent long waiting times for devices, ensure scheduling fairness, and improve device operating efficiency.

[0149] In an optional embodiment, step 4) "adaptive scheduling includes task transfer across floors and corridors" specifically includes the following sub-steps:

[0150] 8a) Based on real-time data from the corridor twin unit, determine whether the corridor is occupied. If occupied, calculate the estimated waiting time for the current waiting device.

[0151] 8b) If the estimated waiting time is greater than the product of the estimated task execution time and the preset trigger multiple, and there are similar devices within the task execution range that meet the preset conditions, then the task transfer process will begin.

[0152] 8c) Filter similar devices. First, filter out devices that do not match the device type or whose load rate is not lower than the preset load threshold to obtain a list of candidate devices;

[0153] 8d) The candidate devices are weighted and scored according to their distance from the task start point, remaining battery power, and historical task matching degree, and the device with the highest score is selected as the task receiving device.

[0154] 8e) Calculate the total cost of task transfer, which is the sum of the transfer instruction issuance time, the time for the receiving device to travel to the task start point, and the task handover time;

[0155] 8f) If the total transfer cost is less than the estimated waiting time, a task transfer instruction is generated and sent to the digital twin of the original device and the receiving device, and the task management unit and waiting management subunit of the merged digital twin are updated synchronously; if the total transfer cost is not less than the estimated waiting time, the task transfer is terminated and the original waiting queue is retained.

[0156] In this embodiment, cross-floor task transfer is a crucial step in adaptive scheduling, requiring six steps to execute. The core objective is to address the issue of idle equipment. It should be noted that the thresholds and weights involved in this process can be adjusted based on equipment resources and task requirements; this application is not limited to these.

[0157] The first step, based on real-time data from the corridor twin unit, determines whether the corridor is occupied. If occupied, it calculates the estimated waiting time for the currently waiting device, which is located in the waiting queue. The real-time data from the corridor twin unit specifically includes the corridor's occupancy status and the current device's movement speed. If a device exists in the corridor and the distance between devices is less than the preset safety distance, it is considered occupied. The estimated waiting time is the anticipated time for the waiting device to wait until the corridor becomes clear and proceeds in queue order. It is calculated based on the remaining travel time of the device currently in the corridor (calculated by combining the distance already traveled with the remaining distance) and the sum of the estimated travel times of the preceding devices in the queue. This step clarifies the waiting status of the devices, providing a basis for whether to trigger a task transfer.

[0158] The second step involves initiating a task transfer process if the estimated waiting time exceeds the estimated task execution time multiplied by a preset trigger multiplier, and if similar devices meeting the preset conditions exist within the task execution range. The preset trigger multiplier is a pre-defined waiting time multiplier used to determine whether to trigger a transfer. It is determined based on task execution efficiency and device waiting tolerance. If the waiting time significantly exceeds the task's duration, triggering a transfer avoids efficiency losses. The task execution range is the device movement range centered on the task's starting point, determined based on device movement efficiency to ensure the receiving device can reach the task's starting point within a reasonable timeframe. Similar devices meeting the preset conditions are those with the same device type as the original (e.g., both being AGV handling robots) and a load rate below a preset load threshold. The same type ensures the ability to perform the same task, and the load rate threshold is determined based on the device's maximum carrying capacity to avoid overload. This step determines whether the basic conditions for transfer are met, preventing blind transfers.

[0159] The third step is to filter similar equipment. First, filter out equipment with incorrect types or load rates ≥ a preset load threshold to obtain a candidate equipment list. Incorrect equipment type specifically means the equipment cannot perform the original task (e.g., the original task requires moving equipment, but the candidate equipment is a sorting device). The filtering criteria are determined by obtaining the equipment type from the attribute data of the equipment's digital twin. Load rate ≥ a preset load threshold specifically means the equipment's current workload has reached or exceeded its maximum capacity. The filtering criteria are determined by calculating the load rate (load rate = number of tasks executed / maximum workload) from the equipment status data. The candidate equipment list is a set of equipment that meets the type and load conditions. The filtering logic ensures that the equipment in the list has the basic capability to undertake the task. The purpose of this step is to initially filter out feasible task-taking equipment, providing a range for subsequent precise selection.

[0160] The fourth step involves weighting the candidate devices based on their distance from the task start point, remaining battery power, and historical task matching accuracy. The device with the highest score is selected as the task-receiving device. The distance from the task start point is specifically the distance from the candidate device's current location to the original task start point; the closer the distance, the shorter the time it takes for the device to reach the task start point, and the higher the score. Remaining battery power refers to the candidate device's current remaining battery capacity; the higher the battery level, the stronger the device's endurance for task execution, and the higher the score. Historical task matching accuracy refers to the number of times and success rate the candidate device has performed similar tasks in the past; the higher the matching accuracy, the higher the device's proficiency in performing tasks, and the higher the score. The weighted scoring involves setting preset weights for the three indicators (e.g., distance 40%, battery power 30%, matching accuracy 30%) and calculating the total score. The purpose of this step is to select the optimal task-receiving device from the candidate devices to ensure efficient task execution.

[0161] The fifth step is to calculate the total cost of the task transfer. This total cost equals the transfer instruction issuance time plus the time it takes for the receiving device to travel to the task start point plus the task handover time. The transfer instruction issuance time is specifically the time from generating the task transfer instruction to the device receiving the instruction, determined based on communication latency (such as the transmission latency of industrial Ethernet). The receiving device's travel time to the task start point is specifically the time it takes for the candidate device to travel from its current location to the task start point, calculated based on distance and the device's preset speed. The task handover time is specifically the time it takes for the original device and the receiving device to exchange task information (such as task parameters and device status), calculated based on the amount of information transmitted and communication speed. The total cost of the transfer reflects the efficiency of the transfer through the total time; if the total cost is too high, the transfer is meaningless. The purpose of this step is to evaluate the efficiency of the task transfer and provide a cost basis for deciding whether to execute the transfer.

[0162] Step 6: If the total transfer cost is less than the estimated waiting time, a task transfer instruction is generated and sent to the digital twins of the original and receiving devices, and the task management unit and waiting list management subunit of the merged digital twins are updated synchronously. If the total transfer cost is greater than or equal to the estimated waiting time, the task transfer is terminated, and the original waiting list is retained. The task transfer instruction is specifically an instruction containing task information, handover requirements, etc., generated based on the original task parameters and the receiving device information. The synchronous update specifically involves marking the original task as transferred in the task management unit, removing the original device in the waiting list management subunit, and inserting the receiving device into the queue according to priority. The purpose of this step is to determine whether to execute the transfer based on the comparison between cost and waiting time, ensuring that the transfer improves overall efficiency rather than increasing losses.

[0163] In an optional embodiment, step 8f) of "generating a task transfer instruction and sending it to the digital twin of the original device and the receiving device" specifically includes the following sub-steps:

[0164] 9a) The task transfer instruction includes basic task information, handover requirements and subsequent task allocation information of the original equipment. The basic task information includes task ID, task type, start and end coordinates, and handover requirements include handover time window and handover location.

[0165] 9b) The task transfer command is sent to the digital twin of the original equipment and the receiving equipment via industrial Ethernet, and the two complete the task information synchronization through near-field communication;

[0166] 9c) After the original equipment completes the task handover at the handover location, it sends a handover confirmation signal to the merged digital twin and then proceeds to the starting point of the subsequent task; after receiving the task information, the receiving equipment updates the task status of its own digital twin sub-entity and prepares to pass through the corridor according to the waiting queue order.

[0167] 9d) If the receiving equipment malfunctions en route to the handover location, repeat steps 8c)-8f) to screen for new candidate equipment; if no equipment meets the criteria in the second screening, terminate the transfer, add the original equipment back to the waiting queue, and temporarily increase its priority.

[0168] In this embodiment, the issuance and execution of the task transfer instruction are carried out in four steps, with the core being to ensure orderly task handover and that the original equipment is not idle. It should be noted that the communication methods and handover rules involved in this process can be adjusted according to the factory network environment, and this application is not limited to them.

[0169] The first step, the task transfer instruction, includes basic task information, handover requirements, and subsequent task allocation information for the original equipment. The basic task information includes the task ID, task type, and start and end coordinates. The handover requirements include the handover time window and handover location. The basic task information is extracted from the core parameters of the original task from the task management unit of the merged digital twin. The task ID is a unique identifier, the task type is a classification specifying the nature of the task (e.g., raw material transportation, finished product handling), and the coordinates are location data specifying the spatial range of the task. The handover requirements are determined based on the original task's execution progress and the receiving equipment's movement time. The handover time window is the time range within which the original and receiving equipment must complete the handover, and the handover location is set by default to the original task's start point (for easy equipment docking). The subsequent task allocation information for the original equipment consists of newly allocated tasks from the smart factory's task pool, ensuring that the original equipment can immediately execute new tasks after the handover. The purpose of this step is to generate a task transfer instruction containing complete information, providing clear guidance to both parties involved in the handover.

[0170] The second step involves sending the task transfer command to the digital twins of both the original and receiving devices via Industrial Ethernet. The two devices then synchronize task information via Near Field Communication (NFC). Industrial Ethernet is a network used for device communication within the factory, characterized by low latency and high reliability, making it suitable for transmitting scheduling commands. NFC is a data transmission method between the original and receiving devices at close range (such as Bluetooth or NFC), chosen to avoid latency caused by cloud transmission and ensure rapid synchronization of task information. Task information synchronization involves the receiving device obtaining detailed parameters of the original task (such as cargo securing methods and handling speed limits), and the original device obtaining handover confirmation information. This step ensures accurate transmission of the task transfer command and synchronization of task information between the two parties, avoiding information discrepancies.

[0171] The third step involves the original device completing the task handover at the handover location and sending a handover confirmation signal to the merged digital twin before proceeding to the starting point of the subsequent task. Upon receiving the task information, the receiving device updates the task status of its own digital twin and prepares to use the corridor according to the waiting queue order. Specifically, the handover confirmation signal is the original device informing the system that the handover is complete, sent when the device's sensors detect that the task information has been synchronized and there are no anomalies. The original device proceeding to the starting point of the subsequent task avoids idleness after the handover, allowing full utilization of its capabilities through new tasks. The receiving device updating its task status involves changing its own task status from idle to executing a transfer task, updated after receiving and confirming the task information. Preparing according to the waiting queue order means the receiving device inserts itself into the waiting queue based on its priority, waiting to use the corridor. This step ensures that both the original and receiving devices can execute subsequent operations in an orderly manner after the task handover is completed, preventing process interruptions.

[0172] The fourth step involves re-executing the screening and scoring process for similar equipment to select new candidate equipment. If no suitable equipment is found in the second screening, the transfer is terminated, and the original equipment is re-added to the waiting queue with a temporary priority increase. Equipment failures are identified through fault signals (such as motor failure or sudden power drop) obtained from the digital twin of the receiving equipment. Re-screening candidate equipment aims to complete the task transfer as much as possible, avoiding transfer failure due to a single equipment failure. The absence of suitable equipment in the second screening indicates that there are no more devices within the task execution scope that meet the type and load conditions; terminating the transfer in this case avoids meaningless repeated screening. Temporarily increasing the priority of the original equipment occurs when the original equipment has been waiting for a period and the transfer has failed; increasing its priority reduces its subsequent waiting time and ensures fairness. This step addresses equipment failures during the transfer process, ensuring the robustness of the transfer process and preventing overall scheduling disruptions due to failures.

[0173] In an optional embodiment, step 4) "adaptively scheduling the automated devices corresponding to each digital twin sub-sub ...

[0174] 10a) Adjust the movement speed of the device in the corridor according to the task type data of the task attribute unit and the occupancy status data of the corridor twin unit; if the task is of high urgency and the corridor is not occupied, set it to the preset maximum movement speed; if the task is of low urgency or there are multiple devices intersecting in the corridor, set it to the preset safe movement speed.

[0175] 10b) Based on the queue length data of the waiting management subunit, switch the corridor access mode; if the queue length is not lower than the preset switching threshold, switch to one-way access mode; if the queue length is less than the preset switching threshold, switch to two-way access mode.

[0176] 10c) Based on the estimated task duration data, allocate the corridor time slice length, wherein the time slice length is the product of the estimated corridor passage time and the preset buffer multiple, to ensure that the equipment completes corridor passage within the time slice.

[0177] In this embodiment, the dynamic adaptation of corridor parameters is performed in three steps, the core of which is to optimize corridor access conditions according to scenario and task requirements. It should be noted that the speed and mode involved in this process can be adjusted according to the corridor structure, and this application is not limited to this.

[0178] The first step is to adjust the device's movement speed in the corridor based on the task type data of the task attribute unit and the occupancy status data of the corridor twin unit. If the task is of high urgency and the corridor is unoccupied, the preset maximum movement speed is set. If the task is of low urgency or multiple devices intersect in the corridor, the preset safe movement speed is set. The task type data specifically refers to the urgency classification of the task (e.g., high urgency, low urgency), determined based on the urgency score of the task attribute unit (e.g., a score ≥ 8 indicates high urgency). The corridor occupancy status data indicates whether there is equipment in the corridor and whether the equipment intersects, determined based on the location data of the corridor twin unit. The preset maximum movement speed is the maximum safe speed that the device can reach in the corridor, determined based on the corridor width, device performance, and safety requirements. The preset safe movement speed is the safe speed of the device in complex scenarios (e.g., multiple devices intersecting), determined based on the device's braking distance and reaction time. The purpose of this step is to adjust the movement speed according to the urgency of the task and the status of the corridor, balancing efficiency and safety.

[0179] The second step involves switching the corridor access mode based on the queue length data from the waiting management subunit. If the queue length is greater than or equal to a preset switching threshold, the mode switches to one-way access; if the queue length is less than the preset switching threshold, the mode switches to two-way access. The queue length data specifically represents the number of devices waiting in the waiting queue, obtained from statistics collected by the waiting management subunit. The preset switching threshold is a pre-set threshold for the number of devices that determines whether to switch access modes, determined based on corridor capacity and the risk of device convergence. If the queue is too long, one-way access can prevent congestion caused by device convergence. One-way access mode involves devices entering from one entrance and exiting from the other; two-way access mode allows devices to enter from both entrances. This step adjusts the access mode based on the number of devices in the waiting queue, preventing congestion caused by multiple devices converging and improving corridor access efficiency.

[0180] The third step involves allocating a time slice for the corridor based on the estimated task duration. This time slice length is calculated as: estimated corridor travel time × preset buffer multiple. This ensures that the equipment completes its corridor travel within the time slice. The estimated corridor travel time is the estimated time for the equipment to travel from the corridor entrance to the exit, calculated based on the corridor length and the adjusted movement speed. The preset buffer multiple is a pre-set multiple used to reserve buffer time to handle unexpected situations such as temporary deceleration or obstacle avoidance (e.g., the equipment needs to avoid obstacles in the corridor). The time slice length is the specific time range allocated to the equipment for travel within the corridor, ensuring that each piece of equipment completes its travel within the time slice and avoiding excessive time occupation that could affect subsequent equipment. The purpose of this step is to allocate reasonable travel time for the equipment, prevent equipment from exceeding its time limit in the corridor, and ensure that subsequent equipment can travel in sequence.

[0181] In an optional embodiment, step 4) is followed by the following steps:

[0182] 11a) The first scheduling strategy is parsed into control instructions corresponding to each digital twin sub-entity and sent to the corresponding physical devices, including automatic devices and common devices;

[0183] 11b) Collect execution data of physical equipment in real time, including actual movement path, operation completion time, equipment fault information and corridor access data, and feed the execution data back to the merged digital twin;

[0184] 11c) The digital twin is merged to update the status data of the corridor twin unit, waiting management sub-unit and task attribute unit according to the execution data. If the device position deviation exceeds the preset deviation threshold or the corridor congestion coefficient exceeds the preset congestion threshold, the adaptive scheduling in step 4) is re-executed to generate a second scheduling strategy, replace the first scheduling strategy and send it to the corresponding device.

[0185] In this embodiment, the execution and feedback of the scheduling strategy need to be carried out in three steps, the core of which is to form a closed loop of scheduling, execution, and adjustment. It should be noted that the thresholds and adjustment logic involved in this process can be adjusted according to the scheduling accuracy requirements, and this application is not limited to this.

[0186] The first step involves parsing the first scheduling strategy into control commands corresponding to each digital twin and distributing them to the corresponding physical automated devices and common equipment. These control commands are specifically device-recognizable operation commands (such as movement path commands and speed adjustment commands), transforming the abstract schemes in the first scheduling strategy (such as path coordinates and time nodes) into concrete operational parameters for the devices. Distributing these commands to physical devices involves the device controller transmitting the commands to the actual automated devices and common equipment (such as elevators and conveyor belts), with the transmission method determined by the device's communication protocol (such as Modbus or Profinet). The purpose of this step is to translate the digital scheduling strategy into actual device operations, driving task execution.

[0187] The second step involves real-time collection of execution data from physical devices, including actual movement paths, task completion times, equipment fault information, and corridor access data. This execution data is then fed back to the merged digital twin. Execution data is acquired through the devices' built-in sensors (such as position and speed sensors) and the controller. The actual movement path is obtained by stitching together coordinates collected by the position sensors. The task completion time is calculated based on the time difference between the task's start and end. Equipment fault information is obtained through fault sensors, and corridor access data is synchronized through sensors in the corridor twin unit. Feeding back to the merged digital twin involves transmitting the collected data to the corresponding unit within the twin (e.g., equipment status data is fed back to the digital twin sub-unit, and corridor data is fed back to the corridor twin unit). The purpose of this step is to obtain the actual situation of the devices performing tasks, providing a basis for subsequent judgments on whether to adjust the scheduling strategy.

[0188] The third step involves merging the digital twin and updating the status data of the corridor twin unit, waiting management subunit, and task attribute unit based on the execution data. If the device location deviation exceeds a preset deviation threshold or the corridor congestion coefficient exceeds a preset congestion threshold, adaptive scheduling is re-executed, generating a second scheduling strategy to replace the first scheduling strategy and distributing it to the corresponding device. Specifically, updating the status data involves synchronizing the latest status in the execution data (such as the new device location and the new corridor congestion coefficient) to each unit to ensure consistency between the twin and the physical scenario. The preset deviation threshold is the maximum allowable deviation for determining whether the device location deviates from the planned path, determined based on scheduling accuracy requirements (e.g., excessive deviation could lead to collision risks). The preset congestion threshold is a preset coefficient threshold for determining whether the corridor is excessively congested, determined based on corridor traffic efficiency requirements. The second scheduling strategy is a new scheduling strategy generated in response to execution anomalies, with the generation logic consistent with the first scheduling strategy, but the input data is the updated status data. This step aims to promptly address anomalies during execution, ensuring the task continues to execute efficiently through rescheduling, thus forming a closed-loop management system.

[0189] In an optional embodiment, step 11b) of "collecting execution data of physical devices in real time, including actual movement path, job completion time, equipment fault information and corridor access data" specifically includes the following sub-steps:

[0190] 12a) Obtain the real-time position coordinates, running speed and motor speed data of the automated equipment through the equipment controller, and use them as the basis for calculating the actual movement path and operation completion time;

[0191] 12b) Collect equipment fault signals, including motor faults and low battery alarms, through equipment sensors and mark them as equipment fault information;

[0192] 12c) The actual moving speed and passage time of the device in the corridor are collected by the lidar sensor associated with the corridor twin unit as corridor passage data;

[0193] 12d) According to the preset feedback frequency, the actual movement path, operation completion time, equipment failure information and corridor access data are integrated into execution data and uploaded to the merged digital twin.

[0194] In this embodiment, the data collection process involves four steps, with the core objective of comprehensively and accurately acquiring the actual operational data of the equipment and corridors. It should be noted that the collection methods and frequencies involved in this process can be adjusted according to data requirements, and this application is not limited to these.

[0195] The first step involves acquiring the real-time position coordinates, operating speed, and motor speed data of the automated equipment through the equipment controller. This data serves as the basis for calculating the actual movement path and job completion time. The equipment controller is the core component of the automated equipment used for control, possessing data acquisition and storage capabilities. The real-time position coordinates are the equipment's real-time position within the factory coordinate system, acquired through position sensors (such as GPS or UWB sensors) connected to the controller. The operating speed is the equipment's real-time speed, calculated by the controller based on the relationship between motor speed and wheel diameter. The motor speed is the rotational speed of the equipment's drive motor, acquired through a speed sensor connected to the controller. This step aims to obtain the core parameters of the equipment's operation, providing fundamental data for calculating the actual movement path (coordinate stitching) and job completion time (speed and distance calculation).

[0196] The second step involves collecting equipment fault signals, including motor faults and low battery alarms, through equipment sensors and marking them as equipment fault information. Specifically, these sensors are installed on critical components of the equipment (such as motors and batteries), with their deployment locations determined based on the components prone to failure. Fault signals are alarm signals sent by sensors when they detect abnormal component operation (such as excessively high motor temperature or low battery voltage). Equipment fault information is specifically the classification and labeling of fault signals (such as motor faults or low battery). The labeling logic is based on matching the type of fault signal with a preset fault database. The purpose of this step is to promptly detect equipment faults, providing a basis for subsequent scheduling adjustments (such as task transfer and queue removal).

[0197] The third step involves collecting data on the actual movement speed and travel time of devices within the corridor using lidar sensors linked to the corridor twin unit. These sensors, specifically the lidar sensors previously deployed at corridor entrances, exits, and bends, are used as corridor traffic data. Data linkage ensures that the data collected by these sensors is synchronized to the corridor twin unit in real time. Travel time is the time it takes for a device to enter and leave the corridor; the sensors record the timestamps of entry and exit, and the time difference is calculated. This step aims to obtain actual traffic data within the corridor, providing a basis for evaluating corridor traffic efficiency and adjusting scheduling strategies.

[0198] The fourth step involves integrating the actual movement path, job completion time, equipment fault information, and corridor access data into execution data according to a preset feedback frequency, and uploading this data to the merged digital twin. Data integration specifically involves packaging data from different sources into a preset format (such as JSON) to facilitate rapid data parsing and use by the merged digital twin. Uploading to the merged digital twin involves transmitting the integrated execution data to the corresponding storage unit of the twin via the industrial network. The purpose of this step is to centrally upload the dispersed collected data, ensuring that the merged digital twin can obtain complete execution data in a timely manner, supporting subsequent status updates and scheduling adjustments.

[0199] In an optional embodiment, the following steps are also included:

[0200] 13a) Periodically integrate the execution data collected in step 11b) with the corresponding first scheduling strategy, waiting queue adjustment record, and task transfer record into new training samples;

[0201] 13b) Add new training samples to the sample library of the AI ​​scheduling model and update the sample library according to the preset optimization cycle;

[0202] 13c) The gradient descent algorithm is used to iteratively train the AI ​​scheduling model, and the priority calculation weight, transfer cost evaluation parameters and corridor parameter adaptation threshold are adjusted.

[0203] 13d) Verify the accuracy of the trained model. If the accuracy improvement meets the preset accuracy threshold, replace the original model; if it does not improve, retain the original model and continue to accumulate new training samples.

[0204] In this embodiment, the optimization of the AI ​​scheduling model requires four steps, with the core being to improve model accuracy through iterative improvement using actual data. It should be noted that the algorithms and cycles involved in this process can be adjusted according to model performance requirements, and this application is not limited to these.

[0205] The first step involves periodically integrating the execution data with the corresponding primary scheduling strategy, waiting queue adjustment records, and task transfer records into new training samples. Execution data specifically refers to the previously collected actual operational data of the equipment and corridors; the primary scheduling strategy refers to the original scheduling scheme generated by the model; the waiting queue adjustment records refer to the operation records of insertion, removal, and rearrangement in the waiting queue; and the task transfer records refer to the triggering conditions, receiving equipment, and transfer results of task transfers. The integration logic of the new training samples is to associate the scheduling scheme, execution results, and adjustment records to form a complete data chain for model learning, ensuring that the samples reflect the actual effect of the scheduling scheme. The purpose of this step is to provide the latest actual data support for model optimization and avoid the model relying on old data, which could lead to a decrease in accuracy.

[0206] The second step involves adding new training samples to the AI ​​scheduling model's sample repository and updating the repository according to a preset optimization cycle. The sample repository is a database that stores training samples for centralized management and easy model access. The preset optimization cycle is the time interval between sample repository updates and model training, determined based on the data accumulation rate and model accuracy requirements. If data accumulation is rapid or the model needs to quickly adapt to new scenarios, the cycle needs to be set shorter. Updating the sample repository involves storing new training samples in a preset format (such as TensorFlow's TFRecord format) and deleting outdated or abnormal samples (such as abnormal data caused by equipment failure). This step aims to maintain the timeliness and accuracy of the sample repository, providing a high-quality data foundation for model training.

[0207] The third step involves iteratively training the AI ​​scheduling model using the gradient descent algorithm to adjust the priority calculation weights, transfer cost assessment parameters, and corridor parameter adaptation thresholds. Gradient descent is a commonly used algorithm for optimizing model parameters, characterized by its convergence stability and ease of implementation, making it suitable for adjusting scheduling model parameters. Iterative training involves inputting new training samples into the model, calculating the error between the model's predictions and actual results using the algorithm, and then adjusting the model parameters accordingly. The number of iterations is determined based on the model's convergence (iteration stops if the error is less than a preset threshold). The adjusted parameters include the urgency and duration weights in priority calculation, the time coefficient in transfer cost assessment, and the speed and mode thresholds in corridor parameter adaptation. The adjustment logic is determined based on the actual effectiveness of the scheduling schemes in the samples; if high-urgency tasks have low completion times, the urgency weight is increased. The purpose of this step is to optimize model parameters through iterative training, making the model-generated scheduling schemes more closely aligned with real-world scenario requirements.

[0208] The fourth step is to verify the accuracy of the trained model. If the accuracy improvement meets the preset accuracy threshold, the original model is replaced; otherwise, the original model is retained, and new training samples are accumulated. Specifically, model accuracy verification involves applying the trained model to a test scenario (such as a simulated cross-layer operation scenario), comparing the scheduling scheme generated by the model with the actual optimal scheme, and calculating accuracy metrics (such as traffic efficiency improvement rate and equipment utilization rate). The preset accuracy threshold is a pre-defined standard for judging whether the model optimization is effective, determined according to scheduling requirements. The purpose of model replacement or retention is to avoid a decrease in scheduling efficiency due to model optimization failure. This step evaluates the model optimization effect, ensuring that the model used always has high accuracy to support efficient scheduling.

[0209] In an optional embodiment, step 13c) "adjusting the priority calculation weight, transfer cost assessment parameters, and corridor parameter adaptation threshold" specifically includes the following sub-steps:

[0210] 14a) Based on the task completion time data in the new training samples, adjust the urgency weight and duration weight of priority calculation to improve the completion time of high-urgency tasks.

[0211] 14b) Based on the task transfer success rate data, adjust the time coefficient in the transfer cost assessment parameters to reduce the probability of transfer failure;

[0212] 14c) Based on the data of corridor congestion rate and passage efficiency, adjust the switching threshold and buffer multiple in the corridor parameter adaptation threshold to balance corridor passage efficiency and safety.

[0213] 14d) Synchronize the adjusted parameters to the corresponding sub-units of the merged digital twin to ensure that subsequent scheduling matches the optimized model parameters.

[0214] In this embodiment, the adjustment of AI scheduling model parameters needs to be performed in four steps, the core of which is to accurately optimize the parameters based on the actual scheduling effect. It should be noted that the parameters and adjustment logic involved in this process can be adjusted according to the needs of the scenario, and this application is not limited to this.

[0215] The first step involves adjusting the urgency and duration weights in the priority calculation based on task completion time data from the new training samples, thereby improving the completion time of high-urgency tasks. Task completion time data specifically refers to the actual time from task start to completion. By comparing the completion times of high-urgency and low-urgency tasks, the rationality of the current weights is determined. If the completion time of high-urgency tasks is lower than expected, it indicates that the urgency weight is insufficient. Adjusting the weights involves increasing or decreasing the values ​​of the urgency and duration weights (the sum of their weights is 1). The adjustment range is determined based on the time difference (a larger time difference requires a larger adjustment). This step optimizes the priority calculation logic, ensuring that high-urgency tasks receive resources first, thus improving their completion time.

[0216] The second step involves adjusting the time coefficient in the transfer cost assessment parameters based on the task transfer success rate data to reduce the probability of transfer failure. The task transfer success rate is specifically the ratio of the number of successful transfers to the total number of transfers. A low success rate may indicate inaccurate transfer cost assessment (e.g., underestimating handover time). The time coefficient is the weighting factor for each time item (e.g., handover time) in the transfer cost calculation. If handover time frequently exceeds the estimate, the handover time coefficient is increased to make the transfer cost calculation more accurate. This step optimizes the transfer cost assessment logic, reduces transfer failures caused by cost estimation errors, and improves the transfer success rate.

[0217] The third step involves adjusting the switching threshold and buffer multiplier in the corridor parameter adaptation threshold based on corridor congestion rate and traffic efficiency data to balance corridor traffic efficiency and safety. Corridor congestion rate data specifically refers to the frequency of corridor congestion (e.g., the proportion of congested time to total time); traffic efficiency data specifically refers to the number of devices passing through the corridor per unit time. If the congestion rate is high, the switching threshold for one-way traffic needs to be lowered (switching to one-way mode earlier); if traffic efficiency is low, the buffer multiplier for time slices needs to be appropriately reduced (reducing wasted time). Adjusting the threshold and multiplier involves adjusting the standard number of devices for the switching threshold and the value of the buffer multiplier. The adjustment range is determined based on the target values ​​for congestion rate and efficiency (e.g., if the target congestion rate is <10%, adjust the switching threshold until the target is met). The purpose of this step is to optimize the corridor parameter adaptation logic, improving traffic efficiency while avoiding congestion, achieving a balance between the two.

[0218] The fourth step is to synchronize the adjusted parameters to the corresponding sub-units of the merged digital twin to ensure that subsequent scheduling matches the optimized model parameters. Parameter synchronization specifically involves transmitting the adjusted priority weights, time coefficients, switching thresholds, and other parameters to the waiting list management sub-unit, task transfer sub-unit, and corridor twin unit of the merged digital twin. The synchronization logic ensures that subsequent scheduling requires these parameters (e.g., priority weights for waiting list ranking, switching thresholds for corridor adaptation). If the parameters are not synchronized, the scheduling logic will be inconsistent with the optimized model logic. This step ensures parameter uniformity throughout the entire scheduling system, avoids scheduling deviations due to parameter mismatches, and guarantees that the model optimization effect can be implemented.

[0219] This application provides a three-dimensional mapping adaptive scheduling system for smart factories, such as... Figure 2 As shown, it includes:

[0220] Panoramic Image Acquisition Module 1: Acquires panoramic images of each floor in the smart factory to be scheduled. Each panoramic image includes patterns of various automated equipment and background patterns of the automated warehouse environment.

[0221] Digital twin generation module 2: Based on the acquired panoramic image, it generates a digital twin for each floor, wherein the digital twin includes digital twin sub-units of each automated device;

[0222] Twin merging module 3: For each automated device that needs to operate across floors, the two digital twins that operate across floors are merged into the same floor based on the distance relationship to form a merged digital twin;

[0223] Adaptive scheduling module 4: For each generated merged digital twin, based on the determined first scheduling strategy, adaptively schedules the automatic devices corresponding to each digital twin sub-sub ...

[0224] The panoramic image acquisition module is a functional module used to acquire panoramic images of each floor of the smart factory. Its hardware includes a pre-set panoramic camera and image acquisition card, while the software includes an image acquisition control program and image stitching and correction algorithms. This module provides raw image data for the subsequent generation of digital twins. The deployment location needs to cover the acquisition sub-areas of each floor to ensure that panoramic images containing automated equipment, warehouse backgrounds, and corridors can be acquired. Its workflow corresponds to the panoramic image acquisition steps in the method, specifically: first, determine the overall acquisition area and divide it into acquisition sub-areas; then, control the panoramic camera to acquire images according to the sub-areas; finally, stitch and correct the images to output a panoramic image covering the entire floor area and including corridor features. The role of this module is to provide the physical scene image foundation for the entire system, ensuring the accuracy of subsequent digital scene construction.

[0225] The digital twin generation module is a functional module used to convert panoramic images into digital twins. Its hardware carrier includes an industrial server (for data processing and modeling), and its software carrier includes image preprocessing algorithms, feature extraction algorithms, and 3D modeling programs. This module is used to transform physical scenes into digital scenes, supporting subsequent cross-layer merging and scheduling. Its workflow corresponds to the digital twin generation steps in the method, specifically: first, the panoramic image is preprocessed by denoising and enhancing contrast; then, feature data of automated equipment, warehouse background, and corridors are extracted; finally, the feature data is mapped to a 3D spatial coordinate system to generate digital twins of associated task attribute units. The role of this module is to construct a digital reproduction of the physical scene, providing the system with a computable and controllable digital carrier.

[0226] The twin merging module is a functional module used to merge digital twins across floors. Its hardware includes edge computing devices (for rapid data fusion processing), and its software includes distance calculation programs, data fusion algorithms, and corridor twin unit construction programs. This module is used to achieve unified digital management and control of cross-floor operation scenarios, solving the problem that a single-floor twin cannot cover cross-floor needs. Its workflow corresponds to the twin merging steps in the method, specifically: first, determine the current twin and the target twin corresponding to the cross-floor device, calculate the distance between them and determine the merging method, then perform data fusion on twins that meet the conditions, add corridor twin units, and output the merged digital twin. The role of this module is to generate a merged twin containing cross-floor scenario and corridor data, providing unified digital scenario support for subsequent adaptive scheduling.

[0227] The adaptive scheduling module is a functional module used to schedule devices in the merged digital twin. Its hardware includes an AI computing server (for running the AI ​​scheduling model), and its software includes the AI ​​scheduling model, a waiting queue management program, a task transfer decision program, and a corridor parameter adaptation program. This module is used to achieve orderly and efficient operation of devices across floors. Its core is to integrate waiting, transfer, and parameter adaptation logic. Its workflow corresponds to the adaptive scheduling steps in the method, specifically: first, a waiting queue is generated based on task attributes and dynamically adjusted; then, a task transfer is triggered based on the corridor status and waiting time; finally, corridor parameters are adapted based on the task and corridor status, and the first scheduling strategy is output. The role of this module is to generate an optimized scheduling scheme, guide devices to efficiently complete cross-floor operations, and adjust the scheduling logic in response to dynamic changes.

[0228] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A three-dimensional mapping adaptive scheduling method for smart factories, characterized in that, include: Acquire panoramic images of each floor in the smart factory to be scheduled. Each panoramic image includes images of various automated equipment and the background image of the automated warehouse environment. Based on the panoramic image, a digital twin of each floor is generated, the digital twin including digital twin sub-units of each automated device; Based on the distance relationship between the two digital twins that need to operate across floors, the two digital twins that need to operate across floors are merged into the same floor to form a merged digital twin. Based on each merged digital twin, and using a determined first scheduling strategy, adaptive scheduling is performed on the automated devices corresponding to each digital twin sub-unit within the merged digital twin. This involves merging two digital twins that operate across different floors into a single-floor unit based on their distance relationship, forming a merged digital twin, including: The digital twin of the cross-floor device is determined on the current floor and is denoted as the current twin; and the digital twin of the target floor is determined and is denoted as the target twin; the cross-floor device is specifically an automated device that needs to perform operations between different floors, and its identification is based on the start and end floor information obtained from the task instructions of the device; Calculate the vertical distance between the current twin and the target twin at the corresponding floor level. Using the standard floor height of the smart factory as the unit, multiply the difference between the target floor number and the current floor number by the standard floor height to obtain the distance value. Determine if the distance value is less than or equal to the preset merging threshold. If it is less than or equal to the threshold, perform the merging operation. If it is greater than the threshold, first split the target twin into several sub-twins, and then select the sub-twin that is closest to the current twin to perform the merging. Data fusion is performed on two digital twins that meet the merging conditions. During the fusion process, a new corridor digital twin unit is added. The corridor digital twin unit is used to collect corridor occupancy status, movement speed data and corridor congestion coefficient data to form a merged digital twin. Adaptive scheduling includes equipment queuing across floors and corridors, with the following specific steps: Extract cross-floor task attribute data that requires the use of the corridor from the task attribute units of the merged digital twin, including task urgency score and estimated task duration; Calculate task priority, which is determined based on task urgency and estimated task duration; An initial waiting queue is generated in descending order of priority and stored in the waiting queue management subunit of the merged digital twin; At each preset update cycle, based on the real-time status and task attribute changes of the corridor twin units of the merged digital twin, the queue is updated by performing insertion, removal or rearrangement operations. Based on the updated waiting list, send status commands such as waiting, about to wait, or waiting cancelled to the digital twin of the corresponding automated equipment to guide the equipment to prepare for passage through the corridor; Adaptive scheduling includes task transfer across floors and corridors, with the following specific steps: Determine if the corridor is occupied; if so, calculate the estimated waiting time for the current waiting device. If the estimated waiting time is greater than the product of the estimated task execution time and the preset trigger multiple, and there are similar devices within the task execution range that meet the preset conditions, then the task transfer process will begin. A candidate device list is obtained for devices whose filter equipment type is incorrect or whose load rate is not lower than the preset load threshold. Candidate devices are weighted and scored based on their distance from the task start point, remaining battery power, and historical task matching accuracy. The device with the highest score is selected as the task receiving device. Calculate the total cost of task transfer, which is the sum of the transfer instruction issuance time, the time for the receiving device to travel to the task start point, and the task handover time. If the total transfer cost is less than the estimated waiting time, a task transfer instruction is generated and sent to the digital twin of the original device and the receiving device, and the task management unit and waiting management subunit of the merged digital twin are updated synchronously; if the total transfer cost is not less than the estimated waiting time, the task transfer is terminated and the original waiting queue is retained.

2. The method according to claim 1, characterized in that, The process of acquiring panoramic images of each floor in the smart factory to be scheduled includes: Determine the overall data collection area for each floor of the automated warehouse in the smart factory, and divide it into several sub-areas for data collection; Control the preset panoramic cameras on each floor to capture images sequentially according to the acquisition sub-area, and set the acquisition resolution and acquisition angle to the preset acquisition resolution and acquisition angle; Images from each collected sub-area are stitched and corrected to form a panoramic image covering the entire area of ​​the corresponding floor. The panoramic image also includes patterns of stairwells across floors, with the stairwell width, entrance location, and curve features marked.

3. The method according to claim 1, characterized in that, The step of generating a digital twin of each floor based on the panoramic image includes: Preprocessing of panoramic images includes noise reduction and image contrast enhancement; Extract feature data of automated equipment, background features of the automated warehouse environment, and corridor features from the preprocessed panoramic image; The extracted feature data is mapped to a preset three-dimensional spatial coordinate system to construct a digital twin that is proportional to the physical scene of the corresponding floor. The digital twin is also associated with a task attribute unit to store the task urgency and estimated task duration data of the automated equipment.

4. The method according to claim 1, characterized in that, The corridor twin unit is used to collect corridor occupancy status, movement speed data, and corridor congestion coefficient data, including: A preset number of lidar sensors are associated with the entrance, exit and bend of the corridor, and the location data and moving speed data of the equipment in the corridor are collected at a preset collection frequency. The location data collected is used to determine whether the corridor is occupied. If there are devices and the distance between the devices is less than the preset safety distance, it is determined to be occupied. Calculate the corridor congestion coefficient, which is the ratio of the actual travel time of the equipment to the preset theoretical travel time. Synchronize the occupancy status, movement speed data, and corridor congestion coefficient data to the merged digital twin in real time.

5. The method according to claim 1, characterized in that, The task transfer instruction includes basic task information, handover requirements, and subsequent task allocation information for the original device. The basic task information includes the task ID, task type, and start and end coordinates. The handover requirements include the handover time window and handover location. A task transfer instruction is generated and sent to the digital twins of both the original and receiving devices, including: The task transfer command is sent to the digital twins of the original equipment and the receiving equipment via industrial Ethernet, and the two complete the synchronization of task information through near-field communication. After the original equipment completes the task handover at the handover location, it sends a handover confirmation signal to the merged digital twin and then proceeds to the starting point of the subsequent task; after receiving the task information, the receiving equipment updates the task status of its own digital twin sub-entity and prepares to pass through the corridor according to the waiting queue order; If the receiving equipment malfunctions en route to the handover location, new candidate equipment will be selected. If no suitable equipment is found in the second selection, the transfer will be terminated, and the original equipment will be added back to the waiting list with its priority temporarily increased.

6. The method according to claim 1, characterized in that, In each generated merged digital twin, based on a determined first scheduling strategy, adaptive scheduling is performed on the automated devices corresponding to each digital twin sub-unit within the merged digital twin. This process also includes: The first scheduling strategy is parsed into control commands corresponding to each digital twin, and then sent to the corresponding physical devices, which include automatic devices and common devices. Collect execution data from physical equipment, including actual movement paths, job completion time, equipment fault information, and corridor access data, and feed the execution data back to the merged digital twin; The merged digital twin updates the status data of the corridor twin unit, waiting management sub-unit, and task attribute unit based on the execution data. If the device position deviation exceeds the preset deviation threshold or the corridor congestion coefficient exceeds the preset congestion threshold, the adaptive scheduling is re-executed, a second scheduling strategy is generated, the first scheduling strategy is replaced, and the strategy is sent to the corresponding device.

7. A three-dimensional mapping adaptive scheduling system for a smart factory, characterized in that, include: Panoramic Image Acquisition Module: Acquires panoramic images of each floor in the smart factory to be scheduled. Each panoramic image includes patterns of various automated equipment and the background pattern of the automated warehouse environment. Digital twin generation module: Based on the acquired panoramic images, it generates a digital twin for each floor, wherein the digital twin includes digital twin sub-units of each automated device; Twin merging module: For each automated device that needs to operate across floors, the two digital twins that operate across floors are merged into the same floor based on their distance relationship to form a merged digital twin; Adaptive scheduling module: For each generated merged digital twin, based on a determined first scheduling strategy, adaptively schedules the automatic devices corresponding to each digital twin sub-unit in the merged digital twin. This involves merging two digital twins that operate across different floors into a single-floor unit based on their distance relationship, forming a merged digital twin, including: The digital twin of the cross-floor device is determined on the current floor and is denoted as the current twin; and the digital twin of the target floor is determined and is denoted as the target twin; the cross-floor device is specifically an automated device that needs to perform operations between different floors, and its identification is based on the start and end floor information obtained from the task instructions of the device; Calculate the vertical distance between the current twin and the target twin at the corresponding floor level. Using the standard floor height of the smart factory as the unit, multiply the difference between the target floor number and the current floor number by the standard floor height to obtain the distance value. Determine if the distance value is less than or equal to the preset merging threshold. If it is less than or equal to the threshold, perform the merging operation. If it is greater than the threshold, first split the target twin into several sub-twins, and then select the sub-twin that is closest to the current twin to perform the merging. Data fusion is performed on two digital twins that meet the merging conditions. During the fusion process, a new corridor digital twin unit is added. The corridor digital twin unit is used to collect corridor occupancy status, movement speed data and corridor congestion coefficient data to form a merged digital twin. Adaptive scheduling includes equipment queuing across floors and corridors, with the following specific steps: Extract cross-floor task attribute data that requires the use of the corridor from the task attribute units of the merged digital twin, including task urgency score and estimated task duration; Calculate task priority, which is determined based on task urgency and estimated task duration; An initial waiting queue is generated in descending order of priority and stored in the waiting queue management subunit of the merged digital twin; At each preset update cycle, based on the real-time status and task attribute changes of the corridor twin units of the merged digital twin, the queue is updated by performing insertion, removal or rearrangement operations. Based on the updated waiting list, send status commands such as waiting, about to wait, or waiting cancelled to the digital twin of the corresponding automated equipment to guide the equipment to prepare for passage through the corridor; Adaptive scheduling includes task transfer across floors and corridors, with the following specific steps: Determine if the corridor is occupied; if so, calculate the estimated waiting time for the current waiting device. If the estimated waiting time is greater than the product of the estimated task execution time and the preset trigger multiple, and there are similar devices within the task execution range that meet the preset conditions, then the task transfer process will begin. A candidate device list is obtained for devices whose filter equipment type is incorrect or whose load rate is not lower than the preset load threshold. Candidate devices are weighted and scored based on their distance from the task start point, remaining battery power, and historical task matching accuracy. The device with the highest score is selected as the task receiving device. Calculate the total cost of task transfer, which is the sum of the transfer instruction issuance time, the time for the receiving device to travel to the task start point, and the task handover time. If the total transfer cost is less than the estimated waiting time, a task transfer instruction is generated and sent to the digital twin of the original device and the receiving device, and the task management unit and waiting management subunit of the merged digital twin are updated synchronously; if the total transfer cost is not less than the estimated waiting time, the task transfer is terminated and the original waiting queue is retained.

Images