Dynamic shelf level intelligent scheduling system

Abstract

The invention relates to the technical field of intelligent control, in particular to a dynamic shelf level intelligent scheduling system. The system comprises a goods shelf distribution module, a mechanical element friction optimization module, a goods gravity center shift evaluation module and a goods handling simulation module. Obtaining storage structure data, and carrying out dynamic shelf storage space detection to obtain dynamic shelf storage space data; obtaining cargo information; performing cargo shelf distribution simulation according to the cargo information and the dynamic shelf storage space data to obtain cargo shelf distribution data; performing electric push rod stroke anomaly detection according to the goods shelf distribution data to obtain electric push rod stroke anomaly data; based on the abnormal stroke data of the electric push rod, driving motor locked-rotor detection is carried out to obtain driving motor locked-rotor data; and performing mechanical element friction optimization according to the locked-rotor data of the driving motor to obtain element friction optimization data. The warehousing operation efficiency and the system operation stability are improved based on the intelligent control technology.

Description

Technical Field

[0001] The present invention relates to the field of intelligent control technology, and in particular to a dynamic shelf-level intelligent scheduling system. Background Art

[0002] Traditional dynamic shelf-level intelligent scheduling lacks high-precision modeling of storage structure space, making it difficult to achieve accurate matching of cargo appearance with shelf space, resulting in low shelf utilization efficiency; lacks a dynamic priority evaluation mechanism based on cargo access frequency, making it impossible to achieve intelligent hierarchical management of cargo in spatial layout, affecting handling efficiency; fails to introduce a mechanical friction optimization mechanism, and ignores the wear and stroke abnormalities of actuators such as push rods and motors in long-term operation, resulting in poor equipment operation stability and high maintenance costs; fails to conduct a comprehensive assessment of cargo center of gravity offset, and ignores the impact of stacking methods on center of gravity stability, which easily leads to safety hazards such as tipping and collision; lacks comprehensive optimization based on center of gravity offset and energy consumption cost in the cargo handling path planning process, resulting in unreasonable handling paths and high energy consumption; the system does not respond in a timely manner to abnormal situations such as cable mis-pressure and cargo collision during the handling process, making it difficult to form an intelligent feedback and scheduling closed loop for the handling results, limiting the improvement of the overall intelligence level of the system. Summary of the Invention

[0003] Based on this, it is necessary for the present invention to provide a dynamic shelf-level intelligent scheduling system to solve at least one of the above technical problems.

[0004] To achieve the above objectives, a dynamic shelf-level intelligent scheduling system is proposed, which includes the following modules:

[0005] The cargo shelf allocation module is used to obtain storage structure data, perform dynamic shelf storage space detection, and obtain dynamic shelf storage space data; obtain cargo information; perform cargo shelf allocation simulation based on the cargo information and dynamic shelf storage space data to obtain cargo shelf allocation data;

[0006] The mechanical component friction optimization module is used to detect electric push rod stroke anomalies based on the cargo shelf allocation data to obtain electric push rod stroke anomaly data; based on the electric push rod stroke anomaly data, drive motor stall detection to obtain drive motor stall data; and optimize mechanical component friction based on the drive motor stall data to obtain component friction optimization data;

[0007] The cargo center of gravity deviation assessment module is used to count cargo access frequency based on cargo information; assess cargo priority based on cargo access frequency; recommend shelf stacking areas based on cargo priority; and assess the degree of cargo center of gravity deviation based on shelf stacking areas.

[0008] The cargo handling simulation module is used to allocate intelligent cargo handling paths based on the degree of cargo center of gravity offset and component friction optimization data to obtain the cargo handling path; perform cargo handling simulation based on the cargo handling path to generate cargo handling data; trigger cargo redistribution signals based on the cargo handling data, and transmit the cargo redistribution signals to the shelf intelligent scheduling system to execute dynamic shelf intelligent scheduling tasks.

[0009] The present invention can realize accurate detection and modeling of storage structure space by constructing a cargo shelf allocation module, improve shelf space utilization efficiency, and avoid space waste caused by mismatch between cargo shape and shelf space; combine cargo information to perform shelf allocation simulation, and realize optimal position allocation for cargo of different volumes, weights and shapes, thereby improving stacking rationality and access efficiency; by introducing a mechanical element friction optimization module, real-time monitoring of the electric push rod stroke and the driving motor working state is realized, and stalls and abnormal stroke problems are discovered in time, and the friction structure design is optimized, thereby reducing the failure rate caused by mechanical wear, extending the system service life and reducing maintenance costs; further combining cargo information to count access frequency and establish a dynamic priority evaluation mechanism, effectively improving the access convenience of high-frequency access cargo, reducing operation path redundancy, and improving handling efficiency. Task response efficiency; by coupling the priority results with the shelf stacking area and evaluating the impact of stacking on the overall center of gravity offset based on the law of spatial center of gravity change, the risk of cargo overturning and collision caused by center of gravity imbalance can be effectively avoided, thereby improving the safety of system operation; at the same time, the center of gravity offset degree and component friction status data are introduced into the cargo handling path planning to realize intelligent path allocation, reduce the system's operating energy consumption and mechanical load in high friction areas and under unstable center of gravity conditions, and improve handling stability; through handling simulation, path execution feedback can be generated in real time, and the cargo redistribution signal can be triggered in combination with the simulation results, and the shelf scheduling system can be linked to complete the dynamic redistribution process, forming a complete closed-loop control system for handling tasks, strengthening the system's ability to identify and feedback abnormal conditions, and improving the intelligence, adaptability and operational stability of the overall dynamic scheduling system. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Other features, objects and advantages of the present invention will become more apparent upon reading the detailed description of non-limiting embodiments thereof made with reference to the following drawings:

[0011] Figure 1 This is a module diagram of a dynamic shelf-level intelligent scheduling system of the present invention;

[0012] Figure 2 Detailed functional flow chart of the cargo shelf allocation module in the present invention;

[0013] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0014] The following is a clear and complete description of the technical method of the present invention in conjunction with the accompanying drawings. Obviously, the embodiments described are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making any creative work are within the scope of protection of the present invention.

[0015] In addition, the accompanying drawings are merely schematic illustrations of the present invention and are not necessarily drawn to scale. Identical reference numerals in the figures denote identical or similar parts, and thus repetitive descriptions thereof will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities that do not necessarily correspond to physically or logically separate entities. These functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor and / or microcontroller approaches.

[0016] It should be understood that although the terms "first," "second," and the like may be used herein to describe various elements, these elements should not be limited by these terms. These terms are used solely to distinguish one element from another. For example, a first element may be referred to as a second element, and similarly, a second element may be referred to as a first element, without departing from the scope of the exemplary embodiments. The term "and / or" as used herein includes any and all combinations of one or more of the listed associated items.

[0017] To achieve this, please refer to Figures 1 to 2 The present invention provides a dynamic shelf-level intelligent scheduling system, which includes the following modules:

[0018] S1: Cargo rack allocation module, used to obtain storage structure data, perform dynamic rack storage space detection, and obtain dynamic rack storage space data; obtain cargo information; perform cargo rack allocation simulation based on the cargo information and dynamic rack storage space data to obtain cargo rack allocation data;

[0019] In this embodiment, a three-dimensional laser scanning device (such as the FARO Focus S70) is used to collect point cloud data of the storage area. During the collection process, a resolution of 1.2 million points per second is used for scanning, and the complete three-dimensional geometric data of the storage structure space is obtained with millimeter-level accuracy (±1mm). The data obtained from multiple angles are aligned by a point cloud stitching algorithm (using the ICP iterative closest point algorithm) to construct a complete three-dimensional space model of the warehouse. Based on the three-dimensional model, a spatial partitioning algorithm (such as a voxel grid partitioning algorithm, setting the voxel side length to 0.5m) is called to number and mark all dynamic shelves in layers, and the spatial coordinates, length, width, height and maximum load data of the available storage units on each layer are extracted based on the shelf hierarchical structure and boundary range to generate a dynamic shelf storage space data table. The data table stores the spatial volume, load limit (unit: kg), availability identifier (0 or 1) and space ID of each shelf unit in a JSON structure. An RFID scanning system (using a UHF RFID reader / writer, frequency band 920–925 MHz) reads the unique code of each incoming item. The system then integrates with the WMS system to obtain the item's length, width, and height (in mm), weight (in g), material, storage requirements (temperature, humidity, and safety), and inbound and outbound frequency. This information is structured into a product information table. A multi-constraint adaptive algorithm is used to simulate product shelf assignment. This algorithm prioritizes the matching of product dimensions and load tolerance (with an acceptable tolerance of ±5%), and the principle of minimizing space waste. The algorithm traverses all available shelf units, calculates the volume and load difference ratios between each item and the shelf, and assigns a comprehensive score based on the weight of the priority. A threshold score of 80 points (out of 100) is set, and the item with the highest score is assigned to the shelf. This generates the final product shelf assignment data, recording each item's assigned shelf ID, its starting coordinates (X, Y, Z), and its alignment.

[0020] S2: Mechanical component friction optimization module, used to detect electric push rod stroke anomalies based on the cargo shelf allocation data to obtain electric push rod stroke anomaly data; based on the electric push rod stroke anomaly data, drive motor stall detection to obtain drive motor stall data; and optimize mechanical component friction based on the drive motor stall data to obtain component friction optimization data;

[0021] In this embodiment, after receiving the cargo shelf allocation data output by S1, the displacement data of the electric push rod during the cargo placement operation is collected through the built-in position sensor (using the Hall effect position detection module). The sampling frequency is set to 100Hz, the start and end positions of the stroke are set to 0mm to 400mm, and the expected error is within the range of ±1mm. If the actual push rod end stroke is less than 390mm detected in the collected data, and there is a sudden drop in the position feedback speed (less than 50mm / s) during operation, it is recorded as a stroke abnormality. If the cumulative number of occurrences is more than 20 times, it is determined to be a stroke abnormality device and the electric push rod stroke abnormality data is output. Subsequently, the drive motor stall detection module is triggered according to the abnormal push rod number, and a standard load command (set to 50% of the full load torque at the rated current) is applied to the controlled motor to detect the stall current data. The current sensor module (accuracy ±0.1A) is used to monitor the actual current curve in the stall test. If the current duration exceeds 300ms and the current value exceeds 1.5 times the rated current (for example, if the rated current is 2A and exceeds 3A, it will be recorded as a stall), the drive motor stall data is output. Combined with the abnormal record of the electric push rod and the abnormal data of the motor stall, the friction compensation parameter adjustment module is called, and the mechanical structure friction modeling tool (using a mixed model of Coulomb friction and viscous friction) is used to calibrate the friction factor μ of each actuator. During the calibration process, by applying a standard force (100N) to each electric element, recording the actual displacement delay time, and inverting the friction factor μ, if μ>0.35, the lubrication scheme is replaced or the stroke speed is adjusted (the upper limit of the speed is adjusted from 20mm / s to 15mm / s), and the element friction optimization data table is output for the subsequent limiting parameters of the handling path.

[0022] S3: Cargo center of gravity deviation assessment module is used to count cargo access frequency based on cargo information; assess cargo priority based on cargo access frequency; recommend shelf stacking areas based on cargo priority; and assess the degree of cargo center of gravity deviation based on shelf stacking areas;

[0023] In this embodiment, based on the cargo information obtained in S1, a fixed-period statistical method is used to summarize the inbound and outbound records of each type of cargo within a 7-day period. Specifically, each daily cargo inbound or outbound operation is recorded as a visit, and the number of daily visits is accumulated one by one to obtain the total visit frequency for each type of cargo within that period. Based on this, a grading standard for visit frequency is established: a high-frequency access level is defined as 20 or more visits within 7 days, a medium-frequency access level is defined as 10 to 19 visits, and a low-frequency access level is defined as fewer than 10 visits. Each type of cargo is assigned a corresponding priority level based on its corresponding visit frequency, forming a structured cargo priority assessment table containing three fields: cargo number, visit frequency, and priority level. After generating a cargo priority evaluation table, the storage space is divided into multiple functional areas, combined with the existing shelf structure layout data in the current warehousing system. Shelf ID number ranges are assigned based on the area's function. For example, the area near the main aisle is defined as ID001 to ID050, serving as the recommended area for high-priority cargo; the central warehouse area or the area within a 3-meter radius of the robotic arm is designated as the recommended area for medium-priority cargo, with shelf IDs ranging from ID051 to ID200; and the deep warehouse area is designated as ID201 to ID300, corresponding to low-priority cargo. Based on the level field in the cargo priority evaluation table, a mapping rule is used to recommend high-priority cargo to shelf units near the main aisle or within the short path of the robotic arm. Medium-priority cargo is mapped to secondary aisles, and low-priority cargo is recommended to deep shelves away from aisles. Next, the center of gravity of the stacked cargo within each shelf unit is estimated based on a combination of the cargo's three-dimensional dimensions (length, width, and height), its stacking method (vertical or horizontal), and its weight. The weight parameters of each cargo unit are collected through the incoming weighing system. The stacking order is determined by the chronological order of incoming goods, with the lowest-height cargo corresponding to the bottom layer, and the weight is calculated upwards. The system combines the weight of each layer with the corresponding layer height to calculate the center of gravity height of the stacked cargo. This center of gravity height is then compared with the vertical center height of the shelf unit, whose total height is known in the system structure data. For example, if the total height of a shelf unit is 2 meters, and the assessed stacking center of gravity height is higher than 1 meter, the shelf unit is identified as having a severe center of gravity shift and is further marked as an unstable location. The obtained cargo priority data, recommended stacking area information, and the center of gravity shift of the cargo on the shelf are integrated into a set of cargo center of gravity shift assessment data. This data records each cargo unit's priority level, corresponding shelf number, stacking center of gravity height, and center of gravity shift identification information. This data will serve as input parameters for subsequent handling route allocation, used to comprehensively evaluate stacking stability and handling efficiency when optimizing handling routes.The entire process relies on static and dynamic parameters such as warehousing operation logs, warehouse structure data, cargo size and weight information, and warehousing time sequence, and uses data screening, area mapping, cargo location structure calculation and other means to achieve cargo center of gravity offset assessment.

[0024] S4: Cargo handling simulation module, which is used to allocate intelligent cargo handling paths based on the degree of cargo center of gravity offset and component friction optimization data to obtain the cargo handling path; perform cargo handling simulation based on the cargo handling path to generate cargo handling data; trigger cargo redistribution signals based on the cargo handling data, and transmit the cargo redistribution signals to the shelf intelligent scheduling system to perform dynamic shelf intelligent scheduling tasks.

[0025] In this embodiment, based on the component friction optimization data and cargo center of gravity offset assessment data output in the previous stage, the path planning module is used to construct an overall transport path graph model. A center of gravity offset correction mechanism is introduced during the path planning process, so that the path selection not only considers physical distance and friction resistance, but also fully weighs the transport risk caused by the unstable center of gravity of the cargo, thereby generating a complete transport path instruction set containing the speed, angle and push rod control parameters of each path point. This instruction set is then input into the transport simulation module, and the virtual actuator library is called in the simulation environment to simulate the propulsion action of the electric push rod, the rotation response of the motor, and the linkage operation of the shelf slide rails, and the action status of each frame is recorded in real time, including whether a collision occurs, whether there is a path offset, and whether the cargo is significantly tilted. During the simulation process, if it is detected that the cargo tilt angle in a certain frame exceeds the set threshold, or the path offset exceeds the allowable range, the system automatically marks it as a transport abnormality and counts the frequency of the abnormality and the cumulative error. When the cumulative number of exceptions exceeds the set value or the total amount of path errors exceeds the standard, the system immediately triggers the cargo redistribution mechanism, generates a redistribution signal containing the cargo number, failed path number, exception type and recommended backup path, and sends it to the intelligent scheduling controller through the internal communication bus, forcing the refresh of the current cargo distribution plan and path planning strategy to execute new dynamic shelf scheduling operations to ensure that the entire handling process achieves the optimal balance between efficiency and stability.

[0026] As an example of the present invention, refer to Figure 2 As shown, Figure 1 The functional flow diagram of the cargo shelf allocation module in this embodiment includes the following functions:

[0027] S11: Obtain storage structure data and extract structural dimension data;

[0028] In this embodiment, a high-precision laser ranging device (with an accuracy of not less than ±2mm) is used to perform a spatial scan of the internal structure of the storage environment. The scanning frequency is set to 20Hz, and the sampling density is set to not less than 10,000 points per cubic meter. During the scanning process, full-space imaging is performed from the four corners of the warehouse with a coverage angle of 360 degrees horizontally and 90 degrees vertically to obtain structural point cloud data including walls, columns, shelf foundations, aisle spacing, ceiling height, etc. The acquired point cloud data is projected onto a two-dimensional plane by the coordinate projection method, and the effective boundary coordinates between each group of shelf columns and aisles are extracted in combination with a spatial reconstruction algorithm (such as the ICP registration algorithm). When extracting structural dimension data, the main parameters calculated are shelf column length, shelf column spacing, layer height position, ground to ceiling height, layer thickness, aisle width, etc., where the shelf layer height is defined by the difference in layer spacing, and the accuracy error shall not exceed 5mm.

[0029] S12: Construct a three-dimensional model of the shelf based on the structural dimension data;

[0030] In this embodiment, a CAD modeling tool was used to perform 3D modeling on the extracted structural dimension data. The 3D modeling resolution was set to a 5mm grid unit, and the modeling objects included each shelf layer, supporting structure, shelf frame, and adjacent storage structure. During the modeling process, the Y-axis was used as the height reference, and the initial height point of each shelf layer was set. Hierarchical modeling was performed from the ground upward to ensure that the height of each shelf layer corresponded to the actual physical dimensions. Each shelf model was annotated with a number, boundary coordinates, and available volume. The 3D model output format was OBJ to ensure data structure compatibility. In the output model, each facet must contain three-point coordinates, a normal vector, and material thickness attributes to ensure accurate spatial docking when subsequently combined with point cloud data.

[0031] S13: Obtaining the volume data of the stored goods and converting it into an external shape point cloud type to obtain external shape point cloud data;

[0032] In this embodiment, each batch of goods is scanned 360 degrees using a three-dimensional scanner with an accuracy of ±1mm. The scanning speed is not less than 300,000 points / second, and the scanning distance is controlled in the range of 0.5 meters to 1 meter to ensure that the surface contour of the goods is completely captured. The scan data is output in PLY format, and the scan results are denoised using a point cloud filtering algorithm (such as voxel downsampling) to compress the original point cloud to no more than 100,000 points. The processed point cloud data is then subjected to a cluster analysis method (using the DBSCAN algorithm, with the minimum number of samples set to 10 and the radius parameter set to 5mm) to extract the outer contour point set, thereby obtaining the outer shape point cloud data that can be used for shelf space adaptation. All point cloud data must be converted into positions in the global coordinate system, and the coordinates of the extreme points of length, width and height must be recorded as volume bounding box information.

[0033] S14: Mapping the shape point cloud data to the shelf 3D model and calculating the dynamic shelf storage space to obtain dynamic shelf storage space data; obtaining cargo information; performing adaptive shelf layer height analysis based on the cargo information to obtain adaptive shelf layer height data;

[0034] In this embodiment, the point cloud data for each item is mapped to the three-dimensional shelf model generated by S12 using a coordinate alignment algorithm. A spatial grid matching method is used to align the item boundary points to the available grid cells within the shelf, with a grid granularity of 10mm. The outer contour of each item's point cloud is then cross-checked with the shelf level space to determine if there is a risk of overhang, overflow, or compression deformation. Spaces that meet the nesting criteria are marked as valid dynamic shelf storage spaces. Dynamic shelf storage spaces must record attributes such as the shelf number, level, height, width, and maximum item size. A database is then accessed to retrieve basic item information, including item number, type, frequency, material, minimum container size, and maximum transport size. This information is used to calculate the appropriate shelf height for each item. The calculation rules require that the item height must be less than 5% of the level height to avoid conflicts caused by installation errors. For example, if the item height is 460mm, the appropriate shelf height must be at least 483mm. Finally, the output is the appropriate shelf level height data, where each record includes parameters such as the item number, corresponding shelf level number, level height, and remaining space.

[0035] S15: Cargo shelf allocation simulation is performed according to the adapted shelf height data and the dynamic shelf storage space data to obtain cargo shelf allocation data.

[0036] In this embodiment, based on the adaptive shelf height data and dynamic shelf storage space data output in S14, a depth-first search algorithm is used to traverse all combinations of goods and shelf spaces. During the search process, the volume ratio of the goods is used as the main sorting basis, and high-frequency used goods are preferentially allocated to the shelf layer close to the ground, and the minimum safety distance between the left and right goods is ensured to be no less than 30mm. In the allocation simulation operation, each allocation action records parameters such as the goods number, shelf number, layer number, goods placement posture (XYZ axis rotation angle), and the minimum distance between the goods and the shelf boundary. To avoid repeated occupation, the corresponding shelf space is immediately marked as occupied after the simulated allocation, and the remaining allocable space is updated. Finally, the goods shelf allocation data is output. Each item in the data structure contains the goods number, target shelf number, target layer, height position, corresponding coordinate range, transportation direction and path initial point information.

[0037] Preferably, the cargo shelf allocation simulation includes:

[0038] Import the adaptive shelf height data and dynamic shelf storage space data into the cargo shelf allocation simulation module;

[0039] In this embodiment, the adapted shelf height data is derived from a structural laser ranging scanning system. This system, based on a laser measuring head (Keyence LJ-V7060), scans the shelf along the X, Y, and Z axes, outputting the net height between each shelf layer and the upper layer. The scan results are output as a two-dimensional array. Each data set contains the fields: "Shelf Number," "Layer Number," "Scan Timestamp," and "Layer Height" (in mm). The layer height measurement accuracy is ±1 mm, and the system error is controlled within ±0.5 mm after calibration. This data is written to the edge control node in standard CSV format and transmitted to the simulation module input buffer via a TCP / IP data bus interface. Dynamic shelf storage space data includes the current shelf space percentage occupied by each layer, the total load mass of the current layer, the remaining load capacity, and bounding coordinate frame information. This data is collected by a matching shelf status sensor system, which includes an infrared ranging module, a piezoelectric weighing sensor, and an encoder-based position monitor. Each sensor is connected to the industrial control master PLC via the Modbus-TCP protocol and uploads data at one-second intervals. The simulation module receives and classifies the shelves by number and establishes an index table based on a hash structure to quickly find the layer height and capacity information.

[0040] In the cargo shelf allocation simulation module, the cargo mass range is set to 0.5kg-50kg;

[0041] In this embodiment, during the initialization phase of the simulation module, the cargo mass range is set. This range is provided by the system global parameter configuration file and is set to a lower limit of 0.5kg and an upper limit of 50kg. The cargo mass data is collected from a high-precision electronic weighing device (Mettler-Toledo XPR205) with an accuracy level of ±0.01kg. The goods are weighed once before entering the warehouse, and the data is directly written into the attribute table under the corresponding cargo code. After reading the cargo mass, the simulation module determines whether it is within the [0.5,50]kg range. If it is not satisfied, it is marked as not eligible for simulation in the filtering process set in the simulation module. Data filtering is based on the logical expression 0.5≤M≤50. After filtering, the compliant cargo data set is retained for the next allocation operation.

[0042] In the cargo rack allocation simulation module, the cargo center of gravity stability coefficient range is set to 0.1-1.0, the cargo stacking coefficient range is set to 0-1, where non-stackable is 0, and the layer height tolerance range is set to ±10mm;

[0043] In this embodiment, the determination of the cargo center of gravity stability coefficient relies on the ratio of the cargo bottom contact area to its center of mass height. Each cargo is scanned by a 3D scanning device (such as the FARO Focus 350 series) to obtain a complete structural point cloud model. The bottom contact area is calculated using a projection algorithm on the Z = 0 plane to obtain the minimum circumscribed rectangular area A of the cargo bottom, in mm.2 The center of mass height, H, is determined by identifying the central voxel after 3D point cloud reconstruction, in units of mm. The stability coefficient is defined as S = A / H and is limited by the simulation module to a range of 0.1 to 1.0. If S < 0.1, the center of gravity is considered too high and the item is removed from the stacking process. If S > 1.0, it is limited to 1.0 for normalization. The stacking coefficient is determined by the material type and structural strength of the item. The compressive load limit is obtained through static pressure testing and normalized to a floating-point value between 0 and 1. The definition rules are as follows: the stacking coefficient for metal components is 1.0, for rigid plastics it is 0.85, for cardboard it is 0.4, for flexible packaging such as foam it is 0.2, and for fragile or structurally unstable items such as ceramics it is 0. The stacking coefficient S_stack is used as follows: if the bottom item, S_stack, is less than 0.5, no other items are allowed to be stacked on top of it. When items need to be stacked, the simulation module traverses the stacking coefficient table for the current layer and selects items with S_stack ≥ 0.5 as candidate support layers. The height tolerance is set to ±10mm. This parameter controls the acceptable error between the actual height of the cargo and the clearance of the target shelf level. The specific logic is: if the cargo height is H and the target shelf height is L, then the legal placement condition is L-10mm≤H≤L+10mm. This judgment logic is embedded in the space allocation function. If the tolerance limit is not met, the current level is skipped and not included as an optional level in the subsequent placement plan. This judgment statement is executed in the system using the expression abs(HL)≤10.

[0044] Run the cargo shelf allocation simulation module, perform cargo simulation placement operations, update shelf storage status, and output as cargo shelf allocation data.

[0045] In this embodiment, after completing all parameter settings, the cargo shelf allocation simulation module initiates simulation task processing through a unified data entry. The input parameters involved include: imported shelf height data, dynamic shelf storage space data, the size parameters of each item (length, width, height, in millimeters), the cargo mass parameter (in kilograms, ranging from 0.5kg to 50kg), the cargo center of gravity stability coefficient S (ranging from 0.1 to 1.0), the stacking coefficient S_stack (ranging from 0 to 1), and the layer height tolerance value (fixed to ±10mm). The simulation module uses a three-dimensional space partitioning tool to convert the three-dimensional size information of each item into a cubic voxel unit with a side length of 10mm. The cargo model is divided into X×Y×Z unit cubes through axial segmentation. Each voxel unit is assigned a unique three-dimensional index (i, j, k) and encoded using a unified voxel numbering system. The shelf space is converted into a 3D voxel array of the same size. All available shelf levels are scanned once, generating a corresponding 3D voxel occupancy matrix. The IDs of occupied voxels are set to 1 using a Boolean matrix, while vacant voxels are set to 0. The system then initiates a simulated placement loop, attempting to place each item in turn. The placement attempt involves three logical checks: the first step is voxel nesting. The system sequentially searches for nestable regions within the 3D voxel array of the current shelf layer and performs a sliding window match between the product voxel matrix and the current spatial matrix. If the target product voxel is completely within the spatial boundary limits within the candidate region (i.e., all voxel center coordinates satisfy Xmin≤x≤Xmax, Ymin≤y≤Ymax, Zmin≤z≤Zmax), and does not overlap with existing placed product voxels (i.e., the sum of voxel values ​​does not exceed 1), the voxel nesting is considered successful. The second step is a stability check. The stability coefficient S of the product is input. If S is less than 0.1, the current layer is skipped and the next layer is attempted. If S is greater than or equal to 0.1, the next layer is attempted. The third step is a stacking suitability check. If the current product needs to be placed on top of another product (i.e., there is a contact surface in the Z direction), the system checks whether the stacking coefficient S_stack of the underlying product is greater than or equal to 0.5. If the stacking coefficient is less than 0.5, the stacking support conditions are not met, and the system cancels the current placement attempt and moves on to the next candidate layer. If the current placement candidate position is still retained after passing the above triple judgment logic, the system will perform a simulated placement operation and write the coordinates, center point position, and center of gravity height (determined by the Y coordinate of the geometric center of the cargo and the mass distribution of the cargo) of the cargo voxel in three-dimensional space into the status record of the current shelf layer.Each time a simulation placement operation is performed, the system updates the following information in real time: the set of occupied voxel IDs on the current level, the total weight of the load on the current level (the original value plus the current cargo mass, in kg), and the stacking height of the current level (in mm). This update operation is written into the real-time mapping table of the shelf status. The mapping table is constructed with a double-key HashMap data structure. The first key is the shelf number, the second key is the level number, and the corresponding values ​​are the cargo voxel set, the remaining voxel distribution map, and the current level load data set. All of the above placement processes are completed through three-dimensional voxel map operations and Boolean matrix logic processing, and do not involve abstract models. After all cargo placement tasks are completed, the simulation module performs a unified data export operation to generate a cargo shelf allocation data file. The file format is a standardized JSON structure, with fields including: Cargo ID (string), Shelf Number (string), Storage Level Number (integer), Cargo Center Coordinates (floating-point X / Y / Z coordinates, in mm), Center of Gravity Height (floating-point, in mm), Stability Factor S (floating-point), Stacking (Boolean, True / False), Stacking Level Number (integer, -1 if not available), and Placement Legal Flag (Boolean, True if all checks pass). The exported data is then transferred to the edge server's cache forwarding mechanism via the control module, which then uploads the JSON file to the central dispatch system. The central system then executes the path scheduling algorithm and task allocation logic based on the file, completing path planning and generating task instructions for the automated handling equipment.

[0046] Preferably, the electric push rod stroke abnormality detection in the mechanical element friction optimization module includes:

[0047] Calculate the push rod target stroke according to the cargo shelf allocation data to obtain the push rod target stroke data;

[0048] In this embodiment, the calculation of the push rod target stroke is directly input using the "Cargo Placement Center Coordinates" field in the cargo shelf allocation data. This field is three-axis floating-point coordinate data in mm, representing the absolute position of the target cargo center point in the three-dimensional coordinate system of the shelf (X_target, Y_target, Z_target). The initial position of the push rod is a fixed reference coordinate (X0, Y0, Z0). This coordinate is obtained by the mechanical assembly station through a 3D laser positioning tool during system initialization, with an accuracy of ±0.05mm. The three-axis coordinate data is written to the system cache via the PLC interface. The calculation logic of the push rod target stroke is the three-dimensional coordinate difference, that is, the stroke vector ΔX = X_target–X0, ΔY = Y_target–Y0, ΔZ = Z_target–Z0. The final push rod target stroke data is the three-axis stroke vector L = √(ΔX2+ΔY2+ΔZ2). This data is stored in a structure array in the form of single-precision floating-point numbers, and each structure corresponds to the push rod path task of a piece of cargo. The entire travel calculation process is polled and processed by the three-axis coordinate calculation module built into the embedded controller within a 20ms cycle. The calculation results are stored in the "push rod path scheduling table", which is a two-dimensional index table with the index key being "goods ID + shelf number + placement layer number" for subsequent path scheduling and instruction issuance.

[0049] Identify abnormal push rod jitter and bounce based on push rod target stroke data to obtain abnormal push rod jitter and bounce data;

[0050] In this embodiment, the system collects motion curve data of the push rod execution phase. The data comes from a high-resolution grating encoder installed on the push rod spindle, with a sampling frequency of 10kHz and a resolution of 0.01mm. The encoder data is compared with the push rod target stroke data in real time, and the current position error ε(t) = actual position (t) – theoretical target path (t) is calculated in every 10ms cycle. The system performs frequency domain analysis on the error sequence through the Fourier transform module and extracts its spectrum amplitude. If there is a peak frequency band higher than 20Hz in the frequency domain and the amplitude is greater than 5mm, it is judged that the current push rod has a jitter anomaly, and the anomaly type is written as "multi-frequency jitter" in the anomaly record table. The record fields include timestamp, cargo number, target stroke, jitter amplitude, and main frequency component. If there is a low-frequency band (less than 3Hz) in the spectrum that changes slowly and is accompanied by periodic overshoot, it is judged as a "stroke jitter anomaly" and the corresponding field is written synchronously. The anomaly detection module completes frequency domain calculations through the fast Fourier transform library (using the Cooley-Tukey algorithm) in the DSP chip. The anomaly threshold is fixed at an amplitude of 5 mm and a frequency of 20 Hz. All data out of the range are recorded as Boolean flags in the anomaly database.

[0051] Extract putter beating frequency based on putter beating and jitter abnormality data;

[0052] In this embodiment, the system retrieves the frequency band containing the main peak in the frequency domain of the stored jitter anomaly data and extracts the corresponding main frequency value f_peak in Hz. The main frequency identification logic is used to extract the push rod jump frequency. Specifically, the frequency point corresponding to the maximum amplitude, f_max = argmax(|FFT(ε(t))|), is found in the spectrum vector that has completed the FFT analysis. If there are multiple main peaks, the first three main frequencies are extracted in descending order of amplitude, denoted as f1, f2, and f3, respectively, to form the main frequency vector F_jump = [f1, f2, f3]. The peak amplitudes A1, A2, and A3 corresponding to each frequency are recorded as a reference for determining the jump level. This data is stored in the "Push Rod Jump Frequency Analysis Table" and serves as the input parameter for the next stage of the push rod servo system's telescopic behavior simulation. This process does not use a filter, the frequency resolution is fixed at 1 Hz, the frequency domain window length is 1024 points, and the sampling window is weighted by a Hanning window function to reduce spectral leakage.

[0053] Perform push rod extension simulation according to the push rod beating frequency to obtain push rod extension data;

[0054] In this embodiment, the push rod telescopic simulation uses a jump frequency vector F_jump and a corresponding amplitude vector A_jump as inputs, executing a resonant telescopic simulation within a three-dimensional virtual push rod model. The simulation target is the mechanical response curve under servo drive. Each jump frequency component is input as a sinusoidal drive source, defining the telescopic displacement S(t) = ΣAi·sin(2πfi·t), where t is the simulation time (set to a range of 0 to 2 seconds, with a sampling interval of 1 ms), and Ai and fi are the amplitude and frequency of the i-th main frequency, respectively. The push rod telescopic simulation module is implemented using the physical modeling tool SimStruct platform, with bearing stiffness K = 1500 N / mm, damping C = 20 Ns / mm, and mass M = 8 kg. During the simulation, the maximum and average displacement response values, period length, and stability deviation are recorded, ultimately generating a push rod telescopic data table with fields including maximum telescopic displacement S_max, average telescopic amplitude S_avg, main period T_main, and instantaneous deviation δ(t). This table serves as input for subsequent guide rail slot clearance analysis.

[0055] Calculate the guide rail groove structure clearance based on the push rod telescopic data;

[0056] In this embodiment, the calculation of the guide rail groove structure gap is based on the difference between the maximum telescopic displacement S_max of the push rod and the theoretical rated stroke L_nominal. The calculation formula is Δ_gap = S_max – L_nominal, and the unit is mm. L_nominal is the standard value of the push rod working stroke set by the mechanical structure drawing during the system initialization phase (fixed to 520mm), and this value of each push rod is stored in the device static configuration file. If Δ_gap>2mm, it is recorded as "overtravel gap", Δ_gap<–2mm is recorded as "blocking gap", and –2mm≤Δ_gap≤2mm is "normal gap". This judgment logic is completed in the embedded processor of the execution unit. It is calculated once each set of telescopic simulations is completed, and the data is written into the "guide rail gap structure mapping table", where the index is "push rod number + level number" and the value is the Δ_gap real number.

[0057] Calculate the assembly offset based on the structural clearance of the guide rail groove;

[0058] In this embodiment, the assembly offset is determined by the guide rail groove gap data and the actual installation inclination angle of the push rod. The inclination angle is collected by the laser alignment device during installation. The equipment model is KEYENCE LJ-V7000, with an accuracy of ±0.01°, and the three-axis angle (αx, αy, αz) is automatically recorded by the system sensor. The assembly offset calculation formula is: offset O = Δ_gap / tan(θ), where θ is the angle between the actual inclination angle of the push rod and the ideal vertical direction, in radians. The unit of O is mm. In order to accurately identify the displacement deviation, each offset calculation must be verified. If O>1mm, it is recorded as "abnormal assembly", otherwise it is "allowed offset". The calculation logic is completed in the upper computer structure error analysis module, and the final result is written into the "push rod structure offset record table". The fields include assembly angle, structural gap, and offset value O.

[0059] The electric push rod stroke abnormality detection is performed according to the assembly offset to obtain the electric push rod stroke abnormality data.

[0060] In this embodiment, after obtaining the offset O, the electric push rod stroke abnormality detection module calculates the actual maximum allowable stroke range as [L_target-O, ​​L_target+O] based on the current push rod target stroke L_target. The system then performs boundary matching on the actual execution stroke L_actual (obtained from the grating encoder reading). If It is then determined as "abnormal stroke". The abnormal types are further divided into "overshoot abnormality" (L_actual > L_target + O) and "lag abnormality" (L_actual < L_target – O) according to the offset direction, and are respectively written into the abnormal data record table. The detection period is fixed at 10 ms and is completed using a hard real-time kernel. The fields of the abnormal data table include the pusher number, abnormal type, target stroke, actual stroke, assembly offset value, and detection timestamp.

[0061] Preferably, the detection of the drive motor stalling in the mechanical component friction optimization module includes:

[0062] Collecting gear engagement images based on the abnormal data of the electric pusher stroke;

[0063] In this embodiment, when there is a displacement deviation of the pusher operation in the abnormal data of the electric pusher stroke greater than 2.5 mm and the deviation duration exceeds 300 ms, the gear state acquisition module is started. An industrial camera with 5 million pixels of model MV-CA050-20GM is arranged above the meshing area of the electric pusher and the gear, and the shooting angle is set at a 45° oblique angle to ensure that the gear meshing interface completely enters the picture range. Infrared backlight (wavelength 850 nm) is used for illumination, and the exposure time is set at 3 ms and the gain is 5 dB to ensure clear gear edge contours. 10 frames of images are synchronously collected per second during the pusher operation, the image resolution is saved as 2592×2048, and the saving format is uncompressed BMP format. The acquisition timing is synchronized with the pusher drive motor through a pulse encoder, and the trigger logic is set as abnormal position trigger + 5-frame window buffer mechanism. The image number is bound to the abnormal data for subsequent trajectory analysis.

[0064] Identifying the gear tooth profile edge according to the gear engagement image; <00001​​​​​​​In this embodiment, the tooth edge points extracted from multiple frame images are aligned in chronological order. The SIFT algorithm in the feature point matching technology is used to match the tooth top feature points between consecutive frames. After eliminating the mismatched points, the least squares method is used to solve the pose transformation matrix to obtain the gear rotation vector and displacement vector sequence. The gear motion trajectory is inverted through the transformation matrix, and the relative offset and rotation angle of each gear meshing cycle are calculated in combination with the timestamp. The trajectory data recording format is: {timestamp t_i, tooth vertex coordinates x_i, y_i, rotation angle θ_i, radial runout r_i}, the sampling period is 100ms, and the recording time is not less than 3s. If the r_i fluctuation in the trajectory is greater than 0.15mm, it is automatically marked as a meshing fluctuation area.

[0068] Calculate the occlusal offset based on the occlusal trajectory data;

[0069] In this embodiment, the radial positions r_i of all tooth vertices are extracted from the obtained meshing trajectory data. A trajectory fit is performed for each tooth, using a least-squares circle fit to obtain the theoretical tooth trajectory circle r_avg. The actual radial offset Δr_i of each tooth vertex is calculated as |r_i - r_avg|. A maximum offset threshold of 0.1 mm is set for each tooth, and any excess is considered the meshing offset. A statistical analysis of the offsets of all teeth is performed, recording the average offset μ_r, maximum offset max_r, and offset standard deviation σ_r. The results are output in the format {tooth number i, Δr_i, μ_r, σ_r}. This offset data is used to determine the severity of problems such as meshing eccentricity and gear shaft offset.

[0070] Evaluate gear wear based on mesh offset;

[0071] In this embodiment, the offset Δr_i of each tooth is compared with the theoretical tooth profile, and the difference between the arc radius of the tooth top and the standard radius of each tooth is analyzed. When the difference is greater than 0.12mm, the tooth top wear is considered significant; the proportion of worn teeth is calculated. If the proportion is greater than 30%, it is considered to be in a moderate wear state, and if it exceeds 60%, it is defined as a severe wear state. The standard tooth profile is set by the design drawing parameters, with the base circle radius set to 12.5mm, the tooth top height to 2.25mm, the tooth root height to 1.85mm, and the tooth thickness to 3.5mm. The wear degree results are output as tooth number, wear value, and wear type label, which includes three categories: "mild", "moderate", and "severe". This step is completely based on geometric comparison and analysis, and there is no need to introduce a prediction model.

[0072] Predict gear instantaneous sticking according to gear wear degree and obtain gear instantaneous sticking data;

[0073] In this embodiment, under severe wear conditions, the bite angle change rate Δθ_i / Δt in the bite trajectory fluctuation area is analyzed. If the instantaneous angular velocity decrease rate is greater than 10 rad / s 2 , and lasting longer than 20ms, is flagged as a sign of gear jamming. If this phenomenon occurs on three consecutive teeth within the same 15° angular range, it is recorded as a high-risk area for gear instantaneous jamming. Jamming data is output in the format of {time t_i, tooth number i, zone angle θ_start–θ_end, angular velocity change Δω, duration Δt}. This data is derived from the instantaneous derivative calculation after trajectory export, eliminating the need for external measurement devices.

[0074] The drive motor stall risk is evaluated based on the gear instantaneous sticking data to obtain the drive motor stall data.

[0075] In this embodiment, after the high-risk area of ​​instantaneous gear jam occurs, the motor current waveform data is collected (through a current sensor, model LEM LA 25-NP), and the motor torque change rate dT / dt is detected. When the torque change rate during the jam period is greater than 50Nm / s, and the current peak exceeds the rated value (set to 5A) by more than 20%, that is, exceeds 6A, it is regarded as a sign of stall. The recorded data is {stuck time t_i, current peak I_max, torque change rate dT / dt, duration Δt}, and the data is uploaded to the scheduling control module for the issuance of the drive device shutdown protection instruction. The jam risk calculation does not require the construction of a prediction model, and is completely dependent on the real-time current, voltage and mechanical behavior quantitative parameter analysis.

[0076] Preferably, the mechanical element friction optimization in the mechanical element friction optimization module includes:

[0077] Calculate the push rod transmission friction force based on the drive motor stall data;

[0078] In this embodiment, a high-frequency current sampling sensor samples the motor's stall current in real time at a fixed interval of 1ms, ensuring that peak current changes are captured every millisecond after the stall condition begins. Simultaneously, a highly sensitive dynamic torque sensor is installed at the connection between the motor output shaft and the push rod to acquire torque data at the moment of stall in real time. Torque data is measured in N·m, with a sampling accuracy within ±0.05 N·m. The stall duration is determined by an encoder that detects changes in motor speed. When the motor stops, the encoder output pulse value remains continuously zero. The duration of the continuous zero pulses is used to determine the duration of the stall, ensuring synchronization between the stall event and mechanical data collection. The push rod drive system is driven by a ball screw mechanism. The ball screw radius is measured with high precision using a three-dimensional coordinate measuring machine with a resolution of 0.01mm. Multiple axial and radial sampling points are sampled during measurement, and the average value is used as the final input parameter to ensure that the r value accuracy is within ±1%. Based on the measured stall torque T_b (unit: N·m) and the screw radius r (unit: m), the push rod transmission friction force F_f is calculated using the formula F_f = T_b / r. The result is output in Newtons (N). All calculations are uniformly based on the International System of Units to avoid errors introduced by unit conversion. The final push rod transmission friction force serves as an important basis for subsequent assembly clearance adjustment and transmission mechanism optimization. In the control system, it is stored in the friction force recording unit of the PLC main control module as floating-point data.

[0079] Adjust the assembly clearance based on the push rod transmission friction force to obtain the assembly clearance data;

[0080] In this embodiment, the assembly gap adjustment is performed using a micron-level automatic adjustment assembly platform, and the input parameter is the push rod transmission friction value in N. The friction value is compared with the theoretical friction force range of the ball screw and the push rod connecting shaft, which is set to 20-35N. If it exceeds this range, it is determined to be a gap abnormality. If the friction force exceeds 35N, the assembly gap between the push rod and the ball screw is adjusted to reduce the actual gap, and the adjustment step is 0.01mm; if it is less than 20N, the gap is increased. The gap data is measured in real time by a laser interferometer, using a double-ended laser probe, scanning every 10ms, outputting the current gap value and automatically recording it in the industrial controller storage unit. After the gap adjustment is completed, the final gap data is confirmed with a standard tolerance of ±0.005mm.

[0081] Performing nitriding treatment on the surface of the push rod based on the assembly clearance data to obtain nitriding data on the surface of the push rod;

[0082] In this embodiment, the push rod surface treatment adopts an ion nitriding process, and the processing equipment is a medium frequency plasma nitriding furnace. The preset temperature inside the furnace body is 520°C, the nitrogen flow rate is 40L / min, and the processing time is 6 hours. The assembly gap data is directly used to control the nitriding thickness: when the gap is in the range of 0.02-0.03mm, the nitriding layer depth is set to 0.3mm; when the gap is less than 0.02mm, it is set to 0.2mm; when it is greater than 0.03mm, it is set to 0.4mm. The nitriding depth control adopts an in-process plasma discharge monitoring system, and the voltage is controlled within the range of 550V±10V. After treatment, the thickness of the nitride layer is measured by scanning electron microscopy (SEM), and the data error shall not exceed ±3μm. Finally, the nitriding thickness, processing time, and processing temperature are recorded as the push rod surface nitriding data.

[0083] Perform motor shaft deformation detection based on the drive motor stall data to obtain motor shaft deformation data;

[0084] In this embodiment, the stall current and stall torque in the stall data of the driving motor are used as detection input signals, and the axial force is reconstructed through a static load loading experiment. The loading platform applies an equivalent torque, and a high-precision laser interferometer probe (resolution of 0.1 μm) is used to simultaneously detect along the radial and axial directions of the shaft. The spacing between the measuring points is fixed at 5 mm, and the total length is set according to the length of the motor shaft to determine the detection point. The maximum deflection change of the shaft bending amount is calculated using the three-point method. If the offset of any measuring point exceeds 10 μm, it is recorded as deformation. The detection process ensures that the ambient temperature is constant at 20 ± 1 ° C to avoid errors introduced by thermal expansion and contraction. Finally, the offset value of each measuring point is combined with the axial position coordinates to form the motor shaft deformation data, which is saved in matrix format for subsequent processing.

[0085] Optimize the bearing support spacing based on the motor shaft deformation data to obtain the bearing support optimized spacing data;

[0086] In this embodiment, the deformation data of the rotating shaft is input into the multi-point support structure calculation module, and optimized and adjusted according to the position of the maximum deformation point. The bearing support spacing in the original support structure is set to 300mm, and the minimum deflection configuration is identified by performing nonlinear fitting between the maximum deformation point and the support point spacing. The calculation method uses the finite difference method to discretely model the rotating shaft, with the step size set to 10mm, and the optimal support position is determined by solving the static force balance equation. The support spacing adjustment range is set to 200-400mm, and the step adjustment interval is 10mm. Finally, the deformation minimization scheme is selected, and the coordinate difference of the left and right support positions after optimization is recorded, which is the bearing support optimization spacing data.

[0087] The push rod surface nitriding data and the bearing support optimization spacing data are integrated to obtain the component friction optimization data.

[0088] In this embodiment, the correlation analysis of the push rod nitride layer thickness, processing temperature and bearing support spacing data is performed. The data integration process is first normalized, and the nitride layer thickness is normalized in units of 0.1mm, and the corresponding push rod hardness variation range is calibrated. The hardness value is obtained according to the Rockwell hardness tester; the support spacing is divided into 10mm grades in units of mm. A mapping relationship is established in the controller in a two-dimensional matrix format: the nitride thickness (row) and the support spacing (column) are combined to form an index, and each group corresponds to a friction factor (calibrated with a granularity of 0.01), which is set according to the median value of the actual friction test data. The friction coefficient of the push rod is tested under a stable load (30N) and a constant stroke speed (10mm / s) using a friction test bench. The friction factor matrix finally obtained is the component friction optimization data, which serves as the standard input for subsequent push rod material processing and structural support design. The data is stored in the industrial edge computing module external to the PLC for real-time call and update.

[0089] Preferably, the cargo center of gravity offset assessment module includes the following functions:

[0090] Statistics of historical cargo in and out records based on cargo information;

[0091] In this example, the warehouse management system database retrieves the inbound and outbound log table associated with each barcode number and extracts the corresponding field data for each record, including the entry and exit time, item number, operator, and workstation number. The log table is grouped by item number, and the number of inbound and outbound transactions for each item number over the past 30 consecutive days is counted. The time window is fixed at 30 days to ensure consistent statistical coverage. All original record data must ensure that the timestamp field is accurate to the second. Time synchronization relies on a unified time server, using the NTP protocol to avoid data bias.

[0092] Calculate the frequency of cargo access based on the cargo historical inbound and outbound records;

[0093] In this embodiment, after obtaining the historical in-and-out records of goods, the frequency of goods access needs to be calculated based on these records. The access frequency is calculated on a daily basis. The sum of the number of in-and-out times of a certain numbered goods in a unit time window is divided by the length of the time window to obtain the average daily access frequency F, which is times / day. The formula is F=N / T, where N is the total number of in-and-out times during the statistical period, and T is the number of days (fixed at 30). For example, if a certain good is in and out of the warehouse a total of 90 times within 30 days, then its F=90 / 30=3 times / day. This value is an important parameter for evaluating priority and is recorded in the data cache of the scheduling system.

[0094] Evaluate cargo priority based on cargo access frequency;

[0095] In this example, a fixed threshold F0 of 2.5 times per day is set as the access frequency threshold. If the access frequency F of a particular item exceeds F0, it is assigned high priority; if F ≤ F0, it is assigned low priority. Priority is represented by a Boolean variable P, where high priority P = 1 and low priority P = 0. After the priority fields of all items are updated, they are written into the scheduling system's real-time scheduling list for direct access by subsequent scheduling modules.

[0096] Extract high-priority cargo data based on cargo priority;

[0097] In this example, the real-time dispatch list is traversed and all cargoes with a priority of 1 are combined into a high-priority cargo dataset, including their ID, specifications, dimensions (length, width, and height), weight, unit volume mass, and packaging shape. This dataset is stored as a two-dimensional array, with each row representing a category of cargo. The data is sorted from highest to lowest by weight to facilitate weight allocation during spatial layout matching.

[0098] Obtain shelf space layout data; perform feasibility analysis of cargo stacking areas on the shelf space layout data based on high-priority cargo data to obtain cargo stacking feasibility data;

[0099] In this embodiment, data is derived from a shelf sensor array and a laser rangefinder. By scanning the available space coordinate range of each shelf layer, the X, Y, and Z coordinates of each layer are obtained, and the boundary position of the currently stacked goods is recorded. The spatial data for each shelf layer is stored in a JSON structure format. Fields include parameters such as "shelf number," "layer number," "starting coordinate," "ending coordinate," "remaining space volume," and "maximum load capacity," with precision in millimeters and kilograms. Based on the high-priority goods data, the acquired shelf space layout data is analyzed for the feasibility of the goods stacking area. The analysis process matches the goods size with the available shelf space, and determines whether each unit of goods meets the three constraints of the placement area: 1) size compatibility (goods size ≤ remaining space size); 2) weight constraint (goods weight ≤ maximum remaining load capacity for the layer); and 3) volume ratio (no more than 90% of the remaining space for the layer). The analysis results are represented as a Boolean matrix, with each item corresponding to the stacking feasibility of each shelf layer, with 1 indicating feasible and 0 indicating infeasible. This matrix serves as the goods stacking feasibility data input to the next step.

[0100] Assess cargo transport accessibility based on cargo stacking feasibility data;

[0101] In this embodiment, by reading the AGV path map and its corresponding dynamic obstacle avoidance path information, the shortest path distance between high-priority goods and feasible shelf positions, the number of turns required, whether there are obstacles above the path and other data are analyzed. Three parameters are defined: shortest path L (unit meter), average passage width W (unit meter), minimum passage height H (unit meter). The transport accessibility standard is set to: L≤10m, W≥1.2m, H≥1.8m. If a feasible shelf position meets the above conditions, it is marked as transport accessible. All shelf layer numbers that meet the conditions and their corresponding cargo numbers are composed of a transport accessibility mapping table.

[0102] Recommend shelf stacking areas based on cargo transportation accessibility;

[0103] In this embodiment, a priority-weighted strategy is employed to select shelves with high priority values ​​and shortest path distances (L) within the reachable area. To avoid duplicate recommendations, the same item is matched to a stacking area only once. If multiple shelves meet the requirements, they are prioritized according to the following order: priority value > space volume utilization > shortest path distance. The final recommendation results are output to the scheduling system's recommended stacking list, which is used to guide the operation of stacker cranes or AGVs.

[0104] Assess the degree of cargo center of gravity deviation based on the shelf stacking area.

[0105] In this embodiment, the actual coordinates of the stacked goods are collected for each recommended shelf layer. Combined with the weight of each item and the center coordinates (X_c, Y_c, Z_c), the three-dimensional center of gravity G of the entire stacking layer is calculated. This is then compared with the geometric center C of the shelf layer to calculate the offset vector D = GC. A critical offset value Δ is set at 50 mm. If any dimensional component of D exceeds Δ, the offset is considered excessive. All calculated data is recorded in the stacking verification module as part of a safety check, used to control whether the task is delivered for execution.

[0106] Preferably, the evaluation of the degree of deviation of the cargo center of gravity includes:

[0107] Extract shelf stacking space coordinates based on shelf stacking area;

[0108] In this embodiment, a panoramic laser scan of the dynamic shelf stacking area is performed with a 5mm point spacing by an industrial-grade 3D laser scanning device (such as a FARO Focus S350), and the structural CAD drawings of each layer of shelves are combined to perform point cloud fitting and coordinate conversion operations. A fixed reference origin is located at the center of the shelf ground to construct a three-dimensional right-handed coordinate system, with the X-axis parallel to the longitudinal direction of the shelf, the Y-axis horizontally, and the Z-axis in the height direction. The center point of the unit cell where the goods are placed is identified by the coordinate fitting method, and the three-dimensional coordinate data of each cell is obtained. The record format of each set of shelf space coordinates is (x, y, z), and the units are all millimeters. The coordinate accuracy is controlled within the range of ±2mm, and the sampling results are exported as a structured table through laser point cloud software (such as Cyclone REGISTER 360).

[0109] Calculate cargo height based on shelf stacking space coordinates;

[0110] In this embodiment, the top coordinate z_top and bottom coordinate z_bottom in the obtained cargo stacking coordinate data are subtracted to obtain the actual height of the cargo H = z_top - z_bottom, in millimeters. The bottom coordinate is calculated by adding the standard height of the shelf pallet to the shelf shelf coordinate. For example, if the pallet thickness is 145mm and the shelf is located at z = 600mm, the bottom surface is z_bottom = 745mm. The top coordinate is measured by the three-dimensional image contour recognition system to measure the top surface of the object. The contour height is inferred from the z-axis height at the highest point of the contour pixel, and the error correction value is set to ±2mm. All data is uniformly output in millimeters after coordinate consistency standardization and recorded in the database field height_mm.

[0111] Determine the attributes of goods based on the shelf stacking space coordinates;

[0112] In this embodiment, the corresponding cargo attributes are matched in the warehouse database based on the binding relationship between the cargo coordinates and cargo numbers. The database contains the coordinate index of each cargo in the three-dimensional coordinate system and its attribute fields, including material type (such as metal, plastic, cardboard, etc.), density (kg / m 3 ), packaging shape (cube, cylinder, etc.), volume (mm 3 ) etc. Attribute records are retrieved using coordinates as keywords, with multi-field cross-validation ensuring uniqueness. If multiple records with the same coordinates exist, the one with the latest entry time is considered valid. The timestamp field must be accurate to the second, using the UTC+8 standard. All attribute information is aggregated into a structured array for subsequent weight calculations.

[0113] Calculate cargo weight based on cargo attributes and cargo height;

[0114] In this embodiment, when the density of the cargo material ρ (unit: kg / m 3 ) and the overall dimensions, calculate the cargo weight. The cargo volume V is obtained by multiplying the length, width, and height. The length and width are measured by the spatial distribution in the X and Y directions of the shelf coordinate system, and the height is taken from the previous step. The unit is uniformly meter. The mass calculation formula is W = ρ × V. The density ρ is 7850 kg / m for metals. 3 , plastic is 950kg / m 3 , the specific value is set according to the database attribute field. Actual calculation example: If the cargo is metal, 500mm long, 300mm wide, and 700mm high, then V = 0.5 × 0.3 × 0.7 = 0.105m 3 , W = 7850 × 0.105 = 824.25 kg. Record the result in the weight field weight_kg, and set the precision to two decimal places.

[0115] Determine the center of gravity of the cargo based on the cargo weight and the coordinates of the shelf stacking space;

[0116] In this embodiment, the center of gravity coordinates of the cargo are determined based on a regular shape model. If the cargo is a cube, the center of gravity is located at the geometric center point, and its spatial coordinates are (x_cg, y_cg, z_cg) = (x + L / 2, y + W / 2, z + H / 2), where L, W, and H are the length, width, and height of the cargo, and x, y, and z are the coordinates of the lower left front corner of the bottom surface. If the cargo has an asymmetric or irregular structure, the center of gravity position is calculated using the 3D modeling data and the mass distribution function using an integral method. The integral method uses a density-weighted geometric center algorithm, and the "Center of Mass" function in SolidWorks Simulation is used to obtain the precise center of gravity coordinates for each type of object. After the calculation is completed, the center of gravity data of each cargo is stored in the 3D center of gravity field cg_position.

[0117] Identify the center axis of the shelf according to the shelf stacking area;

[0118] In this embodiment, the identification of the shelf center axis is based on a symmetry analysis of the shelf structure. Based on the shelf unit length L_total, width W_total, and height H_total, the center axis is defined as the Z axis through the shelf's geometric center, with coordinates (x_center, y_center), where x_center = X_min + L_total / 2, y_center = Y_min + W_total / 2, and the Z axis direction remains unchanged. This coordinate is calculated by combining the extreme positions X_min and Y_min marked on the structural drawings and the total width data measured by laser ranging. The center axis data for all shelves is uniformly written into the axis reference field center_axis.

[0119] The center of gravity offset of the cargo is calculated based on the center of gravity of the cargo and the center axis of the shelf to obtain the degree of center of gravity offset of the cargo.

[0120] In this embodiment, the Euclidean distance between the center of gravity coordinates of the cargo (x_cg, y_cg) and the center axis coordinates of the shelf (x_center, y_center) is calculated, and the offset distance D = √[(x_cg-x_center)2+(y_cg-y_center)2], in millimeters. To quantify the degree of offset, an offset ratio K = D / R_max is introduced, where R_max is the maximum offset radius allowed by the shelf, generally set to half the shelf width minus the safety margin. For example, if W_total = 1200mm and the safety margin = 100mm, then R_max = 500mm. If K>0.8, it is marked as a serious center of gravity offset. All center of gravity offset data is written into the offset degree field cg_offset_ratio and used as an input parameter for subsequent dynamic adjustment.

[0121] Preferably, the cargo handling simulation module includes the following functions:

[0122] Identify the direction of the center of gravity offset according to the degree of center of gravity offset of the cargo, and build a handling posture constraint model based on the center of gravity offset direction;

[0123] In this embodiment, a high-precision two-dimensional pressure sensor array is deployed at the bottom of the shelf to perform static testing on the current cargo and obtain a force distribution diagram of the cargo. The coordinates of the cargo's center of gravity are calculated based on the cargo's external dimensional data. The coordinates of the shelf's central axis are read from the shelf's structural design drawing and serve as a fixed reference point on the shelf's geometric structure. The center of gravity offset is determined by comparing the coordinate difference between the cargo's center of gravity and the shelf's central axis. Specifically, the center of gravity offset recognition threshold is set to 30 mm. If the absolute value of the lateral offset is greater than 30 mm and greater than the absolute value of the longitudinal offset, it is determined to be a left-right offset, with the offset direction being left or right. If the absolute value of the longitudinal offset is greater than 30 mm and greater than the absolute value of the lateral offset, it is determined to be a front-to-back offset, with the offset direction being front or back. Direction classification uses an enumeration system, including four types: front, back, left, and right. The offset direction is recorded as the "direction vector" field and used as an index condition for subsequent handling posture restrictions. After determining the direction vector, the posture restriction parameter retrieval phase begins. The identified offset direction is searched against the attitude parameter table of the end effector of the transport arm to retrieve the corresponding pitch, roll, and yaw angle setting ranges. This parameter table is a pre-set static table derived from the control instruction configuration document for the industrial transport arm and covers motion limits and recommended operating angles for different directions. For example, when the offset direction is forward, the corresponding transport attitude is set to a pitch angle range of -15 to -5 degrees, a roll angle of 0 degrees, and a yaw angle of -10 to +10 degrees. When the offset direction is left, the pitch angle is set to -10 to +10 degrees, the roll angle to -10 to -5 degrees, and the yaw angle to 5 to 15 degrees. All angle values ​​are stored in degrees and in floating-point format. Each direction and angle constraint combination is recorded in a structure with fields such as the direction type, the name of each angle, the upper and lower limits, the angle unit, and the application module number. The recorded results are stored in the "Transport Attitude Constraint Model" table to provide a basis for subsequent dynamic adjustments and path control. The entire process does not involve any modeling process. Parameter acquisition relies on sensor hardware measurement, structure diagram extraction and standard database calls to ensure that the implementation process is stable and reproducible.

[0124] Cargo dynamic adjustment grasping simulation is performed based on the handling posture constraint model to obtain cargo dynamic adjustment grasping data;

[0125] In this embodiment, the shelf environment model and the three-dimensional CAD model of the goods are imported into the industrial simulation platform (such as ABB RobotStudio), the robot arm parameters are set to a six-axis industrial arm (such as IRB 6700), and the posture constraint model generated in the previous step is input. The grasping action set in the simulation system includes four stages: "approach", "positioning", "grasping", and "lifting". The dynamic adjustment process control parameters include: terminal speed 0.2m / s, acceleration 1.0m / s 2, and a gripper pressure of 250N. Simulation path interpolation is performed, inserting three nodes within the control arm's attitude angle range for trajectory smoothing. The system records data such as the end effector position coordinates (in mm), attitude angle (in degrees), and gripper opening and closing status for each action node, forming a "grasping data set" that is output as the dynamic adjustment result.

[0126] Adjust the grabbing data according to the dynamics of the cargo to identify the cargo posture angle;

[0127] In this embodiment, the end-effector's six-degree-of-freedom motion data recorded in each frame is extracted from the collected dynamic grasping dataset, including displacement and posture information. The posture information, collected by an industrial-grade three-dimensional inertial measurement unit (IMU), includes triaxial data from the angular velocity and accelerometer and is synchronized to the control system via a fieldbus interface. In the data processing module, the frame corresponding to the moment the end-effector contacts the cargo is selected as the analysis reference frame. The timing of this frame is determined by the trigger signal fed back by the contact sensor. The contact sensor is a highly sensitive capacitive touch device with a trigger delay of less than 1 millisecond, ensuring high-precision identification of the grasping moment. Within the analysis reference frame, the end-effector's spatial posture is calculated using Euler angles. The calculation method uses a preset ZYX axis sequence to resolve a clockwise rotation sequence. During the analysis process, the posture calculation module first reads the rotation matrix parameters for the frame and then converts them into three corresponding angle values, representing the current pitch, roll, and yaw angles. The posture angle data is uniformly stored in degrees, with accuracy retained to three decimal places, ensuring the stability and resolution of the angle information. The above three angles constitute the spatial posture information of the end effector at the moment of grasping, serving as the initial reference angles for the cargo's posture. Simultaneously, the 3D model data for the current cargo is retrieved from a standard cargo structure model library. The model is stored in CAD format and indexed by the item code. The principal axis direction vector in the model is extracted. This vector represents the principal reference axis from the cargo's center of gravity to its geometric centerline and is then compared with the horizontal unit vector in the global coordinate system using the direction cosines. Using the posture difference function, the directional deviation between the cargo's posture and the horizontal posture is calculated to identify any non-ideal placement conditions such as forward tilt, side tilt, or rotation. The resulting pitch, roll, and yaw angles are stored as 3D vectors in a "Cargo Posture Angle Table." This table structure contains information such as a unique cargo identification code, three fields for posture angle, an identification timestamp, and a reference actuator number. The recorded data will serve as an important foundation for subsequent energy cost analysis and path posture correction. The entire process relies strictly on sensor data and analytical algorithms, without the use of any machine learning models or estimation logic, ensuring process control, traceability, and reproducible operations.

[0128] It is particularly important to identify the cargo posture angles including:

[0129] Adjust the grabbing data according to the dynamics of the goods and extract the three-dimensional coordinates of the gripping point;

[0130] In this embodiment, binocular stereo vision equipment deployed within the dynamic racking system captures the entire process of product grasping, acquiring a sequence of image data during the grasping process. Combined with the end-effector pose data, point cloud data of the gripping area is extracted simultaneously at the moment of grasping. After voxel filtering, the point cloud data is then density-weighted to define a region of interest based on the known gripping area's positional boundaries. Contact points are extracted from this region's point cloud. A 3D coordinate conversion algorithm is then used to convert the image coordinates into 3D Cartesian coordinates based on the camera calibration matrix. The final output is the 3D coordinate data of the gripping points before and after the grasping process.

[0131] Calculate the cargo displacement vector based on the three-dimensional coordinates of the clamping point;

[0132] In this embodiment, an index of gripping point numbers is established. The coordinates of the gripping points with corresponding numbers before and after the gripping are compared. A three-dimensional coordinate difference calculation is performed to obtain the displacement of each gripping point. Outliers exceeding 0.2 mm are removed from the displacement data of all gripping points. A weighted average is then calculated from the remaining data after these removals to ultimately determine the overall cargo displacement vector. This vector describes the actual direction and magnitude of the cargo displacement in three-dimensional space caused by the gripping action, providing basic spatial variation data for gesture recognition.

[0133] Solve the rigid body transformation matrix based on the cargo displacement vector to obtain the cargo posture change matrix data;

[0134] In this example, the 3D coordinates of the gripping points and their displacement vectors are used to construct the registration input. A singular value decomposition algorithm is then employed to perform optimal rigid body alignment on the set of gripping points before and after grasping. This results in a four-dimensional homogeneous transformation matrix representing the object's spatial rotation and translation. The rotational component represents the change in posture in three dimensions, while the translational component represents the overall displacement. This matrix calculation eliminates any approximations and directly obtains precise rotational transformation data through numerical matrix decomposition.

[0135] Perform Euler angle decomposition calculation based on the cargo posture change matrix data to obtain the cargo three-axis rotation angle data;

[0136] In this embodiment, the direction cosine matrix representing rotation in the transformation matrix is ​​parsed and extracted. Combined with the sensor coordinate system configuration and the cargo placement benchmark definition, the rotation matrix is ​​gradually decomposed into the rotation angles around the X, Y, and Z axes, i.e., the three-axis Euler angles, in a ZYX order. All angles are output in degrees. The angle range during the calculation is limited to -180 to +180 degrees to ensure angle continuity, and any jump angles that occur are normalized to their boundaries.

[0137] The posture filtering and fusion processing is performed based on the three-axis rotation angle data of the cargo to obtain the cargo posture angle.

[0138] In this embodiment, in order to eliminate the influence of instantaneous jitter caused by noise and short-term mechanical vibration during the multi-frame acquisition process, the three-axis Euler angle data output above is processed with a five-frame sliding window, and the median filtering method is used to eliminate instantaneous abnormal angle values. At the same time, the filtered angle value is fused with the acceleration and angular velocity data collected by the IMU sensor deployed in the execution layer of the dynamic shelf system. The fusion method adopts the Mahony algorithm. Under the conditions of setting the scale factor to 0.5 and the integration factor to 0.01, the sensor's instantaneous angular velocity and the visually calculated Euler angle are fused and optimized to output stable three-axis posture angle data. This data is used as the key parameter for the dynamic shelf scheduling system to determine whether the cargo posture meets the standard stacking posture after adjustment.

[0139] Perform energy cost analysis on cargo posture angles based on component friction optimization data to obtain energy cost data;

[0140] In this embodiment, the cargo posture angle data is used as input and combined with the friction performance database preset in the end effector to perform an energy cost analysis. The friction database contains a table of material pairings for the contact surface between the gripper and the cargo (e.g., rubber-metal μ = 0.6, plastic-cardboard μ = 0.3, etc.). The friction coefficient μ corresponding to each posture is obtained by indexing the contact material type. The friction work Wf = μ·F·s is then calculated using the force F (unit: N) applied by the end effector and the cargo displacement s (unit: m), and the energy cost C is then calculated. W_f is used as the primary indicator, and the effect of the posture angle on the energy consumption of the gripper adjustment mechanism is weighted to account (e.g., a change in posture angle exceeding 15° increases adjustment energy consumption by 20%). The final energy consumption data is output in kJ, retained to two decimal places, and recorded in the energy cost table.

[0141] Obtain cargo handling channel data;

[0142] In this embodiment, the warehousing scene is scanned by the AGV vehicle-mounted laser SLAM system (such as the Velodyne VLP-16 laser radar) to reconstruct a two-dimensional map at a resolution of 0.01m. The map is stored in an occupancy grid format and read by the path management system. The channel width, slope, curvature radius, and obstacle position are all extracted from the map. Each channel information is composed of a structured array of path start and end coordinates, channel width (unit: mm), maximum allowable turning angle, load limit, and real-time passable status. This channel information is used as the basic input for cargo path planning and is recorded in the "handling channel structure table" for use by the path generation function.

[0143] Generate a preliminary cargo handling path based on the cargo gravity center deviation degree and cargo handling channel data;

[0144] In this embodiment, the cargo center of gravity offset data (expressed in the form of offset direction + offset ratio) and channel data are used as input to generate a preliminary transportation path. The heuristic function in the algorithm is set to the Euclidean distance from the current position to the target position, and the cost function includes three parts: basic movement cost, center of gravity offset penalty, and channel restriction factor. The center of gravity offset penalty is weighted according to the K value. For example, when K>0.6, the weight path corner penalty is increased to prevent the generation of a path with a sharp offset direction. Each node records the moving direction, expected distance, and curve angle. The generated path is a sequence of continuous coordinate points with a spacing of no more than 0.2m between each point. The result is output in JSON format and stored as a "preliminary transportation path table."

[0145] Intelligently optimize the preliminary cargo handling path based on energy consumption cost data to obtain the cargo handling path;

[0146] In this embodiment, based on the preliminary path, energy cost data is used as the objective function for path optimization using an improved ant colony optimization (IACO) algorithm. The energy cost function is set as the sum of the weighted energy consumption of each segment along the path. The path node sequence is used as the ant path code, the number of iterations is set to 200, the pheromone evaporation rate is 0.1, and the initial pheromone value τ = 0.5. The energy consumption value of each path segment is recalculated based on the attitude angle and channel parameters. The current path cost is recorded for each iteration, and the path with the lowest energy consumption is retained as the optimal result. The resulting path coordinate sequence is used as the final cargo handling path and stored in the "Final Handling Path Table."

[0147] Cargo handling simulation is performed based on the cargo handling path to generate cargo handling data;

[0148] In this embodiment, the optimal transport path is loaded with the AGV model and the robotic arm model in the industrial simulation system. The execution speed of each path point is set to 0.3m / s and the acceleration is 0.5m / s. 2 During the simulation, data such as the AGV's position coordinates, attitude angle, interaction time with the cargo, and gripper pressure changes are recorded at a resolution of 10Hz. The simulation results are output as a timestamped "handling trajectory data table" with fields such as time, AGV position, attitude, gripping status, and cargo stability indicators (such as whether there has been a sudden shift in the center of gravity). At least 500 records are generated, all of which are exported in CSV format for access by the real-time scheduling system.

[0149] The cargo redistribution signal is triggered based on the cargo handling data, and the cargo redistribution signal is transmitted to the shelf intelligent scheduling system to perform dynamic shelf intelligent scheduling tasks.

[0150] In this embodiment, after the cargo handling data is completed, the scheduling control module determines whether it is necessary to perform a reallocation operation based on the "degree of deviation of the cargo center of gravity" and the "occupancy status of the target shelf layer at the end of the path". If the current target layer is overloaded (according to the shelf load database, the maximum allowable load of a single layer does not exceed 500kg), a "redistribution request signal" is issued. The signal encoding format is instruction code + cargo ID + center of gravity information + current path coordinates + target spare shelf number, which is sent to the scheduling system main controller through the MQTT industrial communication protocol. After receiving it, the controller executes the reallocation command and drives the actuator to complete the relocation of the cargo to the spare position. When the entire reallocation process is completed, the operation code is marked and recorded in the log file.

[0151] Preferably, the signal triggering the redistribution of goods includes:

[0152] Cargo stacking simulation is performed based on cargo handling data, where the handling speed range is set to 0.1m / s-1.5m / s and the handling path length range is set to 0.5m-20m to obtain cargo stacking data;

[0153] In this embodiment, during the cargo stacking simulation, the key handling data of the industrial handling robot arm during operation is first collected through the high-precision inertial navigation system deployed in the shelf system. The handling data includes the current timestamp, actuator displacement vector, velocity vector, acceleration vector, and cargo coordinate trajectory. The range of the handling speed is obtained by converting the actuator drive current and is limited to between 0.1m / s and 1.5m / s. Data below 0.1m / s is treated as a stationary state, and data above 1.5m / s is treated as invalid handling data and eliminated. The handling path length is obtained by integrating the trajectory vector. The simulation path length range is set between 0.5 meters and 20 meters. Trajectory records exceeding this range are directly discarded to ensure that the stacking simulation is based on reasonable working conditions. Subsequently, a three-dimensional geometric modeling module is used to construct a position model of the cargo evolving over time. The modeling method uses the center of gravity of the cargo bottom surface to be projected onto the shelf surface to form a stacking position plane. Each position is marked with a unique stacking number, and the size boundary box, center coordinates, and attitude angle of the cargo are recorded. The above modeling data forms a "cargo stacking data set" and is stored in RAM in the form of a structure. The fields include: cargo number, three-dimensional coordinates of the final position, handling speed, path length, posture trigonometric function value, etc. The data is updated every 200 milliseconds.

[0154] Conduct cable mis-pressure detection based on cargo stacking data to obtain cable mis-pressure data;

[0155] In this embodiment, a static cable layout diagram is called based on the generated cargo stacking data. This diagram is pre-constructed using industrial 3D modeling software and converted into a grid coordinate point set in CAD format. The cable centerline coordinate sequence is constructed at 0.5 mm intervals and recorded in a "cable basic bitmap matrix" in millimeters. During the detection phase, the system reads the two-dimensional contour coordinates of the projected area of ​​the cargo bottom surface one by one and compares them with the cable centerline data using a point-by-point scanning method. If the Euclidean distance between the coordinates of any corner point or boundary point on the cargo bottom and any point on the cable centerline is less than the set mispressure identification threshold of 10 mm, the cargo is recorded as "mispressure cargo." This mispressure identification process is completed through Boolean matrix calculation. A Boolean value of 1 indicates the presence of a pressure line. The position coordinates and cargo number are simultaneously written into the "cable mispressure data table." The fields include: mispressure location coordinates, cable number, cargo number, number of contact points, contact area boundary, mispressure timestamp, etc.

[0156] It is particularly important that cable misvoltage detection includes:

[0157] Extract the cargo bottom contact area based on cargo stacking data to obtain cargo bottom area data;

[0158] In this embodiment, during the execution of the dynamic shelf-level intelligent scheduling task, it is necessary to obtain the cargo stacking data, which includes the placement coordinates of each cargo on the shelf, the three-dimensional size data of the cargo (length L, width W, height H), the stacking attitude angle information (pitch angle, roll angle, yaw angle), and the timestamp of the placement time. Using the three-dimensional geometric calculation method, the bottom surface of the three-dimensional envelope of each cargo is affine transformed, and the projection surface coordinates of the actual contact surface are calculated in combination with its attitude angle. The axis-aligned bounding box (AABB) algorithm is used to perform boundary clipping on the bottom panel of the cargo, extract the coordinates of the four vertices of the bottom surface, and use these vertices as input to form a closed polygonal area, thereby forming the cargo bottom contact area data. In order to improve processing efficiency, all boundary coordinates are recorded and stored in double-precision floating-point format with millimeter-level accuracy.

[0159] Generate grid determination points within the area based on cargo bottom area data;

[0160] In this embodiment, based on the cargo bottom contact area data obtained above, a regular grid division is performed within each contact area. The division rule adopts a two-dimensional Cartesian coordinate grid method, with the side length of each grid unit set to 10 mm. The grid judgment point is the two-dimensional coordinate position of each grid center point. The division process scans the contact area boundary bounding box and samples at equal intervals on the horizontal axis (x direction) and the vertical axis (y direction) in sequence to form a uniform judgment point set. To exclude grid points outside the polygon, ray casting is used for point in-point judgment, retaining only the grid point set with the judgment result of "inside", ultimately forming a grid judgment point data set within the effective contact area of ​​the cargo bottom surface. All points are stored in a structured array in the form of two-dimensional coordinates (x, y), with the coordinate unit being millimeters.

[0161] Obtain cable layout path coordinate data;

[0162] In this embodiment, in the dynamic shelf scheduling scenario, the cable laying path is a preset fixed layout, and the data comes from the static scene modeling process. The cable laying path coordinate data is a set of continuous two-dimensional coordinate point sequences, each coordinate point represents the actual passing position of the cable on the ground plane, and the spacing is controlled within 5mm to ensure the continuity and accuracy of the path. The cable path is expressed in the form of a set of line segments, where each line segment is represented by the starting point coordinates (x1, y1) and the end point coordinates (x2, y2), and are arranged in sequence according to the path direction. The coordinate point format is CSV structured data and is recorded in the cable path information table. The cable diameter is set to 20mm, which serves as a reference parameter for the effective coverage radius of the cable in subsequent calculations.

[0163] The spatial coincidence is calculated based on the grid decision points and the coordinates of the cable laying path in the area to obtain the spatial coincidence;

[0164] In this embodiment, the grid judgment point data set and the cable layout path coordinate set are used to determine the spatial overlap through the Euclidean distance calculation method. The specific method is: for each grid judgment point, calculate the distance between it and all cable path points in turn, select the minimum distance Dmin, and if Dmin is less than or equal to the cable radius plus a 10mm buffer distance (i.e., Dmin≤30mm), then the point is determined to be an overlap point. Finally, the ratio of the number of all overlap points to the total number of judgment points is counted as the spatial overlap between the bottom surface area of ​​the cargo and the cable path. The ratio is expressed in decimal form, retaining three decimal places, and recorded in the cable pressure risk table.

[0165] Cable mis-pressure detection is performed based on spatial overlap to obtain cable mis-pressure data.

[0166] In this embodiment, the spatial overlap data is compared with the set overlap threshold. The threshold is set to 0.15, which means that if the overlap point ratio is greater than 15%, the cargo is judged to have a cable mis-pressure risk. All cargo numbers that exceed the threshold and their corresponding bottom area coordinates, the number of cable overlap points, and the cable path segment information are recorded as cable mis-pressure events. All mis-pressure events are organized in the form of structured records, and each record contains the following fields: cargo ID, cable segment ID, mis-pressure area center coordinates, overlap value, number of overlap points, and timestamp. Finally, all cable mis-pressure event data are exported as cable mis-pressure data for subsequent collision analysis or cargo redistribution operations.

[0167] Identify cable mis-pressure areas based on cable mis-pressure data;

[0168] In this embodiment, the operation of identifying the mispressure area is based on the contact point and contact area boundary information in the "cable mispressure data table". The system uses the minimum envelope rectangle algorithm to perform geometric processing on the mispressure point group to form a minimum closed area, which is the mispressure area. The specific calculation method adopts the minimum circumscribed matrix method, which scans the boundaries of all mispressure points and generates a rectangle according to the maximum axial projection range. The mispressure area is represented by a matrix coordinate boundary quadruple, including the starting row, ending row, starting column, and ending column (in millimeters). At the same time, in order to prevent error interference, the system sets the minimum boundary size threshold of the mispressure area to 10 mm × 10 mm. If the set of mispressure points is not enough to form the area, it is not identified as a mispressure area. The recognition operation cycle is once per second, and the recognition results are written into the "cable mispressure area table". The fields include area number, boundary coordinates, cable number, associated goods number, mispressure intensity value (calculated by the number of contact points), etc.

[0169] Cargo collision analysis is performed based on the cable mis-pressing area, with the minimum collision distance threshold set to 15mm-50mm and the collision speed difference range set to 0.01m / s-0.5m / s to obtain cargo collision data.

[0170] In this embodiment, during the collision analysis phase, the system retrieves the cargo trajectory data within 500 mm of the mispressed area and analyzes the minimum relative spacing and relative speed of any two cargo items near this area. Spacing is calculated using the three-dimensional envelope box shortest side method, with the minimum collision spacing threshold set between 15 mm and 50 mm. If the distance between any two cargo items is less than the lower threshold of 15 mm, it is marked as a physical collision event. Relative velocity difference is calculated by calculating the Euclidean modulus length of the instantaneous velocity vector difference between the two cargo items. The system sets the collision velocity difference recognition range to 0.01 m / s to 0.5 m / s. Only events with a spacing less than the threshold within this range are recorded as valid collision events. All collision events are recorded in the "Cargo Collision Data Table," which includes fields such as: colliding cargo numbers A and B, the three-dimensional coordinates of the collision location, relative velocity, minimum spacing, and collision time.

[0171] Generate cargo redistribution signals based on cargo collision data, and transmit the cargo redistribution signals to the shelf intelligent scheduling system to perform dynamic shelf intelligent scheduling tasks.

[0172] In this embodiment, based on the cargo collision data table, the system extracts the IDs of all collided cargoes and retrieves the corresponding cargo handling records, task priorities, and shelf numbers. The system weights the cargo based on its current shelf level, number of collisions, and collision severity score (calculated by superimposing speed difference and spacing scores), prioritizing the repositioning of cargo with high collision scores. Each group of cargo generates a "cargo reallocation signal packet" consisting of the following fields: source shelf level number, target shelf level number, cargo unique identifier, recommended handling path vector, timestamp, executing robot ID, and recommended handling posture constraint values. The target shelf level number is dynamically assigned by the system based on the free shelf table to ensure that cargo positions are not duplicated after reallocation. The reallocation signal packet is transmitted to the main control module of the intelligent shelf scheduling system via the CAN bus. The transmission rate is fixed at 1 Mbps, the frame length is 64 bytes, and the communication protocol complies with the CANopen standard for industrial scheduling systems, with an 8-bit cyclic redundancy check bit. The receiving module parses the signal packet in real time, immediately issues the handling task, and initiates the rescheduling process.

[0173] The present invention is therefore intended to be illustrative and non-restrictive in all respects, with the scope of the invention being defined by the appended claims rather than the foregoing description, and all changes that come within the meaning and range of equivalents of the application documents are intended to be embraced therein.

[0174] The foregoing description is intended only to provide specific embodiments of the present invention, which will enable those skilled in the art to understand and implement the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not intended to be limited to the embodiments shown herein, but is to be construed in the widest possible manner consistent with the principles and novel features disclosed herein.

Claims

1. A dynamic shelf level intelligent scheduling system, characterized by: Includes the following modules: The cargo shelf allocation module is used to obtain storage structure data, perform dynamic shelf storage space detection, obtain dynamic shelf storage space data, and obtain cargo information; Cargo shelf allocation simulation is performed based on cargo information and dynamic shelf storage space data to obtain cargo shelf allocation data; The mechanical component friction optimization module is used to detect electric push rod stroke anomaly based on the cargo shelf allocation data and obtain electric push rod stroke anomaly data; Based on the abnormal data of the electric push rod stroke, the motor stall detection is driven to obtain the drive motor stall data; Optimize the friction of mechanical components based on the stall data of the drive motor to obtain component friction optimization data; The cargo center of gravity deviation assessment module is used to count cargo access frequency based on cargo information; assess cargo priority based on cargo access frequency; recommend shelf stacking areas based on cargo priority; and assess the degree of cargo center of gravity deviation based on shelf stacking areas. The cargo handling simulation module is used to allocate intelligent cargo handling paths based on the degree of cargo center of gravity offset and component friction optimization data to obtain the cargo handling path; perform cargo handling simulation based on the cargo handling path to generate cargo handling data; trigger cargo redistribution signals based on the cargo handling data, and transmit the cargo redistribution signals to the shelf intelligent scheduling system to execute dynamic shelf intelligent scheduling tasks.

2. The dynamic shelf level intelligent scheduling system according to claim 1 is characterized in that: The cargo shelf allocation module includes the following functions: Obtain storage structure data and extract structural dimension data; Construct a three-dimensional model of the shelf based on the structural dimension data; Obtain the volume data of stored goods and convert it into the shape point cloud type to obtain the shape point cloud data; Mapping the shape point cloud data to the shelf 3D model and calculating the dynamic shelf storage space to obtain the dynamic shelf storage space data; Obtain cargo information; Perform adaptive shelf height analysis based on cargo information to obtain adaptive shelf height data; Cargo shelf allocation simulation is performed based on the adapted shelf height data and the dynamic shelf storage space data to obtain cargo shelf allocation data.

3. The dynamic shelf level intelligent scheduling system according to claim 2 is characterized in that: The cargo shelf allocation simulation includes: Import the adaptive shelf height data and dynamic shelf storage space data into the cargo shelf allocation simulation module; In the cargo shelf allocation simulation module, the cargo mass range is set to 0.5kg-50kg; In the cargo rack allocation simulation module, the cargo center of gravity stability coefficient range is set to 0.1-1.0, the cargo stacking coefficient range is set to 0-1, where non-stackable is 0, and the layer height tolerance range is set to ±10mm; Run the cargo shelf allocation simulation module, perform cargo simulation placement operations, update shelf storage status, and output as cargo shelf allocation data.

4. The dynamic shelf level intelligent scheduling system according to claim 1 is characterized in that: The electric push rod stroke abnormality detection in the mechanical element friction optimization module includes: Calculate the push rod target stroke according to the cargo shelf allocation data to obtain the push rod target stroke data; Identify abnormal push rod jitter and bounce based on push rod target stroke data to obtain abnormal push rod jitter and bounce data; Extract putter beating frequency based on putter beating and jitter abnormality data; Perform push rod extension simulation according to the push rod beating frequency to obtain push rod extension data; Calculate the guide rail groove structure clearance based on the push rod telescopic data; Calculate the assembly offset based on the structural clearance of the guide rail groove; The electric push rod stroke abnormality detection is performed according to the assembly offset to obtain the electric push rod stroke abnormality data.

5. The dynamic shelf level intelligent scheduling system according to claim 1 is characterized in that: The drive motor stall detection in the mechanical element friction optimization module includes: Gear meshing images are collected based on abnormal data of electric push rod stroke; Identify gear tooth edges based on gear meshing images; Perform bite trajectory analysis based on the gear tooth profile edge to obtain bite trajectory data; Calculate the occlusal offset based on the occlusal trajectory data; Evaluate gear wear based on mesh offset; Predict gear instantaneous sticking according to gear wear degree and obtain gear instantaneous sticking data; The drive motor stall risk is evaluated based on the gear instantaneous sticking data to obtain the drive motor stall data.

6. The dynamic shelf level intelligent scheduling system according to claim 1 is characterized in that: The mechanical element friction optimization in the mechanical element friction optimization module includes: Calculate the push rod transmission friction force based on the drive motor stall data; Adjust the assembly clearance based on the push rod transmission friction force to obtain the assembly clearance data; Performing nitriding treatment on the push rod surface based on the assembly clearance data to obtain the push rod surface nitriding data; Perform motor shaft deformation detection based on the drive motor stall data to obtain motor shaft deformation data; Optimize the bearing support spacing based on the motor shaft deformation data to obtain the bearing support optimized spacing data; The push rod surface nitriding data and the bearing support optimization spacing data are integrated to obtain the component friction optimization data.

7. The dynamic shelf level intelligent scheduling system according to claim 1 is characterized in that: The cargo center of gravity offset assessment module includes the following functions: Statistics of historical cargo in and out records based on cargo information; Calculate the frequency of cargo access based on the cargo historical inbound and outbound records; Assess cargo priority based on cargo access frequency; Extract high-priority cargo data based on cargo priority; Obtain shelf space layout data; perform feasibility analysis of cargo stacking areas on the shelf space layout data based on high-priority cargo data to obtain cargo stacking feasibility data; Assess cargo transport accessibility based on cargo stacking feasibility data; Recommend shelf stacking areas based on cargo transportation accessibility; Assess the degree of cargo center of gravity deviation based on the shelf stacking area.

8. The dynamic shelf level intelligent scheduling system according to claim 7 is characterized in that: The evaluation of the degree of deviation of the cargo center of gravity includes: Extract shelf stacking space coordinates based on shelf stacking area; Calculate cargo height based on shelf stacking space coordinates; Determine the attributes of goods based on the shelf stacking space coordinates; Calculate cargo weight based on cargo attributes and cargo height; Determine the center of gravity of the cargo based on the cargo weight and the coordinates of the shelf stacking space; Identify the center axis of the shelf according to the shelf stacking area; The center of gravity offset of the cargo is calculated based on the center of gravity of the cargo and the center axis of the shelf to obtain the degree of center of gravity offset of the cargo.

9. The dynamic shelf level intelligent scheduling system according to claim 1 is characterized in that: The cargo handling simulation module includes the following functions: Identify the direction of the center of gravity offset according to the degree of center of gravity offset of the cargo, and build a handling posture constraint model based on the center of gravity offset direction; Cargo dynamic adjustment grasping simulation is performed based on the handling posture constraint model to obtain cargo dynamic adjustment grasping data; Adjust the grabbing data according to the dynamics of the cargo to identify the cargo posture angle; Perform energy cost analysis on cargo posture angles based on component friction optimization data to obtain energy cost data; Obtain cargo handling channel data; Generate a preliminary cargo handling path based on the cargo gravity center deviation degree and cargo handling channel data; Intelligently optimize the preliminary cargo handling path based on energy consumption cost data to obtain the cargo handling path; Cargo handling simulation is performed based on the cargo handling path to generate cargo handling data; The cargo redistribution signal is triggered based on the cargo handling data, and the cargo redistribution signal is transmitted to the shelf intelligent scheduling system to perform dynamic shelf intelligent scheduling tasks.

10. The dynamic shelf level intelligent scheduling system according to claim 9 is characterized in that: The signal triggering the redistribution of goods includes: Cargo stacking simulation is performed based on cargo handling data, where the handling speed range is set to 0.1m / s-1.5m / s and the handling path length range is set to 0.5m-20m to obtain cargo stacking data; Conduct cable mis-pressure detection based on cargo stacking data to obtain cable mis-pressure data; Identify cable mis-pressure areas based on cable mis-pressure data; Cargo collision analysis is performed based on the cable mis-pressing area, with the minimum collision distance threshold set to 15mm-50mm and the collision speed difference range set to 0.01m / s-0.5m / s to obtain cargo collision data. Generate cargo redistribution signals based on cargo collision data, and transmit the cargo redistribution signals to the shelf intelligent scheduling system to perform dynamic shelf intelligent scheduling tasks.

Images