A remote real-time monitoring method and system for AGV forklifts based on edge computing

By performing multi-dimensional data spatiotemporal alignment and transmission strategy optimization within the local edge node of the AGV forklift, the problems of inaccurate trajectory determination and network latency in the remote monitoring of AGV forklifts in the existing technology are solved, and efficient real-time monitoring and anomaly identification are achieved.

CN123117982APending Publication Date: 2026-11-17JIANGXI LONGEN INTELLIGENT EQUIP CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611229717.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-13
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing remote monitoring solutions for AGV forklifts fail to effectively perform spatiotemporal registration processing of multi-dimensional sensor data, resulting in inaccurate trajectory determination, independent obstacle recognition and trajectory deviation detection, fixed and simplistic data transmission strategies that cannot meet real-time requirements, excessive cloud load, and frequent network latency and packet loss.

Method used

Within the local edge node of the AGV forklift, multi-dimensional state feature information is aggregated and spatiotemporally aligned to construct a real-time motion trajectory vector, obstacles are identified and abnormal warning signals are generated, transmission strategies are determined based on network quality and battery status, monitoring data packets are encapsulated and sent to the remote monitoring center.

Benefits of technology

It achieves high-precision trajectory deviation and path blockage determination, reduces cloud computing load, improves the real-time performance and stability of data transmission, reduces resource consumption, and ensures the accuracy of real-time identification and remote monitoring of AGV forklift operating status.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123117982A_ABST
    Figure CN123117982A_ABST
Patent Text Reader

Abstract

This invention relates to the field of intelligent warehousing technology, and provides a method and system for remote real-time monitoring of AGV forklifts based on edge computing. The method includes: performing spatiotemporal alignment processing on real-time operating status data to obtain multi-dimensional status feature information; constructing a real-time motion trajectory vector based on the multi-dimensional status feature information, and comparing and analyzing the real-time motion trajectory vector with a preset standard trajectory vector to obtain trajectory deviation information; identifying obstacles on the travel path to obtain path blockage information, generating an abnormal warning signal, and annotating the abnormal warning signal into the multi-dimensional status feature information to obtain annotated status data; determining a target transmission strategy based on real-time network quality data and battery status data, encapsulating the annotated status data to obtain a monitoring data packet; and sending the monitoring data packet to a remote monitoring center to obtain operating status display data. This invention can improve the efficiency of remote real-time monitoring of AGV forklifts.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of intelligent warehousing technology, and in particular to a method and system for remote real-time monitoring of AGV forklifts based on edge computing. Background Technology

[0002] Currently, remote monitoring of AGV forklifts in warehousing scenarios mostly adopts a centralized cloud-based data processing architecture. All raw data from various sensors and images must be uploaded to the cloud server for unified parsing and processing. The simultaneous transmission of massive amounts of heterogeneous data from multiple sources continuously consumes a large amount of wireless communication bandwidth, resulting in frequent network latency and packet loss. The centralized computing power in the cloud is overloaded, making it difficult to perform low-latency real-time analysis of information such as forklift posture, obstacles, and trajectory deviations. Remote monitoring and early warning have significant lags, failing to meet the real-time requirements for dynamic scheduling and safety management of forklifts in the warehouse.

[0003] Existing monitoring solutions lack unified spatiotemporal registration processing for multi-dimensional sensor data. Pose, speed, battery, and image data are analyzed independently without a correlation and fusion mechanism. Trajectory determination relies solely on single-point coordinate comparison, failing to incorporate heading angle and acceleration for trajectory smoothing correction. Obstacle recognition and trajectory deviation detection are independent of each other, making it difficult to comprehensively judge forklift operation anomalies. Furthermore, the data transmission strategy is fixed and singular, failing to dynamically adjust data packet encapsulation and transmission rules based on on-site network fluctuations and the forklift's remaining battery power, further exacerbating the waste of transmission resources and the distortion of monitoring data. Therefore, improving the accuracy of remote real-time monitoring of AGV forklifts has become an urgent problem to be solved. Summary of the Invention

[0004] This invention provides a remote real-time monitoring method and system for AGV forklifts based on edge computing, in order to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides a remote real-time monitoring method for AGV forklifts based on edge computing, comprising:

[0006] A. Summarize the spatial pose data, movement speed data, battery status data, and surrounding environment image data of the AGV forklift to obtain the real-time operating status data of the AGV forklift;

[0007] B. Within the local edge node of the AGV forklift, perform spatiotemporal alignment processing on the real-time operating status data to obtain multi-dimensional status feature information of the AGV forklift;

[0008] C. Within the local edge node, based on the pose data and velocity data in the multi-dimensional state feature information, a real-time motion trajectory vector is constructed, and the real-time motion trajectory vector is compared and analyzed with a preset standard trajectory vector to obtain the trajectory deviation information of the AGV forklift.

[0009] D. Based on the surrounding environment image data, identify obstacles on the travel path, obtain path blockage information, and generate an abnormal warning signal within the local edge node based on the trajectory deviation information and the path blockage information. Mark the abnormal warning signal into the multi-dimensional state feature information to obtain the marked state data of the AGV forklift.

[0010] E. Obtain the real-time network quality data of the local edge node, determine the target transmission strategy based on the real-time network quality data and the battery status data, and encapsulate the labeled status data according to the target transmission strategy to obtain the monitoring data packet of the AGV forklift;

[0011] F. Send the monitoring data packet to the remote monitoring center and parse the monitoring data packet to obtain the operating status display data of the AGV forklift.

[0012] Preferably, the real-time operating status data is spatiotemporally aligned within the local edge node of the AGV forklift to obtain multi-dimensional status feature information of the AGV forklift, including:

[0013] Acquire spatial pose data, motion speed data, battery status data, and surrounding environment image data of the AGV forklift;

[0014] Using the timestamp of the spatial pose data as the reference time axis, time interpolation synchronization processing is performed on the motion speed data, the battery status data and the surrounding environment image data to obtain the time-aligned state dataset of the AGV forklift.

[0015] Based on the spatial coordinate system parameters of the spatial pose data after centralized alignment in the time-aligned state dataset, coordinate projection transformation is performed on the pixel coordinates detected in the surrounding environment image data to obtain the environmental association data of the AGV forklift.

[0016] According to the coordinate axis direction in the spatial coordinate system parameters, the aligned motion speed data in the time-aligned state dataset is decomposed to obtain the speed component data of the AGV forklift.

[0017] The environmental correlation data, the speed component data, and the battery status data are correlated and fused to obtain the multi-dimensional status feature information of the AGV forklift.

[0018] Preferably, the step of constructing a real-time motion trajectory vector within the local edge node based on the pose data and velocity data in the multi-dimensional state feature information includes:

[0019] Real-time position coordinate sequence and real-time attitude angle data, as well as real-time velocity component values ​​and real-time acceleration component values ​​along different coordinate axes at different sampling times are extracted from the multi-dimensional state feature information.

[0020] Using the real-time position coordinates at the current sampling moment as the trajectory starting point, and calculating the displacement component increments of the AGV forklift along different coordinate axes within a preset sampling time interval, the predicted position coordinates at the next sampling moment are obtained.

[0021] Using the predicted position coordinates as the new trajectory starting point, the displacement component increment calculation operation is repeated to obtain the predicted position coordinate sequence of the AGV forklift.

[0022] Based on the real-time heading angle in the real-time attitude angle data, a heading angle offset correction is applied to the predicted position coordinates in the predicted position coordinate sequence to obtain the corrected predicted position coordinate sequence of the AGV forklift.

[0023] The real-time position coordinate sequence and the corrected predicted position coordinate sequence are weighted and fused to obtain the smoothed position coordinate sequence of the AGV forklift.

[0024] By sequentially connecting the smoothed position coordinate sequence according to the sampling time order, the real-time motion trajectory vector of the AGV forklift is constructed.

[0025] Preferably, the formula for calculating the displacement component increment is:

[0026] ;

[0027] in, Indicates the first The sampling time along the first sampling moment The displacement component increment in each coordinate axis direction, Indicates the first The sampling time along the first sampling moment Real-time velocity component values ​​in each coordinate axis direction. Indicates the first The sampling time along the first sampling moment Real-time acceleration component values ​​in each coordinate axis direction. This indicates the preset sampling time interval. Indicates the first The sampling time along the first sampling moment Preset heading offset compensation coefficients in each coordinate axis direction Indicates the first The real-time heading angle at the sampling time along the first sampling point The direction cosine components of each coordinate axis.

[0028] Preferably, the step of comparing and analyzing the real-time motion trajectory vector with a preset standard trajectory vector to obtain the trajectory deviation information of the AGV forklift includes:

[0029] The real-time trajectory position coordinates at the sampling time are extracted from the real-time motion trajectory vector, and the standard trajectory position coordinates corresponding to the sampling time are extracted from the preset standard trajectory vector.

[0030] The difference between the coordinates of the real-time trajectory position point and the coordinates of the standard trajectory position point is analyzed to obtain the instantaneous position deviation value at the sampling time;

[0031] Based on the real-time trajectory position coordinates at continuous sampling times, the real-time travel direction angle of the AGV forklift between adjacent sampling times is determined, and based on the corresponding standard trajectory position coordinates, the standard travel direction angle of the AGV forklift is determined.

[0032] A differential analysis is performed on the real-time travel direction angle and the standard travel direction angle to obtain the instantaneous direction deviation value at the sampling time;

[0033] The instantaneous position deviation value and instantaneous direction deviation value are arranged according to the sampling time sequence to obtain the trajectory deviation information of the AGV forklift.

[0034] Preferably, the step of identifying obstacles on the travel path and obtaining path obstruction information based on the surrounding environment image data includes:

[0035] The surrounding environment image data is subjected to grayscale processing and noise filtering to obtain the preprocessed environmental image data of the AGV forklift.

[0036] Based on the gray-level gradient change characteristics of pixels in the preprocessed environmental image data, the contour edge information in the preprocessed environmental image data is extracted to obtain the environmental contour edge image of the AGV forklift.

[0037] From the environmental contour edge image, the region that overlaps with the AGV forklift's travel path is selected to obtain the obstacle contour set of the AGV forklift;

[0038] Based on the geometric size and positional distribution characteristics of the obstacle contours in the obstacle contour set, the obstacle contours are marked as path-blocking obstacles;

[0039] Based on the distribution and extent of the obstacles blocking the path on the AGV forklift's travel path, the path blocking information of the AGV forklift is obtained.

[0040] Preferably, generating an anomaly warning signal within the local edge node based on the trajectory deviation information and the path congestion information includes:

[0041] Instantaneous position deviation value and instantaneous direction deviation value are extracted from the trajectory deviation information, and time differentiation processing is performed on the instantaneous position deviation value and the instantaneous direction deviation value respectively to obtain the position deviation change rate and direction deviation change rate of the AGV forklift;

[0042] When the sign of the change rate of position deviation or the change rate of direction deviation remains unchanged within consecutive sampling times, it is determined that the AGV forklift is in a continuous deviation state, and a path deviation warning signal for the AGV forklift is generated.

[0043] The distribution location and occupied area of ​​the path-blocking obstacles are obtained. Based on the relative distance between the distribution location and the center line of the AGV forklift travel path, and the relative relationship between the occupied area and the width of the AGV forklift travel path, the remaining passable width of the AGV forklift travel path is determined.

[0044] When the remaining passable width is less than the width of the AGV forklift body, it is determined that the AGV forklift's travel path is completely blocked, and a path blockage warning signal for the AGV forklift is generated.

[0045] The path deviation warning signal or the path blockage warning signal shall be used as the abnormal warning signal of the AGV forklift.

[0046] Preferably, the step of acquiring real-time network quality data of the local edge node and determining the target transmission strategy based on the real-time network quality data and the battery status data includes:

[0047] Real-time network quality detection results are extracted from the local edge node of the AGV forklift. The real-time network quality detection results include signal strength indication value, transmission delay measurement value and packet loss rate statistics.

[0048] Read the remaining power indicator value from the battery status data, and determine the power adequacy level of the AGV forklift based on the remaining power indicator value;

[0049] By integrating the signal strength indication value, transmission delay measurement value, and packet loss rate statistics, a comprehensive network quality score for the local edge node is obtained.

[0050] Based on the power adequacy level corresponding to the power weight coefficient, the network quality comprehensive score is corrected to obtain the transmission strategy selection evaluation value of the AGV forklift.

[0051] Based on the transmission strategy, an evaluation value is selected, and a matching target transmission strategy is chosen from the pre-built transmission strategy library.

[0052] Preferably, sending the monitoring data packet to the remote monitoring center and parsing the monitoring data packet to obtain the operating status display data of the AGV forklift includes:

[0053] The monitoring data packets are transmitted to the data receiving port of the remote monitoring center via a wireless communication link.

[0054] Within the remote monitoring center, the monitoring data packets are decapsulated according to the data encapsulation format corresponding to the target transmission strategy to restore the labeled status data.

[0055] Extract the multi-dimensional state feature information, the trajectory deviation information, the path blockage information, and the anomaly warning signal from the labeled state data;

[0056] The spatial pose data in the multi-dimensional state feature information is mapped to the coordinate grid of the virtual monitoring interface to obtain the real-time pose display elements of the AGV forklift.

[0057] The trajectory deviation information, the path blockage information, and the abnormal warning signal are superimposed on the real-time pose display element to obtain the operating status display data of the AGV forklift.

[0058] To address the aforementioned problems, the present invention also provides a remote real-time monitoring system for AGV forklifts based on edge computing, the system comprising:

[0059] The data acquisition module is used to summarize the spatial pose data, movement speed data, battery status data and surrounding environment image data of the AGV forklift to obtain the real-time operating status data of the AGV forklift.

[0060] The data alignment module is used to perform spatiotemporal alignment processing on the real-time operating status data within the local edge node of the AGV forklift to obtain multi-dimensional status feature information of the AGV forklift.

[0061] The deviation analysis module is used to construct a real-time motion trajectory vector within the local edge node based on the pose data and velocity data in the multi-dimensional state feature information, and compare and analyze the real-time motion trajectory vector with the preset standard trajectory vector to obtain the trajectory deviation information of the AGV forklift.

[0062] The status labeling module is used to identify obstacles on the travel path based on the surrounding environment image data, obtain path blocking information, and generate an abnormal warning signal in the local edge node based on the trajectory deviation information and the path blocking information. The abnormal warning signal is then labeled in the multi-dimensional status feature information to obtain the labeled status data of the AGV forklift.

[0063] The data encapsulation module is used to acquire the real-time network quality data of the local edge node, determine the target transmission strategy based on the real-time network quality data and the battery status data, and encapsulate the labeled status data according to the target transmission strategy to obtain the monitoring data packet of the AGV forklift.

[0064] The status display module is used to send the monitoring data packet to the remote monitoring center and parse the monitoring data packet to obtain the operating status display data of the AGV forklift.

[0065] Compared with the prior art, the present invention has the following beneficial effects:

[0066] 1. This invention completes the entire process of data preprocessing, trajectory calculation and obstacle recognition through local edge nodes. It performs spatiotemporal alignment, coordinate fusion and trajectory smoothing on multi-source heterogeneous operating data. It can simultaneously output high-precision trajectory deviation and path blockage judgment results, greatly reduce the data processing latency of cloud computing power intervention, ensure the efficiency of local real-time analysis of AGV forklift operating status and abnormal information, and realize the instant identification of vehicle operation risks.

[0067] 2. This invention uses a dedicated data transmission strategy that dynamically matches real-time network quality with the vehicle's remaining battery power. It adaptively encapsulates and transmits labeled multi-dimensional status data, allowing the remote monitoring center to quickly reconstruct complete vehicle operation information and generate visualized data. While stabilizing the transmission link, it reduces resource consumption caused by invalid data transmission, continuously improving the timeliness of remote monitoring data synchronization and overall operational stability. Attached Figure Description

[0068] Figure 1 This is a flowchart illustrating a remote real-time monitoring method for AGV forklifts based on edge computing, provided in an embodiment of the present invention.

[0069] Figure 2 A functional block diagram of an AGV forklift remote real-time monitoring system based on edge computing provided in an embodiment of the present invention;

[0070] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0071] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0072] This application provides a remote real-time monitoring method for AGV forklifts based on edge computing. The execution entity of this edge computing-based AGV forklift remote real-time monitoring method includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application embodiment: a server, a terminal, etc. In other words, the edge computing-based AGV forklift remote real-time monitoring method can be executed by software or hardware installed on a terminal device or a server device. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster. The server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0073] Reference Figure 1 The diagram shown is a flowchart illustrating a remote real-time monitoring method for AGV forklifts based on edge computing, according to an embodiment of the present invention. In this embodiment, the remote real-time monitoring method for AGV forklifts based on edge computing includes:

[0074] A. Summarize the spatial pose data, movement speed data, battery status data, and surrounding environment image data of the AGV forklift to obtain the real-time operating status data of the AGV forklift;

[0075] In this embodiment of the invention, the positioning acquisition hardware deployed in the AGV forklift body positioning module continuously collects real-time spatial pose data of the vehicle. The positioning acquisition hardware relies on the on-board positioning sensing element to capture the current site plane coordinates and body orientation information of the forklift. Each sensing capture generates a complete set of original spatial pose data and directly transmits it to the built-in temporary storage area of ​​the on-board local edge node for complete storage.

[0076] The speed acquisition hardware, which is fixedly mounted on the AGV forklift's travel drive mechanism, synchronously collects vehicle movement speed data. The speed acquisition hardware relies on the speed sensing element built into the drive wheel to capture the motion rate information generated by the wheel rotation in real time. Each time the sensing element completes a speed capture, it generates a set of raw motion speed data and pushes it synchronously to the vehicle's local edge node's built-in temporary storage area and the raw spatial pose data for centralized storage.

[0077] The battery acquisition hardware installed in the AGV forklift power supply unit continuously completes the battery status data acquisition operation. The battery acquisition hardware continuously captures three types of battery operating information through the built-in sensing element in the power supply circuit: the remaining power of the power battery, the real-time operating temperature of the power load, etc. The sensing element captures a complete set of battery status raw data in a single capture and transmits it in real time to the built-in temporary storage area of ​​the vehicle's local edge node for unified collection.

[0078] Image acquisition hardware mounted on the four directions of the AGV forklift continuously collects image data of the vehicle's surrounding environment. The image acquisition hardware relies on high-definition imaging elements to continuously capture real-time images of the entire forklift's travel path. Each complete image captured by the imaging element generates a frame of the original image of the surrounding environment and transmits it frame by frame to the on-board local edge node's built-in temporary storage area, where it is combined with the other three types of original data.

[0079] The onboard local edge node's built-in data collection and reading unit synchronously retrieves all four types of raw data—spatial pose, motion speed, battery status, and surrounding environment images—stored in the temporary storage area. The data collection and reading unit then integrates each of the four types of raw data into a single independent data file according to a unified data storage format, ensuring no data is missed. After integration, it generates unique real-time operating status data corresponding to the AGV forklift, which is the final output of this step and is stored in the local edge node for subsequent steps.

[0080] The beneficial effects are that by using dedicated multi-position on-board acquisition hardware to collect four independent types of operational information for the AGV forklift: spatial pose, motion speed, battery status, and surrounding environment images, all the collected raw data is then uniformly integrated by a dedicated aggregation unit built into the local edge node to generate complete and unified real-time operational status data. All data acquisition and aggregation operations are completed locally on the forklift's on-board hardware, eliminating the need to transmit scattered raw sensor data to the external cloud in advance. This ensures that all dimensions of the forklift's original operational information are completely preserved, avoiding the problem of incomplete information for subsequent vehicle operational status judgment caused by the absence of single-type data. At the same time, the unified aggregation of real-time operational status data has a fixed and complete data structure, which can directly adapt to the data reading requirements of the subsequent spatiotemporal alignment processing flow of the edge node, eliminating the data reading and parsing lag problem caused by inconsistent formats of multiple heterogeneous raw data.

[0081] B. Within the local edge node of the AGV forklift, perform spatiotemporal alignment processing on the real-time operating status data to obtain multi-dimensional status feature information of the AGV forklift;

[0082] In this embodiment of the invention, the real-time operating status data is spatiotemporally aligned within the local edge node of the AGV forklift to obtain multi-dimensional status feature information of the AGV forklift, including:

[0083] Acquire spatial pose data, motion speed data, battery status data, and surrounding environment image data of the AGV forklift;

[0084] Using the timestamp of the spatial pose data as the reference time axis, time interpolation synchronization processing is performed on the motion speed data, the battery status data and the surrounding environment image data to obtain the time-aligned state dataset of the AGV forklift.

[0085] Based on the spatial coordinate system parameters of the spatial pose data after centralized alignment in the time-aligned state dataset, coordinate projection transformation is performed on the pixel coordinates detected in the surrounding environment image data to obtain the environmental association data of the AGV forklift.

[0086] According to the coordinate axis direction in the spatial coordinate system parameters, the aligned motion speed data in the time-aligned state dataset is decomposed to obtain the speed component data of the AGV forklift.

[0087] The environmental correlation data, the speed component data, and the battery status data are correlated and fused to obtain the multi-dimensional status feature information of the AGV forklift.

[0088] The onboard local edge node has a built-in data reading unit that retrieves the real-time running status data generated in the previous steps. The data reading unit splits the internal storage blocks of the real-time running status data and extracts four independent data types from the blocks: spatial pose data, motion speed data, battery status data, and surrounding environment image data. These data are then cached in the high-speed storage area inside the edge node for subsequent synchronization processing.

[0089] The onboard local edge node's built-in time reference locking unit reads the timestamp information inherent in all spatial pose data, arranges all timestamps in the order of acquisition to generate a unified reference timeline, and maintains the same sampling interval as the native sampling interval of the spatial pose data, set to 50ms. The edge node's built-in time interpolation unit reads the timestamps inherent in the motion speed data group by group and matches them with each time node on the reference timeline. For motion speed data, battery status data, and surrounding environment image data that cannot accurately match the reference time nodes, a unified linear interpolation algorithm is used to complete time synchronization, generating synchronized speed data, synchronized battery data, and synchronized image data corresponding to the reference time. After all four types of data have completed time interpolation and synchronization processing, they are integrated into a single data file, generating a complete AGV forklift time-aligned status dataset with perfectly aligned timestamps across all dimensions, which is then stored.

[0090] The onboard local edge node's built-in coordinate system extraction unit reads all aligned spatial pose data from the time-aligned state dataset. It extracts the spatial pose data and caches the complete global spatial coordinate system parameters of the warehouse within the coordinate transformation unit. The coordinate transformation unit performs coordinate transformation based on a pinhole camera projection model: It pre-obtains the camera's intrinsic parameter matrix K (including focal length and principal point coordinates) and extrinsic parameter matrix [R|t] (including the camera's rotation matrix R and translation vector t relative to the vehicle coordinate system) from the onboard image acquisition hardware using the Zhang calibration method. The coordinate transformation unit reads frame-by-frame the surrounding environment image data that has been synchronously matched to the time alignment state dataset. It identifies the pixel coordinates of all obstacles in each frame, first converting the pixel coordinates to 3D coordinates in the camera coordinate system using the intrinsic parameter matrix, and then converting them to spatial coordinates in the warehouse global coordinate system using the vehicle-to-global coordinate system transformation matrix. After completing the pixel coordinate transformation of all images, the spatial coordinates are bound and integrated with the corresponding image information to obtain AGV forklift environment-related data in the same coordinate system as the pose data, which is stored separately.

[0091] The onboard local edge node's built-in speed decomposition unit reads the motion speed data after centralized alignment of the time-aligned state data, retrieves all coordinate axis division rules within the previously cached global spatial coordinate system parameters of the warehouse, and decomposes the speed components based on the global coordinate system of the warehouse site: first, it extracts the forward speed, lateral speed, and angular velocity in the vehicle coordinate system from the raw speed data collected by the encoder, combines the vehicle heading angle at the current sampling time, and projects the speed vector in the vehicle coordinate system to the X, Y, and Z axes of the global coordinate system through a coordinate rotation matrix to obtain the speed component data along the global coordinate axis direction, ensuring that the speed data and spatial pose data are under the same spatial reference, and uniformly classifies and stores the multiple sets of independent speed values ​​of the coordinate axes to generate the speed component data of the AGV forklift.

[0092] The onboard local edge node's built-in data fusion unit synchronously retrieves three types of data stored separately: environmental correlation data, speed component data, and battery status data. The fusion unit reads the unified matching reference time identifier within each group of data, and splices and integrates the environmental correlation information, speed component information, and battery operation information corresponding to the same reference time identifier into the same data entry. After all time entries are spliced ​​and integrated, a complete and unified multi-dimensional status feature information of the AGV forklift containing multi-dimensional operation information is generated as the final output product of this step for subsequent trajectory analysis processes.

[0093] The beneficial effects are that it relies on the vehicle's local edge nodes to complete the unified spatiotemporal registration of multi-source sensor data. First, a unified time reference is constructed using the timestamps of spatial pose data, and interpolation is used to complete the synchronization of the time dimensions of four types of data to generate a time-aligned state dataset. Then, the spatial coordinate system parameters are used to complete the projection transformation from image pixels to global site coordinates to obtain environmental correlation data. The velocity data is simultaneously split to obtain velocity component data. Finally, the three types of derived data are fused to generate multi-dimensional state feature information. The entire data registration, conversion and fusion operation is completed locally at the edge without uploading the original images and sensor data to the cloud. This eliminates the problems of trajectory deviation and obstacle recognition distortion caused by the time misalignment of the acquisition of multiple heterogeneous data and the lack of a unified spatial reference. The multi-dimensional state feature information obtained after fusion has a unified time and spatial reference, which can directly support the high-precision calculation of subsequent trajectory vector construction and path blockage recognition, and avoid the misjudgment of vehicle operation anomalies caused by spatiotemporal asynchrony.

[0094] C. Within the local edge node, based on the pose data and velocity data in the multi-dimensional state feature information, a real-time motion trajectory vector is constructed, and the real-time motion trajectory vector is compared and analyzed with a preset standard trajectory vector to obtain the trajectory deviation information of the AGV forklift.

[0095] In this embodiment of the invention, the step of constructing a real-time motion trajectory vector within the local edge node based on the pose data and velocity data in the multi-dimensional state feature information includes:

[0096] Real-time position coordinate sequence and real-time attitude angle data, as well as real-time velocity component values ​​and real-time acceleration component values ​​along different coordinate axes at different sampling times are extracted from the multi-dimensional state feature information.

[0097] Using the real-time position coordinates at the current sampling moment as the trajectory starting point, and calculating the displacement component increments of the AGV forklift along different coordinate axes within a preset sampling time interval, the predicted position coordinates at the next sampling moment are obtained.

[0098] Using the predicted position coordinates as the new trajectory starting point, the displacement component increment calculation operation is repeated to obtain the predicted position coordinate sequence of the AGV forklift.

[0099] Based on the real-time heading angle in the real-time attitude angle data, a heading angle offset correction is applied to the predicted position coordinates in the predicted position coordinate sequence to obtain the corrected predicted position coordinate sequence of the AGV forklift.

[0100] The real-time position coordinate sequence and the corrected predicted position coordinate sequence are weighted and fused to obtain the smoothed position coordinate sequence of the AGV forklift.

[0101] By sequentially connecting the smoothed position coordinate sequence according to the sampling time order, the real-time motion trajectory vector of the AGV forklift is constructed.

[0102] The formula for calculating the displacement component increment is:

[0103] ;

[0104] in, Indicates the first The sampling time along the first sampling moment The displacement component increment in each coordinate axis direction, Indicates the first The sampling time along the first sampling moment Real-time velocity component values ​​in each coordinate axis direction. Indicates the first The sampling time along the first sampling moment Real-time acceleration component values ​​in each coordinate axis direction. This indicates the preset sampling time interval. Indicates the first The sampling time along the first sampling moment Preset heading offset compensation coefficients in each coordinate axis direction Indicates the first The real-time heading angle at the sampling time along the first sampling point The direction cosine components of each coordinate axis.

[0105] The step of comparing and analyzing the real-time motion trajectory vector with a preset standard trajectory vector to obtain the trajectory deviation information of the AGV forklift includes:

[0106] The real-time trajectory position coordinates at the sampling time are extracted from the real-time motion trajectory vector, and the standard trajectory position coordinates corresponding to the sampling time are extracted from the preset standard trajectory vector.

[0107] The difference between the coordinates of the real-time trajectory position point and the coordinates of the standard trajectory position point is analyzed to obtain the instantaneous position deviation value at the sampling time;

[0108] Based on the real-time trajectory position coordinates at continuous sampling times, the real-time travel direction angle of the AGV forklift between adjacent sampling times is determined, and based on the corresponding standard trajectory position coordinates, the standard travel direction angle of the AGV forklift is determined.

[0109] A differential analysis is performed on the real-time travel direction angle and the standard travel direction angle to obtain the instantaneous direction deviation value at the sampling time;

[0110] The instantaneous position deviation value and instantaneous direction deviation value are arranged according to the sampling time sequence to obtain the trajectory deviation information of the AGV forklift.

[0111] The local edge node has a built-in feature extraction unit that reads the multi-dimensional state feature information output from the previous steps. The feature extraction unit divides the multi-dimensional state feature information into independent data storage partitions, and reads and extracts all real-time position coordinate sequences, all real-time attitude angle data, all real-time velocity component values ​​and all real-time acceleration component values ​​corresponding to each coordinate axis at different sampling times stored in the partition. The four types of motion-related data extracted are uniformly cached in the high-speed read and write storage area of ​​the edge node to await trajectory construction calculation.

[0112] The local edge node's built-in displacement estimation unit reads the real-time position coordinates corresponding to the current sampling time in the cache area and marks this set of coordinates as the trajectory start point. The displacement estimation unit retrieves the standard duration of the sampling time interval pre-written in the local fixed storage area, and sequentially retrieves the real-time velocity component value and real-time acceleration component value corresponding to each coordinate axis at the current sampling time. Combined with the weighted correction information related to the vehicle's travel direction, it completes the calculation of the travel distance of a single coordinate axis. After the calculation of each coordinate axis is completed, the complete displacement component increment within the sampling interval is obtained. The displacement component increment is superimposed on the coordinate value corresponding to the trajectory start point to generate the exclusive predicted position coordinates for the next sampling time.

[0113] The local edge node has a built-in loop calculation unit that updates the predicted position coordinates generated in the previous step to the new trajectory starting point. It cyclically retrieves the real-time velocity component value and real-time acceleration component value at the corresponding sampling time to complete a new round of displacement component increment calculation and generate the predicted position coordinates for the next time. The loop calculation unit continues to repeat the point calculation operation according to the order of sampling time until it covers all sampling times in the current data window. Then it summarizes all the calculated points to generate a complete sequence of predicted position coordinates for the AGV forklift.

[0114] Real-time velocity component values ​​are extracted from corresponding data at different sampling times along different coordinate axes contained in multi-dimensional state feature information; real-time acceleration component values ​​are extracted from corresponding data at different sampling times along different coordinate axes contained in multi-dimensional state feature information.

[0115] The preset sampling time interval is set by the system in advance and is the sampling time step uniformly adopted in the process of constructing the real-time motion trajectory vector. The preset heading offset compensation coefficient is used to adjust the influence of heading angle related calculation terms on the displacement calculation results. This coefficient is determined in advance by the system through on-site calibration experiments: multiple sets of actual displacement and calculated displacement data are collected on standard straight paths and turning paths respectively, and the optimal compensation coefficient values ​​under different coordinate axes are obtained by least squares fitting. In typical scenarios, the X-axis forward direction compensation coefficient is 0.15m / s, and the Y-axis lateral direction compensation coefficient is 0.3m / s.

[0116] The real-time heading angle is extracted from the real-time attitude angle data contained in the multi-dimensional state feature information; the direction cosine component is obtained by projecting the real-time heading angle along the corresponding coordinate axis.

[0117] This calculation yields the displacement component increment along a single coordinate axis at a single sampling moment. This displacement component increment, combined with the real-time position coordinates at the current sampling moment, is used to deduce the predicted position coordinates at the next sampling moment. Repeating the corresponding calculation with the predicted position coordinates as the new trajectory starting point generates a complete sequence of predicted position coordinates. This calculation incorporates factors related to velocity, acceleration, and heading angle, ensuring the accuracy of displacement prediction and providing fundamental data for subsequent heading angle offset correction and trajectory smoothing.

[0118] Assuming all other calculation conditions remain constant, an increase in the real-time velocity component value will result in a corresponding increase in the displacement component increment. Similarly, an increase in the real-time acceleration component value will result in a corresponding increase in the displacement component increment.

[0119] With other calculation conditions remaining constant, increasing the preset sampling time interval will increase the corresponding displacement component increment, and the impact of acceleration will increase at a faster rate with the increase of the time interval. With other calculation conditions remaining constant, changes in the direction cosine component corresponding to the real-time heading angle will correspondingly change the value of the displacement component increment, thereby achieving quantitative correction of the impact of heading deviation on displacement.

[0120] The local edge node has a built-in heading correction unit that reads all real-time attitude angle data from the cache area and extracts the real-time heading angle information of each set of points. The heading correction unit reads each set of predicted position coordinates in the predicted position coordinate sequence line by line, and determines the degree of deviation between the actual orientation of the forklift body and the site coordinate axis based on the real-time heading angle at the corresponding sampling time. The coordinate axis values ​​of a single set of predicted position coordinates are compensated and adjusted accordingly according to the degree of deviation. After all points have completed the heading deviation compensation and adjustment, they are integrated to generate the corrected predicted position coordinate sequence of the AGV forklift.

[0121] The local edge node's built-in coordinate fusion unit synchronously retrieves the complete and stored real-time position coordinate sequence and the corrected predicted position coordinate sequence. The coordinate fusion unit reads the two sets of coordinate points corresponding to the same sampling time in the two sets of coordinate sequences, and according to the local preset fixed weight allocation rule: the weight of the real-time position coordinate is 0.6, and the weight of the corrected predicted position coordinate is 0.4. The corresponding calculated weights are assigned to the real-time coordinate points and the corrected predicted coordinate points respectively. The coordinate axis values ​​of the two sets of coordinate points are respectively superimposed with the corresponding weights to generate a single set of smooth coordinate points. After the weighted integration of all sampling time points, they are collected in chronological order to generate the smoothed position coordinate sequence of the AGV forklift.

[0122] The local edge node has a built-in trajectory vector construction unit that reads the complete smoothed position coordinate sequence. The trajectory vector construction unit reads each set of smoothed coordinate points in the sequence in the order of sampling time from first to last. It establishes a continuous point connection relationship between the smoothed coordinate points corresponding to adjacent sampling times. After all points are connected in time sequence, a complete and continuous point association structure is generated as the real-time motion trajectory vector of the AGV forklift and stored in the edge node for deviation comparison calculation.

[0123] The local edge node's built-in point matching unit retrieves the stored real-time motion trajectory vector and extracts the coordinates of the real-time trajectory position points corresponding to each sampling time within the vector. The point matching unit reads the pre-stored standard trajectory vector from the local edge node. The preset standard trajectory vector is generated by the warehouse's pre-planned operation path: the planned path in the global map is discretized into a sequence of standard points at fixed time intervals ΔT. Each point contains position coordinates and a standard heading angle, corresponding one-to-one with the sampling interval of the real-time trajectory. The trajectory comparison uses the nearest neighbor time matching rule: the standard trajectory point with the smallest time difference from the real-time sampling time is taken as the comparison benchmark. If the time difference is greater than ΔT / 2, linear interpolation is performed on the standard trajectory point before deviation calculation. The real-time point coordinates matched at the same sampling time are paired and cached with the standard point coordinates for difference comparison.

[0124] The local edge node has a built-in position difference calculation unit that reads the coordinates of the real-time trajectory position points and the standard trajectory position points that were paired at the same sampling time. The position difference calculation unit performs a corresponding subtraction operation on the coordinate values ​​corresponding to each coordinate axis in the two sets of coordinates. After all the coordinate axis differences are calculated, they are integrated and summarized to obtain the instantaneous position deviation value corresponding to the sampling time and bound to the corresponding sampling time identifier for separate storage.

[0125] The local edge node has a built-in direction angle calculation unit that reads the coordinates of the real-time trajectory position points at two adjacent sampling times within the real-time motion trajectory vector. The direction angle calculation unit determines the actual vehicle tilt angle corresponding to the line connecting the two points based on the distribution of the two sets of points in the field coordinate system, which is used as the real-time travel direction angle. The direction angle calculation unit retrieves the coordinates of two sets of standard trajectory position points within the standard trajectory vector that completely match the sampling times of the above two sets of real-time points. Based on the distribution of the two sets of standard points, it determines the standard travel tilt angle corresponding to the line connecting the two points, which is used as the standard travel direction angle. The real-time travel direction angle and the standard travel direction angle obtained in the same time period are paired and stored.

[0126] The local edge node has a built-in angle difference calculation unit that reads the paired real-time travel direction angle and standard travel direction angle. The angle difference calculation unit performs a numerical subtraction operation on the two sets of angle values ​​to obtain the instantaneous direction deviation value corresponding to the sampling time and binds it to the corresponding sampling time identifier to complete the separate storage.

[0127] The local edge node has a built-in deviation collection unit that retrieves the instantaneous position deviation value and instantaneous direction deviation value stored at all sampling times. The deviation collection unit arranges and integrates the two sets of deviation values ​​into the same independent data file in a fixed order from the sampling time. After completely collecting the deviation data at all times, it generates the trajectory deviation information of the AGV forklift as the final output product of this process for subsequent abnormal warning steps.

[0128] The beneficial effects are that all trajectory construction and trajectory deviation comparison calculations are completed within the local edge node of the AGV. Relying on motion data with a unified spatiotemporal reference within multi-dimensional state feature information, the point-to-point calculation, heading correction, coordinate smoothing, and time-series connection are completed step by step to generate a real-time motion trajectory vector. Then, through time-by-time point-to-point pairing coordinate difference calculation and angle difference calculation, the position and direction deviation data are completely extracted to generate trajectory deviation information. The entire process does not require the transmission of massive amounts of raw motion sensor data to the external cloud. Relying on local hardware to complete the entire process of numerical calculation can eliminate the calculation delay caused by cloud transmission. At the same time, the trajectory vector after heading correction and weighted smoothing can avoid the point-to-point jump problem caused by single positioning acquisition. Simultaneously distinguishing between position deviation and direction deviation information can completely restore all the details of the forklift deviating from the standard path, avoiding the path deviation omission caused by relying solely on single-point coordinate comparison, and improving the accuracy of subsequent edge node anomaly warning judgment.

[0129] D. Based on the surrounding environment image data, identify obstacles on the travel path, obtain path blockage information, and generate an abnormal warning signal within the local edge node based on the trajectory deviation information and the path blockage information. Mark the abnormal warning signal into the multi-dimensional state feature information to obtain the marked state data of the AGV forklift.

[0130] In this embodiment of the invention, the step of identifying obstacles on the travel path and obtaining path obstruction information based on the surrounding environment image data includes:

[0131] The surrounding environment image data is subjected to grayscale processing and noise filtering to obtain the preprocessed environmental image data of the AGV forklift.

[0132] Based on the gray-level gradient change characteristics of pixels in the preprocessed environmental image data, the contour edge information in the preprocessed environmental image data is extracted to obtain the environmental contour edge image of the AGV forklift.

[0133] From the environmental contour edge image, the region that overlaps with the AGV forklift's travel path is selected to obtain the obstacle contour set of the AGV forklift;

[0134] Based on the geometric size and positional distribution characteristics of the obstacle contours in the obstacle contour set, the obstacle contours are marked as path-blocking obstacles;

[0135] Based on the distribution and extent of the obstacles blocking the path on the AGV forklift's travel path, the path blocking information of the AGV forklift is obtained.

[0136] The step of generating an anomaly warning signal within the local edge node based on the trajectory deviation information and the path congestion information includes:

[0137] Instantaneous position deviation value and instantaneous direction deviation value are extracted from the trajectory deviation information, and time differentiation processing is performed on the instantaneous position deviation value and the instantaneous direction deviation value respectively to obtain the position deviation change rate and direction deviation change rate of the AGV forklift;

[0138] When the sign of the change rate of position deviation or the change rate of direction deviation remains unchanged within consecutive sampling times, it is determined that the AGV forklift is in a continuous deviation state, and a path deviation warning signal for the AGV forklift is generated.

[0139] The distribution location and occupied area of ​​the path-blocking obstacles are obtained. Based on the relative distance between the distribution location and the center line of the AGV forklift travel path, and the relative relationship between the occupied area and the width of the AGV forklift travel path, the remaining passable width of the AGV forklift travel path is determined.

[0140] When the remaining passable width is less than the width of the AGV forklift body, it is determined that the AGV forklift's travel path is completely blocked, and a path blockage warning signal for the AGV forklift is generated.

[0141] The path deviation warning signal or the path blockage warning signal shall be used as the abnormal warning signal of the AGV forklift.

[0142] The local edge node's built-in image preprocessing unit retrieves the surrounding environment image data output from the previous steps. The preprocessing unit traverses all color pixels in the image, extracts the built-in three-color brightness values ​​of a single pixel, and integrates them into a single grayscale brightness value according to a fixed conversion rule to replace the original three-color pixel information. After all pixels in the entire frame have completed the grayscale value replacement, a grayscale image is generated. The preprocessing unit uses a fixed-size pixel window to sequentially cover each pixel in the grayscale image, extracts all pixel grayscale values ​​enclosed by the window and sorts them in ascending order, selects the grayscale value in the middle of the sorted sequence to replace the original pixel grayscale value in the center of the window, and after the entire grayscale image has completed full-pixel window traversal and noise reduction, isolated black and white noise points generated during the shooting and transmission process are eliminated. After the grayscale processing and noise filtering processes are fully executed, the preprocessed environmental image data of the AGV forklift is generated and completely cached in the local storage area of ​​the edge node.

[0143] The local edge node has a built-in gradient extraction unit that traverses all pixels in the preprocessed environmental image data row by row and column by column. The gradient extraction unit calculates the grayscale difference between the current pixel and its horizontally adjacent pixel, and the grayscale difference between the current pixel and its vertically adjacent pixel. The two grayscale differences are merged and summarized to obtain the grayscale gradient change amplitude corresponding to the current pixel. The gradient extraction unit marks all pixels whose gradient change amplitude exceeds the local preset fixed judgment standard as edge-specific pixels. The gradient extraction unit splices the interconnected edge-specific pixels in the image according to the spatial distribution relationship to form continuous closed lines. All closed lines are completely stored in the same image carrier to generate the environmental contour edge image of the AGV forklift for contour filtering operation.

[0144] The local edge node's built-in path matching unit retrieves the pre-stored global coordinate boundary information of the AGV forklift's travel path. The travel path width is preset to 1.5 times the vehicle body width. The path matching unit reads all pixel coordinates corresponding to each complete closed contour line within the environmental contour edge image, and uses the intersection-union ratio (IOU) rule to determine overlap: it calculates the ratio of the intersection area to the union area of ​​the obstacle contour's minimum bounding rectangle and the path region. When the IOU is greater than 0.1, it is determined that the obstacle contour overlaps with the travel path. The path matching unit separately collects and stores all closed contour lines that meet the overlap determination conditions, and then summarizes and stores all the collected closed contours to generate the AGV forklift's obstacle contour set.

[0145] The local edge node's built-in contour discrimination unit reads each group of closed obstacle contours within the obstacle contour set line by line. The unit counts the total number of pixels covered horizontally and vertically by a single contour line, and calculates the overall geometric dimensions of the contour by combining this with the pixel physical dimensions. The unit extracts the pixel coordinates of the center points of each contour line and converts them to global spatial coordinates to obtain the contour's positional distribution features. Obstacle contours that simultaneously meet the following two conditions are marked as path-blocking obstacles: first, the geometric dimension condition—the physical dimension corresponding to the shortest side of the contour's smallest bounding rectangle is not less than 10cm; second, the positional condition—the lateral distance from the obstacle's centroid to the path's centerline is less than half the path width, meaning the obstacle intrudes into the path's effective passage area. The contour discrimination unit compares the geometric dimension features and positional distribution features with the above criteria. Contours that simultaneously meet both conditions are uniformly marked with a path-blocking obstacle identification tag. After all contours are categorized and marked, the obstacle contour information with the unique tag is retained.

[0146] The local edge node has a built-in blocking information collection unit that retrieves the center point coordinates and overall outline coverage pixel range of all marked path blocking obstacles. The collection unit converts the center point pixel coordinates of each group of obstacles into the distribution position in the global coordinate system of the warehouse site, and converts the outline coverage pixel range into the site space occupation range. The collection unit integrates the distribution position and occupation range of all path blocking obstacles in the same frame image into a unified independent data file. After the integration operation is completed, the path blocking information of the AGV forklift is generated for subsequent abnormal warning and judgment processes.

[0147] The local edge node has a built-in deviation extraction unit that retrieves the trajectory deviation information generated by the previous steps. The deviation extraction unit splits the trajectory deviation information into independent storage entries divided by sampling time, extracts the instantaneous position deviation value and instantaneous direction deviation value corresponding to each sampling time, and caches them into two independent storage areas respectively. The deviation extraction unit sequentially selects the instantaneous position deviation value corresponding to two adjacent sampling times to compare the numerical changes and obtains the position deviation change trend. It also selects the instantaneous direction deviation value corresponding to two adjacent sampling times to compare the numerical changes and obtains the direction deviation change trend. After the two change trends are completely extracted, the position deviation change rate and direction deviation change rate of the AGV forklift are generated respectively and bound to the corresponding sampling time identifier to complete the storage.

[0148] The local edge node has a built-in deviation determination unit that reads the positive and negative labels of the position deviation change rate and the direction deviation change rate at each time step. The deviation determination unit continuously monitors the positive and negative labels of the change rate corresponding to multiple consecutive sets of adjacent sampling times. When the positive and negative labels of the position deviation change rate or the positive and negative labels of the direction deviation change rate remain completely consistent within consecutive sampling times, it is directly determined that the current AGV forklift is in a state of continuous deviation from the standard travel path. After the determination operation is completed, a dedicated path deviation warning signal is generated and the corresponding sampling time information is bound and stored separately.

[0149] The local edge node's built-in width calculation unit retrieves the distribution location and occupied range of all path blocking obstacles within the path congestion information. The width calculation unit reads the pre-entered coordinates of the AGV forklift's travel path centerline and the overall passage width standard of the path in the global coordinate system of the storage area. The width calculation unit calculates the straight-line distance between the center point coordinates of a single set of path blocking obstacles and the centerline coordinates of the travel path to obtain the relative distance between the two. The width calculation unit calculates the spatial length covered by the obstacle outline in the lateral passage direction of the path to obtain the path range occupied by the obstacle. The width calculation unit subtracts the lateral occupied length of the obstacle from the overall passage width of the travel path to obtain the remaining passable width value of the AGV forklift's travel path.

[0150] The local edge node has a built-in blocking determination unit that reads the remaining passable width value after the measurement is completed. The blocking determination unit retrieves the fixed width standard of the AGV forklift body that is pre-entered locally. When the remaining passable width value is less than the fixed width standard of the body, it directly determines that the current AGV forklift travel path is completely blocked by an obstacle and cannot pass normally. After the determination operation is completed, a dedicated path blocking warning signal is generated and the corresponding sampling time information is bound and stored separately.

[0151] The local edge node has a built-in signal collection unit that reads path deviation warning signals or path blocking warning signals generated at the same sampling time. The collection unit classifies any type of warning signal as an abnormal warning signal of the AGV forklift and binds it to the corresponding sampling time identifier for separate caching.

[0152] The local edge node has a built-in data annotation unit that retrieves the multi-dimensional status feature information output from the previous spatiotemporal alignment process. The data annotation unit reads the sampling time identifier of the cached abnormal warning signal. The data annotation unit retrieves the complete data entry corresponding to the same sampling time within the multi-dimensional status feature information. The data annotation unit attaches the abnormal warning signal to the end storage block of the entry to complete the annotation binding. After all sampling time matching and signal annotation operations are completed, the AGV forklift's annotated status data is generated and completely stored for subsequent transmission strategy matching and data packet encapsulation processes.

[0153] The beneficial effect is that the entire obstacle recognition, anomaly detection, and data annotation process is deployed and executed within the AGV's onboard local edge node. First, image noise interference is removed through grayscale conversion and median filtering. Then, obstacle contours are accurately extracted based on pixel grayscale gradients, and obstructions within the path range are filtered to generate path obstruction information. Next, time-series change analysis is performed on trajectory deviation data to determine the risk of continuous vehicle deviation, and the remaining passage width of the path is calculated based on the obstacle space occupancy range to determine the risk of complete obstruction. Two types of warning signals are generated and integrated into a unified anomaly warning signal, which is directly annotated to multi-dimensional state feature information to obtain labeled state data. The entire process does not require transmitting complete original image data to the cloud, significantly reducing the data volume and transmission latency of wireless communication. It relies on both contour geometry and position to filter real obstructions and determines deviation risks based on the trend of deviation changes over time rather than a single instantaneous deviation. This avoids the problems of false anomaly judgments caused by site light and shadow interference and instantaneous positioning jitter. The labeled status data simultaneously carries multi-dimensional vehicle operation data and abnormal risk indicators. Remote visualization can be completed without secondary data association and matching in the cloud, continuously improving the real-time response speed and judgment accuracy of AGV forklift operation safety warnings.

[0154] E. Obtain the real-time network quality data of the local edge node, determine the target transmission strategy based on the real-time network quality data and the battery status data, and encapsulate the labeled status data according to the target transmission strategy to obtain the monitoring data packet of the AGV forklift;

[0155] In this embodiment of the invention, the step of acquiring the real-time network quality data of the local edge node and determining the target transmission strategy based on the real-time network quality data and the battery status data includes:

[0156] Real-time network quality detection results are extracted from the local edge node of the AGV forklift. The real-time network quality detection results include signal strength indication value, transmission delay measurement value and packet loss rate statistics.

[0157] Read the remaining power indicator value from the battery status data, and determine the power adequacy level of the AGV forklift based on the remaining power indicator value;

[0158] By integrating the signal strength indication value, transmission delay measurement value, and packet loss rate statistics, a comprehensive network quality score for the local edge node is obtained.

[0159] Based on the power adequacy level corresponding to the power weight coefficient, the network quality comprehensive score is corrected to obtain the transmission strategy selection evaluation value of the AGV forklift.

[0160] Based on the transmission strategy, an evaluation value is selected, and a matching target transmission strategy is chosen from the pre-built transmission strategy library.

[0161] The local edge node has a built-in network detection and reading unit that continuously reads the real-time network quality detection results collected by the vehicle communication hardware and stored in the local storage area of ​​the edge node. The network detection and reading unit divides the detection results into independent storage partitions and extracts three types of network monitoring data stored in the partitions: signal strength indication value, transmission delay measurement value, and packet loss rate statistics value. These data are then uniformly stored in the high-speed cache area of ​​the edge node.

[0162] The local edge node's built-in power reading unit retrieves previously generated multi-dimensional status feature information and internally stored battery status data. The power reading unit separately extracts the remaining power indication value recorded in the battery status data storage block. It then retrieves the four pre-defined power range judgment criteria pre-defined by the edge node and compares the currently acquired remaining power indication value with each range criterion: remaining power ≥ 60% is considered ample, 30% ≤ remaining power < 60% is considered moderate, 15% ≤ remaining power < 30% is considered critical, and remaining power < 15% is considered alarm. After matching, a corresponding level identifier is assigned as a separate cache for the AGV forklift's power adequacy level. The local edge node's built-in scoring correction unit retrieves the calculated comprehensive network quality score. The scoring correction unit reads the pre-defined power weight coefficients specific to the currently matched power adequacy level: ample level corresponds to a weight of 1.0, moderate level to 0.85, critical level to 0.7, and alarm level to 0.5. The scoring correction unit performs the integration and correction calculation according to the formula: transmission strategy selection evaluation value = network quality comprehensive score × power weight coefficient. After the correction calculation is completed, an independent cache of the transmission strategy selection evaluation value of the AGV forklift is generated, which takes into account both the on-site network communication conditions and the forklift's power supply capacity.

[0163] The local edge node's built-in network scoring and integration unit synchronously retrieves cached signal strength indication values, transmission delay measurement values, and packet loss rate statistics. It then calculates the comprehensive network quality score using a three-step method: First, normalization is performed, normalizing the RSSI range (-100dBm to -30dBm), transmission delay range (0ms to 500ms), and packet loss rate range (0 to 1) to the [0,1] interval. Higher values ​​indicate worse network quality. Second, a weighted summation is performed, with signal strength weighted at 0.4, transmission delay at 0.3, and packet loss rate at 0.3, to calculate the network quality loss score. Third, a comprehensive score conversion is performed: Comprehensive Network Quality Score = 100 × (1 - Network Quality Loss Score), with a score range of 0 to 100. Higher scores indicate better network quality. After the aggregation calculation, a complete comprehensive network quality score reflecting the current wireless communication link's operational status is generated and stored.

[0164] The local edge node has a built-in scoring correction unit that retrieves the calculated comprehensive network quality score. The scoring correction unit reads the preset power weight coefficient that is specifically matched to the current power adequacy level. The scoring correction unit integrates the comprehensive network quality score with the corresponding power weight coefficient to complete the correction calculation. After the correction calculation is completed, an independent cache of the AGV forklift's transmission strategy selection evaluation value is generated, which takes into account both the on-site network communication conditions and the forklift's power supply capacity.

[0165] The local edge node has a built-in policy retrieval unit that retrieves a pre-stored transmission policy library. This library is divided into numerical ranges and binds various exclusive data encapsulation transmission rules. The pre-built transmission policy library includes three basic policies: The first is a full-capacity high-definition policy, matching scenarios with an evaluation score ≥ 80 points, with a transmission interval of 100ms. The data packet contains full multi-dimensional state features, complete trajectory deviation, obstacle contour data, and warning signals, and the data uses lossless compression. The second is a balanced transmission policy, matching scenarios with an evaluation score ≤ 50 points and < 80 points, with a transmission interval of 200ms. The data packet contains core pose and velocity data, deviation extremes, blockage locations, and warning signals, and redundant image contour details are cropped, using lossy compression. The third is a low-power keep-alive policy, matching scenarios with an evaluation score < 50 points, with a transmission interval of 500ms. It only transmits pose keyframes, abnormal warning signals, and battery status; when there are no abnormalities, it only sends heartbeat packets, minimizing bandwidth and power consumption. The strategy retrieval unit sequentially matches the cached transmission strategy selection evaluation values ​​with all preset value ranges in the transmission strategy library. After locating the corresponding range to which the evaluation value belongs, it retrieves the complete set of transmission control rules bound to that range as the matching target transmission strategy and stores them separately.

[0166] The local edge node's built-in data encapsulation unit retrieves the labeled status data generated in the previous steps and matches the target transmission strategy. The data encapsulation unit reads the data segmentation rules, verification rules, and additional identification rules specified in the target transmission strategy. The data encapsulation unit splits the complete labeled status data into fixed-length data segments according to the segmentation rules. The data encapsulation unit adds the verification identifier and timing identifier specified by the strategy to each data segment. The data encapsulation unit integrates all the segmented data with added identifiers into a unified data carrier in chronological order. After all encapsulation operations are completed, a monitoring data packet for the AGV forklift is generated and awaits transmission to the remote monitoring center.

[0167] The beneficial effects are that the entire process of network detection, power level classification, network score calculation, score correction, strategy matching, and data encapsulation is completed independently within the AGV's on-board local edge node, without waiting for transmission control commands from the cloud. The communication link quality is comprehensively evaluated based on three types of measured data: signal strength, transmission delay, and packet loss rate. The score is then weighted and corrected by combining the remaining power level of the forklift to achieve strategy selection under the dual constraints of network and battery life. According to the adapted target transmission strategy, the labeled status data is standardized and encapsulated to generate monitoring data packets. When the network communication quality is poor, the size of the transmitted data can be reduced to avoid data packet loss. When the battery power is low, the transmission frequency can be reduced to reduce power consumption. The fully encapsulated monitoring data packets have unified verification and timing identifiers. The remote monitoring center can directly and quickly decapsulate and restore the complete vehicle operation and abnormal warning information according to the corresponding strategy, effectively balancing the relationship between wireless communication resources, on-board battery energy consumption, and the integrity of monitoring data transmission.

[0168] F. Send the monitoring data packet to the remote monitoring center and parse the monitoring data packet to obtain the operating status display data of the AGV forklift.

[0169] In this embodiment of the invention, sending the monitoring data packet to a remote monitoring center and parsing the monitoring data packet to obtain the operating status display data of the AGV forklift includes:

[0170] The monitoring data packets are transmitted to the data receiving port of the remote monitoring center via a wireless communication link.

[0171] Within the remote monitoring center, the monitoring data packets are decapsulated according to the data encapsulation format corresponding to the target transmission strategy to restore the labeled status data.

[0172] Extract the multi-dimensional state feature information, the trajectory deviation information, the path blockage information, and the anomaly warning signal from the labeled state data;

[0173] The spatial pose data in the multi-dimensional state feature information is mapped to the coordinate grid of the virtual monitoring interface to obtain the real-time pose display elements of the AGV forklift.

[0174] The trajectory deviation information, the path blockage information, and the abnormal warning signal are superimposed on the real-time pose display element to obtain the operating status display data of the AGV forklift.

[0175] The on-board local edge node has a built-in wireless transmission unit that retrieves the encapsulated monitoring data packets of the AGV forklift. The wireless transmission unit relies on the on-board wireless communication hardware to build a stable wireless communication link. The wireless transmission unit pushes the monitoring data packets one by one according to the transmission interval set by the target transmission strategy. All data packets are completely transmitted to the preset data receiving port of the remote monitoring center through the wireless communication link to complete unified reception and buffering.

[0176] The data parsing unit deployed inside the remote monitoring center reads all monitoring data packets cached at the data receiving port. The data parsing unit retrieves the data encapsulation format standard recorded inside the target transmission strategy matched during the generation stage of the monitoring data packet. The data parsing unit peels off the outer layer of the monitoring data packet layer by layer according to the encapsulation format standard, removing the additional verification mark, timing mark and block encapsulation carrier. After the layer-by-layer peeling operation is completed, the labeled status data without additional encapsulation structure is completely restored and stored separately on the monitoring center's local storage device.

[0177] The remote monitoring center has a built-in information splitting unit that retrieves the fully restored labeled status data. The information splitting unit divides the labeled status data into independent data storage blocks, and reads and extracts four types of independent operating data from the blocks: multi-dimensional status feature information, trajectory deviation information, path blockage information, and abnormal warning signals. The extracted four types of information are uniformly cached in the monitoring center's visualization computing partition for interface mapping and processing.

[0178] The remote monitoring center's built-in coordinate mapping unit retrieves and extracts multi-dimensional state feature information and separately captures internally stored spatial pose data. The coordinate mapping unit reads the pre-drawn site coordinate grid corresponding to the conversion rules in the virtual monitoring interface. The virtual monitoring interface uses a 1:100 scale to map the warehouse's global coordinates to interface pixel coordinates, with the lower left corner of the site as the origin. The coordinate mapping unit converts the warehouse's global spatial coordinates embedded in the spatial pose data into interface pixel coordinates within the virtual monitoring interface grid according to the conversion rules. Based on the converted interface pixel coordinates, the coordinate mapping unit generates a forklift graphic element with a heading arrow at the corresponding grid position in the virtual monitoring interface, overlaying real-time speed and remaining battery power labels. This element serves as the real-time pose display element for the AGV forklift and is permanently stored at the bottom layer of the interface layer.

[0179] The remote monitoring center's built-in layer overlay unit synchronously retrieves three types of auxiliary operational data from the cached data: trajectory deviation information, path blockage information, and abnormal warning signals. The layer overlay unit reads the sampling time matching identifier inherent in each of the three data types. It then searches the virtual monitoring interface for the real-time pose display element corresponding to the same sampling time. Trajectory deviation is displayed as a deviation value label with a path offset shadow, highlighted in red when the deviation exceeds a threshold. Path blockages are marked with red blocks at the corresponding path positions, indicating the obstacle size and remaining passage width. When an abnormal warning is triggered, the corresponding forklift graphic element flashes and a warning type pop-up window appears, simultaneously recording the warning log. The layer overlay unit mounts all three types of auxiliary information modules to the interface layers surrounding the corresponding real-time pose display elements. After all auxiliary information layer mounting and overlay operations are completed, the data is integrated to generate AGV forklift operation status display data containing vehicle position, driving deviation, path obstacles, and safety warnings for intuitive viewing by monitoring personnel.

[0180] The beneficial effects are that the monitoring data packets are transmitted directionally to the receiving port of the remote monitoring center via the vehicle-mounted wireless link. The monitoring center completes the decapsulation of the data packets according to the transmission strategy standard corresponding to the edge node encapsulation, accurately restores the labeled status data, extracts all vehicle operation and abnormal related data step by step, and maps the spatial pose data to the virtual monitoring coordinate grid to generate basic vehicle display elements. Then, all trajectory deviation, path blockage and abnormal warning information are superimposed on the basic display elements to form an integrated operation status display data. The entire analysis and visualization process is standardized and matches the edge end encapsulation rules, so there will be no data parsing misalignment or information loss. Various types of operational risk information are directly bound to the corresponding forklift graphic carrier for centralized display, saving monitoring personnel from comparing multiple independent data across modules, and greatly improving the intuitiveness and work efficiency of remote monitoring and risk identification of AGV forklifts in the warehouse.

[0181] like Figure 2 The diagram shown is a functional block diagram of an AGV forklift remote real-time monitoring system based on edge computing, provided in an embodiment of the present invention.

[0182] The edge computing-based remote real-time monitoring system for AGV forklifts described in this invention can be installed in an electronic device. Depending on the functions implemented, the edge computing-based remote real-time monitoring system for AGV forklifts may include a data acquisition module, a data alignment module, a deviation analysis module, a status labeling module, a data encapsulation module, and a status display module. The module described in this invention can also be referred to as a unit, which refers to a series of computer program segments that can be executed by the processor of an electronic device and perform a fixed function, and are stored in the memory of the electronic device.

[0183] In this embodiment, the functions of each module / unit are as follows:

[0184] The data acquisition module is used to summarize the spatial pose data, movement speed data, battery status data and surrounding environment image data of the AGV forklift to obtain the real-time operating status data of the AGV forklift.

[0185] The data alignment module is used to perform spatiotemporal alignment processing on the real-time operating status data within the local edge node of the AGV forklift to obtain multi-dimensional status feature information of the AGV forklift.

[0186] The deviation analysis module is used to construct a real-time motion trajectory vector within the local edge node based on the pose data and velocity data in the multi-dimensional state feature information, and compare and analyze the real-time motion trajectory vector with a preset standard trajectory vector to obtain the trajectory deviation information of the AGV forklift.

[0187] The status labeling module is used to identify obstacles on the travel path based on the surrounding environment image data, obtain path blocking information, and generate an abnormal warning signal in the local edge node based on the trajectory deviation information and the path blocking information. The abnormal warning signal is then labeled in the multi-dimensional status feature information to obtain the labeled status data of the AGV forklift.

[0188] The data encapsulation module is used to acquire real-time network quality data of the local edge node, determine the target transmission strategy based on the real-time network quality data and the battery status data, and encapsulate the labeled status data according to the target transmission strategy to obtain the monitoring data packet of the AGV forklift.

[0189] The status display module is used to send the monitoring data packet to the remote monitoring center and parse the monitoring data packet to obtain the operating status display data of the AGV forklift.

[0190] In the several embodiments provided by this invention, it should be understood that the disclosed methods and systems can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0191] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0192] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0193] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0194] This application embodiment can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0195] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for remote real-time monitoring of AGV forklifts based on edge computing, characterized in that, The method includes: A. Summarize the spatial pose data, movement speed data, battery status data, and surrounding environment image data of the AGV forklift to obtain the real-time operating status data of the AGV forklift; B. Within the local edge node of the AGV forklift, perform spatiotemporal alignment processing on the real-time operating status data to obtain multi-dimensional status feature information of the AGV forklift; C. Within the local edge node, based on the pose data and velocity data in the multi-dimensional state feature information, a real-time motion trajectory vector is constructed, and the real-time motion trajectory vector is compared and analyzed with a preset standard trajectory vector to obtain the trajectory deviation information of the AGV forklift. D. Based on the surrounding environment image data, identify obstacles on the travel path, obtain path blockage information, and generate an abnormal warning signal within the local edge node based on the trajectory deviation information and the path blockage information. Mark the abnormal warning signal into the multi-dimensional state feature information to obtain the marked state data of the AGV forklift. E. Obtain the real-time network quality data of the local edge node, determine the target transmission strategy based on the real-time network quality data and the battery status data, and encapsulate the labeled status data according to the target transmission strategy to obtain the monitoring data packet of the AGV forklift; F. Send the monitoring data packet to the remote monitoring center and parse the monitoring data packet to obtain the operating status display data of the AGV forklift.

2. The remote real-time monitoring method for AGV forklifts based on edge computing as described in claim 1, characterized in that, Within the local edge node of the AGV forklift, the real-time operating status data is subjected to spatiotemporal alignment processing to obtain multi-dimensional status feature information of the AGV forklift, including: Acquire spatial pose data, motion speed data, battery status data, and surrounding environment image data of the AGV forklift; Using the timestamp of the spatial pose data as the reference time axis, time interpolation synchronization processing is performed on the motion speed data, the battery status data and the surrounding environment image data to obtain the time-aligned state dataset of the AGV forklift. Based on the spatial coordinate system parameters of the spatial pose data after centralized alignment in the time-aligned state dataset, coordinate projection transformation is performed on the pixel coordinates detected in the surrounding environment image data to obtain the environmental association data of the AGV forklift. According to the coordinate axis direction in the spatial coordinate system parameters, the aligned motion speed data in the time-aligned state dataset is decomposed to obtain the speed component data of the AGV forklift. The environmental correlation data, the speed component data, and the battery status data are correlated and fused to obtain the multi-dimensional status feature information of the AGV forklift.

3. The remote real-time monitoring method for AGV forklifts based on edge computing as described in claim 1, characterized in that, Within the local edge node, the real-time motion trajectory vector is constructed based on the pose data and velocity data in the multi-dimensional state feature information, including: Real-time position coordinate sequence and real-time attitude angle data, as well as real-time velocity component values ​​and real-time acceleration component values ​​along different coordinate axes at different sampling times are extracted from the multi-dimensional state feature information. Using the real-time position coordinates at the current sampling moment as the trajectory starting point, and calculating the displacement component increments of the AGV forklift along different coordinate axes within a preset sampling time interval, the predicted position coordinates at the next sampling moment are obtained. Using the predicted position coordinates as the new trajectory starting point, the displacement component increment calculation operation is repeated to obtain the predicted position coordinate sequence of the AGV forklift. Based on the real-time heading angle in the real-time attitude angle data, a heading angle offset correction is applied to the predicted position coordinates in the predicted position coordinate sequence to obtain the corrected predicted position coordinate sequence of the AGV forklift. The real-time position coordinate sequence and the corrected predicted position coordinate sequence are weighted and fused to obtain the smoothed position coordinate sequence of the AGV forklift. By sequentially connecting the smoothed position coordinate sequence according to the sampling time order, the real-time motion trajectory vector of the AGV forklift is constructed.

4. The remote real-time monitoring method for AGV forklifts based on edge computing as described in claim 3, characterized in that, The formula for calculating the displacement component increment is: ; in, Indicates the first The sampling time along the first sampling moment The displacement component increment in each coordinate axis direction, Indicates the first The sampling time along the first sampling moment Real-time velocity component values ​​in each coordinate axis direction. Indicates the first The sampling time along the first sampling moment Real-time acceleration component values ​​in each coordinate axis direction. This indicates the preset sampling time interval. Indicates the first The sampling time along the first sampling moment Preset heading offset compensation coefficients in each coordinate axis direction Indicates the first The real-time heading angle at the sampling time along the first sampling point The direction cosine components of each coordinate axis.

5. The remote real-time monitoring method for AGV forklifts based on edge computing as described in claim 3, characterized in that, The step of comparing and analyzing the real-time motion trajectory vector with a preset standard trajectory vector to obtain the trajectory deviation information of the AGV forklift includes: The real-time trajectory position coordinates at the sampling time are extracted from the real-time motion trajectory vector, and the standard trajectory position coordinates corresponding to the sampling time are extracted from the preset standard trajectory vector. The difference between the coordinates of the real-time trajectory position point and the coordinates of the standard trajectory position point is analyzed to obtain the instantaneous position deviation value at the sampling time; Based on the real-time trajectory position coordinates at continuous sampling times, the real-time travel direction angle of the AGV forklift between adjacent sampling times is determined, and based on the corresponding standard trajectory position coordinates, the standard travel direction angle of the AGV forklift is determined. A differential analysis is performed on the real-time travel direction angle and the standard travel direction angle to obtain the instantaneous direction deviation value at the sampling time; The instantaneous position deviation value and instantaneous direction deviation value are arranged according to the sampling time sequence to obtain the trajectory deviation information of the AGV forklift.

6. The remote real-time monitoring method for AGV forklifts based on edge computing as described in claim 1, characterized in that, The step of identifying obstacles on the travel path and obtaining path obstruction information based on the surrounding environment image data includes: The surrounding environment image data is subjected to grayscale processing and noise filtering to obtain the preprocessed environmental image data of the AGV forklift. Based on the gray-level gradient change characteristics of pixels in the preprocessed environmental image data, the contour edge information in the preprocessed environmental image data is extracted to obtain the environmental contour edge image of the AGV forklift. From the environmental contour edge image, the region that overlaps with the AGV forklift's travel path is selected to obtain the obstacle contour set of the AGV forklift; Based on the geometric size and positional distribution characteristics of the obstacle contours in the obstacle contour set, the obstacle contours are marked as path-blocking obstacles; Based on the distribution and extent of the obstacles blocking the path on the AGV forklift's travel path, the path blocking information of the AGV forklift is obtained.

7. The remote real-time monitoring method for AGV forklifts based on edge computing as described in claim 6, characterized in that, The step of generating an anomaly warning signal within the local edge node based on the trajectory deviation information and the path congestion information includes: Instantaneous position deviation value and instantaneous direction deviation value are extracted from the trajectory deviation information, and time differentiation processing is performed on the instantaneous position deviation value and the instantaneous direction deviation value respectively to obtain the position deviation change rate and direction deviation change rate of the AGV forklift; When the sign of the change rate of position deviation or the change rate of direction deviation remains unchanged within consecutive sampling times, it is determined that the AGV forklift is in a continuous deviation state, and a path deviation warning signal for the AGV forklift is generated. The distribution location and occupied area of ​​the path-blocking obstacles are obtained. Based on the relative distance between the distribution location and the center line of the AGV forklift travel path, and the relative relationship between the occupied area and the width of the AGV forklift travel path, the remaining passable width of the AGV forklift travel path is determined. When the remaining passable width is less than the width of the AGV forklift body, it is determined that the AGV forklift's travel path is completely blocked, and a path blockage warning signal for the AGV forklift is generated. The path deviation warning signal or the path blockage warning signal shall be used as the abnormal warning signal of the AGV forklift.

8. The remote real-time monitoring method for AGV forklifts based on edge computing as described in claim 1, characterized in that, The step of acquiring real-time network quality data of the local edge node and determining a target transmission strategy based on the real-time network quality data and the battery status data includes: Real-time network quality detection results are extracted from the local edge node of the AGV forklift. The real-time network quality detection results include signal strength indication value, transmission delay measurement value and packet loss rate statistics. Read the remaining power indicator value from the battery status data, and determine the power adequacy level of the AGV forklift based on the remaining power indicator value; By integrating the signal strength indication value, transmission delay measurement value, and packet loss rate statistics, a comprehensive network quality score for the local edge node is obtained. Based on the power adequacy level corresponding to the power weight coefficient, the network quality comprehensive score is corrected to obtain the transmission strategy selection evaluation value of the AGV forklift. Based on the transmission strategy, an evaluation value is selected, and a matching target transmission strategy is chosen from the pre-built transmission strategy library.

9. The remote real-time monitoring method for AGV forklifts based on edge computing as described in claim 1, characterized in that, The step of sending the monitoring data packet to the remote monitoring center and parsing the monitoring data packet to obtain the operating status display data of the AGV forklift includes: The monitoring data packets are transmitted to the data receiving port of the remote monitoring center via a wireless communication link. Within the remote monitoring center, the monitoring data packets are decapsulated according to the data encapsulation format corresponding to the target transmission strategy to restore the labeled status data. Extract the multi-dimensional state feature information, the trajectory deviation information, the path blockage information, and the anomaly warning signal from the labeled state data; The spatial pose data in the multi-dimensional state feature information is mapped to the coordinate grid of the virtual monitoring interface to obtain the real-time pose display elements of the AGV forklift. The trajectory deviation information, the path blockage information, and the abnormal warning signal are superimposed on the real-time pose display element to obtain the operating status display data of the AGV forklift.

10. A remote real-time monitoring system for AGV forklifts based on edge computing, characterized in that, The system for implementing the edge computing-based remote real-time monitoring method for AGV forklifts as described in claim 1 includes: The data acquisition module is used to summarize the spatial pose data, movement speed data, battery status data and surrounding environment image data of the AGV forklift to obtain the real-time operating status data of the AGV forklift. The data alignment module is used to perform spatiotemporal alignment processing on the real-time operating status data within the local edge node of the AGV forklift to obtain multi-dimensional status feature information of the AGV forklift. The deviation analysis module is used to construct a real-time motion trajectory vector within the local edge node based on the pose data and velocity data in the multi-dimensional state feature information, and compare and analyze the real-time motion trajectory vector with the preset standard trajectory vector to obtain the trajectory deviation information of the AGV forklift. The status labeling module is used to identify obstacles on the travel path based on the surrounding environment image data, obtain path blocking information, and generate an abnormal warning signal in the local edge node based on the trajectory deviation information and the path blocking information. The abnormal warning signal is then labeled in the multi-dimensional status feature information to obtain the labeled status data of the AGV forklift. The data encapsulation module is used to acquire the real-time network quality data of the local edge node, determine the target transmission strategy based on the real-time network quality data and the battery status data, and encapsulate the labeled status data according to the target transmission strategy to obtain the monitoring data packet of the AGV forklift. The status display module is used to send the monitoring data packet to the remote monitoring center and parse the monitoring data packet to obtain the operating status display data of the AGV forklift.