Edge computing communication methods and systems for industrial IoT

Abstract

This invention relates to the field of Industrial Internet of Things (IIoT), and particularly to an edge computing communication method and system for IIoT. The method includes the following steps: Step 1: A user issues a task to a service platform through a user platform. Upon receiving the task request, the service platform classifies and labels the task based on the associated device type and task type, determining the task's real-time level and the importance weight of its associated production links. Step 2: The service platform transmits the classified tasks to the corresponding service sub-platforms. The service sub-platforms obtain historical operating data of the associated devices and basic edge node information, and generate a first computing task carrying task attribute tags based on the classification results, which is then sent to a management platform. Step 3: The management platform manages multiple first computing tasks uniformly. The technical solution of this invention is well-suited to the characteristics of industrial scenarios, taking into account task priority, resource dynamism, and network adaptability.

Description

Technical Field

[0001] This invention relates to the field of industrial Internet of Things (IoT), and particularly to edge computing communication methods and systems for industrial IoT. Background Technology

[0002] With the deepening of industrial development, the Industrial Internet of Things (IIoT) has become a core infrastructure for smart factories and intelligent manufacturing. Through the interconnection of massive sensors, intelligent devices, and control systems, it has realized the digitalization and intelligentization of the entire production process. In this context, edge computing, as a key technology connecting terminal devices and the cloud, is widely used in the real-time processing of industrial data due to its low latency and high reliability. For example, scenarios such as robot collaborative control in automotive welding production lines, closed-loop temperature regulation in chemical reactors, and AGV path optimization in warehousing and logistics all rely on the instantaneous response of edge nodes to data.

[0003] However, the complexity of industrial scenarios presents unique challenges to edge computing: on the one hand, industrial equipment is diverse (from high-precision CNC machine tools to low-power sensors), resulting in significantly heterogeneous computing tasks, ranging from real-time control tasks requiring millisecond-level response (such as pressure feedback adjustment of stamping equipment) to data statistics tasks with minute-level latency tolerance (such as energy consumption report generation); on the other hand, the industrial environment is highly dynamic, with the movement of production equipment (such as AGV robots), interference from wireless signals (such as electromagnetic noise generated by motor operation), and temporary offline status of edge nodes (such as equipment maintenance) all leading to an imbalance between computing resources and task requirements.

[0004] Existing edge computing task allocation schemes have significant limitations: most schemes schedule tasks based on only a single dimension (such as physical distance or network latency), ignoring the fact that critical equipment (such as sensors in steelmaking blast furnaces) has a higher task priority than auxiliary equipment (such as lighting control systems), and also failing to consider the real-time load of edge nodes (e.g., if an edge gateway is already processing 100 video streams, allocating high-computing tasks will lead to congestion). This directly results in production interruptions caused by scheduling delays for high-priority tasks, waste of critical resources due to low-priority tasks, and duplicate allocation of tasks to mobile devices (such as inspection robots) due to dynamic changes in location, ultimately reducing the operational efficiency and reliability of the entire industrial system.

[0005] Furthermore, industrial data demands stringent security and integrity requirements. Existing solutions lack adaptation optimizations for industrial protocols (such as Modbus and Profinet) during task transmission, making them prone to data loss or errors when the network fluctuates, further exacerbating the uncertainty of task processing. Therefore, there is an urgent need for an edge computing communication method and system for the Industrial Internet of Things that fits the characteristics of industrial scenarios and takes into account task priority, resource dynamism, and network adaptability. Summary of the Invention

[0006] This invention provides an edge computing communication method for the Industrial Internet of Things (IIoT), which can be tailored to the characteristics of industrial scenarios and take into account task priority, resource dynamism, and network adaptability.

[0007] To solve the above-mentioned technical problems, this application provides the following technical solution:

[0008] An edge computing communication method for the Industrial Internet of Things includes the following steps:

[0009] Step 1: The user submits a task to the service platform through the user platform. After receiving the task request, the service platform classifies and marks the task based on the device type and task type associated with the task, and determines the real-time level and importance weight of the associated production links of the task.

[0010] Step 2: The service platform transmits the categorized tasks to the corresponding service sub-platforms. The service sub-platforms obtain the historical operating data of the devices associated with the tasks and the basic information of the edge nodes, and generate the first computing task carrying the task attribute tags according to the categorization results, and send it to the management platform.

[0011] Step 3: The management platform manages multiple first computing tasks in a unified manner, and uses a multi-parameter balancing algorithm to calculate task priority coefficients. The balancing algorithm comprehensively evaluates the following factors to determine the priority coefficients by weighted summation: the logical distance between the device to which the task belongs and the service sub-platform, the physical distance between the device and the center of the edge computing area, the weight corresponding to the task's real-time level, and the current load rate of the edge node; then, the tasks are sorted according to the priority coefficients, generating second computing tasks with computing sequence numbers and sending them to the sensor network platform.

[0012] Step 4: After receiving the second computing task, the edge computing area center of the sensor network platform first verifies the data integrity, then adjusts the transmission strategy based on the real-time network quality, organizes it into an independent third computing task, and sends it to the target platform.

[0013] Step 5: The edge nodes of the object platform execute the third calculation task according to the calculation sequence number, and at the same time report their own load status to the management platform in real time. If the load exceeds the preset threshold, the management platform dynamically adjusts the priority coefficient calculation parameters of subsequent tasks according to the load status reported by the object platform.

[0014] The basic principle and beneficial effects of the solution are as follows: Based on the core characteristics of heterogeneity of equipment, timeliness of tasks and complexity of networks in industrial scenarios, this invention constructs a full-link communication mechanism that includes task classification, dynamic priority calculation, adaptive network transmission and resource load feedback.

[0015] Step one involves classifying and labeling tasks based on the type attributes of the associated equipment (such as production execution equipment and status monitoring equipment) and the task's own type (such as real-time control and data acquisition). Simultaneously, the importance weight of the production process to which the task belongs and the real-time requirements are combined to bind the task to the core needs of industrial production from the task initiation stage, ensuring that the task processing priority matches the criticality of the production process.

[0016] Step 2: Obtain historical operating data of the devices associated with the task and basic information of the edge nodes through the service sub-platform, and assign attribute tags to the task; Step 3: Based on four core parameters, the logical distance between the device to which the task belongs and the service sub-platform (reflecting network transmission costs), the physical distance between the device and the edge computing area center (reflecting edge node coverage efficiency), the weight corresponding to the task's real-time level (reflecting industrial latency requirements), and the current load rate of the edge node (reflecting resource availability), a weighted summation algorithm is used to calculate the priority coefficient, so as to achieve accurate adaptation of task priority to the dynamic changes of distance, timeliness, and resources in the industrial scenario, and avoid resource mismatch caused by single-dimensional decision-making.

[0017] Step four addresses the issues of electromagnetic interference and signal fluctuations in industrial environments by linking data integrity verification (such as CRC check and signature verification) with real-time network quality (packet loss rate and transmission latency) and dynamically adjusting transmission strategies (such as fragmented transmission, priority scheduling, and compression algorithm adaptation) to ensure reliable transmission of task data in complex industrial network environments.

[0018] Step 5 uses edge nodes to provide real-time feedback on load status (CPU utilization, memory usage, etc.), forming a closed-loop control of task allocation, load monitoring, and parameter adjustment. When the load exceeds a preset threshold, the management platform dynamically optimizes the priority calculation parameters of subsequent tasks to avoid task blocking caused by edge node overload, adapting to the dynamic changes in equipment load in industrial scenarios.

[0019] By binding tasks to the importance of production processes and equipment types, the communication mechanism is tailored to industrial production workflows (e.g., control tasks for critical production line equipment take precedence over statistical tasks for auxiliary equipment). This addresses the issue of insufficient adaptation of general edge computing methods to specific industrial scenarios and reduces ineffective resource consumption. A sorting mechanism based on real-time performance levels and dynamic priority coefficients ensures that high-priority tasks (such as equipment fault warning responses) are processed first, reducing the risk of production interruptions due to task delays and enhancing the anti-interference capability and operational security of industrial systems.

[0020] By providing real-time load rate feedback and dynamically adjusting priority parameters, performance degradation caused by overload of edge nodes can be avoided, while resource idleness is reduced, edge computing resource utilization is improved, and the lifespan of equipment is extended.

[0021] The closed-loop design of the entire process from task initiation to execution enables coordinated linkage of classification, calculation, transmission and execution. Compared with traditional methods, the average processing latency of tasks is reduced and the system response speed is significantly improved in high-concurrency scenarios, meeting the core requirements of industrial IoT for high reliability and low latency communication.

[0022] In summary, this invention achieves the goal of fitting the characteristics of industrial scenarios while taking into account task priority, resource dynamism, and network adaptability.

[0023] Furthermore, the balancing algorithm dynamically adjusts the logical distance between the task's device and the service sub-platform, the physical distance between the device and the edge computing area center, the weight corresponding to the task's real-time level, and the weight ratio of the current load rate of the edge node based on the real-time status of the industrial scenario, with the sum of all weight coefficients being 1. In step one, before classifying and labeling the task, the service platform first obtains the real-time operating parameters of the task-related devices. These real-time operating parameters include the device's current load rate, temperature fluctuation value, and fault warning code. Based on a preset parameter threshold range, the device status is divided into three levels: normal, alert, and emergency. The service platform performs matrix matching between the device status level and the task type, and performs weighted correction on the initially determined real-time level. At the same time, the service platform also obtains the difference between the planned completion time and the current progress of the production link to which the task belongs, as well as the order delivery cycle corresponding to the production link. Based on the rule that the larger the progress difference, the higher the importance weight, and the delivery cycle ≤ the preset warning, the importance weight is increased, the importance weight is dynamically calibrated for the importance weight of the associated production link.

[0024] Furthermore, in step two, when the service platform obtains the historical operating data of the device associated with the task, it simultaneously calls the stored fault handling records, average fault-free runtime, and historical load peak data of the edge node within a preset time period for the device, and embeds the dynamically corrected real-time level and the secondary calibrated importance weight from step one into the associated fields of the historical data. The attribute tags generated by the service sub-platform include a task priority identifier based on the real-time level, a resource requirement threshold based on the importance weight and the historical load peak of the edge node, and a data sampling frequency. At the same time, the service sub-platform performs dual verification on the generated first calculation task. First, it verifies the legality of the task by matching the device's unique code with the timestamp, and then it ensures the integrity of data transmission by performing CRC redundancy verification. Only after the verification is passed is it sent to the management platform.

[0025] Furthermore, in step three, before the management platform calculates the priority coefficient using the balancing algorithm, it first parses the attribute tags of the first calculation task. If the device status level in the tag is urgent and the resource demand threshold is greater than or equal to the preset high value, then the weight ratio corresponding to the task's real-time level will be increased during the weighted summation, while the weight ratio of the current load rate of the edge node will be reduced. If the device status level is normal and the resource demand threshold is less than the preset low value, then the weight ratio of logical distance and physical distance will be increased to prioritize allocation to the closer edge node.

[0026] Furthermore, in step three, when the management platform performs unified management of the first computing task, if the device associated with the task is a mobile industrial device, it calls the historical motion trajectory data of the device provided by the service sub-platform to calculate the predicted location value within a preset time period in the future, and uses the distance between the predicted value and the center of the edge computing area as the dynamic correction value of the physical distance; if it is a static device, it uses the average value of the historical physical distance as the physical distance input value.

[0027] Furthermore, in step three, when the management platform sorts tasks according to priority coefficients, if the difference in priority coefficients between two or more tasks is less than or equal to a preset threshold, a secondary sorting mechanism is triggered: first, the importance weights of the production links associated with the tasks are compared, and tasks with higher weights are sorted first; if the weights are the same, the historical fault handling records of the devices associated with the tasks are called, and the tasks corresponding to the devices with more faults in the recent preset time are sorted first; at the same time, the management platform performs edge node load pre-verification on the generated second computing tasks. If the total resource demand of the tasks to be allocated on a certain edge node exceeds 80% of its current remaining capacity, the priority coefficients of the excess tasks are reduced proportionally to the remaining capacity and reordered.

[0028] Furthermore, in step four, when the edge computing area center of the sensor network platform verifies data integrity, a hierarchical verification mechanism is adopted: for tasks with an urgent device status level or a high real-time level in the attribute label, in addition to the basic CRC check, data signature verification is performed using the device's unique key pre-stored in the service sub-platform to prevent data tampering; for tasks in a normal state with a low real-time level, only the basic CRC check is performed; when adjusting the transmission strategy based on real-time network quality, if the network packet loss rate is greater than the first preset delay, the industrial data-specific compression algorithm is automatically enabled; if the network latency is greater than the second preset delay, the third computing task is split into the smallest execution unit for fragmented transmission, and the fragment priority is marked.

[0029] Furthermore, in step four, when organizing the third computing task, the sensor network platform needs to embed real-time network quality parameters and data compression / fragmentation identifiers, and synchronously feed these identifiers back to the management platform. If three consecutive verification failures or single-task transmission timeouts are detected during transmission, the transmission strategy will be dynamically switched: the backup communication link will be automatically enabled, and the transmission priority of non-urgent tasks will be reduced to reserve bandwidth.

[0030] Furthermore, in step five, the load status reported by the edge nodes of the object platform to the management platform includes real-time CPU utilization, memory usage, and task queue length. After receiving the data, the management platform first classifies the edge node status into three levels: normal, slightly overloaded, and severely overloaded based on the load data.

[0031] If the overload is mild, the management platform will increase the weight of the current load rate of the edge node and decrease the weight of logical distance when adjusting the priority coefficient calculation parameters of subsequent tasks, and prioritize the allocation of tasks to nodes with lower load. If the overload is severe, the allocation of new tasks to the node will be suspended, and the weight of physical distance will be increased to guide subsequent tasks to the next closest edge node with normal load. At the same time, the real-time network quality data fed back by the sensor network platform will be called to select the transmission path with the lowest packet loss rate for the diversion tasks. Attached Figure Description

[0032] Figure 1 This is a flowchart illustrating an embodiment of an edge computing communication method for the Industrial Internet of Things (IIoT). Detailed Implementation

[0033] The following detailed description illustrates the specific implementation method:

[0034] Edge computing communication methods for the Industrial Internet of Things (such as) Figure 1 (As shown), including the following steps:

[0035] Step 1: The user submits a task to the service platform through the user platform. After receiving the task request, the service platform classifies and marks the task based on the device type and task type associated with the task, and determines the real-time level and importance weight of the associated production links of the task.

[0036] Step 2: The service platform transmits the categorized tasks to the corresponding service sub-platforms. The service sub-platforms obtain the historical operating data of the devices associated with the tasks and the basic information of the edge nodes, and generate the first computing task carrying the task attribute tags according to the categorization results, and send it to the management platform.

[0037] Step 3: The management platform manages multiple first computing tasks in a unified manner, and uses a multi-parameter balancing algorithm to calculate task priority coefficients. The balancing algorithm comprehensively evaluates the following factors to determine the priority coefficients by weighted summation: the logical distance between the device to which the task belongs and the service sub-platform, the physical distance between the device and the center of the edge computing area, the weight corresponding to the task's real-time level, and the current load rate of the edge node; then, the tasks are sorted according to the priority coefficients, generating second computing tasks with computing sequence numbers and sending them to the sensor network platform.

[0038] Step 4: After receiving the second computing task, the edge computing area center of the sensor network platform first verifies the data integrity, then adjusts the transmission strategy based on the real-time network quality, organizes it into an independent third computing task, and sends it to the target platform.

[0039] Step 5: The edge nodes of the object platform execute the third calculation task according to the calculation sequence number, and at the same time report their own load status to the management platform in real time. If the load exceeds the preset threshold, the management platform dynamically adjusts the priority coefficient calculation parameters of subsequent tasks according to the load status reported by the object platform.

[0040] In practical applications, the following is a detailed explanation using an industrial IoT scenario in an automotive welding workshop. In this scenario, the workshop is equipped with welding robots (production equipment), temperature sensors (monitoring equipment), AGV handling robots (mobile devices), and edge nodes, including edge gateways in the welding area and edge servers in the logistics area. The service platform is linked with the Manufacturing Execution System (MES) and the equipment management system to achieve real-time control and data management of the welding process.

[0041] The workshop manager needs to issue three tasks through the user platform: real-time adjustment of the current parameters of welding robot A, hourly data summary of temperature sensor B, and path optimization of AGV robot C.

[0042] After receiving the tasks, the service platform categorizes them as follows: Task A of welding robot: associated equipment type is production equipment, task type is real-time control, real-time performance level is set to high, and the initial importance weight of the associated production process (welding procedure) is set to 0.8; Task B of temperature sensor: associated equipment type is monitoring equipment (the specific type can be set according to requirements, such as subsequent mobile equipment), task type is data acquisition, real-time performance level is set to low, and the initial importance weight is set to 0.3; Task C of AGV robot: associated equipment type is mobile equipment, task type is analysis and decision-making, real-time performance level is set to medium, and the initial importance weight is set to 0.5.

[0043] The service platform transmits the task to the corresponding service sub-platform. Taking the task of welding robot A as an example, the service sub-platform calls the sub-service database to obtain its historical operation data (average current fluctuation value in the past 30 days) and basic information of the edge node (edge ​​gateway of the welding area), generates the first calculation task, and sets the attribute tags as real-time control, high real-time performance, and welding process, and sends it to the management platform.

[0044] The management platform uniformly manages three tasks and uses a balancing algorithm to calculate priority coefficients: Welding Robot A: logical distance (e.g., network hop count 2 with the service sub-platform), physical distance (e.g., 5 meters from the center of the welding area), real-time weight (high = 1.2), edge gateway load rate (e.g., 30%), and the priority coefficient after weighted summation is, for example, 0.85; Temperature Sensor B: logical distance (e.g., network hop count 3), physical distance (e.g., 8 meters from the center of the welding area), real-time weight (e.g., low = 0.5), edge gateway load rate (e.g., 30%), and the priority coefficient is, for example, 0.32; AGV Robot C: logical distance (e.g., network hop count 4), physical distance (e.g., 10 meters from the center of the logistics area), real-time weight (e.g., medium = 0.8), edge server load rate (e.g., 40%), and the priority coefficient is, for example, 0.58. After sorting, a second calculation task is generated with the sequence number: Welding Robot A (1) → AGV Robot C (2) → Temperature Sensor B (3), and sent to the sensor network platform.

[0045] The welding area center of the sensor network platform receives the task, verifies the data integrity (CRC check passes), detects the real-time network quality (packet loss rate 1%, latency 20ms), and organizes it into a third calculation task before sending it to the target platform (the control module of welding robot A) without adjusting the transmission strategy.

[0046] The edge nodes of the welding area on the object platform execute tasks according to their sequence numbers and provide real-time feedback on the load status (e.g., CPU utilization of 45%, not exceeding the threshold of 60%); the management platform maintains the original priority calculation parameters.

[0047] The balancing algorithm dynamically adjusts the logical distance between the task's device and the service sub-platform, the physical distance between the device and the edge computing area center, the weight corresponding to the task's real-time level, and the weight ratio of the current load rate of the edge node based on the real-time status of the industrial scenario, with the sum of all weight coefficients being 1. In step one, before classifying and labeling the task, the service platform first obtains the real-time operating parameters of the task-related devices. These real-time operating parameters include the device's current load rate, temperature fluctuation value, and fault warning code. Based on a preset parameter threshold range, the device status is divided into three levels: normal, alert, and emergency. The service platform performs matrix matching between the device status level and the task type, and performs weighted correction on the initially determined real-time level. At the same time, the service platform also obtains the difference between the planned completion time and the current progress of the production link to which the task belongs, as well as the order delivery cycle corresponding to the production link. Based on the rule that the larger the progress difference, the higher the importance weight, and the delivery cycle ≤ the preset warning, the importance weight is increased, the importance weight is dynamically calibrated for the importance weight of the associated production link.

[0048] In actual use, the service platform obtains the real-time operating status of welding robot A through the equipment management system: it is currently in a high-load operation mode (for example, the load rate is 90% > 80% threshold), which is determined to be a warning state;

[0049] Adjusting the real-time performance level based on the alert status: The original high-level performance is weighted by 1.2 times, and the equivalent real-time performance weight after adjustment is 1.44;

[0050] The production progress of the welding process was obtained by connecting to the MES system: 500 welding points were planned to be completed, 400 were actually completed, the progress lag factor was 20% ((500-400) / 500), and the order delivery cycle was 48 hours (≤72 hours).

[0051] Secondary calibration of importance weights: 20% delay coefficient, weight +0.2; expedited orders, weight increased by 20%, original 0.8, 0.8 + 0.2 = 1.0, 1.0 × 1.2 = 1.2.

[0052] In step two, when the service platform obtains the historical operating data of the device associated with the task, it simultaneously calls the stored fault handling records, average fault-free runtime, and historical load peak data of the edge node within a preset time period for the device. It also embeds the dynamically corrected real-time level and the secondary calibrated importance weight from step one into the associated fields of the historical data. The attribute tags generated by the service sub-platform include the task priority identifier based on the real-time level, the resource requirement threshold based on the importance weight and the historical load peak of the edge node, and the data sampling frequency. At the same time, the service sub-platform performs dual verification on the generated first calculation task. First, it verifies the legality of the task by matching the device's unique code with the timestamp, and then it ensures the integrity of data transmission by performing CRC redundancy verification. Only after the verification is passed is the task sent to the management platform.

[0053] Specifically, when the welding area service sub-platform interacts with the sub-service database, in addition to obtaining the historical operating data of welding robot A (the last 3 faults were all abnormal currents) and basic information of the edge gateway, it also calls the real-time level (equivalent to 1.44) corrected in step one and the calibrated importance weight (1.2); it correlates and matches historical fault records with the load fluctuation curve of the edge node (the load peak in the last hour was 40%), and generates preprocessing suggestions: adjust the data sampling frequency to 10ms / time (high frequency), and reserve 60% of computing resources (due to high importance); embed the preprocessing suggestions into the first computing task, with the additional attributes of sampling frequency = 10ms, reserved resources = 60%, and verify the task identifier (Weld-20231001-001), the unique device code (Robot-A-007), and the timestamp (2023-10-01 08:30:00). After the verification is passed, it is sent to the management platform.

[0054] In step three, before the management platform calculates the priority coefficient using the balancing algorithm, it first parses the attribute tags of the first calculation task. If the device status level in the tag is urgent and the resource demand threshold is greater than or equal to the preset high value, the weight ratio corresponding to the task's real-time level will be increased during the weighted summation, while the weight ratio of the current load rate of the edge node will be reduced. If the device status level is normal and the resource demand threshold is less than the preset low value, the weight ratio of logical distance and physical distance will be increased to prioritize allocation to the closer edge node.

[0055] Specifically, the attribute tags of the first calculation task are analyzed, revealing hidden warnings in the device status (generated in step two), a resource reservation ratio of 60% ≥ the preset high value (50%), and an importance weight of 1.2 ≥ 0.8. The dynamic weight adjustment of the balancing algorithm is initiated: the weight ratio corresponding to the task's real-time level is increased from the original 25% to 55% (+30%), and the weight of the current load rate of the edge node is reduced from 25% to 5% (-20%). When calculating the priority coefficient, the high proportion of the real-time weight ensures that the task of welding robot A is allocated first, and even if the edge gateway load rate rises to 50% in the short term, it still gets priority in obtaining resources, avoiding welding quality problems caused by abnormal current.

[0056] In step three, when the management platform manages the first computing task in a unified manner, if the device associated with the task is a mobile industrial device, it calls the historical motion trajectory data of the device provided by the service sub-platform to calculate the predicted location value within a preset time period in the future, and uses the distance between the predicted value and the center of the edge computing area as the dynamic correction value of the physical distance; if it is a static device, it uses the average value of the historical physical distance as the physical distance input value.

[0057] For the task of AGV robot C (mobile device), the management platform processes it as follows in step three: it calls the historical motion trajectory data of AGV robot C provided by the service sub-platform (average speed of 0.5m / s over the past 3 days, mainly moving back and forth in the logistics channel) to predict its position in the next 10 minutes (moving from the current point P1 to P2, with the distance from the center of the edge computing area changing from 10 meters to 8 meters); the physical distance is dynamically corrected to the current distance (10 meters) × 70% + predicted distance (8 meters) × 30% = 9.4 meters, replacing the original static distance of 10 meters; after the correction, the priority coefficient of the AGV task is increased from 0.58 to 0.62, and the order is moved forward to after welding robot A and before temperature sensor B, avoiding the deviation in subsequent task allocation caused by the movement of the AGV.

[0058] In step three, when the management platform sorts tasks according to priority coefficients, if the difference in priority coefficients between two or more tasks is less than or equal to a preset threshold, a secondary sorting mechanism is triggered: first, the importance weights of the production links associated with the tasks are compared, and tasks with higher weights are sorted first; if the weights are the same, the historical fault handling records of the devices associated with the tasks are called, and the tasks corresponding to the devices with the most faults in the recent preset time are sorted first; at the same time, the management platform performs edge node load pre-verification on the generated second computing tasks. If the total resource demand of the tasks to be allocated on a certain edge node exceeds 80% of its current remaining capacity, the priority coefficients of the excess tasks are reduced proportionally to the remaining capacity and reordered.

[0059] Specifically, if the task of welding robot A (priority coefficient 0.85) differs from the task of another welding robot D (priority coefficient 0.83, difference 2% ≤ 5%), the sorting process of the management platform in step three is as follows:

[0060] First level of detection: compare importance weights. A's weight of 1.2 is greater than D's weight of 0.9, so A takes priority.

[0061] If the weights are the same (assuming they are both 1.0), the second detection is triggered: historical fault records are called. A has failed 2 times in the past 30 days, and D has failed 5 times in the past 30 days. D's task takes priority (due to the frequent failures, more timely path optimization is needed).

[0062] Meanwhile, the edge server load pre-check found that the total resource demand of the tasks to be assigned reached 90% of the remaining capacity (exceeding the 80% threshold). The priority coefficient of the 10% of tasks exceeding this limit (such as the sub-task of temperature sensor B) was reduced to 0.32 × 90% = 0.29 to ensure that the edge nodes are not overloaded.

[0063] In step four, when the edge computing area center of the sensor network platform verifies data integrity, a hierarchical verification mechanism is adopted: for tasks with an urgent or high real-time level in the attribute label, in addition to the basic CRC check, data signature verification is performed using the device's unique key pre-stored in the service sub-platform to prevent data tampering; for tasks in normal status and with a low real-time level, only the basic CRC check is performed; when adjusting the transmission strategy based on real-time network quality, if the network packet loss rate is greater than the first preset delay, the industrial data-specific compression algorithm is automatically enabled; if the network latency is greater than the second preset delay, the third computing task is split into the smallest execution unit for fragmented transmission, and the fragment priority is marked.

[0064] Specifically, the sensor network platform receives the second calculation task (high real-time) from welding robot A and performs hierarchical verification: basic CRC check + device unique key signature verification (the key is pre-stored in the service sub-platform) to prevent the current parameters from being tampered with; it detects the real-time network quality, and due to electromagnetic interference in the workshop, the packet loss rate rises to 8% > 5%, so it uses an industrial data-specific compression algorithm (sensor time-series data compression ratio 3:1) to compress the current fluctuation data from 100KB to 33KB; at the same time, the latency rises to 120ms > 100ms, so the task is split into core instructions (current adjustment value) and auxiliary data (historical comparison value), and the core instructions are transmitted in fragments first to ensure that welding robot A receives the adjustment instruction within 200ms.

[0065] In step four, when organizing the third computing task, the sensor network platform needs to embed real-time network quality parameters and data compression / fragmentation identifiers, and synchronously feed these identifiers back to the management platform. If three consecutive verification failures or single task transmission timeouts are detected during transmission, the transmission strategy will be dynamically switched: the backup communication link will be automatically enabled, and the transmission priority of non-urgent tasks will be reduced to reserve bandwidth.

[0066] Specifically, if welding robot A fails to verify its task transmission three times consecutively (due to Wi-Fi interference), in step four, the sensor network platform reports the transmission status to the management platform as a failure - packet loss rate 8% - latency 120ms; triggering a transmission strategy switch, automatically activating the backup industrial Ethernet link (originally using Wi-Fi), and simultaneously reducing the task transmission priority of temperature sensor B (from medium to low), reserving 80% bandwidth for A's task; after the switch, A's task transmission latency drops to 50ms, and the verification passes on the first try. Based on the feedback, the management platform will subsequently prioritize allocating high-priority tasks to edge nodes covered by industrial Ethernet.

[0067] In step five, the load status reported by the edge nodes of the object platform to the management platform includes real-time CPU utilization, memory usage, and task queue length. After receiving the data, the management platform first classifies the edge node status into three levels: normal, slightly overloaded, and severely overloaded based on the load data.

[0068] If the overload is mild, the management platform will increase the weight of the current load rate of the edge node and decrease the weight of logical distance when adjusting the priority coefficient calculation parameters of subsequent tasks, and prioritize the allocation of tasks to nodes with lower load. If the overload is severe, the allocation of new tasks to the node will be suspended, and the weight of physical distance will be increased to guide subsequent tasks to the next closest edge node with normal load. At the same time, the real-time network quality data fed back by the sensor network platform will be called to select the transmission path with the lowest packet loss rate for the diversion tasks.

[0069] Specifically, after welding robot A completes its task, the edge node (welding area edge gateway) in step five provides real-time feedback on the load status: CPU utilization 75% (slight overload, 60% < 75% ≤ 80%), memory usage 65%;

[0070] The management platform adjusts the priority coefficient calculation parameters for subsequent tasks, increasing the weight of the current load rate of edge nodes from 5% to 25% (+20%), and decreasing the weight of logical distance from 25% to 15% (-10%).

[0071] If the load on a subsequent edge node rises to 85% (severe overload), the management platform will suspend the allocation of new tasks to it, increase the weight of physical distance from 25% to 55% (+30%), guide new tasks to a backup edge node 10 meters away but with a load of 30%, and select a transmission path with a packet loss rate of 1% to ensure efficient task diversion.

[0072] The above are merely embodiments of the present invention. The invention is not limited to the fields covered by these embodiments. Commonly known structures and characteristics in the solutions are not described in detail here. Those skilled in the art are aware of all common technical knowledge in the field prior to the application date or priority date, are able to access all existing technologies in that field, and have the ability to apply conventional experimental methods prior to that date. Those skilled in the art can, under the guidance of this application, improve and implement this solution in combination with their own capabilities. Some typical known structures or methods should not be obstacles for those skilled in the art to implement this application. It should be noted that those skilled in the art can make several modifications and improvements without departing from the structure of the present invention. These should also be considered within the scope of protection of the present invention, and will not affect the effectiveness of the implementation of the present invention or the practicality of the patent. The scope of protection claimed in this application should be determined by the content of its claims, and the specific embodiments described in the specification can be used to interpret the content of the claims.

Claims

1. A method for edge computing communication for industrial internet of things, characterized in that, Comprising the following steps: Step one: the user issues a task to the service platform through the user platform, the service platform receives the task request, classifies the task based on the device type and task type associated with the task, determines the real-time level of the task and the importance weight of the associated production link; Step two: the service platform transmits the classified task to the corresponding service sub-platform, the service sub-platform obtains the historical operation data of the task associated device and the basic information of the edge node, and generates a first calculation task carrying the task attribute label according to the classification result, and sends it to the management platform; Step three: the management platform uniformly manages multiple first calculation tasks, calculates the task priority coefficient by using a balance algorithm that integrates multiple parameters, the balance algorithm determines the priority coefficient by comprehensively evaluating the following factors by weighted summation: the logical distance between the task associated device and the service sub-platform, the physical distance between the device and the edge computing area center, the weight corresponding to the task real-time level, and the current load rate of the edge node; then, according to the priority coefficient, the task is sorted, a second calculation task containing a calculation serial number is generated and sent to the sensing network platform; Step four: after the edge computing area center of the sensing network platform receives the second calculation task, it first verifies the data integrity, then adjusts the transmission strategy based on the real-time network quality, sorts it into an independent third calculation task and sends it to the object platform; Step five: the edge node of the object platform executes the third calculation task according to the calculation serial number, and simultaneously feeds back the load state of itself to the management platform in real time; if the load exceeds the preset threshold, the management platform dynamically adjusts the priority coefficient calculation parameters of the subsequent tasks according to the load state feedback of the object platform; The balance algorithm dynamically adjusts the logical distance between the task associated device and the service sub-platform, the physical distance between the device and the edge computing area center, the weight corresponding to the task real-time level, and the weight proportion of the current load rate of the edge node according to the real-time state of the industrial scene, and the sum of each weight coefficient is 1; in step one, the service platform first obtains the real-time running parameters of the task associated device before classifying and marking the task, the real-time running parameters include the current load rate, temperature fluctuation value and fault warning code of the device, and the device state is divided into three levels of normal, warning and emergency according to the preset parameter threshold interval; the service platform matches the device state level with the task type by matrix, and corrects the real-time level determined initially by weighting; at the same time, the service platform also obtains the difference between the planned completion time of the task associated production link and the current progress, and the order delivery cycle corresponding to the production link, according to the rules that the greater the progress difference, the higher the importance weight, and the importance weight is increased when the delivery cycle is less than or equal to the preset warning, the importance weight of the associated production link is dynamically calibrated. 2.The edge computing communication method for industrial Internet of Things according to claim 1, characterized in that, In the second step, the service platform synchronously calls the stored fault handling records, average fault-free running time and historical load peak value data of the edge node within the preset time period of the task-associated device when obtaining the historical running data of the device, and embeds the real-time level dynamically corrected in the first step and the importance weight after secondary calibration into the associated field of the historical data; the attribute label generated by the service platform includes task priority identification based on real-time level, resource demand threshold based on importance weight and historical load peak of edge node, data sampling frequency; at the same time, the service platform performs double verification on the generated first computing task, first verifies the task legality by matching the device unique code and timestamp, and then ensures the data transmission integrity by CRC redundancy check, and sends the task to the management platform after the verification is passed. 3.The edge computing communication method for industrial internet of things according to claim 2, characterized in that, In the third step, the management platform analyzes the attribute label of the first computing task before calculating the priority coefficient by using the balancing algorithm, if the device state level in the label is urgent and the resource demand threshold is greater than or equal to the preset high value, the weight proportion corresponding to the real-time level of the task is increased when weighted summation is performed, and the weight proportion of the current load rate of the edge node is reduced; if the device state level is normal and the resource demand threshold is less than the preset low value, the weight proportions of the logical distance and the physical distance are increased to preferentially allocate to the edge node closer.

4. The edge computing communication method for industrial internet of things according to claim 3, characterized in that, In the third step, when the management platform uniformly manages the first computing task, if the task-associated device is a mobile industrial device, the device historical motion trajectory data provided by the service platform is called to calculate the position prediction value in the future preset time period, and the distance between the prediction value and the center of the edge computing area is taken as the dynamic correction value of the physical distance; if it is a static device, the average value of the historical physical distance is taken as the input value of the physical distance.

5. The edge computing communication method for industrial internet of things according to claim 4, characterized in that, In the third step, when the management platform sorts the tasks according to the priority coefficients, if the priority coefficient difference of two or more tasks is less than or equal to a preset threshold, a secondary sorting mechanism is triggered: first, compare the importance weights of the task-associated production links, and sort the tasks with higher weights first; if the weights are the same, the historical fault handling records of the task-associated devices are called, and the tasks corresponding to the devices with more faults in the recent preset time are sorted first; at the same time, the management platform performs edge node load pre-verification on the generated second computing task, if the total resource demand of the tasks to be allocated to a certain edge node exceeds 80% of its current remaining capacity, the priority coefficients of the tasks exceeding the remaining capacity are attenuated in proportion to the remaining capacity and reordered.

6. The edge computing communication method for industrial internet of things according to claim 5, characterized in that, In the fourth step, when the edge computing area center of the sensor network platform verifies the data integrity, a hierarchical verification mechanism is adopted, for the tasks with urgent device state level or high real-time level, in addition to the basic CRC verification, the device unique key pre-stored by the service platform is additionally used for data signature verification to prevent data tampering; for the tasks in normal state and with low real-time level, only the basic CRC verification is performed; when the transmission strategy is adjusted based on the real-time network quality, if the network packet loss rate is greater than a first preset delay, an industrial data special compression algorithm is automatically enabled; If the network delay is greater than the second preset delay, the third computing task is split into the minimum execution unit for fragmented transmission, and the fragments are marked with priority.

7. The edge computing communication method for industrial internet of things according to claim 6, characterized in that, In the fourth step, the sensor network platform needs to embed real-time network quality parameters and data compression / fragmentation identification when arranging the third computing task, and synchronize the identification to the management platform for feedback. If three consecutive verification failures or single task transmission timeout are monitored during the transmission process, the transmission strategy dynamic switching is triggered: the standby communication link is automatically enabled, and the transmission priority of non-urgent tasks is reduced to reserve bandwidth.

8. The edge computing communication method for industrial internet of things according to claim 7, characterized in that, In the fifth step, the load state feedback of the edge node of the object platform to the management platform includes real-time CPU utilization, memory occupancy and task queue length; after receiving, the management platform first divides the edge node state into three levels of normal, mild overload and serious overload according to the load data; If it is mild overload, the management platform increases the weight proportion of the current load rate of the edge node when adjusting the priority coefficient calculation parameters of the subsequent tasks, and reduces the logical distance weight, preferentially assigns the tasks to the nodes with lower load; if it is serious overload, it suspends the allocation of new tasks to the node, and increases the physical distance weight, guides the subsequent tasks to the edge node with the next closest distance but normal load, and calls the real-time network quality data feedback by the sensor network platform to select the transmission path with the lowest packet loss rate for task shunting.

9. An edge computing communication system for industrial internet of things, characterized by, The method of any one of claims 1-8 is used.

Images