Multi-platform concurrent transmission edge computing data acquisition method

Abstract

The invention specifically relates to the technical field of edge computing, and discloses a multi-platform concurrent transmission edge computing data acquisition method, which comprises the following steps: S1, concurrent access of multi-platform equipment: establishing a mapping relationship between equipment identity and a communication link; s2, data concurrent collection and preprocessing: obtaining a preprocessed data stream; s3, concurrent transmission priority scheduling: allocating a transmission bandwidth for each data stream; s4, edge node parallel computing: distributing the subtasks to a plurality of computing nodes; s5, cloud collaboration and global optimization: generating a global optimization strategy; s6, carrying out full-link monitoring and dynamic regulation and control; concurrent access of heterogeneous equipment is realized by generating data streams in a uniform format, multi-platform compatibility is effectively enhanced, bandwidth is dynamically allocated to the data streams according to priority weight indexes by calculating the priority weight indexes of the data streams, data transmission efficiency is improved, and the data transmission efficiency is improved by allocating sub-tasks to a plurality of computing nodes. And the resource utilization rate of the edge computing gateway is improved.

Description

Technical Field

[0001] The present invention relates to the field of edge computing technology, and more specifically, to an edge computing data acquisition method for multi-platform concurrent transmission. Background Art

[0002] With the popularization of Internet of Things technology, a large number of sensors and smart devices are connected to the network, generating a large amount of data. The traditional cloud computing model faces the problems of high latency and bandwidth pressure. Emerging applications have extremely high requirements for the timeliness of data processing. Uploading a large amount of data to the cloud will cause network congestion, affecting the stability and efficiency of data transmission. With the development of edge computing, by sinking computing resources to the edge of the network, local data processing can be achieved, which helps to protect data security and reduce data transmission volume and latency. Existing edge computing data collection methods include device connection and configuration, data collection, data filtering and preprocessing, local data analysis and decision-making, data transmission and storage, security and privacy protection, as well as management and monitoring. Local preprocessing and analysis significantly reduce latency and alleviate the pressure on cloud bandwidth.

[0003] However, in actual use, it still has some shortcomings. First, it lacks multi-platform compatibility. The device connection and configuration technology of existing edge computing data collection methods are mostly targeted at a single type of device, with weak data format adaptability, making it difficult to cope with concurrent access of heterogeneous devices.

[0004] Second, data transmission efficiency is low. Existing edge computing data collection methods mostly use single-link serial transmission, which is prone to transmission congestion when data is generated concurrently on multiple platforms. The transmission scheduling strategy is not optimized for concurrent scenarios and lacks dynamic bandwidth allocation and traffic priority management capabilities, making it unable to meet the real-time requirements of high-concurrency scenarios.

[0005] Third, the resource utilization rate of edge computing nodes is low. The local data analysis and decision-making modules of existing edge computing data collection methods mostly adopt a single-task exclusive resource mode. The computing and storage resources of edge nodes are difficult to dynamically allocate among multi-platform data, and the resource management mechanism lacks the ability to coordinate the scheduling of concurrent tasks on multiple platforms, resulting in low resource utilization rate of edge computing nodes. Summary of the Invention

[0006] In view of this, an embodiment of the present invention provides an edge computing data collection method for concurrent transmission on multiple platforms. By generating data streams in a unified format, concurrent access of heterogeneous devices is achieved. By calculating the priority weight index of each data stream, bandwidth is dynamically allocated to the data stream according to the priority weight index. By allocating subtasks to multiple computing nodes, efficient execution of multiple subtasks is ensured, effectively solving the problems of insufficient multi-platform compatibility, low data transmission efficiency and low resource utilization of edge computing nodes raised in the background technology.

[0007] To achieve the above object, the present application provides the following technical solutions: a multi-platform concurrent transmission edge computing data acquisition method, comprising a data acquisition terminal, a relay terminal, an edge computing gateway and a cloud server, each terminal is bidirectionally connected, and the specific steps are as follows:

[0008] S1: Multi-platform device concurrent access: a protocol adaptation layer is constructed through the data acquisition terminal, a unified interface format is unified, the relay terminal synchronously maintains a device registration table, a unique logical address is allocated for the access device, and a mapping relationship between the device identity and the communication link is established;

[0009] S2: Data concurrent acquisition and preprocessing: based on the device logical address, the data acquisition terminal acquires multi-device data through a distributed concurrent sampling mechanism, and performs abnormal detection and truncation, and delivers the data to the relay terminal for data merging and compression to obtain a preprocessed data stream;

[0010] S3: Concurrent transmission priority scheduling: the relay terminal calculates the priority weight index of each data stream through a preset weight model, the edge computing gateway allocates transmission bandwidth for each data stream according to the priority weight index, and feeds back the scheduling result to the relay terminal;

[0011] S4: Edge node parallel computing: the data stream attribute triggers a preset rule engine to generate a computing task, the edge computing gateway decomposes the computing task into parallel subtasks based on the data type, and distributes the parallel subtasks to multiple computing nodes through a load balancing algorithm;

[0012] S5: Cloud coordination and global optimization: the cloud server receives the concurrent data streams transmitted by multiple edge computing gateways through a distributed message queue, generates a global optimization strategy through a preset AI model, generates and issues an optimization control instruction;

[0013] S6: Full-link monitoring and dynamic regulation: each terminal acquires its own running parameters in real time, uploads the running parameters to the edge computing gateway through a unified protocol, calculates a full-link running state index, judges the full-link running state level, and executes a dynamic regulation strategy.

[0014] The technical effects and advantages of the present application are as follows:

[0015] 1. The present application pre-integrates mainstream device communication protocols through a unified protocol adaptation layer, the edge gateway automatically identifies the protocol type of the access device and completes protocol conversion, and pre-processes the real-time acquired data to generate a unified format data stream, realizes concurrent access of heterogeneous devices, and effectively enhances multi-platform compatibility;

[0016] 2. The present invention distributes concurrent data to multiple physical links based on data flow attributes through the edge computing gateway, calculates the priority weight index of each data flow, dynamically allocates bandwidth to the data flow according to the priority weight index, reserves bandwidth resources for high-priority data, ensures that critical data is transmitted first, improves the efficiency of data transmission, and meets the real-time requirements of high-concurrency scenarios;

[0017] 3. The present invention decomposes the computing task into multiple subtasks through the edge computing gateway, and calculates the node performance weight based on the resource status of each computing node. Combined with the priority weight index of the computing task and the subtask type, the subtask is distributed to the corresponding computing node to avoid idle resources, ensure the efficient execution of multiple subtasks, and improve the resource utilization of the edge computing gateway. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 Schematic diagram of the method steps of the present invention.

[0019] Figure 2 It is a schematic diagram of the overall structure of the present invention.

[0020] Figure 3 Schematic diagram of the steps for determining the full link operation status level of the present invention. DETAILED DESCRIPTION

[0021] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0022] As attached Figure 1 The edge computing data collection method shown in the figure is a multi-platform concurrent transmission method, including a data collection terminal, a relay terminal, an edge computing gateway and a cloud server, and each terminal is bidirectionally connected.

[0023] In a more specific application of the present invention, it needs to be specifically explained that the data acquisition terminal is used to collect multi-source heterogeneous data in industry. Each data acquisition terminal is used for a unique device ID. The data acquisition terminal is set with a communication agent, which can abstract Bluetooth, RFID, WIFI, Ethernet, serial port and 4 / 5G multiple communication methods into a uniform resource locator. The data acquisition terminal accesses devices with different communication protocols through the communication agent. The data acquisition terminal can be connected to the relay terminal via industrial Ethernet or Bluetooth communication to transmit the collected data to the relay terminal.

[0024] The relay terminal is autonomously elected by a plurality of data collection terminals, the resource usage of the relay terminal is good and the relay terminal is close to the edge computing gateway and has less delay, is used for merging the same data of the plurality of data collection terminals, reducing the data transmission amount, and uniformly converting the data of different sources into a protocol supported by the edge gateway, the relay terminal is connected with the edge computing gateway through an Ethernet / 5G network, and compressed data is transmitted.

[0025] The edge computing gateway is used for converting an edge gateway protocol of MQTT or CoAP into an HTTP / HTTPS protocol, realizing cross-layer communication, scheduling data based on priority, realizing multi-path transmission, and executing a predefined rule to reduce the transmission of data to a cloud server, the edge computing gateway is connected with the cloud server through the Internet, supports breakpoint continuation and load balancing, and long-term stores historical data.

[0026] The cloud server is used for storing full historical data, performing deep analysis based on the historical data, predicting device failure, making a global optimization decision, and issuing a configuration update, and remotely monitoring the state of the edge node.

[0027] The connection mode of the data collection terminal, the relay terminal, the edge computing gateway and the cloud server is described in Figure 2 .

[0028] The specific embodiment of the application includes the following steps:

[0029] S1: Multi-platform device concurrent access: a protocol adaptation layer is constructed through the data collection terminal, a unified interface format is constructed, the relay terminal synchronously maintains a device registration table, a unique logical address is allocated for the access device, and a mapping relationship between the device identity and the communication link is established;

[0030] Further, the protocol adaptation layer includes a protocol analysis layer, a format conversion layer and an interface abstraction layer three-layer architecture, the protocol analysis layer includes protocol identification and access, distributed parser deployment and original data extraction, the format conversion layer includes data model unification, type mapping and cleaning and compression and encryption preprocessing, and the interface abstraction layer includes URL template definition, interface parameter adaptation and interface registration and routing.

[0031] In this embodiment, it needs to be specifically pointed out that the protocol identification and access in the protocol analysis layer means that the initial communication data of the access device is matched with the preset protocol feature library, the underlying protocol adopted by the device is automatically identified, the distributed parser deployment means that the corresponding lightweight parser is built in the data collection terminal for different protocol types, and the original data extraction is stripping the protocol redundant information through the parser, extracting the core business data, and converting it into structured data.

[0032] The data model unification in the format conversion layer refers to defining a general data element model, including device identification, data type, timestamp, data value, unit, and integrity identification, to ensure consistent data format. Type mapping and cleaning refer to type conversion and cleaning of abnormal data based on the parsed structured data according to the data element model. Compression and encryption preprocessing refers to lightweight compression of standardized data and encryption of data to ensure the security of data transmission within the adaptation layer.

[0033] The URL template definition in the interface abstraction layer refers to designing a standardized URL interface format. Interface parameter adaptation refers to mapping control instructions of different protocols into query parameters of a unified URL to achieve standardized control of devices through URL parameters. Interface registration and routing refer to the interface abstraction layer binding the generated unified URL with device logical addresses and original protocol parsers to form an interface-device mapping table. When the upper layer calls the URL, it is automatically routed to the corresponding parser and device.

[0034] Furthermore, the device registry adopts a distributed architecture, supporting real-time updates and fast queries. The maintenance of the device registry includes online status updates, information change synchronization, and offline device cleaning.

[0035] In this embodiment, it needs to be specifically explained that the online status update in the device registry refers to receiving the heartbeat packet sent by the device every 30 seconds, and marking the device that has not responded for more than 5 minutes as offline. Information change synchronization refers to automatically updating the corresponding device information in the registry and synchronously adjusting the mapping relationship when the device has attribute changes. Offline device cleaning refers to automatically migrating the information of the device that is marked as offline and has not reconnected for more than 24 hours to the historical registry for archiving, and releasing the logical address resource.

[0036] After the relay terminal receives the device access request sent by the data collection terminal, it starts the detection mechanism, triggers the registry registration process through the verified device, automatically collects and stores the device core information to the temporary buffer, forms the to-be-registered record, and generates a globally unique logical address based on the preset coding rule.

[0037] In this embodiment, it needs to be specifically explained that the detection mechanism refers to detecting whether the communication link between the device and the relay terminal is smooth through the device physical address feedback by the data collection terminal, calling the interface of the protocol adaptation layer to verify whether the device communication protocol has completed standardized conversion, and checking the identity identification of the device and the pre-stored device whitelist to reject unauthorized device access. The to-be-registered record should include basic attributes, communication attributes, and state attributes.

[0038] The mapping relationship of device identity and communication link includes a logical address-physical address mapping, a logical address-communication link mapping, and a logical address-device attribute mapping, and the relay terminal stores the three-layer mapping relationship in a distributed hash table.

[0039] In this embodiment, it needs to be specifically explained that the logical address-physical address mapping is used to record the correspondence between the logical address and the original physical address of the device, which is the bottom-layer routing basis during data forwarding; the logical address-communication link mapping is used to associate the logical address with the uniform URL interface generated by the protocol adaptation layer, and to explicitly determine the communication path of the data collection terminal and the device; and the logical address-device attribute mapping is used to bind the attribute information of the device type and the sampling frequency, and to provide basic data for subsequent data collection priority judgment.

[0040] S2: data concurrent collection and preprocessing: based on the device logical address, the data collection terminal collects multi-device data through a distributed concurrent sampling mechanism, and performs abnormality detection and truncation, and delivers the preprocessed data stream to the relay terminal for data merging and compression;

[0041] Further, the distributed concurrent sampling mechanism refers to constructing a concurrent sampling framework based on the device logical address, allocating a sampling task, and converting the logical address into physical communication parameters according to the mapping relationship of the device identity and the communication link, and concurrently accessing multiple device interfaces to collect multi-device data by using an asynchronous I / O model;

[0042] In this embodiment, it needs to be specifically explained that the allocation of the sampling task needs the data collection terminal to decompose the collection task into independent threads or coroutines according to the device type and the importance level, to execute the threads in parallel through a thread pool, and to collect multi-device data by using an asynchronous I / O model, so as to release thread resources when waiting for a device response and to improve concurrent efficiency.

[0043] Threshold detection and time sequence consistency detection are performed on the collected data through an abnormality detection algorithm, single-point abnormality and continuous abnormality are intelligently truncated, and a quality label is added to each data, the data collection terminal encapsulates the preprocessed data into a unified format, and delivers the data to the relay terminal through an internal message queue;

[0044] In this embodiment, it needs to be specifically explained that the threshold detection specifically refers to presetting a normal range for each device data item, and marking the data points exceeding the range as abnormal, and the time sequence consistency detection refers to predicting a data trend through a Kalman filter, and triggering an abnormality mark for the data points deviating from the predicted value; for single-point abnormality, the previous value+linear interpolation is used to replace the abnormal data point, and for continuous abnormality, the data stream needs to be truncated, the abnormal start time is recorded, and an alarm is sent to the edge computing gateway.

[0045] The relay terminal groups data based on logical addresses, aligns data in the same group by time stamp sliding window, merges multi-device parameters in the window into the same batch, and performs data compression.

[0046] In this embodiment, it needs to be specifically pointed out that the relay terminal divides data into logical groups according to the area identifier and type code of the device logical address, and the data compression specifically refers to calculating the difference value of adjacent points for time series data, saving only the difference value, establishing a dictionary mapping for device state data, converting a string into a single-byte integer, and performing block-level compression on the merged batch data.

[0047] The data stream refers to various data sets delivered in the process of multi-platform concurrent transmission of device data collection, and the data stream runs through device access, collection, transmission, processing, and cooperation of the whole link, and is the core data carrier of interaction between terminals. The content of the data stream changes constantly as the link advances.

[0048] S3: Concurrent transmission priority scheduling: The relay terminal calculates the priority weight index of each data stream through a preset weight model, the edge computing gateway allocates transmission bandwidth to each data stream according to the priority weight index, and feeds back the scheduling result to the relay terminal;

[0049] Further, the calculation of the priority weight index needs to extract multi-dimensional priority parameters in each data stream by the relay terminal, specifically including data urgency D1, time sensitivity D2, data integrity D3, device importance D4, and link stability D5, using the formula:

[0050]

[0051] The priority weight index P corresponding to each data stream is calculated, wherein α1, α2, α3, α4, and α5 represent the weight coefficients of data urgency, time sensitivity, data integrity, device importance, and link stability respectively, and K is a normalization coefficient.

[0052] In this embodiment, it needs to be specifically pointed out that the data urgency is based on a preset rule of data type, classifies the business scenarios to which the data belongs, assigns values to the urgency of the data stream, and the value range is 1-5. When an abnormal event is detected, the data urgency of the associated data stream is automatically increased to 5;

[0053] The time sensitivity is set according to the timeliness requirement and transmission delay threshold of the data in the data stream, and the value range is 1-4. For example, when a control instruction needs to be delivered within 10 seconds, otherwise it will be invalid, the time sensitivity D2 of the associated data stream is 3, and when the preset delay threshold is less than 1 second, the time sensitivity D2 of the associated data stream is 4;

[0054] Data integrity requires referencing the quality labels added to the data during concurrent data collection and preprocessing to assign a value to the data integrity of the associated data stream. The value range is 1-3. For example, the data integrity of complete data is D3=3, the data integrity of interpolated and repaired data is D3=2, and the data integrity of truncated data is D3=1.

[0055] The importance of equipment needs to be assigned to the associated data flow based on the core degree of the equipment's physical role. The value range is 1-5, and the impact weight is dynamically adjusted based on historical fault data. The equipment importance of core control equipment is D4=5, the equipment importance of key execution equipment is D4=4, the equipment importance of auxiliary monitoring equipment is D4=2 / 3, and the equipment importance of redundant standby equipment is D4=1;

[0056] Link stability requires real-time monitoring of link status data, calculating the packet loss rate, establishing a mapping relationship between packet loss rate and link stability, and assigning a link stability value to the associated data flow. The value range is 1-3. Link stability D5 = 3 for a packet loss rate less than 1%, D5 = 2 for a packet loss rate greater than 1% and less than 10%, and D5 = 1 for a packet loss rate greater than 10%.

[0057] The setting of weight coefficients requires determining the weight coefficients corresponding to each parameter through the AHP hierarchical analysis method. The sum of the weights is 1, and a dynamic calibration mechanism is set. When a certain type of parameter fluctuates abnormally, the weight adjustment is automatically triggered. The normalization coefficient K is to ensure that the value range of the priority weight index is 0-100. The maximum value calibration method is used to take the maximum value of data urgency, time sensitivity, data integrity, equipment importance and link stability, and the normalization coefficient K is calculated using the formula K = 100 / P.

[0058] Furthermore, the allocation of transmission bandwidth requires the edge computing gateway to build a mapping relationship between the priority weight index of the data flow and the bandwidth, monitor the available network bandwidth of the current edge computing gateway in real time, use 80% of the network bandwidth as the allocable network bandwidth, divide the priority corresponding to the data flow based on the priority weight index of the data flow, and allocate transmission bandwidth according to the priority of the data flow. The edge computing node will feed back the bandwidth allocation results to the relay terminal in real time, and the relay terminal will optimize the subsequent data flow processing based on the feedback results.

[0059] In this embodiment, it should be specifically explained that the priority of data flows is divided into high priority, medium priority and low priority. A data flow with a priority weight index greater than 70 is a high priority and is allocated 50% of the allocatable bandwidth. A data flow with a priority weight index greater than 30 and less than 70 is a medium priority and is allocated 30% of the allocatable bandwidth. A data flow with a priority weight index less than 30 is a low priority and is allocated 20% of the allocatable bandwidth. When the bandwidth demand of a high-priority data flow exceeds the quota, bandwidth is preempted from the low-priority data flow first, and the preemption record is marked. The WFQ algorithm is used to manage the transmission queue, and time slices are allocated to each data flow according to the weight index to ensure that the average delay of the high-priority data flow is controlled within 100ms.

[0060] S4: Edge node parallel computing: Data stream attributes trigger the preset rule engine to generate computing tasks. The edge computing gateway decomposes the computing tasks into parallel subtasks based on the data type and distributes them to multiple computing nodes through a load balancing algorithm.

[0061] Furthermore, the generation of computing tasks requires the edge computing gateway to extract the core attributes of the received data stream, including data type, data size, timestamp density and priority weight index, trigger the preset rule engine, and map the generated computing tasks. The edge computing gateway decomposes the computing tasks into subtasks according to the task type and data characteristics. The edge computing gateway calculates the node performance weight by real-time monitoring of the resource status of each computing node, and assigns the subtask execution priority based on the priority weight index of the computing task and the subtask type. Each subtask is assigned to the computing node for execution, and the execution progress is reported to the edge computing gateway in real time.

[0062] In this embodiment, the generation of computing tasks needs to be specifically explained. For example, when the data type is alarm data and P is greater than 80, an abnormality root cause location task is generated. When the data size is greater than 500KB and P is less than 30, a historical data compression and archiving task is generated. The rule engine assigns a unique ID to each computing task and defines the task type, input data address, output result format, and maximum execution time.

[0063] Decomposing computing tasks into subtasks includes data sharding and functional splitting. Data sharding refers to sharding multiple continuous time series data by time windows. Each subtask processes the time slice independently, and there is no dependency between subtasks. Functional splitting refers to splitting complex tasks into parallel subtasks according to functional modules. After the subtasks are decomposed, the edge computing gateway should mark the order of the subtasks.

[0064] The calculation of the node performance weight is performed by obtaining the CPU usage rate η1, the memory occupancy rate η2 and the historical task completion rate η3 of each node of the edge computing gateway, and calculating the node performance weight W of each computing node by using the formula W = β1(1-η1) + β2(1-η2) + β3η3, wherein β1, β2 and β3 represent the weight coefficients of the CPU usage rate, the memory occupancy rate and the historical task completion rate, and the sum is 1.

[0065] In the process of subtask allocation, a dynamic allocation strategy needs to be followed. Specifically, for subtasks without dependency, they are allocated in proportion to the node performance weight; for subtasks with dependency, they are allocated to the same node as the previous subtask. If it is monitored in real time that the CPU usage rate of the node is greater than 85%, it is automatically excluded from the redistribution range to avoid node overload.

[0066] S5: Cloud collaboration and global optimization: The cloud server receives the concurrent data stream transmitted by multiple edge computing gateways through the distributed message queue, generates a global optimization strategy through a preset AI model, generates and issues an optimization control instruction;

[0067] Further, the generation of the global optimization strategy needs to pass through the multi-level AI model architecture of the real-time monitoring model, the offline training model and the reinforcement learning model preset by the cloud server, generate the global optimization strategy based on the constraint condition, the strategy dimension and the strategy priority, and the cloud server generates a standardized optimization control instruction according to the global optimization strategy, issues it to the edge computing gateway, and returns the execution result to the cloud server after the edge computing gateway executes the optimization control instruction.

[0068] In this embodiment, it needs to be specifically pointed out that the access of the concurrent data stream needs the edge computing gateway to encapsulate the processed data in a unified format, push it to the Kafka cluster through the TCP protocol, and the partitioner of the message queue performs hash calculation according to the ID of each computing node in the edge computing gateway, routes the data stream of the same node to a fixed partition.

[0069] The constraint condition includes resource constraint, business constraint and cost constraint, the strategy dimension includes edge resource scheduling strategy, device sampling optimization strategy and transmission parameter adjustment strategy, and the strategy priority order is, for example, the priority of the economic repair strategy is greater than that of the performance optimization strategy, and the priority of the performance optimization strategy is greater than that of the resource expansion strategy.

[0070] S6: Full-link monitoring and dynamic regulation: Each terminal collects its own running parameters in real time, uploads them to the edge computing gateway through a unified protocol, calculates the full-link running state index, judges the full-link running state level, and executes the dynamic regulation strategy.

[0071] Further, the calculation of the full-link running state index includes the following steps:

[0072] A1: Set a time window and collect the real-time data transmission delay T of each terminal within the time window in real time. i and data transmission packet loss rate L i ;

[0073] In this embodiment, it should be specifically explained that the real-time data transmission delay refers to the one-way time from the generation of terminal data to the transmission to the next node, which can directly reflect the real-time response capability of the link. Delay spikes indicate network congestion, node failure or signal interference; the data transmission packet loss rate refers to the proportion of data packets sent by the terminal that are not successfully received, which can directly reflect the reliability of the link. A high packet loss rate will lead to a decrease in data integrity and increase the delay.

[0074] A2: Standardize the parameters of each terminal collected in real time to obtain the standardized real-time data transmission delay T Bi and data transmission packet loss rate L Bi , and by the formula:

[0075]

[0076] Calculate the operating status index S of the i-th terminal i ;

[0077] In this embodiment, it should be specifically explained that the standardization of the real-time data transmission delay is to map the delay to the interval [0,1], and use the reverse mapping formula The standardized real-time data transmission delay T Bi , k1 is the adjustment coefficient, and the standardization of the data transmission packet loss rate is to map it to the interval [0,1], and the reverse mapping is used through the formula Get the standardized data transmission packet loss rate L Bi , k2 is the adjustment coefficient.

[0078] A3: The operating status indexes of n terminals are calculated using the formula The full link operation status index S is calculated, where ω i represents the weight coefficient of the i-th terminal, and m is the nonlinear adjustment factor.

[0079] In this embodiment, it should be specifically noted that the setting of the weight coefficient of each terminal should be obtained based on historical data analysis. The nonlinear adjustment factor is usually taken as 2. The power function is used to amplify the contribution of terminals with larger operating status indexes to the overall link index, weaken the negative impact of terminals with smaller operating status indexes, and avoid the failure of a single terminal excessively lowering the value of the overall link operating status index.

[0080] Further, the full-link operation state level includes a first state level, a second state level, a third state level and a fourth state level, the full-link operation state level is determined according to the full-link operation state index, the graded dynamic regulation strategy is formulated based on the full-link operation state level, and the parameters after the strategy is executed are collected in real time to optimize the regulation strategy.

[0081] In the embodiment, it is specifically pointed out that, as shown in Figure 3 The judgment step of the full-link operation state level is as follows:

[0082] B1: when the full-link operation state index is greater than 80 and less than 100, the first state level is in, which indicates that the full-link parameter is abnormal, and there is no abnormal index;

[0083] B2: when the full-link operation state index is greater than 60 and less than 80, the second state level is in, which indicates that the local parameter slightly fluctuates and does not affect the core function;

[0084] B3: when the full-link operation state index is greater than 40 and less than 60, the third state level is in, which indicates that the key index is abnormal and affects data transmission or calculation;

[0085] B4: when the full-link operation state index is less than 40, the fourth state level is in, which indicates that the full-link appears a serious bottleneck and faces the risk of function interruption.

[0086] The formulation of the graded dynamic regulation strategy is that, for the first state level, the current configuration is maintained, and preventive monitoring is started; for the second state level, local fine tuning and transmission optimization are performed to reduce bandwidth occupation; for the third state level, the edge computing gateway migrates the sub-tasks of the overloaded node to the idle node and releases the bandwidth resource; and for the fourth state level, the data transmission of non-key equipment is cut off, the core equipment is preferentially guaranteed, the edge computing gateway suspends the non-real-time task calculation, the cloud starts global resource scheduling, and alarm information is pushed to the relevant staff.

[0087] Secondly, in the drawings of the embodiment, only the structures related to the embodiment are involved, other structures can be referred to the general design, and in the case of no conflict, the same embodiment and different embodiments of the application can be combined with each other;

[0088] Finally, the above only describes the preferred embodiments of the application and is not used to limit the application, and any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the application shall be included in the protection scope of the application.

Claims

1. A method for collecting edge computing data with concurrent transmission on multiple platforms, characterized in that: It includes data collection terminals, relay terminals, edge computing gateways, and cloud servers. Each terminal is bidirectionally connected. The specific steps are as follows: S1: Concurrent access of multi-platform devices: Build a protocol adaptation layer through the data acquisition terminal, unify the interface format, and synchronize the maintenance of the device registry by the relay terminal. It assigns a unique logical address to the access device and establishes a mapping relationship between the device identity and the communication link. S2: Concurrent data collection and preprocessing: Based on the device logical address, the data collection terminal collects data from multiple devices through a distributed concurrent sampling mechanism, performs anomaly detection and truncation, and transmits the data to the relay terminal for data merging and compression to obtain the preprocessed data stream; S3: Concurrent transmission priority scheduling: The relay terminal calculates the priority weight index of each data flow through a preset weight model. The edge computing gateway allocates transmission bandwidth to each data flow based on the priority weight index and feeds back the scheduling result to the relay terminal. S4: Edge node parallel computing: Data stream attributes trigger the preset rule engine to generate computing tasks. The edge computing gateway decomposes the computing tasks into parallel subtasks based on the data type and distributes them to multiple computing nodes through a load balancing algorithm. S5: Cloud collaboration and global optimization: The cloud server receives concurrent data streams from multiple edge computing gateways through a distributed message queue, generates a global optimization strategy based on a preset AI model, and generates and issues optimization control instructions. S6: Full-link monitoring and dynamic control: Each terminal collects its own operating parameters in real time, uploads them to the edge computing gateway through a unified protocol, calculates the full-link operating status index, determines the full-link operating status level, and executes dynamic control strategies.

2. The edge computing data collection method for multi-platform concurrent transmission according to claim 1 is characterized in that: The protocol adaptation layer includes a three-layer architecture of protocol parsing layer, format conversion layer and interface abstraction layer. The protocol parsing layer includes protocol identification and access, distributed parser deployment and raw data extraction. The format conversion layer includes data model unification, type mapping and cleaning, and compression and encryption preprocessing. The interface abstraction layer includes URL template definition, interface parameter adaptation, and interface registration and routing.

3. The edge computing data collection method for multi-platform concurrent transmission according to claim 1 is characterized in that: The device registry adopts a distributed architecture, supports real-time updates and fast queries, and the maintenance of the device registry includes online status updates, information change synchronization, and offline device cleanup; After receiving the device access request from the data collection terminal, the relay terminal starts the detection mechanism. The verified device triggers the registration process. The relay terminal automatically collects and stores the device core information in a temporary buffer to form a record to be registered, and generates a globally unique logical address based on the preset encoding rules. The mapping relationship between device identity and communication link includes logical address-physical address mapping, logical address-communication link mapping and logical address-device attribute mapping. The relay terminal stores the three-layer mapping relationship in a distributed hash table.

4. The edge computing data collection method for multi-platform concurrent transmission according to claim 1 is characterized in that: The distributed concurrent sampling mechanism refers to building a concurrent sampling framework based on the device logical address, assigning sampling tasks, converting the logical address into physical communication parameters based on the mapping relationship between the device identity and the communication link, concurrently accessing multiple device interfaces, and using an asynchronous I / O model to collect multi-device data; The anomaly detection algorithm performs threshold detection and time series consistency detection during data collection, intelligently cuts off single-point anomalies and continuous anomalies, and adds quality labels to each data. The data collection terminal encapsulates the pre-processed data into a unified format and transmits it to the relay terminal through the internal message queue. The relay terminal groups data based on logical addresses, performs sliding window alignment on data in the same group according to timestamps, merges parameters of multiple devices in the window into the same batch, and performs data compression.

5. The edge computing data collection method for multi-platform concurrent transmission according to claim 1 is characterized in that: The calculation of the priority weight index requires extracting multi-dimensional priority parameters from each data stream through the relay terminal, specifically including data urgency D1, time sensitivity D2, data integrity D3, device importance D4, and link stability D5, using the formula: The priority weight index P corresponding to each data flow is calculated, where α1, α2, α3, α4, and α5 represent the weight coefficients of data urgency, time sensitivity, data integrity, device importance, and link stability, respectively, and K is the normalization coefficient.

6. The edge computing data collection method for multi-platform concurrent transmission according to claim 1 is characterized in that: The allocation of transmission bandwidth requires the edge computing gateway to construct a mapping relationship between the priority weight index of the data flow and the bandwidth, monitor the available network bandwidth of the current edge computing gateway in real time, use 80% of the network bandwidth as the allocable network bandwidth, divide the priority corresponding to the data flow based on the priority weight index of the data flow, and allocate transmission bandwidth according to the priority of the data flow. The edge computing node will feed back the bandwidth allocation result to the relay terminal in real time, and the relay terminal will optimize the subsequent data flow processing based on the feedback result.

7. The edge computing data collection method for multi-platform concurrent transmission according to claim 1 is characterized in that: The generation of the computing task requires the edge computing gateway to extract the core attributes of the received data stream, including data type, data scale, timestamp density and priority weight index, trigger the preset rule engine, map and generate computing tasks, and the edge computing gateway decomposes the computing task into subtasks according to the task type and data characteristics. The edge computing gateway calculates the node performance weight by monitoring the resource status of each computing node in real time, and assigns the subtask execution priority based on the priority weight index of the computing task and the subtask type. Each subtask is assigned to the computing node for execution, and the execution progress is reported to the edge computing gateway in real time.

8. The edge computing data collection method for multi-platform concurrent transmission according to claim 1 is characterized in that: The generation of the global optimization strategy requires a multi-level AI model architecture of a real-time monitoring model, an offline training model, and a reinforcement learning model preset by the cloud server. The global optimization strategy is generated based on constraints, strategy dimensions, and strategy priorities. The cloud server generates standardized optimization control instructions according to the global optimization strategy and sends them to the edge computing gateway. The edge computing gateway executes the optimization control instructions and returns the execution results to the cloud server.

9. The edge computing data collection method for multi-platform concurrent transmission according to claim 1 is characterized in that: The calculation of the full link operation status index includes the following steps: A1: Set a time window and collect the real-time data transmission delay T of each terminal within the time window in real time. i and data transmission packet loss rate L i ; A2: Standardize the parameters of each terminal collected in real time to obtain the standardized real-time data transmission delay T Bi and data transmission packet loss rate L Bi , and by the formula: Calculate the operating status index S of the i-th terminal i ; A3: The operating status indexes of n terminals are calculated using the formula The full link operation status index S is calculated, where ω i represents the weight coefficient of the i-th terminal, and m is the nonlinear adjustment factor.

10. The edge computing data collection method for multi-platform concurrent transmission according to claim 1, characterized in that: The full-link operation status level includes the first status level, the second status level, the third status level and the fourth status level. The full-link operation status level is judged according to the full-link operation status index. A hierarchical dynamic control strategy is formulated based on the full-link operation status level, and the parameters after the strategy is executed are collected in real time to optimize the control strategy.

Images