An industrial edge gateway configuration-free communication method, device and system based on high-dimensional semantic alignment

Abstract

This invention belongs to the field of Industrial Internet of Things (IIoT) technology, and particularly relates to a configuration-free communication method, device, and system for industrial edge gateways based on high-dimensional semantic alignment. The method includes: using a graph embedding algorithm engine to extract data behavior and structural dependency features of the underlying discrete communication protocol, generating a first high-dimensional structure vector set Vector A; simultaneously receiving an abstract industrial standard information model provided by a higher layer, generating a second high-dimensional structure vector set Vector B; calculating the spatial distance metric between elements in the first and second high-dimensional structure vector sets in local memory space without relying on a manually configured reference table; and performing real-time interception, payload extraction, semantic reassembly, and protocol encapsulation and forwarding of underlying heterogeneous packets according to the communication conversion routing table. This invention reduces the gateway debugging time, which traditionally requires calculations over days or weeks, to minutes.

Description

Technical Field

[0001] This invention belongs to, but is not limited to, the field of industrial Internet of Things (IoT) technology, and particularly relates to a configuration-free communication method, device, and system for industrial edge gateways based on high-dimensional semantic alignment. Background Technology

[0002] With the advancement of the Industrial Internet, intelligent manufacturing, and digital transformation of production sites, industrial edge gateways are gradually becoming data conversion nodes between underlying automated equipment and upper-level information systems. These gateways typically connect one end to field devices such as PLCs, CNC machine tools, sensors, and frequency converters, and the other end to MES systems, SCADA systems, industrial cloud platforms, or data middleware platforms. They are used to complete field data acquisition, protocol conversion, edge preprocessing, and uploading. Because many devices in existing industrial sites still use Modbus, proprietary serial protocols, or other communication methods centered on register addresses, while upper-level systems typically require data to be organized using semantic objects such as devices, components, measurement points, and status variables, edge gateways not only need to perform message format conversion but also need to establish the correspondence between underlying address data and upper-level semantic paths.

[0003] The closest existing technology to this invention is an industrial edge gateway communication conversion scheme based on manually configured point tables. This scheme typically involves first loading the corresponding underlying protocol driver into the gateway, and then engineers manually entering the device address, function code, register address, data type, scaling factor, unit, variable name, acquisition cycle, and upper-layer publishing topic or OPC UA node path into the gateway configuration interface according to the equipment manual, PLC register table, or process point table. During actual operation, the gateway polls the underlying address according to the pre-configured acquisition cycle, parses the data type, performs scaling conversion and format encapsulation of the raw payload, and then forwards it according to preset MQTT topics, OPC UA nodes, or other cloud interface rules. Publicly available industrial gateway application examples also typically include configuring RS-485 parameters, adding Modbus polling points, specifying device addresses and register addresses, setting polling intervals, and establishing mapping rules between Modbus register values ​​and MQTT topics and JSON payload formats.

[0004] While the existing solutions described above can achieve basic protocol conversion, their core remains a static mapping mechanism driven by manual location tables. For lower-level devices, data in protocols such as Modbus is typically represented as discrete registers or address sets, and the addresses themselves do not directly carry business semantics such as "temperature, pressure, speed, alarm status, equipment components." For upper-level systems, industrial information models such as OPC UA emphasize the hierarchical structure and semantic constraints expressed by objects, variables, reference relationships, and type definitions. Existing gateways only convert based on manually entered address-path correspondences, failing to proactively extract behavioral characteristics such as access relationships, change patterns, adjacent structures, and data dependencies between lower-level addresses, nor can they automatically align these characteristics with the object structure in upper-level standard information models. Therefore, in a workshop with tens of thousands of locations, multiple equipment brands, incomplete field documentation, or inconsistent register table versions, manual configuration is not only labor-intensive but also prone to problems such as address mismatches, unit mismatches, and misjudgments of variable semantics.

[0005] Furthermore, the existing point configuration table is strongly coupled with specific equipment models, register arrangements, communication parameters, and upper-level path names. Even if two devices contain similar data such as motor temperature, bearing temperature, operating current, and alarm status in their process functions, the original point table usually cannot be directly reused if the underlying register address, function code, or data arrangement changes. When changing PLC brands, adjusting equipment programs, expanding I / O modules, or replicating production lines on-site, engineers still need to re-check the manuals and rebuild a large number of mapping rules. This static configuration method cannot utilize the potential similarities between the underlying data structure and the upper-level semantic model, and lacks the ability to automatically discover the "source address - target path" correspondence at the edge, resulting in high costs for rapid industrial site transformation and batch deployment.

[0006] Meanwhile, the data collection and forwarding logic of existing edge gateways typically relies on preset polling cycles and fixed conversion rules. For actual data in industrial settings, the data behavior varies significantly across different locations: some variables are high-frequency changing operational status quantities, while others are low-frequency changing process parameters or static equipment attributes. Existing solutions primarily rely on manually configured cycles for data collection, making it difficult to form adaptive data organization and conversion routes based on data behavior characteristics, structural dependencies, or semantic importance. This easily leads to low-change-rate data repeatedly occupying fieldbus bandwidth, while the real-time performance of high-value dynamic data is negatively impacted. Because gateways lack an automatic matching mechanism from underlying data behavior to upper-layer semantic models, their protocol conversion process essentially remains a linear flow of "reading address—applying rules—encapsulating and sending," making it difficult to form an adaptive closed loop oriented towards changes in field equipment.

[0007] Therefore, in the process of industrial edge gateways transforming from underlying address-based protocols to upper-layer semantic industrial information models, there is an urgent need for a configuration-free communication method that can be executed independently at the edge. This method would enable the gateway to extract the data behavior and structural dependency features of the underlying discrete communication protocols without relying on manual explicit configuration of reference tables, and to automatically perform semantic alignment with the higher-level industrial standard information model. Furthermore, based on the similarity judgment results, it would adaptively generate a communication conversion routing table between the source address and the target path, thereby achieving real-time interception, payload extraction, semantic reorganization, and encapsulation and forwarding of underlying heterogeneous packets in accordance with cloud protocol rules. This would reduce on-site configuration costs, improve adaptability after equipment replacement, and enhance the real-time communication efficiency under traditional fixed mapping and fixed polling mechanisms. Summary of the Invention

[0008] To address the problems existing in the prior art, this invention provides a configuration-free communication method, device, and system for industrial edge gateways based on high-dimensional semantic alignment.

[0009] This invention is implemented as follows: a configuration-free communication method for industrial edge gateways based on high-dimensional semantic alignment. This method is executed independently by the edge device system deployed in the industrial field and includes the following steps:

[0010] S1: Using a graph embedding algorithm engine, extract the data behavior and structural dependency features of the underlying discrete communication protocol to generate the first high-dimensional structure vector set Vector A; at the same time, receive the abstract industrial standard information model provided by the higher layer to generate the second high-dimensional structure vector set Vector B.

[0011] S2: Without relying on manual explicit configuration of the reference table, calculate the spatial distance metric between the elements in the first high-dimensional structure vector set and the second high-dimensional structure vector set in the local memory space;

[0012] S3: When the feature similarity between the source vector metadata and the target vector metadata exceeds the automatic matching threshold based on the distance metric, an adaptive communication conversion routing table associated with the source address and the target path is generated at the edge.

[0013] S4: According to the communication conversion routing table, perform real-time interception, payload extraction, semantic reassembly, and protocol encapsulation and forwarding of underlying heterogeneous packets in accordance with cloud protocol rules.

[0014] Furthermore, the basis for feature generation in step S1 includes: the access frequency of data addresses, the encoding format of data types, and the causal jump constraints generated between different physical addresses within the same underlying device within a time window.

[0015] Furthermore, the spatial distance metric in step S2 is evaluated and calculated using cosine similarity or Euclidean spatial distance.

[0016] Furthermore, it also includes a bus bandwidth adaptive scheduling step based on latent space semantic features:

[0017] The dynamic scheduling engine of the edge gateway obtains the first high-dimensional structure vector set, extracts the latent space features that represent the importance of node services in the vector space, and calculates the semantic priority index corresponding to each underlying physical node accordingly.

[0018] The underlying bus controller of the edge gateway establishes a resource allocation mapping relationship based on the semantic priority index of each node, dynamically outputs and adjusts the non-uniform polling time slices for different underlying physical address segments to replace the global fixed-period bus scanning mechanism; the higher the priority index of the node, the higher the polling frequency is allocated.

[0019] Furthermore, the calculation of the semantic priority index relies on dimensionality reduction analysis of the multidimensional features of the high-dimensional structure vector set, and its evaluation dimensions include at least:

[0020] The distance metric between the cluster centers of the node's high-dimensional vector in the unified common semantic space;

[0021] And the size of the norm of the feature vector mapped by the number of time-dependent edges associated with the node in the original graph network;

[0022] When a node represents a rapidly changing interlocking control signal, its corresponding latent space features are amplified, causing the semantic priority index to exceed the set high-frequency preemption threshold.

[0023] Furthermore, it also includes an adaptive scheduling mechanism for reclaiming bandwidth from the underlying bus:

[0024] The gateway reads the industrial semantic importance and data jump activity features contained in the first high-dimensional structure vector, and uses the corresponding features as the state input of the deep reinforcement learning RL bus scheduling agent.

[0025] The proxy model's action strategy includes: a time slice allocation strategy for continuously outputting the communication bus, controlling the gateway's communication interface to perform high-frequency intensive sampling polling on low-level targets that exhibit high semantic importance and activity, and forcing long-cycle reduced-frequency polling on statically configured low-level targets that exhibit weak semantic association.

[0026] Furthermore, the dynamic scheduling engine includes a deep reinforcement learning scheduling model based on Markov decision processes.

[0027] The state input of the model includes: the real-time blocking rate of the current physical network interface, the semantic priority index of the node vector, and the node update silence time.

[0028] The action space output Action of the model is the specific dynamic polling time slice length for different physical address segments;

[0029] The reward function of the model is set to maximize the effective data throughput of the bus physical layer while ensuring that the communication delay of high semantic priority nodes does not exceed the safety timeliness penalty threshold.

[0030] Another objective of this invention is to provide an intelligent edge gateway device with rule-free configuration capabilities, comprising:

[0031] Southbound fieldbus interface array and northbound cloud communication interface;

[0032] Built-in computing unit that runs embedded programs, including a spatial vector mapping matching module, an automatic routing engine, and a bandwidth scheduling module;

[0033] The spatial vector mapping matching module collaborates with the automatic routing engine to execute the method described in any one of claims 1 to 3; the bandwidth scheduling module executes the scheduling mechanism described in any one of claims 4 to 6.

[0034] Based on the above technical solutions and the technical problems solved, the advantages and positive effects of the technical solution to be protected by this invention are as follows:

[0035] This invention boasts exceptionally high implementation efficiency (plug and play): reducing gateway setup time, which traditionally required days or weeks, to minutes. Once the gateway is plugged in, it automatically identifies the underlying device and connects to cloud objects.

[0036] This invention boasts exceptional anti-interference and device migration capabilities: Even when devices from different manufacturers are replaced (address tables change), as long as the fundamental physical characteristics of the device remain unchanged, the gateway's vector matching system can still compare these characteristics and automatically reconstruct the routing table in a very short time. This solves the serious engineering problems of manually configuring point mapping tables when deploying gateways in industrial settings, which are extremely time-consuming, error-prone, and have poor generalization capabilities.

[0037] This invention eliminates network congestion by using vector calculations to identify critical dynamic data and redundant static data. It allocates limited bus bandwidth to higher-priority, more important tasks, effectively ensuring the real-time performance of industrial control systems. This solves the problem of traditional buses polling at fixed intervals regardless of data importance, leading to the consumption of valuable fieldbus resources by a large amount of redundant static data and resulting in low data acquisition efficiency.

[0038] The expected benefits and commercial value of the technical solution of this invention after its transformation are as follows: It revolutionizes the industry's delivery costs, shortening the point alignment work that originally required engineers to spend several weeks on-site to a highly efficient and automatic matching process involving simply plugging in a network cable. Implementation costs are significantly reduced, saving companies substantial manpower expenses.

[0039] The technical solution of this invention fills a technical gap in the industry both domestically and internationally: it fills the technical gap of using semantic geometric distance metrics of high-dimensional vector space to replace manually compiling hardwired communication dictionary tables; and it realizes an adaptive learning framework that crosses the OSI model and uses the semantic importance of the application layer to guide the time slice scheduling of the physical layer (MAC layer) bus.

[0040] The technical solution of this invention solves a long-standing but unresolved technical problem: in recent years, the industry has been eager to achieve plug-and-play functionality for heterogeneous electrical devices at the underlying level. However, due to the differences in protocols among manufacturers, this barrier has remained unbroken. This patent aims to bypass the protocol barrier by reducing the protocol dimension to a mathematical vector comparison, thus overcoming this problem through mathematical operations.

[0041] The technical solution of this invention overcomes technical bias: it overcomes the bias in the automation engineering field that communication protocols from different manufacturers and standards must rely on human engineers to establish mapping rules word by word by referring to point tables and manuals in order to communicate. This invention proves that by relying on AI comparative learning space, machines can completely find physical correspondences in messy data on their own. Attached Figure Description

[0042] Figure 1 This is a flowchart of a configuration-free communication method for industrial edge gateways based on high-dimensional semantic alignment provided in an embodiment of the present invention.

[0043] Figure 2 This invention provides a configuration-free communication architecture for industrial edge gateways based on high-dimensional semantic alignment.

[0044] Figure 3 This is a comparison chart of commercial delivery time and core data efficiency provided in an embodiment of the present invention. Detailed Implementation

[0045] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0046] like Figure 1As shown, this embodiment of the invention provides a configuration-free communication method for industrial edge gateways based on high-dimensional semantic alignment. This method is executed independently by the edge device system deployed in the industrial field and includes the following steps:

[0047] S1: Using a graph embedding algorithm engine, extract the data behavior and structural dependency features of the underlying discrete communication protocol to generate the first high-dimensional structure vector set Vector A; at the same time, receive the abstract industrial standard information model provided by the higher layer to generate the second high-dimensional structure vector set Vector B.

[0048] S2: Without relying on manual explicit configuration of the reference table, calculate the spatial distance metric between the elements in the first high-dimensional structure vector set and the second high-dimensional structure vector set in the local memory space;

[0049] S3: When the feature similarity between the source vector metadata and the target vector metadata exceeds the automatic matching threshold based on the distance metric, an adaptive communication conversion routing table associated with the source address and the target path is generated at the edge.

[0050] S4: According to the communication conversion routing table, perform real-time interception, payload extraction, semantic reassembly, and protocol encapsulation and forwarding of underlying heterogeneous packets in accordance with cloud protocol rules.

[0051] The basis for feature generation in step S1 includes: the access frequency of data addresses, the encoding format of data types, and the causal jump constraints generated between different physical addresses within the same underlying device within a time window.

[0052] In step S2, the spatial distance metric is evaluated and calculated using cosine similarity or Euclidean spatial distance.

[0053] It also includes a bus bandwidth adaptive scheduling step based on latent space semantic features:

[0054] The dynamic scheduling engine of the edge gateway obtains the first high-dimensional structure vector set, extracts the latent space features that represent the importance of node services in the vector space, and calculates the semantic priority index corresponding to each underlying physical node accordingly.

[0055] The underlying bus controller of the edge gateway establishes a resource allocation mapping relationship based on the semantic priority index of each node, dynamically outputs and adjusts the non-uniform polling time slices for different underlying physical address segments to replace the global fixed-period bus scanning mechanism; the higher the priority index of the node, the higher the polling frequency is allocated.

[0056] The calculation of the semantic priority index relies on dimensionality reduction analysis of the multidimensional features of the high-dimensional structure vector set, and its evaluation dimensions include at least:

[0057] The distance metric between the cluster centers of the node's high-dimensional vector in the unified common semantic space;

[0058] And the size of the norm of the feature vector mapped by the number of time-dependent edges associated with the node in the original graph network;

[0059] When a node represents a rapidly changing interlocking control signal, its corresponding latent space features are amplified, causing the semantic priority index to exceed the set high-frequency preemption threshold.

[0060] It also includes an adaptive scheduling mechanism for reclaiming bandwidth from the underlying bus:

[0061] The gateway reads the industrial semantic importance and data jump activity features contained in the first high-dimensional structure vector, and uses the corresponding features as the state input of the deep reinforcement learning RL bus scheduling agent.

[0062] The proxy model's action strategy includes: a time slice allocation strategy for continuously outputting the communication bus, controlling the gateway's communication interface to perform high-frequency intensive sampling polling on low-level targets that exhibit high semantic importance and activity, and forcing long-cycle reduced-frequency polling on statically configured low-level targets that exhibit weak semantic association.

[0063] The dynamic scheduling engine includes a deep reinforcement learning scheduling model based on Markov decision processes.

[0064] The state input of the model includes: the real-time blocking rate of the current physical network interface, the semantic priority index of the node vector, and the node update silence time.

[0065] The action space output Action of the model is the specific dynamic polling time slice length for different physical address segments;

[0066] The reward function of the model is set to maximize the effective data throughput of the bus physical layer while ensuring that the communication delay of high semantic priority nodes does not exceed the safety timeliness penalty threshold.

[0067] like Figure 2 As shown, this embodiment of the invention provides an intelligent edge gateway device with rule-free configuration capabilities, comprising:

[0068] Southbound fieldbus interface array and northbound cloud communication interface;

[0069] Built-in computing unit that runs embedded programs, including a spatial vector mapping matching module, an automatic routing engine, and a bandwidth scheduling module;

[0070] The spatial vector mapping matching module collaborates with the automatic routing engine to execute the method described in any one of claims 1 to 3; the bandwidth scheduling module executes the scheduling mechanism described in any one of claims 4 to 6.

[0071] This invention can be applied to the automated integration of flexible production lines, the rapid construction of digital twin bases, and the low-code intelligent transformation of traditional old workshops.

[0072] The products involved include, but are not limited to: plug-and-play zero-code gateways, digital twin cloud-edge collaboration platforms, and transparent bus controllers.

[0073] The specific implementation phase encompasses two main hardware control aspects: on-site deployment and underlying link control. First, during deployment and power-on, the spatial metric model within the gateway automatically compares the semantic vector similarity between the underlying physical data stream and the upper-layer cloud model, automatically generating a communication mapping routing table. This completely replaces the time-consuming manual table lookup process traditionally performed by engineers. Second, in the runtime scheduling phase, adaptive control algorithms, such as reinforcement learning (RL), handle the underlying polling drive of the gateway's physical network ports (e.g., RS485 / CAN bus). The algorithm dynamically adjusts the time slice allocation of the bus control chip based on the semantic priority of the data points, performing intensive sampling on high-frequency, high-risk points and forcibly reducing the frequency of static data points, thereby significantly improving the bus's utilization efficiency.

[0074] This embodiment uses the edge access and cloud data fusion of a newly added injection molding machine in a manufacturing workshop as an example. The injection molding machine on site has a built-in PLC controller, which is connected to the intelligent edge gateway of this invention via an RS485 physical bus, using the Modbus RTU flat communication protocol at the underlying level. The top-level MES cloud system in the workshop requires the gateway to report data in a standard object format based on the MQTT protocol and conforming to the OPC UA specification. Due to the lack of a site equipment location table (Mapping List), the gateway needs to complete the matching and routing forwarding of thousands of unknown register addresses at the underlying level to the standard model in the cloud without any preset manual rules.

[0075] The specific implementation steps are as follows:

[0076] Step S1: Bidirectional data stream interception and high-dimensional structural feature vectorization

[0077] After establishing a physical connection via the southbound interface, the edge gateway initiates a listening mode to capture the underlying Modbus data stream sequence and extracts the data behavior characteristics and spatiotemporal dependency characteristics of each underlying address. For example:

[0078] The system detected that the data at address 40050 has the following physical characteristics: the data format is floating point (Float), the value shows a smooth jump at the level of 500ms, and the jump of this data shows a strong temporal causal relationship with address 40010 (device start / stop status control word) (that is, within the time window after 40010 is set to 1, 40050 shows a significant upward gradient).

[0079] Based on the above features, the gateway's built-in graph embedding computing engine generates a 256-dimensional low-level structural feature vector (Vector A) for address 40050. Simultaneously, the gateway loads the standard information model of the injection molding machine distributed from the cloud, transforming its standard node, extruder temperature (Extruder_Temperature), into a global standard feature vector of the same dimension (denoted as Vector_B).

[0080] Step S2: High-dimensional semantic space alignment calculation based on local memory

[0081] Without human intervention, the edge gateway executes a high-dimensional vector space mapping algorithm in its isolated local memory computing space. The system calculates the pairwise distances between the vector set generated by the underlying device (containing Vector_A) and the upper-layer standard dictionary vector set (containing Vector B). In this embodiment, cosine similarity is used for calculation, and the cosine similarity value between Vector A and Vector B in the feature space is found to be 0.982.

[0082] Step S3: Generate an automated mapping routing table that requires no human intervention.

[0083] The system internally sets the automatic judgment threshold to 0.95. Since the calculated similarity of 0.982 is greater than the set threshold, the gateway's intelligent matching engine determines that Modbus underlying register 40050 and the extruder temperature node in the cloud standard model have a high degree of physical semantic overlap.

[0084] The gateway's automatic routing system immediately establishes hard-wired rules between the underlying network interface stack and the northbound forwarding protocol stack, automatically generating communication mapping relationships: Route: [Modbus Reg 40050] points to [MQTT Payload:Extruder_Temperature]. This enables zero-code access and plug-and-play functionality for heterogeneous device locations, reducing the implementation cycle of manual on-site debugging.

[0085] Step S4: Adaptive scheduling of underlying physical bus bandwidth based on reinforcement learning

[0086] After establishing a communication connection, the edge gateway synchronously enables the bandwidth adaptive scheduling mechanism.

[0087] The gateway's listening and analysis module discovered that the data at register address 40001 exhibited extremely low-frequency fluctuations within the continuous listening window, and that this data had no causal correlation with other physical state quantities (it was determined by feature comparison to be static configuration parameters such as the device's factory serial number SN).

[0088] Based on the weak semantic association and static activity attributes of its vector, the system marks it as a low-priority polling node. The scheduling agent inside the edge gateway, according to the reward function control strategy, automatically reduces the RS485 hardware polling request cycle for address 40001 from the default 100ms to 3600s (once per hour). Simultaneously, the released underlying bus bandwidth is dynamically allocated by the system to key process control variables such as 40050 (extruder temperature), which exhibit strong semantic association and high-frequency fluctuations, to support higher-frequency (e.g., 10ms / time) intensive sampling. This mechanism effectively avoids the communication network storms caused by traditional fixed-polling gateways and greatly optimizes the effective load rate of the industrial fieldbus.

[0089] The diagonal texture on the left represents the gateway configuration duration, using a logarithmic coordinate system. Figure 3 As can be seen, traditional gateway systems require 21 days for manual configuration at specific points, while the configuration-free gateway of this invention can automatically complete the gateway configuration in just 45 seconds. The grid texture bar chart on the right represents the core control data efficiency graph. Through RL reinforcement learning scheduling agent, the bus, which was originally filled with garbage static data packets (including 20% ​​high-frequency data), is allocated more bus bandwidth for higher priority and more important core data, significantly improving bus data efficiency to 85%.

[0090] It should be noted that embodiments of the present invention can be implemented in hardware, software, or a combination of both. The hardware portion can be implemented using dedicated logic; the software portion can be stored in memory and executed by a suitable instruction execution system, such as a microprocessor or dedicated-design hardware. Those skilled in the art will understand that the above-described devices and methods can be implemented using computer-executable instructions and / or included in processor control code, for example, such code provided on a carrier medium such as a disk, CD, or DVD-ROM, a programmable memory such as read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. The devices and modules of the present invention can be implemented by hardware circuitry such as very large-scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, or programmable hardware devices such as field-programmable gate arrays, programmable logic devices, etc., or by software executed by various types of processors, or by a combination of the above-described hardware circuitry and software, such as firmware.

[0091] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications, equivalent substitutions, and improvements made by those skilled in the art within the scope of the technology disclosed in the present invention, and within the spirit and principles of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A configuration-free communication method for industrial edge gateways based on high-dimensional semantic alignment, characterized in that, This method is executed independently by an edge device system deployed in an industrial field and includes the following steps: S1: Using a graph embedding algorithm engine, extract the data behavior and structural dependency features of the underlying discrete communication protocol to generate the first high-dimensional structure vector set Vector A; at the same time, receive the abstract industrial standard information model provided by the higher layer to generate the second high-dimensional structure vector set Vector B. S2: Without relying on manual explicit configuration of the reference table, calculate the spatial distance metric between the elements in the first high-dimensional structure vector set and the second high-dimensional structure vector set in the local memory space; S3: When the feature similarity between the source vector metadata and the target vector metadata exceeds the automatic matching threshold based on the distance metric, an adaptive communication conversion routing table associated with the source address and the target path is generated at the edge. S4: According to the communication conversion routing table, perform real-time interception, payload extraction, semantic reassembly, and protocol encapsulation and forwarding of underlying heterogeneous packets in accordance with cloud protocol rules.

2. The configuration-free communication method for industrial edge gateways based on high-dimensional semantic alignment according to claim 1, characterized in that, The basis for feature generation in step S1 includes: the access frequency of data addresses, the encoding format of data types, and the causal jump constraints generated between different physical addresses within the same underlying device within a time window.

3. The configuration-free communication method for industrial edge gateways based on high-dimensional semantic alignment according to claim 1, characterized in that, In step S2, the spatial distance metric is evaluated and calculated using cosine similarity or Euclidean spatial distance.

4. The configuration-free communication method for industrial edge gateways based on high-dimensional semantic alignment according to claim 1, characterized in that, It also includes a bus bandwidth adaptive scheduling step based on latent space semantic features: The dynamic scheduling engine of the edge gateway obtains the first high-dimensional structure vector set, extracts the latent space features that represent the importance of node services in the vector space, and calculates the semantic priority index corresponding to each underlying physical node accordingly. The underlying bus controller of the edge gateway establishes a resource allocation mapping relationship based on the semantic priority index of each node, dynamically outputs and adjusts the non-uniform polling time slices for different underlying physical address segments to replace the global fixed-period bus scanning mechanism; the higher the priority index of the node, the higher the polling frequency is allocated.

5. The configuration-free communication method for industrial edge gateways based on high-dimensional semantic alignment according to claim 4, characterized in that, The calculation of the semantic priority index relies on dimensionality reduction analysis of the multidimensional features of the high-dimensional structure vector set, and its evaluation dimensions include at least: The distance metric between the cluster centers of the node's high-dimensional vector in the unified common semantic space; And the size of the norm of the feature vector mapped by the number of time-dependent edges associated with the node in the original graph network; When a node represents a rapidly changing interlocking control signal, its corresponding latent space features are amplified, causing the semantic priority index to exceed the set high-frequency preemption threshold.

6. The configuration-free communication method for industrial edge gateways based on high-dimensional semantic alignment according to claim 1, characterized in that, It also includes an adaptive scheduling mechanism for reclaiming bandwidth from the underlying bus: The gateway reads the industrial semantic importance and data jump activity features contained in the first high-dimensional structure vector, and uses the corresponding features as the state input of the deep reinforcement learning RL bus scheduling agent. The proxy model's action strategy includes: a time slice allocation strategy for continuously outputting the communication bus, controlling the gateway's communication interface to perform high-frequency intensive sampling polling on low-level targets that exhibit high semantic importance and are active, and forcing long-cycle reduced-frequency polling on statically configured low-level targets that exhibit weak semantic association.

7. The configuration-free communication method for industrial edge gateways based on high-dimensional semantic alignment according to claim 4, characterized in that, The dynamic scheduling engine includes a deep reinforcement learning scheduling model based on Markov decision processes. The state input of the model includes: the real-time blocking rate of the current physical network interface, the semantic priority index of the node vector, and the node update silence time. The action space output Action of the model is the specific dynamic polling time slice length for different physical address segments; The reward function of the model is set to maximize the effective data throughput of the bus physical layer while ensuring that the communication delay of high semantic priority nodes does not exceed the safety timeliness penalty threshold.

8. An intelligent edge gateway device with rule-free configuration capability implementing the method as described in any one of claims 1-6, characterized in that, include: Southbound fieldbus interface array and northbound cloud communication interface; Built-in computing unit that runs embedded programs, including a spatial vector mapping matching module, an automatic routing engine, and a bandwidth scheduling module; The spatial vector mapping matching module collaborates with the automatic routing engine to execute the method described in any one of claims 1 to 3; the bandwidth scheduling module executes the scheduling mechanism described in any one of claims 4 to 6.

Images