Multi-protocol intelligent interaction heterogeneous device full interconnection method and system

Abstract

The invention discloses a multi-protocol intelligent interaction heterogeneous device full interconnection method and system, and belongs to the technical field of computers, and the method comprises the steps: S1, collecting data, and deploying edge computing nodes; s2, preprocessing the collected data; s3, performing classification according to task requirements of the data; s4, after classification, determining an optimal communication protocol suitable for the current network condition; s5, realizing data format standardization and communication protocol unification by adopting an OPC UA protocol; s6, on the basis of an OPC UA publishing / subscribing mechanism and a QoS priority mechanism, allowing the equipment to publish or subscribe data in real time and ensuring priority transmission of high-priority data; s7, performing data matching and association, and fusing the data; s8, realizing a global decision; according to the multi-protocol intelligent interaction heterogeneous equipment full interconnection method and system provided by the invention, by introducing protocol unification, intelligent interaction and distributed optimization technologies, the interoperability of multi-protocol equipment and the overall management efficiency of a traffic system can be effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a method and system for fully interconnecting heterogeneous devices with multi-protocol intelligent interaction. Background Art

[0002] With the continuous advancement of technology, communication and coordination between heterogeneous devices have become a key issue in the development of intelligent transportation systems. In order to meet the challenges of device interoperability in complex traffic environments, many technical solutions have emerged to promote the intelligent development of transportation systems. Existing technologies mainly use Modbus technology, CANopen technology and Gateway (protocol gateway). Although some results have been achieved in improving the intelligent management of transportation systems, the limitations of current technologies in terms of heterogeneous device interoperability, data unification in multi-protocol environments, and global system optimization restrict the overall efficiency and real-time response capabilities of the system, as follows: (1) In the current transportation system, heterogeneous devices use different communication protocols, resulting in poor interoperability between devices. Real-time data exchange between devices still faces challenges, especially when the protocols are incompatible, which reduces the efficiency of data transmission and interaction; (2) In the existing system, the data formats generated by heterogeneous devices are diverse and difficult to quickly integrate and standardize. In the traffic management system, the data sharing efficiency between devices is low, which hinders the system's processing and response to real-time information. Especially in high-load environments or emergencies, the data fusion capabilities of the existing system are insufficient; (3) Although multi-agent technology has shown potential in other fields, in traffic management, existing systems still mostly use centralized control and lack the ability to coordinate optimization between agents, resulting in poor global optimization and adaptability of the system in complex environments. Summary of the invention

[0003] The purpose of the present invention is to provide a method and system for fully interconnecting heterogeneous devices with multi-protocol intelligent interaction to solve the problems existing in the above-mentioned background technology.

[0004] To achieve the above object, the present invention provides a method for fully interconnecting heterogeneous devices with multi-protocol intelligent interaction, comprising the following steps: S1, collect data and deploy edge computing nodes; S2, preprocessing the collected data; S3, through the data layering mechanism, the data is divided into a high real-time data layer and a low real-time data layer according to the task requirements of the data, and then different transmission processing methods are adopted; S4. After classification, the optimal communication protocol suitable for the current network conditions is determined through the protocol priority mechanism; S5. Use OPC UA protocol to standardize data formats and unify communication protocols; S6, OPC UA-based publish / subscribe mechanism and QoS priority mechanism, allowing devices to publish or subscribe to data in real time and ensure that high-priority data is transmitted first; S7, use graph neural network and Mahalanobis distance to match and associate data, and use Dempster-Shafer evidence theory to fuse data; S8. Make global decisions based on multi-agent technology.

[0005] Preferably, the delay optimization target of the edge computing of the edge computing node deployed in step S1 is: ; In the formula, is the total delay for data to be transmitted from the device to the edge node and processed, is the amount of data to be transmitted, is the network bandwidth, The computation processing delay of the edge node.

[0006] Preferably, the high real-time data in step S3 is transmitted point-to-point quickly using a low-latency protocol dynamically selected by reinforcement learning, and the formula is: ; In the formula, For high real-time data, MQTT is a low-latency communication protocol dynamically selected for reinforcement learning; Low real-time data is standardized through OPC UA, and the data standardized by OPC UA is transmitted through the publish / subscribe mode. The formula is as follows: ; In the formula, OPC UA data format conversion function, converting the original data Convert to standard format .

[0007] Through the data stratification mechanism, the appropriate transmission method and processing means can be selected according to the real-time and priority of the data. The system can ensure that high-real-time tasks are given priority, while low-real-time data is standardized and transmitted with delay.

[0008] Preferably, step S4 specifically includes: The system optimizes the priorities of different communication protocols through the deep Q network (DQN). According to the current network status, bandwidth, latency, etc. of the device, DQN learns and dynamically adjusts the priority of each protocol to select the optimal protocol. The formula is: ; In the formula, The communication protocols that can be selected for the system include OPC UA, MQTT and HTTP protocols; , , The weights of OPC UA, MQTT and HTTP protocols respectively; Status The current network status (bandwidth, latency, packet loss rate) and task requirements; According to the current status Weights learned by DQN.

[0009] Through this formula, DQN can dynamically adjust and select the protocol with the best long-term benefits based on network conditions and historical experience.

[0010] Preferably, step S5 specifically includes: S51. The introduction of OPC UA and the unification of data formats. OPC UA converts all transmitted data into a unified standardized format. S52, data abstraction and standardization, data abstraction is to extract device data in different protocols into common standardized fields, specifically using field mapping abstraction, hierarchical abstraction and type abstraction; S53. After completing data abstraction, the system converts the standardized data into OPC UA format through a protocol conversion algorithm, including: S531, field mapping conversion, mapping the abstracted standardized fields to the OPC UA data model nodes, as follows: ; In the formula, is the abstracted field; It is a standard node in OPC UA; S532, data format conversion, is used to handle the differences in encoding, byte order and data type between different protocols, ensuring that data from different devices in the system can be correctly decoded and transmitted. Assume that the device Usage Agreement , its data format is , after data format conversion, unified into a standardized format , the formula is as follows: ; In the formula, It is a data format conversion function, which is used to convert data formats of different protocols into the standard format of the system; S533, transmission protocol conversion, used to handle the differences in transmission modes of different protocols. Assume that the device The transmission protocol used is , its transmission mode is , the system converts the transmission protocol into a standardized transmission mode , the formula is as follows: ; In the formula, It is the transmission protocol conversion function.

[0011] The protocol identification and conversion mechanism ensures that the data of heterogeneous devices can be processed in a standardized manner by using data abstraction algorithms (field mapping, hierarchical abstraction, and type abstraction). The protocol conversion algorithm ensures that data can be correctly transmitted in the system and participate in intelligent interaction by processing differences in data formats and transmission protocols. The entire conversion mechanism not only supports full interconnection between devices, but also provides a reliable foundation for the intelligent operation of the system.

[0012] Preferably, step S6 specifically includes: S61. OPC UA provides a publish / subscribe model. The subscription model formula is: ; In the formula, is the subscribed dataset; It is the time period; S62, introduce QoS (Quality of Service) priority mechanism to ensure that high-priority data can be transmitted first. It dynamically adjusts the order of data transmission according to the real-time requirements of the task, bandwidth limitations and device load to ensure that the system can maintain stable performance under different network conditions. It is divided into: High priority data : including emergency traffic data (such as accident alarms), which need to be transmitted in real time; Medium priority data : Including real-time traffic flow data, requiring low latency and reliable transmission; Low priority data : Includes common device status data (such as the fuel status of the vehicle), allowing for a certain delay; The QoS scoring formula determines the transmission priority of data in the system: ; In the formula, Scoring the QoS of data determines the transmission priority of the data in the system; The priority of the data, reflecting its importance in the system; The bandwidth allocated to data. The higher the bandwidth, the faster the transmission. The data transmission delay, the lower the better; is the packet loss rate, the lower the better; To transmit jitter, the smaller the delay variation, the better; are the weights of delay, packet loss rate and jitter respectively.

[0013] Preferably, the data matching and associating in step S7 specifically includes: First, the system models all devices and their communication relationships as a graph structure, and learns the data interaction patterns between devices through the graph neural network (GNN). The GNN captures the complex associations between devices by transferring and updating node features layer by layer, and generates high-order features for each node. The feature extraction formula of the GNN is: ;.

[0014] In the formula, is the adjacency matrix; is the degree matrix; For the The node feature matrix of the layer; is the weight matrix; is an activation function used to perform nonlinear transformation on features; Through multi-layer GNN transmission, the system can capture deeper correlation information between devices and generate high-level feature representations for each device. .

[0015] After extracting the high-order features of each device, the system uses the Mahalanobis distance to match these features. The Mahalanobis distance can effectively measure the differences between multidimensional data with covariance relationships. Here, it is used to measure the matching degree between the high-order features of each device and the target device features. Before calculating the Mahalanobis distance, the covariance matrix of the sample feature vector must be calculated. , the covariance matrix is ​​used to describe the correlation between each feature dimension, which is: ; In the formula, For the High-order feature vectors of each device; is the mean vector of the device; is the sample size; The Mahalanobis distance is used to calculate the distance between two data points. and The multidimensional distance between is defined as: ; In the formula, and are the high-order eigenvectors of the two devices respectively; is the covariance matrix The inverse matrix of is the difference between the eigenvectors; By calculating the Mahalanobis distance between two data points ,if <ϵ, then and match; Wherein, the threshold ϵ is a matching criterion set according to system needs.

[0016] Preferably, the data fusion in step S7 specifically adopts the Dempster-Shafer evidence theory to fuse the evidence from multiple data sources. The formula of the Dempster-Shafer evidence theory is: ; In the formula, The reliability of hypothesis A indicates the degree to which multiple pieces of evidence support A; and are the credibility assignments of hypothesis B and C from data source 1 and data source 2 respectively; Assign conflict, which represents the total amount of conflict when there is a complete conflict between the evidences.

[0017] Preferably, step S8 is specifically: S81. Information sharing of multi-agent technology. The information sharing formula is: ; In the formula, For the At the iteration Status updates; For intelligent agents The neighbor set of A collection of intelligent agents that interact with each other; is the weight of information sharing, indicating weight equalization or adaptive allocation in information sharing; Neighbor Agent The state at the kth iteration; S82. Global collaborative optimization of multi-agent technology. The collaborative optimization formula is: ; Where N is the number of agents; For each agent The local objective function of For intelligent agents state variables; For intelligent agents The feasible solution space of ; S83, distributed convex optimization decision, is used to minimize the waiting time of different transportation modes. The objective function for minimizing the waiting time is: ; Where N is the number of agents; For the The waiting time of each agent; S84, using the gradient descent method to optimize and minimize the objective function, each agent The state vector of is updated at the kth iteration by the following formula: ; In the formula, For intelligent agents The state at the kth iteration; is the step length coefficient; The waiting time objective function Status The gradient of For intelligent agents The updated state at the k+1th iteration; S85, distributed collaborative update, the update formula is expanded to: ; In the formula, For intelligent agents The set of neighbors of ; For intelligent agents and neighbors The information exchange weight between them indicates the degree of influence between the two agents; Neighbor Agent The state at the kth iteration; S86, real-time update, the formula is: ; In the formula, Generate new data for each time point; For the agent at time The data state at the time; Gradient It is a gradient update based on the latest data; S87, iterative convergence, the convergence condition is: ; It indicates that the optimization process reaches the global optimal solution through distributed collaboration.

[0018] A multi-protocol intelligent interactive heterogeneous device full interconnection system, including: The access layer includes several heterogeneous devices; The technical layer includes the deployment of data preprocessing modules and edge computing nodes; Protocol conversion layer, including OPC UA protocol conversion module and data fusion and matching module; The business layer also includes data fusion and matching modules, as well as multi-agent distributed optimization modules; The application layer includes the global optimization control module and the real-time decision and feedback module.

[0019] Therefore, the present invention adopts the above-mentioned multi-protocol intelligent interactive heterogeneous device full interconnection method and system, which has the following beneficial effects: (1) The OPC UA protocol is used to achieve unified management of multi-protocol devices, breaking the bottleneck of the inability of devices with different communication protocols to interoperate in the existing technology. Through standardized data formats and publish / subscribe mechanisms, the system can dynamically adapt to the access and exit of heterogeneous devices, achieving seamless interconnection of multiple devices. (2) Through the QoS mechanism, the system dynamically adjusts transmission parameters such as bandwidth and delay according to the importance and urgency of the data to ensure that key data is transmitted first. This mechanism ensures the system's real-time responsiveness in complex traffic scenarios. In particular, in the event of an emergency, the system can give priority to emergency traffic data. (3) By utilizing multi-agent technology and distributed collaborative optimization algorithms, autonomous collaboration among various devices is achieved, which not only optimizes local decision-making but also enables dynamic adjustment of global traffic management. Compared with existing centralized control systems, this solution greatly improves the system's flexibility, scalability, and ability to handle complex traffic conditions.

[0020] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 This is an architecture diagram of a fully interconnected system of heterogeneous devices with multi-protocol intelligent interaction according to the present invention; Figure 2 The present invention is a flowchart of a method for fully interconnecting heterogeneous devices with multi-protocol intelligent interaction. DETAILED DESCRIPTION

[0022] The following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0023] In a large city, the traffic system consists of multiple heterogeneous modes of transportation, and the data between these modes of transportation needs to be exchanged and processed in real time to achieve efficient traffic management and optimization. Under multiple modes of transportation, the system can collect data through multiple heterogeneous devices such as sensors, vehicle terminals, road monitoring equipment, etc. distributed in the traffic network. Different devices may use different communication protocols (such as MQTT, HTTP, CAN bus, etc.), and these data will be preliminarily processed at the edge computing node to reduce the computing burden of the central server and improve the real-time performance of the system.

[0024] See also Figure 1 , a multi-protocol intelligent interactive heterogeneous device fully interconnected system, consists of access layer, technology layer, protocol conversion layer, business layer and application layer, covering key modules such as edge computing, data preprocessing, data stratification, protocol conversion, priority scheduling, and multi-agent global optimization. The system collects heterogeneous device data through edge nodes, uses OPC UA to unify protocols, and dynamically optimizes communication protocol selection through the DQN priority mechanism. Combined with QoS priority scheduling and multi-agent technology, it realizes intelligent interaction and full connectivity between heterogeneous devices, and finally generates the global optimal scheduling decision to ensure efficient collaboration and resource allocation of the system.

[0025] See also Figure 2 , a method for fully interconnecting heterogeneous devices with multi-protocol intelligent interaction, comprising the following steps: S1. Data collection and edge computing deployment 1.1 Edge computing node deployment In a multi-protocol, multi-device transportation system, edge computing node deployment is the key to this system. It reduces the latency of data transmission and improves real-time responsiveness by processing data from different sources on-site. Each edge node directly interacts with heterogeneous devices of different protocols, such as MQTT, HTTP, OPC UA, etc., to process and format data for subsequent steps. The use of edge computing effectively disperses the computing load and avoids bottlenecks in data centers. The latency optimization goals of edge computing are: ; In the formula, It is the total delay for data to be transmitted from the device to the edge node and processed; is the amount of data that needs to be transferred; is the network bandwidth; Indicates the computational processing delay of edge nodes.

[0026] The input is the data source of heterogeneous devices, using data streams of different protocols; the output is standardized data that has undergone preliminary processing and is ready to enter the protocol conversion and preprocessing stage.

[0027] 1.2 Data Collection The system collects data from a variety of devices and sensors through edge computing nodes, and these data are transmitted to edge computing nodes through different communication protocols (such as MQTT, HTTP, Modbus, OPC UA). Edge computing nodes receive and preliminarily process these heterogeneous data to ensure that they can enter the next preprocessing stage. Each edge node is directly connected to the device to ensure that the data is transmitted to the system in a timely and reliable manner.

[0028] The input is a variety of sensor data and heterogeneous protocol data generated by the device; the output is a raw data stream that is ready to enter the preprocessing step.

[0029] 1.2.1 Data Source Since data sources are very extensive and heterogeneous, and different types of devices generate data in different formats, these devices collect various traffic information through different sensors, monitoring systems, and vehicle systems. The main data sources include: 1. Sensor data: Sensors are installed at multiple key locations such as roads, vehicles, and stations.

[0030] 2.GPS location data: vehicle location information, used to track and monitor the vehicle's movement trajectory.

[0031] 3. Speed ​​sensor data: monitor vehicle speed and analyze real-time changes in traffic flow.

[0032] 4. Acceleration sensor data: records the vehicle's acceleration and is used to determine sudden braking or dangerous driving behavior.

[0033] 5. Temperature and humidity sensor data: Collects meteorological data of roads and environment to help determine road safety conditions.

[0034] 6. Monitoring equipment data: Monitoring equipment such as cameras is used to obtain video streams or image data to monitor road conditions, traffic flow, and the behavior of pedestrians and vehicles.

[0035] 7. On-board terminal data: On-board equipment of buses, taxis and private cars (such as on-board GPS, OBD system) collects real-time status data of the vehicle, including the vehicle's fuel usage, braking status, engine operating status, etc.

[0036] The data from these devices are transmitted to edge nodes via wired or wireless networks.

[0037] 1.2.2 Data Transmission Model The unified data transmission model is as follows: ; In the formula, Indicates Devices at time The total amount of data transferred; Indicates the number of different protocols; Indicates The device uses Protocol in time The transmission rate; is the amount of data transferred using the protocol.

[0038] 1.2.3 Data Generation and Collection The data collected by different devices are processed and cached locally. Sensors collect data periodically, the vehicle terminal system collects vehicle status information in real time, and surveillance cameras continuously capture images or videos.

[0039] 1.2.4 Data transmission to edge nodes After data collection is completed, the device transmits the data to the edge computing node through multiple protocols. The edge node has multi-protocol support capabilities and can receive data from multiple communication protocols and process the data locally.

[0040] S2. Data preprocessing In the multi-protocol intelligent interaction and heterogeneous device full interconnection system, the quality of data transmitted by different devices through multiple protocols varies greatly, and there may be problems such as noise, abnormal data, and inconsistent formats. These problems will directly affect subsequent data processing, protocol selection, and conversion. Therefore, this step ensures the accuracy, consistency, and integrity of the data. The intelligentization of data preprocessing is enhanced by introducing a neural network structure. Specifically, an autoencoder is used as the core model to achieve automatic data cleaning, noise filtering, and anomaly detection.

[0041] 2.1 Data Denoising In a multi-protocol environment, the data generated by the device often contains noise. The system first denoises the input data and uses a low-pass filter to suppress high-frequency noise in the data. This stage of processing ensures that the data has higher accuracy when entering the subsequent processing steps. For data of different protocols, the filter will be adjusted according to the characteristics of the protocol to ensure the best denoising effect. The formula is as follows: ; In the formula, is the signal frequency; is the cutoff frequency. The low-pass filter can effectively remove high-frequency noise and only retain important low-frequency information, which is recorded as .

[0042] 2.2 Timestamp Alignment Different devices collect data at different frequencies, so the data timestamps need to be unified and aligned. To this end, linear interpolation is used to align the timestamps: ; In the formula, and The start time and end time are respectively; is the original timestamp; is the aligned timestamp. Through this method, data from different data sources can be analyzed and processed on the same timeline.

[0043] 2.3 Timestamp sorting During data collection and transmission, data from different devices may be inconsistent due to different collection times or transmission delays. To this end, each device carries a timestamp when sending data, and the receiving device sorts the data according to the timestamp to ensure that subsequent data processing and fusion steps are based on the same time node.

[0044] ① Timestamp allocation: Each data packet is assigned a timestamp on the device side , represents the precise time when the data was collected.

[0045] ②Data transmission: Data from different devices are transmitted to the central processing unit or edge computing node according to their respective protocols.

[0046] ③ Timestamp sorting: After the data arrives at the edge computing node, it is sorted by timestamp. Sorting is performed to ensure that subsequent processing steps such as data fusion and matching are based on a consistent time series.

[0047] equipment Sending Data The timestamp is , when data sent by multiple devices is transmitted to the central processing unit, it is sorted: ; By sorting by timestamp, the synchronization of data is guaranteed, avoiding data conflicts caused by delays or different collection times.

[0048] 2.4 Autoencoder Anomaly Detection and Correction In order to further improve data quality, the system uses an auto encoder for intelligent anomaly detection and correction. The auto encoder first converts the input data Compressed into low-dimensional features by the encoder, and then restored to high-dimensional data by the decoder By calculating the difference between the input data and the reconstructed data , the system can automatically identify abnormal data and make corrections to ensure that the data entering the system has a high degree of credibility.

[0049] ; In the formula, Represents the data after timestamp correction and denoising; The weight matrix of the encoder is responsible for mapping the input data into a low-dimensional space. Represents the bias vector, which is used to adjust the linear transformation of the data; As the activation function, the introduction of nonlinear relationships enables the network to represent more complex relationships; It is the output after the encoder processing, which represents the compressed low-dimensional features of the input data. It is the main feature extracted by the neural network from the high-dimensional information of the data, and has the functions of reducing redundancy and denoising.

[0050] Reconstruction error (loss function): The core goal of the autoencoder is to minimize the input data and rebuild data .

[0051] ; In the formula, is the loss function, which is used to measure the difference between the input data and the reconstructed data; Indicates the data after timestamp correction and denoising, that is, the original data; To reconstruct the data, the network tries to make it consistent with Consistency; Represents the Euclidean distance, which is used to calculate the difference between data vectors.

[0052] Anomaly detection and correction: When the reconstruction error Exceeding a certain threshold , the system will determine that the data is abnormal data; for the detected abnormal data, you can use Replace it as corrected data or process it further.

[0053] , ; By using autoencoders, the system can not only efficiently clean data and filter noise, but also automatically detect anomalies in the data, ensuring the quality and consistency of all data as it enters the system.

[0054] 2.4 Data Redundancy Detection In a system of multi-protocol devices, data redundancy may occur frequently, especially when multiple sensors collect information about the same event at the same time. This redundant data wastes system resources and may cause unnecessary repeated calculations. Therefore, this embodiment uses a combination of hash functions and feature vector comparisons. The system detects and removes redundant data through hash functions and feature vector comparisons. Hash functions can quickly identify duplicate data blocks, while feature vector comparisons are used to further accurately detect the similarity of potential redundant data.

[0055] 2.4.1 Feature extraction Feature extraction is a crucial part of the multi-source data collection and processing process. By extracting key features such as timestamp t, location information (x, y), speed v, etc., the system can better match and merge data and provide strong support. The goal of feature extraction is to convert raw data into structured feature information for subsequent processing and analysis.

[0056] Feature extraction uses Fourier transform to convert time domain data into frequency domain signals in order to extract hidden features in the data, especially periodicity and frequency-related information.

[0057] ; In the formula, is the frequency The frequency domain signal under is the time domain signal, i.e. the original data, is the imaginary unit; e is the complex exponential function in Fourier transform, which converts the time domain signal into the frequency domain.

[0058] Through Fourier transform and time domain feature extraction, the system can extract valuable feature information from raw data. These features lay a solid foundation for subsequent data fusion, matching, prediction and optimization.

[0059] 2.4.2 Preliminary screening The hash function is mainly used to quickly determine the uniqueness of data. After each device collects data, it will generate a corresponding hash value through the hash function. The same hash value means that the content of the data is highly similar or exactly the same.

[0060] 2.4.3 Hash Generation When the device Collecting data When the system passes the hash function Generates a hash value for the data.

[0061] ; 2.4.4 Hash comparison After the system receives data sent by multiple devices, it first compares the hash values ​​of all the data.

[0062] ; If the hash values ​​of two data are the same, , it is considered that the two data may be redundant, and the data marked as redundant will skip the subsequent processing steps to save system resources.

[0063] 2.4.5 Comparison based on feature vectors For complex data (such as images, videos, sensor signals, etc.), this system generates feature vectors of the data through feature extraction, and then detects redundant data by comparing the similarity of feature vectors. Perform feature extraction to extract feature values ​​in specific fields.

[0064] ; The system sends feature vectors to two devices and For comparison, if the distance between the feature vectors is less than the set threshold ϵ, the two data are considered redundant.

[0065] ; In the formula represents the Euclidean distance, and ϵ is the similarity threshold set by the system to control the sensitivity of detection.

[0066] 2.4.6 Redundant Data Processing For data marked as redundant , the system can choose to discard it directly or store it as redundant information for further analysis when needed.

[0067] Dynamic Adjustment of Redundancy Detection Strategy Since the data collected by the equipment is of various types and inconsistent frequencies, the system allows the redundant detection strategy to be dynamically adjusted according to the different types of equipment and the complexity of the data. ① Simple data (such as numerical data such as vehicle location and speed) is mainly screened using hash functions. ② Complex data (such as images, sensor signals and other image-based data) uses feature vector comparison for more in-depth detection.

[0068] During system operation, the weight of hash function and feature vector comparison can be dynamically adjusted according to data processing load and accuracy requirements. When the system load is high, the frequency of feature vector comparison is reduced and lighter hash detection is used instead.

[0069] 2.5 Data Standardization Since different devices have different data formats, after the data is transmitted to the edge node, it is necessary to convert the data in different formats (such as JSON, XML, CSV, etc.) into a unified format. The edge node has the function of data format conversion, which can convert data of different protocols and formats into a standardized format that can be used later: ; In the formula, is the raw sensor data; It is the mean value of sensor data within a certain period of time; Represents the standard deviation of the data within the time period.

[0070] S3, data tiering mechanism The data stratification mechanism is mainly used to divide data into high real-time data layer and low real-time data layer according to the task requirements of the data, and adopt different transmission processing methods to ensure that high real-time tasks are processed first, while low real-time tasks can be delayed through standardized processing.

[0071] 3.1 Data Layering Mechanism The system divides data into high-real-time data layers and low-real-time data layers based on the real-time nature of the data and the priority of the task. High-real-time data is transmitted through fast communication protocols (such as MQTT), while low-real-time data is transmitted through more secure and delay-tolerant protocols (such as OPC UA). This layered mechanism ensures the efficient operation of the system, which can not only ensure the rapid response of emergency tasks, but also handle the data transmission needs of long-term tasks.

[0072] 3.2 Data layering process (1) High-real-time data processing: The system identifies high-real-time data that needs to be processed first, and uses a low-latency protocol (such as MQTT) dynamically selected by reinforcement learning for point-to-point rapid transmission. The formula is as follows: ; In the formula, Representing high real-time data, MQTT is a low-latency communication protocol dynamically selected for reinforcement learning.

[0073] (2) Low real-time data processing: Standardized processing is performed through OPC UA to ensure that the data can be used seamlessly by all devices in the system. Data standardized by OPC UA can be transmitted through the publish / subscribe mode to ensure system data consistency. The formula is as follows: ; In the formula, OPC UA data format conversion function, converting the original data Convert to standard format .

[0074] Through the data stratification mechanism, the appropriate transmission method and processing means can be selected according to the real-time and priority of the data. The system can ensure that high-real-time tasks are given priority, while low-real-time data is standardized and transmitted with delay.

[0075] S4. Protocol priority mechanism based on deep Q network After data classification is completed, the system determines the optimal communication protocol suitable for the current network conditions through the protocol priority mechanism. This mechanism dynamically selects the appropriate protocol based on the real-time status of the network (such as bandwidth, delay, jitter, etc.) and the real-time and security requirements of the data task.

[0076] Through reinforcement learning, the most appropriate protocol is dynamically selected for data transmission according to network conditions, device priorities and task requirements. It is mainly used to optimize the use of protocols, so that the system can automatically select communication protocols in different network environments to ensure maximum communication efficiency.

[0077] 4.1 Determination of protocol priority The system optimizes the priorities of different communication protocols through the Deep Q Network (DQN). According to the current network status, bandwidth, latency, etc. of the device, DQN learns and dynamically adjusts the priority of each protocol to select the optimal protocol. The formula is as follows: ; In the formula, Indicates the communication protocols that the system can select, such as MQTT, HTTP, OPC UA, etc.) , , They are the weights of OPC UA, MQTT and HTTP protocols respectively. The system will dynamically adjust these weights according to the current network status and task requirements; Status Indicates the current network status (bandwidth, latency, packet loss rate) and task requirements; According to the current status Weights learned by DQN.

[0078] Through this formula, DQN can dynamically adjust and select the protocol with the best long-term benefits based on network conditions and historical experience.

[0079] S5. Protocol conversion based on OPC UA 5.1 Introduction of OPC UA and unification of data formats After the data transmission is completed, due to the heterogeneity of the devices, there are still differences in the data format and communication protocol. To solve this problem, the OPC UA (Unified Architecture) protocol is used to standardize the data format and unify the communication protocol; OPC UA provides a common data model that can map the data of different devices into a unified format, so that the status, parameters and operations of each device can be accessed through a standardized interface. The system standardizes the data format through OPC UA to ensure seamless interaction between heterogeneous devices under different protocols. OPC UA is responsible for converting all transmitted data into a unified standardized format to ensure interoperability between devices. The specific implementation is listed in the following aspects: Devices are abstracted as nodes: Each device (such as traffic lights, sensors, and vehicles) is abstracted as a node in the OPC UA information model. The node contains the status of the device (such as the red and green light status of the traffic light, the speed and position of the vehicle, the detection data of the sensor, etc.) and the operation method (such as changing the status of the traffic light, controlling the speed of the vehicle, etc.). Intelligent mechanism for protocol identification The primary goal of protocol identification is to accurately determine the communication protocol used by the device through an automated identification algorithm. Since different devices (such as traffic lights, sensors, vehicle terminals, etc.) may use multiple communication protocols (such as MQTT, HTTP, Modbus, CAN bus, etc.), the system must have the ability to quickly identify these protocols.

[0080] Identification process Communication header analysis: When the device is connected, the system analyzes the header information of the communication data packet to extract the characteristic fields of the protocol (such as header format, data structure, flag bit, etc.). These characteristic fields determine the type of protocol used by the device.

[0081] Assume the communication header is , which contains a set of characteristic fields , through the protocol matching algorithm, the system determines the protocol type based on field matching , as detailed below: ; In the formula, A header indicating data transmitted by a device; These are characteristic fields of different protocols, and the system matches the protocols through these fields.

[0082] Whether it is a new device or an existing device, the system can automatically complete protocol identification when it is connected, ensuring that the device can communicate with the platform immediately. By parsing the communication header and matching the algorithm, the system can automatically and quickly identify the protocol type of the device without manual intervention.

[0083] 5.2 Data Abstraction and Standardization The purpose of data abstraction is to extract device data from different protocols into common standardized fields. In order to effectively abstract data from different protocols, this system uses the following common abstraction algorithms: Field mapping abstraction: By defining field mapping tables for different protocols, specific fields in the protocol (such as timestamp, location, and status) are mapped to standardized common fields. For example, the Payload of MQTT or the Value of Modbus are both mapped to Data Value. Hierarchical abstraction: For data with a hierarchical structure (such as MQTT data in JSON format), the system uses a hierarchical abstraction algorithm to parse its hierarchical structure into a standardized node tree. For example, nested JSON data is abstracted into a hierarchical node structure of OPC UA. Type abstraction: This algorithm is used to unify data types (such as integers, floating-point numbers, and strings) in different protocols into standardized data types, and perform data type conversions when necessary. For example, convert a 16-bit integer in Modbus to a standardized floating-point representation of OPC UA. Assume that the original protocol data of the device is , through the data abstraction module After processing, unified standardized data is generated The formula is: ; In the formula, Represents a combination of abstraction algorithms, including field mapping, hierarchical abstraction, and type abstraction, to ensure uniform data formats.

[0084] 5.3 Implementation of protocol conversion algorithm After completing data abstraction, the system converts standardized data into OPC UA format through a protocol conversion algorithm to resolve differences in fields, data formats, and transmission methods among different protocols, ensuring that device data is compatible and can participate in intelligent interactions within the system.

[0085] 5.3.1 Field Mapping Conversion Map the abstracted standardized fields to the OPC UA data model nodes. For example, map DataValue to the OPC UA Value node.

[0086] ; In the formula, It is the abstracted field; It is a standard node in OPC UA.

[0087] 5.3.2 Data Format Conversion Data format conversion deals with the differences in encoding, byte order, and data types between different protocols. It ensures that data from different devices in the system can be correctly decoded and transmitted.

[0088] Assume equipment Usage Agreement , its data format is , after the data format conversion module, unified into a standardized format The formula is as follows: ; In the formula, It is a data format conversion function, which is responsible for converting the data formats of different protocols (such as byte order, data encoding) into the standard format of the system.

[0089] For example: The Modbus device uses the big-endian byte order, while the system requires the little-endian byte order. The data 0x1234 is converted to 0x3412.

[0090] 5.3.3 Transmission Protocol Conversion The transmission protocol conversion deals with the differences in transmission modes of different protocols. For example, MQTT uses the publish / subscribe mode, while Modbus uses the request / response mode. The system uses this conversion mechanism to unify the transmission mode of devices.

[0091] For example, MQTT uses a publish / subscribe model, while Modbus uses a request / response model. Map it into a unified transmission logic.

[0092] The protocol identification and conversion mechanism ensures that the data of heterogeneous devices can be processed in a standardized manner by using data abstraction algorithms (field mapping, hierarchical abstraction, and type abstraction). The protocol conversion algorithm ensures that data can be correctly transmitted in the system and participate in intelligent interaction by processing differences in data formats and transmission protocols. The entire conversion mechanism not only supports full interconnection between devices, but also provides a reliable foundation for the intelligent operation of the system.

[0093] S6. Publish / Subscribe Mechanism and QoS Priority Mechanism Based on OPC UA 6.1 Publish / Subscribe Model OPC UA provides an asynchronous publish / subscribe model that allows devices to publish or subscribe to data of interest in real time. For example, sensors can publish traffic flow information regularly, and vehicles and traffic lights can subscribe to this data and make corresponding adjustments. Through this publish / subscribe mechanism of OPC UA, communication between devices no longer relies on a centralized manager, but obtains the required data through a subscription mechanism, which greatly improves the flexibility and responsiveness of the system; through the OPC UA network, each intelligent agent can subscribe to the required data and interact in real time. Each transportation system (such as a bus system) subscribes to data through the OPC UA client and optimizes control based on the subscribed data. The subscription model formula is as follows: ; In the formula, is the subscribed dataset; It's a time period.

[0094] 6.2QoS Priority Mechanism During data publishing and transmission, the system introduces a QoS (Quality of Service) priority mechanism to ensure that high-priority data can be transmitted first. The QoS priority mechanism dynamically adjusts the order of data transmission according to the real-time requirements of the task, bandwidth limitations, and device load to ensure that the system can maintain stable performance under different network conditions: ①High priority data : Includes emergency traffic condition data (such as accident alarms), which need to be transmitted in real time.

[0095] ② Medium priority data : Includes real-time traffic flow data, requiring low latency and reliable transmission.

[0096] ③Low priority data : Includes common device status data (such as the fuel status of a vehicle), allowing for a certain delay.

[0097] The QoS scoring formula is as follows: ; In the formula, Represents the QoS score of the data, which determines the transmission priority of the data in the system; Indicates the priority of data, reflecting its importance in the system; Represents the bandwidth allocated to data. The higher the bandwidth, the faster the transmission. Represents the data transmission delay, the lower the better; Represents the packet loss rate, the lower the better; Represents transmission jitter. The smaller the delay variation, the better. are the weights of delay, packet loss rate and jitter respectively.

[0098] Through this mechanism, the system can dynamically adjust the data transmission order to ensure that high-priority data (such as traffic accident alarms) can be transmitted in real time, while ordinary data is transmitted when bandwidth is sufficient.

[0099] S7. Data association and data fusion 7.1 Data Matching and Association The system uses a combination of graph neural networks (GNN) and Mahalanobis distance to match and associate data. GNN learns the complex relationships between devices and extracts high-order features, while Mahalanobis distance is used to measure the similarity between devices. This method can efficiently match and associate data between multi-protocol devices.

[0100] The input is heterogeneous data converted through OPC UA, and the output is matched associated data.

[0101] 7.2 Feature Extraction Based on GNN First, the system models all devices and their communication relationships as a graph structure, and learns the data interaction patterns between devices through a graph neural network (GNN). Each device is regarded as a node in the graph, and the communication between different devices is regarded as an edge. The weight of the edge can be determined based on information such as communication frequency and data type.

[0102] The GNN model can capture the complex associations between devices and generate high-order features for each node (device) by transferring and updating node features layer by layer.

[0103] GNN feature extraction formula: ; In the formula, is the adjacency matrix, which represents the communication connection relationship between devices; is the degree matrix, which indicates the number of connections of a node; For the The node feature matrix of the layer; is the weight matrix; It is an activation function used to perform nonlinear transformation on features.

[0104] Through multi-layer GNN transmission, the system can capture deeper correlation information between devices and generate high-level feature representations for each device. .

[0105] 7.3 Data Matching After extracting the high-order features of each device, the system matches these features using the Mahalanobis distance, which can effectively measure the differences between multidimensional data with covariance relationships. Here, it is used to measure the matching degree between the high-order features of each device and the features of the target device.

[0106] 7.4 Calculation of the covariance matrix Before calculating the Mahalanobis distance, we first need to calculate the covariance matrix of the sample eigenvector The covariance matrix is ​​used to describe the correlation between each feature dimension: ; In the formula It is The high-order feature vectors of the devices; is the mean vector of the devices, is the sample size. The covariance matrix It is an n×n symmetric matrix, which represents the correlation between device data features.

[0107] 7.5 Mahalanobis distance algorithm The Mahalanobis distance is used to calculate the distance between two data points. and The multidimensional distance between . It is defined as: ; In the formula, and are the high-order eigenvectors of the two devices, respectively; is the covariance matrix The inverse matrix of ensures that the correlation between features is taken into account; represents the difference between the eigenvectors.

[0108] 7.6 Data Matching Determination By calculating the Mahalanobis distance between two data points , if the distance is less than the preset threshold ϵ, the two data points are considered to be matched, as follows: if <ϵ, then and match.

[0109] In the formula, the threshold ϵ is a matching criterion set according to system needs, which is usually obtained through historical data or model training.

[0110] This combined method can not only learn deeper patterns from the feature relationships of devices, but also accurately match and associate through statistical measurement methods, thereby improving the accuracy and efficiency of data matching and association. This innovation organically combines deep learning with statistical distance, which can better adapt to the data interaction needs in heterogeneous device environments.

[0111] 7.7 Data Fusion In traffic management systems, different data sources often generate redundant or even conflicting data. This conflict may come from equipment measurement errors, data delays, noise interference, etc. Therefore, in the data fusion process, a method is needed to deal with these conflicts and combine evidence from multiple data sources to make more reasonable judgments. Therefore, the Dempster-Shafer evidence theory is used to fuse evidence from multiple data sources to form a more comprehensive and conflict-free comprehensive judgment. Dempster-Shafer evidence theory formula: ; In the formula, It indicates the credibility of hypothesis A, indicating the degree to which multiple pieces of evidence support A; and are the credibility assignments of data source 1 and data source 2 to hypothesis B and C, respectively; is the conflict distribution, which represents the total amount of conflict when there is a complete conflict between the evidences.

[0112] After the above steps, a relatively complete traffic data view has been formed in the system.

[0113] S8. Global decision-making based on multi-agent technology 8.1 Information Sharing in Multi-Agent Technology The application of multi-agent technology realizes autonomous decision-making and collaborative optimization of heterogeneous devices in the traffic system through a distributed control architecture. Each agent (such as vehicles, traffic lights, sensors) has autonomous perception and decision-making capabilities, and can adjust its behavior in real time based on local information and information shared by other agents without relying on a central controller. This distributed system significantly reduces dependence on centralized data processing through local data processing, and improves the response speed and stability of the system. At the same time, through information sharing and collaborative optimization among multiple agents, it ensures that the optimal operating state of the global system is achieved while optimizing local performance, thereby effectively reducing traffic congestion and improving traffic flow efficiency.

[0114] The information sharing formula is as follows: ; In the formula, It is At the iteration Status updates; It is an intelligent agent The neighbor set of A collection of intelligent agents that interact with each other; is the weight of information sharing, which usually satisfies , represents weighted equalization or adaptive allocation in information sharing; Neighbor Agent The state at the kth iteration.

[0115] 8.2 Global Collaborative Optimization of Multi-Agent Technology In order to ensure the optimal operation state of the global system, multiple agents need to work together to achieve the global optimal goal through a distributed optimization algorithm. The collaborative optimization formula is as follows: ; Where N is the number of agents; Each agent Local objective functions (such as minimizing vehicle waiting time, optimizing traffic light timing, etc.); It is an intelligent agent state variables; It is an intelligent agent The feasible solution space.

[0116] 8.3 Distributed Convex Optimization Decision Making Through a distributed convex optimization algorithm, multiple agents collaborate to make optimal traffic decisions, where the goal is to minimize the waiting time of different modes of transportation.

[0117] The objective function to minimize the waiting time is: ; Where N is the number of agents; Indicates The objective function is to minimize the total waiting time of the entire system by optimizing the decision-making of each agent.

[0118] 8.4 Iterative Update of Gradient Descent In order to minimize the objective function, the gradient descent method is used for optimization. There is a state vector , then the global optimal solution can be gradually approached by updating the state vector. The state vector of is updated at the kth iteration by the following formula: ; In the formula, It is an intelligent agent The state at the kth iteration; is the step size coefficient, which controls the update amplitude of each iteration; is the waiting time objective function Status The gradient of It is an intelligent agent The updated state at the k+1th iteration. At each iteration, the agent's state vector is updated along the gradient direction of the objective function, thereby gradually reducing the waiting time.

[0119] 8.5 Distributed Collaborative Update In order to ensure that the update of each agent does not conflict with the decision of other agents, in a multi-agent system, each agent will communicate with its neighboring agents. Share information and achieve global optimization through coordinated optimization.

[0120] The update formula expands to: ; In the formula, It is an intelligent agent The neighbor set of Other agents with which they interact directly; It is an intelligent agent and neighbors The information exchange weight between them indicates the degree of influence between the two agents; Neighbor Agent The state at the kth iteration. In addition to each agent performing gradient descent optimization based on its own objective function, the state of neighboring agents is also considered, and collaborative optimization is performed through information exchange to ensure mutual coordination in the global system.

[0121] 8.6 Real-time Update Mechanism Since data is updated in real time in actual application scenarios, the agent must dynamically adjust its decision based on the latest data. Each agent regularly receives data from the environment (such as vehicle speed, traffic flow, and signal light status) and uses this data to update the state variables in the objective function.

[0122] The real-time update formula is: ; In the formula, Indicates that new data is generated at each time point; is the agent at time The data state at the time; Gradient Represents a gradient update based on the latest data.

[0123] The real-time update mechanism enables the agent to adjust its decisions in a timely manner after receiving new environmental information, ensuring that the entire system can respond to changes in traffic flow, thereby reducing overall congestion and waiting time.

[0124] 8.7 Iterative Convergence In order to ensure that the system can reach the global optimum, it is necessary to ensure that the gradient of each iterative update gradually approaches zero, that is, the state of the agent no longer changes significantly, thus achieving convergence.

[0125] The convergence conditions are: ; This means that the gradient of each agent will approach zero in the later stages of the iteration, indicating that the optimization process has reached the global optimal solution through distributed collaboration.

[0126] Through the distributed convex optimization algorithm combined with the gradient descent method, each agent can continuously update its decision based on real-time traffic information and information from neighboring agents to minimize its own waiting time. At the same time, through information sharing and collaborative optimization between agents, the goal of the global system (i.e., the sum of the waiting time of all agents) can be optimized.

[0127] This embodiment first collects data from heterogeneous devices of multiple protocols through the deployment of edge computing nodes, and applies autoencoders to automatically detect and correct anomalies in data preprocessing, combining hash functions and feature vectors to detect redundant data. In the data stratification mechanism, the system prioritizes data according to the real-time requirements of the task to ensure that important tasks are processed first.

[0128] Next, the priority of the communication protocol is dynamically adjusted through the deep Q network (DQN), optimizing the intelligent processing of protocol selection. The system uses the OPC UA protocol to uniformly format heterogeneous data to achieve seamless connection and intelligent communication of multiple protocols. During the data transmission process, the QoS priority mechanism is used to ensure the data transmission order and service quality.

[0129] In the data matching and association stage, the system combines the graph neural network (GNN) and the Mahalanobis distance algorithm to learn the complex interaction relationship between devices through the graph structure and perform efficient data matching and association. In the global decision-making process, the system shares information through multi-agent technology and uses distributed optimization algorithms to achieve collaborative work between heterogeneous devices, generating globally optimal scheduling and resource allocation decisions.

[0130] Therefore, the present invention adopts the above-mentioned multi-protocol intelligent interactive heterogeneous device full interconnection method and system to solve the key problems in the existing traffic management system, such as the inability of heterogeneous devices to communicate efficiently and the inability of multi-protocol devices to interconnect. Compared with the existing technology, this method has obvious advantages in the intelligent interaction of multi-protocol devices, real-time data processing and global optimization of heterogeneous device cooperation, and improves the overall efficiency and flexibility of the system.

[0131] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solution of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solution to deviate from the spirit and scope of the technical solution of the present invention.

Claims

1. A method for fully interconnecting heterogeneous devices with multi-protocol intelligent interaction, characterized in that: The following steps are involved: S1, collect data and deploy edge computing nodes; S2, preprocessing the collected data; S3, through the data layering mechanism, the data is divided into a high real-time data layer and a low real-time data layer according to the task requirements of the data, and then different transmission processing methods are adopted; S4. After classification, the optimal communication protocol suitable for the current network conditions is determined through the protocol priority mechanism; S5. Use OPC UA protocol to standardize data formats and unify communication protocols; S6, OPC UA-based publish / subscribe mechanism and QoS priority mechanism, allowing devices to publish or subscribe to data in real time and ensure that high-priority data is transmitted first; S7, use graph neural network and Mahalanobis distance to match and associate data, and use Dempster-Shafer evidence theory to fuse data; S8. Make global decisions based on multi-agent technology.

2. The method for fully interconnecting heterogeneous devices with multi-protocol intelligent interaction according to claim 1 is characterized in that: The delay optimization goal of edge computing for deploying edge computing nodes in step S1 is: ； In the formula, is the total delay for data to be transmitted from the device to the edge node and processed, is the amount of data to be transmitted, is the network bandwidth, The computation processing delay of the edge node.

3. The method for fully interconnecting heterogeneous devices with multi-protocol intelligent interaction according to claim 1 is characterized in that: The high-real-time data in step S3 is transmitted point-to-point quickly using a low-latency protocol dynamically selected by reinforcement learning. The formula is: ； In the formula, For high real-time data, MQTT is a low-latency communication protocol dynamically selected for reinforcement learning; Low real-time data is standardized through OPC UA, and the data standardized by OPC UA is transmitted through the publish / subscribe mode. The formula is as follows: ； In the formula, This is the OPC UA data format conversion function, which converts the original data Convert to standard format .

4. The method for fully interconnecting heterogeneous devices with multi-protocol intelligent interaction according to claim 1 is characterized in that: Step S4 specifically includes: The system optimizes different communication protocols through a deep Q network, the formula is: ； In the formula, The communication protocols that can be selected for the system include OPC UA, MQTT and HTTP protocols; , , The weights of OPC UA, MQTT and HTTP protocols respectively; Status For current network conditions and mission requirements; According to the current status Weights learned by DQN.

5. The method for fully interconnecting heterogeneous devices with multi-protocol intelligent interaction according to claim 1 is characterized in that: Step S5 specifically includes: S51. The introduction of OPC UA and the unification of data formats. OPC UA converts all transmitted data into a unified standardized format. S52, data abstraction and standardization, data abstraction is to extract device data in different protocols into common standardized fields, specifically using field mapping abstraction, hierarchical abstraction and type abstraction; S53. After completing data abstraction, the system converts the standardized data into OPC UA format through a protocol conversion algorithm, including: S531, field mapping conversion, mapping the abstracted standardized fields to the OPC UA data model nodes, as follows: ； In the formula, is the abstracted field; It is a standard node in OPC UA; S532, data format conversion, used to handle the differences in encoding, byte order and data type between different protocols. Assume that the device Usage Agreement , its data format is , after data format conversion, unified into a standardized format , the formula is as follows: ； In the formula, It is a data format conversion function, which is used to convert data formats of different protocols into the standard format of the system; S533, transmission protocol conversion, used to handle the differences in transmission modes of different protocols. Assume that the device The transmission protocol used is , its transmission mode is , the system converts the transmission protocol into a standardized transmission mode , the formula is as follows: ； In the formula, It is the transmission protocol conversion function.

6. The method for fully interconnecting heterogeneous devices with multi-protocol intelligent interaction according to claim 1 is characterized in that: Step S6 specifically includes: S61. OPC UA provides a publish / subscribe model. The subscription model formula is: ； In the formula, is the subscribed dataset; It is the time period; S62, introduce QoS priority mechanism, dynamically adjust the order of data transmission according to the real-time requirements of the task, bandwidth limitations and device load, divided into: High priority data : Including emergency traffic data, real-time transmission; Medium priority data : Includes real-time traffic flow data and low-latency transmission; Low priority data : Includes general device status data, allowing for delays; The QoS scoring formula determines the transmission priority of data in the system: ； In the formula, Score the QoS of the data; Prioritize the data; The bandwidth allocated to data; The transmission delay of data; Represents the packet loss rate; For transmission jitter; are the weights of transmission delay, packet loss rate and transmission jitter respectively.

7. The method for fully interconnecting heterogeneous devices with multi-protocol intelligent interaction according to claim 1 is characterized in that: The data matching and association in step S7 specifically includes: First, the system models all devices and their communication relationships as a graph structure, and learns the data interaction patterns between devices through the graph neural network. The graph neural network captures the complex associations between devices by transferring and updating node features layer by layer, and generates high-order features for each node. The feature extraction formula of the graph neural network is: ；。 In the formula, is the adjacency matrix; is the degree matrix; For the The node feature matrix of the layer; is the weight matrix; is the activation function; After extracting the high-order features of each device, the system uses the Mahalanobis distance to match these features. Before calculating the Mahalanobis distance, the covariance matrix of the sample feature vector is calculated. ,for: ； In the formula, For the The high-order feature vector of each device; is the mean vector of the device; is the sample size; The Mahalanobis distance is used to calculate the distance between two data points. and The multidimensional distance between is defined as: ； In the formula, and are the high-order eigenvectors of the two devices respectively; is the covariance matrix The inverse matrix of is the difference between the eigenvectors; By calculating the Mahalanobis distance between two data points ,if <ϵ, then and match; Wherein, the threshold ϵ is a matching criterion set according to system needs.

8. The method for fully interconnecting heterogeneous devices with multi-protocol intelligent interaction according to claim 1, characterized in that: The data fusion in step S7 specifically uses the Dempster-Shafer evidence theory to fuse the evidence from multiple data sources. The Dempster-Shafer evidence theory formula is: ； In the formula, The reliability of hypothesis A indicates the degree to which multiple pieces of evidence support A; and are the credibility assignments of hypothesis B and C from data source 1 and data source 2 respectively; Assign conflict, which represents the total amount of conflict when there is a complete conflict between the evidences.

9. The method for fully interconnecting heterogeneous devices with multi-protocol intelligent interaction according to claim 1, characterized in that: Step S8 is specifically as follows: S81. Information sharing of multi-agent technology. The information sharing formula is: ； In the formula, For the At the iteration Status updates; For intelligent agents The neighbor set of A collection of intelligent agents that interact with each other; is the weight of information sharing, indicating weight equalization or adaptive allocation in information sharing; Neighbor Agent The state at the kth iteration; S82. Global collaborative optimization of multi-agent technology. The collaborative optimization formula is: ； Where N is the number of agents; For each agent The local objective function of For intelligent agents state variables; For intelligent agents The feasible solution space of ; S83, distributed convex optimization decision, is used to minimize the waiting time of different transportation modes. The objective function for minimizing the waiting time is: ； Where N is the number of agents; For the The waiting time of each agent; S84, using the gradient descent method to optimize and minimize the objective function, each agent The state vector of is updated at the kth iteration by the following formula: ； In the formula, For intelligent agents The state at the kth iteration; is the step length coefficient; The waiting time objective function Status The gradient of For intelligent agents The updated state at the k+1th iteration; S85, distributed collaborative update, the update formula is expanded to: ； In the formula, For intelligent agents The set of neighbors of ; For intelligent agents and neighbors The information exchange weight between them indicates the degree of influence between the two agents; Neighbor Agent The state at the kth iteration; S86, real-time update, the formula is: ； In the formula, Generate new data for each time point; For the agent at time The data state at the time; Gradient It is a gradient update based on the latest data; S87, iterative convergence, the convergence condition is: ； It indicates that the optimization process reaches the global optimal solution through distributed collaboration.

10. A multi-protocol intelligent interactive heterogeneous device full interconnection system, characterized in that: include: The access layer includes several heterogeneous devices; The technical layer includes the deployment of data preprocessing modules and edge computing nodes; Protocol conversion layer, including OPC UA protocol conversion module and data fusion and matching module; The business layer also includes data fusion and matching modules, as well as multi-agent distributed optimization modules; The application layer includes the global optimization control module and the real-time decision and feedback module.

Images