An industrial internet of things-oriented high-concurrency semantic transmission method and system

Abstract

The application discloses an industrial internet of things oriented high concurrency semantic transmission method and system. The application realizes data transmission between industrial internet of things device nodes and edge servers through a semantic communication encoding model and a decoding model, reduces industrial data transmission redundancy by transmitting semantic data, ensures transmission delay and reliability of high priority tasks in an industrial scene in a high concurrency priority differentiation conflict resolution process by a pre-set priority identifier ID, and balances the same priority task resource scheduling strategy. By constructing a high concurrency semantic transmission resource scheduling optimization model, the system minimum error information age is output by an average information age minimization model, the real-time performance of semantic transmission is ensured, and the maximum average structural similarity is obtained by an average structural similarity maximization model, the accuracy of semantic transmission is ensured. The system solves the conflict while jointly optimizing the real-time performance and accuracy of semantic transmission.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of industrial Internet of Things communication, and relates to a high-concurrency semantic transmission method and system for industrial Internet of Things. BACKGROUND

[0002] With the vigorous development of intelligent manufacturing, industrial digital reform is gradually deepening, and industrial Internet of Things constructs an interconnected industrial ecological system by connecting a large number of sensors, actuators and industrial devices. However, the simultaneous transmission of data by a large number of connected devices will cause serious network conflicts and resource competition problems. In the traditional industrial Ethernet network, nodes transmit raw data based on a competition mechanism, and the competition between nodes is serious. The system does not distinguish the types of data to be transmitted, and with the continuous generation of data, the total amount of data is large. Direct transmission of these data will cause great waste of resources. Time-sensitive network introduces time synchronization, traffic scheduling and fault-tolerant communication mechanisms, and distinguishes and transmits data according to the priority requirement of real-time performance, providing low-latency and high-reliable data transmission services to guarantee the real-time performance of data with higher priority. However, time-sensitive network does not distinguish the amount of information carried by the transmission information. In industrial systems, newly added data often has a large amount of information entropy, while the total amount of background data has a low semantic information density. Although these data have a higher requirement for real-time performance, the transmission of a large amount of background data will still cause a large waste of resources.

[0003] Semantic communication can significantly improve the transmission efficiency and resource utilization of industrial Internet of Things by transmitting the semantic features of data rather than raw bit data. However, in industrial Internet of Things, when a large number of nodes simultaneously initiate semantic transmission, there will still be serious access conflicts. Traditional communication transmission adopts random access protocols such as Slotted ALOHA, Carrier Sense Multiple Access (CSMA) to manage node concurrency. Existing researches have proposed an analysis framework for semantic communication random access, and designed semantic communication compression and resource allocation algorithms for users with the same service level. However, the requirements of industrial downstream tasks for data transmission real-time performance and reliability are different, and traditional random access protocols cannot effectively distinguish the priority of tasks, making it difficult to meet the urgent needs of industrial applications for low-latency and high-reliable transmission of key tasks.

[0004] Therefore, there is an urgent need for a comprehensive solution that can effectively manage a large number of node accesses, differentially schedule according to task priority, and at the same time guarantee information freshness (real-time performance) and semantic accuracy, reduce redundancy while meeting the low-latency and high-reliable transmission requirements of industrial Internet of Things. SUMMARY

[0005] The purpose of the present application is to overcome the shortcomings of the prior art, provide an industrial internet of things oriented high concurrency semantic transmission method and system, meet the low latency, low redundancy, high reliability transmission requirements of the industrial internet of things.

[0006] In order to achieve the above purpose, the present application provides the following technical scheme:

[0007] An industrial internet of things oriented high concurrency semantic transmission method, specifically comprising the following steps:

[0008] Step one, store the pre-set priority identifier ID in each industrial internet of things device node.

[0009] At the same time, an average error information age AoII based high concurrency semantic transmission resource scheduling optimization model is constructed in the edge server, the high concurrency semantic transmission resource scheduling optimization model includes an average information age AoI minimization model and an average structural similarity index SSIM maximization model. The average AoI minimization model enables the device node to solve the conflict in time and successfully transmit in the conflict resolution process, ensuring the real-time performance of semantic transmission; the average SSIM maximization model enables the semantic data transmitted by the device node to be recovered as much as possible, ensuring the accuracy of semantic transmission.

[0010] Step two, the DeepJSCC encoding model in each industrial internet of things device node extracts the semantic feature information of the industrial data collected by the sensor in the device node, and stores the extracted industrial data semantic feature information in the data packet cache area;

[0011] Step three, all industrial internet of things device nodes with non-empty data packet cache area initiate access edge server request through the transmission module, and send the preamble to the edge server respectively, the preamble sent by each device node contains the priority identifier ID of the device node;

[0012] Step four, after the receiving module of the edge server receives the access request sent by the industrial internet of things device node, the industrial internet of things device node is scheduled for transmission, and the corresponding ACK is replied to all device nodes that send access request, and the semantic transmission scheduling is performed; and the high concurrency semantic transmission resource scheduling optimization model is called to optimize the real-time performance and accuracy of semantic transmission.

[0013] Step five, the industrial internet of things device node receives the ACK sent by the edge server through the transmission module, sends the extracted semantic feature information to the edge server, and the edge server sends the end code to the device node after receiving the semantic feature information sent by the device node, and completes the semantic transmission.

[0014] Repeat the operations of steps two to five to realize the high concurrency semantic transmission of the industrial internet of things.

[0015] Furthermore, in actual industrial processes, the high-concurrency semantic transmission resource scheduling optimization model varies with the number of device nodes. Dynamic changes, dynamic changes in device nodes The probability that the data cache is empty. Equipment Nodes Transmission power Values ​​and data compression ratios yield different results. Probability, maximum average structural similarity index and minimum average information age under different transmission powers and data compression rates.

[0016] Furthermore, the scheduling process described in step four is as follows:

[0017] If there is only one device node with a non-empty data cache area and no conflict occurs, the edge server replies with an ACK that the device node sent semantic information.

[0018] If more than one device node in the data cache area is not empty, a conflict occurs. When there are different priorities among the conflicting nodes, the edge server replies with an ACK indicating that the high-priority device node has sent semantic information, and replies with an ACK indicating that the low-priority node has entered a waiting state.

[0019] When all conflicting nodes have the same priority, the edge server sends an ACK to all conflicting nodes to enter backoff mode. Each node randomly selects a backoff time within the backoff time window to wait. After the waiting time expires, new access requests are initiated sequentially according to the backoff time selected by the device node.

[0020] The backoff time setting process is as follows: Set the number of device nodes. The proportionality coefficient with the size of the retreat time window Based on the number of nodes and proportionality coefficient Set the maximum backoff time window size .

[0021] Furthermore, the optimization process for the real-time performance and accuracy of the high-concurrency semantic transmission resource scheduling optimization model described in step four is as follows:

[0022] Step 4.1: Calculate all device nodes that have issued access requests. Probability of successful transmission within a cycle .

[0023] Step 4.2: Increase step size using a fixed increment. To narrow down the feasible region, based on the obtained Dynamic changes Value linear search yields different Minimum average information age under size The obtained minimum average information age makes the device nodes timely solve conflicts and successfully transmit in the conflict resolution process, thereby guaranteeing real-time performance of semantic transmission;

[0024] Step 4.3, the maximum average structural similarity index of all device nodes sending an access request is calculated, and the obtained maximum average SSIM makes the semantic data transmitted by the device nodes be recovered as much as possible, thereby guaranteeing accuracy of semantic transmission.

[0025] Further, in the step four scheduling process, the total time from initiation of an access request to conflict resolution of each device node is defined as a cycle period time; in the high-concurrency priority conflict resolution process, time is divided according to time slots, and transmission and backoff time are both counted by time slots; if a device node fails to send in a cycle period, the data packet is processed as a lost packet.

[0026] As a preferred, in the step one, priority identification is performed according to a transmission task of the device node. For example, alarm-type device nodes are assigned high priority, and collection-type device nodes are assigned low priority.

[0027] On the other hand, the application provides a system for running the high-concurrency semantic transmission method for the industrial Internet of Things, comprising an edge server and a plurality of industrial Internet of Things device nodes.

[0028] A sensor module for collecting industrial data is arranged in each industrial Internet of Things device node, and a pre-calibrated priority identifier ID is stored.

[0029] A DeepJSCC encoding model is deployed in each industrial Internet of Things device node, for extracting semantic feature information of industrial data; meanwhile, a transmission module for transmitting semantic data is used to transmit the extracted semantic feature information to the edge server; the industrial Internet of Things device node has a data packet buffer area inside, and when the buffer area is not empty, the device node initiates an access request to the edge Internet of Things;

[0030] A high-concurrency semantic transmission resource scheduling optimization model is deployed in the edge server, and when a single or multiple device nodes send an access request to the edge server, the edge server performs transmission scheduling on the industrial Internet of Things device nodes and replies with corresponding ACK to all device nodes sending an access request; meanwhile, the high-concurrency semantic transmission resource scheduling optimization model is called to optimize real-time performance and accuracy of semantic transmission.

[0031] A receiving module for receiving semantic data is deployed in the edge server to receive the semantic feature information extracted by the DeepJSCC encoding model in the industrial Internet of Things device node. A DeepJSCC decoding model is deployed in the edge server to decode the received semantic feature information.

[0032] As preferred, the transmission module and the receiving module of the semantic data operate under the condition of a wireless AWGN channel.

[0033] The present application has the following advantages:

[0034] By deploying a semantic communication encoding model inside the industrial IoT device node to extract semantic information for transmission, and deploying a semantic communication decoding model on the edge server to recover data, compared with transmitting raw bit data, the present application transmits semantic data, thereby reducing the redundancy of industrial data transmission.

[0035] By using a pre-set priority identifier ID, the transmission delay and reliability of high-priority tasks in the industrial scene are ensured in the process of executing high-concurrency priority differentiation conflict resolution, the resource scheduling strategy of tasks with the same priority is balanced, and the differentiated needs of the industrial control process are met.

[0036] A high-concurrency semantic transmission resource scheduling optimization model is constructed, the minimum error information age of the system is output by the average information age minimization model to ensure the real-time performance of semantic transmission, and the maximum average SSIM is obtained by the average SSIM maximization model to ensure the accuracy of semantic transmission.

[0037] The system jointly optimizes the real-time performance and accuracy of semantic transmission while solving the conflict. BRIEF DESCRIPTION OF DRAWINGS

[0038] Figure 1 The structure block diagram of the system of the present application is shown in the figure;

[0039] Figure 2 The flowchart of the method of the present application is shown in the figure;

[0040] Figure 3 The data transmission process diagram between the industrial IoT device node and the edge server is shown in the figure;

[0041] Figure 4 The AoI geometric description provided by the embodiment of the present application is shown in the figure;

[0042] Figure 5 The fitting curve of SSIM and SNR under different compression rates provided by the embodiment of the present application is shown in the figure;

[0043] Figure 6 The minimum average AoII value of the present application under different device node quantities is compared with the minimum average AoII value of the CSMA method and the Slotted ALOHA method;

[0044] Figure 7 The different The minimum average AoII value of the CSMA method under different values and different transmission power conditions;

[0045] Figure 8 The different values provided by the embodiment of the application The minimum average AoII value of the CSMA method under different values and different transmission power conditions;

[0046] Figure 9 The different values provided by the embodiment of the application The minimum average AoII value of the Slotted ALOHA method under different values and different transmission power conditions;

[0047] Figure 10 The average SSIM value of the application under different transmission power and different compression rate conditions provided by the embodiment of the application;

[0048] Figure 11 The average semantic transmission rate value under different transmission power and different compression rate conditions provided by the embodiment of the application. DETAILED DESCRIPTION

[0049] The application will be further described below in combination with the drawings.

[0050] The application overcomes the shortcomings of the prior art and provides a high-concurrency semantic transmission method and system for industrial Internet of Things, which meets the transmission requirements of low redundancy, low latency and high reliability of industrial Internet of Things. The system comprises an edge server and a plurality of industrial Internet of Things device nodes;

[0051] Figure 1 The structure block diagram of the system of the application is shown, a sensor module for collecting industrial data is arranged in each industrial Internet of Things device node, and a pre-calibrated priority identifier ID is stored; in this embodiment, the priority identifier is identified according to the transmission task (type) of the device node, the alarm type device node is assigned a high priority, and the collection type device node is assigned a low priority.

[0052] A DeepJSCC encoding model is deployed in each industrial Internet of Things device node, which is used to extract semantic feature information of industrial data; a transmission module for transmitting semantic data transmits the extracted semantic feature information to the edge server through a wireless AWGN channel; the industrial Internet of Things device node has a data packet buffer area inside, when the buffer area is not empty, the device node initiates an access request to the edge Internet of Things;

[0053] The high-concurrency semantic transmission resource scheduling optimization model is deployed in the edge server. When a single or multiple device nodes send access requests to the edge server, the edge server schedules the transmission of the industrial Internet of Things device nodes and replies to all device nodes sending access requests with corresponding ACK (acknowledgement character). Meanwhile, the high-concurrency semantic transmission resource scheduling optimization model is called to optimize the real-time performance and accuracy of semantic transmission.

[0054] The receiving module for receiving semantic data deployed in the edge server receives the semantic feature information extracted by the DeepJSCC encoding model in the industrial Internet of Things device node through a wireless AWGN channel. The DeepJSCC decoding model deployed in the edge server decodes the received semantic feature information.

[0055] In the scheduling process, the total time from the initiation of an access request to the conflict resolution of each device node is defined as a cycle period time. In the high-concurrency priority conflict resolution process, time is divided according to time slots slot (t), and transmission and backoff time are both counted by time slots. If a device node fails to send in a cycle period, the data packet is processed as a lost packet.

[0056] Based on the above system, the following high-concurrency semantic transmission resource scheduling method for industrial Internet of Things is provided, as shown in Figure 2 The specific steps include the following steps:

[0057] Step 1: Store the pre-set priority identifier ID in each industrial Internet of Things device node. In this embodiment, the alarm device node is assigned a high priority, and the acquisition device node is assigned a low priority.

[0058] Meanwhile, an average error information age AoII (Age of Incorrect Information) based high-concurrency semantic transmission resource scheduling optimization model is constructed in the edge server. The high-concurrency semantic transmission resource scheduling optimization model includes an average information age AoI (Age of Information) minimization model and an average structural similarity index SSIM (Structure Similarity index measure) maximization model. The high-concurrency semantic transmission resource scheduling optimization model is:

[0059] ,

[0060] ,

[0061] ,

[0062] ,

[0063] ,

[0064] ,

[0065] ;

[0066] in It is a device node Average Error Message Age (AoII) is used to describe the real-time performance and accuracy of semantic transmission. The total number of all device nodes. represent Any one of the device nodes; For device nodes The structural similarity index SSIM, This is the SSIM threshold that needs to be met. If it exceeds this threshold, the device node... Only the transmitted data meets the set decoding accuracy. For device nodes Signal-to-noise ratio (SNR) between the transmission point and the edge server. This is the required SNR threshold; if it exceeds this threshold, the device node... Only then can the data be successfully transmitted; For device nodes The transmission power cannot exceed the maximum power. ; For device nodes The probability that the data buffer is empty; this value is between 0 and 1. For device nodes The probability of successful transmission within one cycle, which is between 0 and 1; This is the maximum backoff time window size. To control the ratio (set value) between device node density and backoff time window size, the backoff time window increases proportionally with the increase of device nodes, in order to alleviate node conflict.

[0067] Device Node Instantaneous error information age ,in, For device nodes The instantaneous information age, This refers to the semantic similarity between the data recovered from the edge server and the original data in the device node, which is represented here as structural similarity. ,in For device nodes The SNR value between the transmission and the edge server. For device node The selected data compression rate, , , , are compression fitting parameters, the values of which are determined by , After determination, the parameters and are obtained by fitting the numerical values of , , , specific values.

[0068] The average error information age can be further expressed as , The average information age of the device node , represents the expected value function, represents the expected value, i.e. the average value, of .

[0069] Therefore, the high-concurrency semantic transmission resource scheduling optimization model can be further expressed as:

[0070] ,

[0071] That is, the average error information age is composed of the average information age and the average structural similarity, where the average information age is used to describe the real-time performance of transmission, and the smaller the value is, the better it is, indicating that the data can be transmitted in time; the average structural similarity is used to describe the accuracy of the data after being recovered, and the larger the value is, the better it is, indicating that the data can be transmitted more accurately.

[0072] From the above formula, it can be seen that AoI and SSIM are independent of each other, and the high-concurrency semantic transmission resource scheduling optimization model can be split into an average AoI minimization model and an average SSIM maximization model.

[0073] The split high-concurrency semantic transmission resource scheduling optimization model is obtained as follows:

[0074] The average AoI minimization model is:

[0075] ,

[0076] ,

[0077] ,

[0078] ;

[0079] wherein, is the average information age of the device node average AoI, device node probability of data buffer being empty, the probability value is between 0 and 1, device node probability of successful transmission in one cycle period, the probability value is between 0 and 1, maximum backoff time window size, proportionality coefficient (set value) for controlling the density of device nodes and the size of backoff time window, as the number of device nodes increases, the backoff time window increases in direct proportion to alleviate the node conflict situation.

[0080] average SSIM maximization model:

[0081] ,

[0082] ,

[0083] ,

[0084] ;

[0085] wherein, device node structural similarity index SSIM, minimum SSIM threshold value, greater than the threshold value, the data transmitted by the device node satisfies the required decoding accuracy; SNR between the device node transmission and the edge server, SNR threshold value, greater than the threshold value, the data of the device node can be successfully transmitted; transmission power of the device node , the size of which cannot exceed the maximum power .

[0086] proportionality coefficient , according to the number of nodes and the proportionality coefficient , the maximum backoff time window size is set.

[0087] The average AoII minimization model makes the device node solve the conflict and successfully transmit in time in the conflict resolution process, and guarantees the real-time performance of semantic transmission; the average SSIM maximization model makes the semantic data transmitted by the device node be recovered as much as possible, and guarantees the accuracy of semantic transmission.

[0088] Step 2: The DeepJSCC encoding model in each industrial IoT device node extracts the semantic feature information of the industrial data collected by the sensors in that device node, and stores the extracted semantic feature information of the industrial data in the data packet cache area;

[0089] Step 3: All industrial IoT device nodes with non-empty data packet buffer areas initiate access requests to the edge server via the wireless AWGN channel, sending preambles to the edge server respectively. Each device node's preamble contains its priority identifier (ID). The data transmission process between the industrial IoT device nodes and the edge server is as follows: Figure 3 As shown;

[0090] Step 4: After receiving the access request from the industrial IoT device node, the edge server's receiving module performs transmission scheduling for the industrial IoT device node, replies with the corresponding ACK (acknowledgment character) to all device nodes that have sent access requests, performs semantic transmission scheduling, and calls the high-concurrency semantic transmission resource scheduling optimization model to optimize the real-time performance and accuracy of semantic transmission.

[0091] If there is only one device node with a non-empty data cache area (only a single device node sends an access request), and no conflict occurs, the edge server replies to the device node with an ACK (acknowledgment character) that sends semantic information.

[0092] If more than one device node in the data cache area is not empty, a conflict occurs. When there are different priorities among the conflicting nodes, the edge server replies with an ACK (acknowledgment character) indicating that the higher-priority device node has sent semantic information, and replies with an ACK (acknowledgment character) indicating that the lower-priority node has entered a waiting state.

[0093] When all conflicting nodes have the same priority, the edge server sends an ACK (acknowledgment character) to all conflicting nodes to indicate that they have entered backoff mode. Each node then randomly selects a backoff time within its backoff time window to wait. Once the waiting time expires, new access requests are initiated sequentially according to the backoff time selected by each node. The optimization process for the real-time performance and accuracy of the high-concurrency semantic transmission resource scheduling optimization model is as follows:

[0094] Step 4.1: Calculate all device nodes that have issued access requests. Probability of successful transmission within a cycle The calculation process is as follows:

[0095] When there is only one device node When an access request is initiated, the probability of successful transmission is: , express of This is a series of products.

[0096] When two device nodes with different priorities simultaneously initiate access requests, each device node has a 0.5 probability of being either a high-priority node or a low-priority node. The probability of successful transmission is ; express of This is a series of products.

[0097] When two device nodes of the same priority simultaneously initiate access requests, a conflict occurs. The device node with the shorter backoff waiting time has a higher probability of successful transmission. ,in The total backoff time window size (maximum backoff time). For device nodes Randomly selected retreat time (0~W). This represents the total number of nodes involved in other conflicts. Meanwhile, all other conflicting nodes choose a longer backoff waiting time; the node choosing a shorter backoff waiting time will not engage in conflict in the next round. When a conflict occurs, any one of the conflicting nodes... The probability of successful transmission is .

[0098] When three nodes simultaneously initiate access requests, if the device node The first node is a high-priority node, and the other two are low-priority nodes. Their probability of successful transmission is... If one of the other two nodes is a high-priority node, the device node... The probability of successful transmission is If all three nodes are high-priority nodes, the device node... The probability of successful transmission is ;

[0099] When three nodes simultaneously initiate access requests, if the device node The first node is a low-priority node, and the other two nodes are high-priority nodes; these are device nodes. Transmission can only proceed after the two high-priority nodes have finished transmitting. The probability of successful transmission is If one of the nodes is a low-priority node, the device node... The probability of successful transmission is If all three nodes are low-priority nodes, the device node... The probability of successful transmission is .

[0100] Step 4.2: Increase step size using a fixed increment. To narrow down the feasible region, based on the obtained Dynamic changes Value linear search yields different Minimum average information age under size .

[0101]

[0102] in, Representative before ( Within a cycle time, the device node All transmissions experienced failed transmissions, and the probability of failure for each cycle is expressed as ( ), Indicates the first Within a cycle time, the device node The probability of successful transmission is expressed as: , For device nodes forward Before the start of each cycle time until the [number]th cycle The average AoI after each cycle time, This indicates the number of successful transmissions, i.e. Represents device node No. The average age of information transmitted each time. The above formula can be used as follows: Figure 4 The geometric description shows that the size of the shaded area in the figure is... .

[0103] The obtained minimum average information age (AoI) enables device nodes to resolve conflicts in a timely manner and transmit successfully during the conflict resolution process, ensuring the real-time performance of semantic transmission.

[0104] Step 4.3: Calculate the maximum average structural similarity index of all device nodes that issued access requests;

[0105] like Figure 5 As shown, the fitting curves of SSIM and SNR under different compression ratios (CR) are shown. Under the same compression ratio, SSIM is directly proportional to SNR. Under the same signal-to-noise ratio (SNR), SSIM is directly proportional to the compression ratio (CR). Therefore, selecting the maximum compression ratio will yield the maximum SSIM. The obtained maximum average structural similarity index (SSIM) ensures that the semantic data transmitted by the device nodes is recovered as much as possible, thus guaranteeing the accuracy of semantic transmission.

[0106] Step five, the industrial Internet of Things device node receives the ACK (acknowledgement character) sent by the edge server through the wireless AWGN channel, sends the extracted semantic feature information to the edge server, and the edge server sends an end code to the device node after receiving the semantic feature information sent by the device node, completing the semantic transmission.

[0107] Repeat steps two to five to achieve high-concurrency semantic transmission of the industrial Internet of Things.

[0108] In actual industrial processes, with the dynamic change of the number of device nodes, the dynamic change of the probability, the transmission power of the device node and the compression rate, the maximum average structural similarity index SSIM and the minimum average information age AoI under different probabilities, different transmission powers and different compression rates can be obtained.

[0109] As shown in Figure 6 , under different numbers of device nodes, the circular dotted line is the minimum average AoI of the high-concurrency semantic transmission resource scheduling method and system for the industrial Internet of Things of the present application, the triangular dotted line is the minimum average AoI of the CSMA method, and the square dotted line is the minimum average AoI of the Slotted ALOHA method. The minimum average AoI of the present application is always less than that of the CSMA method. Since AoI is an index that comprehensively describes the transmission real-time performance and accuracy, it indicates that the present application can transmit semantic data more real-time and at the same time can guarantee the transmission accuracy. When the number of nodes is large, the minimum average AoI of the present application is less than that of the Slotted ALOHA method. Therefore, in the scene of a large number of nodes in the industrial Internet of Things, the present application can transmit semantic data more real-time and at the same time can guarantee the transmission accuracy. Therefore, the present application is superior to the CSMA method and the Slotted ALOHA method in terms of transmission real-time performance and accuracy.

[0110] Figure 7 The minimum average AoI obtained by using the system and the method of the present application is shown under different values and different transmission powers. Figure 8 The minimum average AoI obtained by using the CSMA method is shown under different values and different transmission powers. Figure 9 The minimum average AoI obtained by using the Slotted ALOHA method is shown under different values and different transmission powers. Figures 7 to 9 In summary, ​Under different transmission power conditions, the application can obtain the minimum AoII. Since AoII is an index that comprehensively describes the real-time and accuracy of transmission, it indicates that the application can transmit semantic data more real-time and at the same time can guarantee transmission accuracy. Therefore, under the conditions of different probabilities of empty device node data buffer area and different transmission power, the application is superior to the CSMA method and the Slotted ALOHA method in terms of transmission real-time and accuracy.

[0111] Figure 10 The average SSIM value of the application under different transmission power and different compression rate conditions is shown, which is used to describe the transmission accuracy. The value is between 0 and 1, and the closer to 1, the better. When the transmission power is 10dBm, the average SSIM is greater than 0.65. When the compression rate is 1 / 6, the average SSIM can reach 0.97. It indicates that when the device node transmission power is small (0dBm~10dBm), the industrial internet of things device node transmission power is generally small. The application can still maintain a high average SSIM value, that is, a good transmission accuracy.

[0112] Figure 11 The average semantic transmission rate of the application under different transmission power and different compression rate conditions is shown. The semantic transmission rate ; wherein, is the transmission bandwidth, which can be fixed as a constant; represents the expected amount of semantic information to be transmitted, represents the expected number of words to be transmitted, and are usually coupled with the actual transmission situation. Users can set them as constants according to the actual transmission data, is the compression rate selected by the device node for transmission, is the structural similarity of the device node . By dynamically changing the transmission power and the compression rate, the signal-to-noise ratio (SNR) can be dynamically adjusted, which in turn affects the structural similarity, further affecting the semantic transmission rate, obtaining the semantic transmission rate under different transmission power and compression rate conditions. The average semantic transmission rate is used to describe the transmission fluency, and the larger the value is, the better, Figure 11 in the middle, when the transmission power is 10dBm, the average semantic transmission rate is greater than , which indicates that when the device node transmission power is small (0dBm~10dBm), the industrial internet of things device node transmission power is generally small. The application can still maintain a high average semantic transmission rate, that is, a good transmission fluency.

Claims

1. An industrial Internet of Things (IIoT) oriented high concurrency semantic transmission method, characterized in that: comprising the following steps: Step 1: storing a pre-set priority identifier (ID) in each IIoT device node; Meanwhile, constructing a high concurrency semantic transmission resource scheduling optimization model in an edge server based on an average error information age, which includes an average information age minimization model and an average structural similarity index maximization model; The average information age minimization model enables the device nodes to solve conflicts in time and successfully transmit in the conflict resolution process, thereby ensuring the real-time performance of semantic transmission; The average structural similarity index maximization model enables the semantic data transmitted by the device nodes to be recovered as much as possible, thereby ensuring the accuracy of semantic transmission; Step 2: a DeepJSCC encoding model in each IIoT device node extracts semantic feature information of industrial data collected by a sensor in the device node and stores the extracted semantic feature information of industrial data in a data packet buffer area; Step 3: all IIoT device nodes with non-empty data packet buffer areas initiate an access edge server request through a transmission module and send a preamble to the edge server, respectively, wherein the preamble sent by each device node contains the priority identifier (ID) of the device node; Step 4: after receiving the access request sent by the IIoT device nodes, the receiving module of the edge server performs transmission scheduling on the IIoT device nodes, replies to the corresponding ACK to all device nodes that send the access request, performs semantic transmission scheduling, and calls the high concurrency semantic transmission resource scheduling optimization model to optimize the real-time performance and accuracy of semantic transmission; The real-time performance and accuracy optimization process of the high concurrency semantic transmission resource scheduling optimization model is as follows: Step 4.3: calculating the maximum average structural similarity index of all device nodes that send the access request, obtaining the maximum average SSIM, which enables the semantic data transmitted by the device nodes to be recovered as much as possible, thereby ensuring the accuracy of semantic transmission; Step 4.1, Calculate all the device nodes that send access requests Probability of successful transmission in one cycle ; Step 4.2, fixed increase with fixed step size To reduce the feasible region, according to the obtained , dynamically change The value linear search obtains the minimum average information age under different Size ; The obtained minimum average information age makes the device node solve the conflict in time and successfully transmit in the conflict resolution process, and guarantees the real-time performance of semantic transmission; Step 5: the IIoT device nodes receive the ACK sent by the edge server through the transmission module, send the extracted semantic feature information to the edge server, and the edge server sends an end code to the device nodes after receiving the semantic feature information sent by the device nodes, thereby completing the semantic transmission; Repeat steps 2-5 to achieve high concurrency semantic transmission of the IIoT.

3. The IIoT oriented high concurrency semantic transmission method of claim 1, characterized in that:

2. The high concurrency semantic transmission method for industrial Internet of Things of claim 1, wherein: In actual industrial processes, the high-concurrency semantic transmission resource scheduling optimization model varies with the number of device nodes. Dynamic changes, dynamic changes in device nodes The probability that the data cache is empty. Equipment Nodes Transmission power Values ​​and data compression ratios yield different results. Probability, maximum average structural similarity index and minimum average information age under different transmission powers and data compression rates. The scheduling process in step 4 is as follows: If there is only one device node with non-empty data buffer area, no conflict occurs, and the edge server replies to the ACK of the device node to send semantic information; If there are more than one device nodes with non-empty data buffer area, a conflict occurs, and when there are different priorities among the conflicting nodes, the edge server replies to the ACK of the high-priority device node to send semantic information and the ACK of the low-priority node to enter a waiting state. ​ When the conflict nodes are of the same priority, the edge server sends an ACK to all the conflict nodes to enter the backoff mode, and all the nodes randomly select a backoff time in the backoff time window and wait for the end of the waiting time according to the backoff time selected by the device node, and then initiates a new access request in turn according to the backoff time selected by the device node. The backoff time setting process is as follows: setting the number of device nodes a proportional coefficient of the backoff time window size , setting the maximum backoff time window size according to the number of nodes and the proportional coefficient .

4. The high concurrency semantic transmission method for industrial Internet of Things of claim 1, wherein: In the step four scheduling process, the total time from initiating an access request to conflict resolution of each device node is defined as a cycle period time; in the high-concurrency priority conflict resolution process, time is divided according to time slots, and transmission and backoff time are counted by time slots; if a device node fails to send in a cycle period, the data packet is processed as a lost packet.

5. The high concurrency semantic transmission method for industrial Internet of Things of claim 3, wherein: In the step four scheduling process, the total time from initiating an access request to conflict resolution of each device node is defined as a cycle period time; in the high-concurrency priority conflict resolution process, time is divided according to time slots, and transmission and backoff time are counted by time slots; if a device node fails to send in a cycle period, the data packet is processed as a lost packet.

6. The high concurrency semantic transmission method for industrial Internet of Things of claim 1, wherein: In the step one, the priority of the device node is identified according to the transmission task of the device node.

7. System for operating a method as claimed in any one of the claims 1-5, characterized in that The system comprises an edge server and a plurality of industrial Internet of Things device nodes. A sensor module for collecting industrial data is arranged in each industrial Internet of Things device node, and a pre-calibrated priority identifier ID is stored. A DeepJSCC encoding model is deployed in each industrial Internet of Things device node for extracting semantic feature information of industrial data; a transmission module for transmitting semantic data is used to transmit the extracted semantic feature information to the edge server; the industrial Internet of Things device node has a data packet buffer area, and when the buffer area is not empty, the device node initiates an access request to the edge Internet of Things; A high-concurrency semantic transmission resource scheduling optimization model is deployed in the edge server, and when a single or multiple device nodes send an access request to the edge server, the edge server schedules the transmission of the industrial Internet of Things device nodes and replies to the device nodes with corresponding ACKs; at the same time, the high-concurrency semantic transmission resource scheduling optimization model is called to optimize the real-time performance and accuracy of semantic transmission; A receiving module for receiving semantic data is deployed in the edge server to receive the semantic feature information extracted by the DeepJSCC encoding model in the industrial Internet of Things device node. A DeepJSCC decoding model is deployed in the edge server to decode the received semantic feature information.

8. The system of claim 7, wherein: The transmission module and the receiving module of the semantic data operate under the condition of a wireless AWGN channel.

Images