Information timeliness optimization method, device and equipment of communication network and storage medium

Abstract

The application relates to a method and device for optimizing information aging of a communication network, an equipment and a storage medium, and relates to the technical field of wireless communication network resource optimization. The method inputs the obtained system state of each time slot into a two-stage resource optimization model comprising a first-stage model and a second-stage model. The first-stage model is used to output power allocation results of each sensing node, and the second-stage model is used to output decoding orders of all sub-messages. According to the power allocation results and the decoding orders, state update data packets of each sensing node are sent to a monitoring node in a current time slot to update information ages of each sensing node at the monitoring node and battery energies of each sensing node. The technical scheme of the application divides the power and the decoding order into stages for optimization by means of the two-stage model, effectively solves the technical problem that continuous transmission power and high-dimensional discrete decoding order are difficult to be jointly optimized, and realizes continuous optimization of information aging of a wireless energy supply network.

Description

Technical Field

[0001] This application relates to the technical field of wireless communication network resource optimization, and in particular to a method, apparatus, device and storage medium for optimizing the information timeliness of a communication network. Background Technology

[0002] With the increasing prevalence of applications such as the Internet of Things (IoT) and smart monitoring, a large number of distributed sensor nodes need to continuously collect and transmit data. Traditional battery and wired power supply methods face high maintenance costs and security risks in large-scale deployments, leading to the widespread application of wireless power communication networks. This type of service demands extremely high information freshness; information age is a core indicator for judging information timeliness.

[0003] Currently, wireless power supply networks mainly adopt multiple access methods such as time division multiple access and non-orthogonal multiple access. When faced with concurrent transmission of multiple nodes and complex channel interference, the access flexibility and anti-interference ability are insufficient, making it difficult to guarantee information timeliness. In contrast, rate division multiple access has stronger interference management and resource allocation capabilities, and is an effective technical direction for improving network timeliness performance.

[0004] Applying Rate Segmentation Multiple Access (SIC) to wireless power networks still faces technical bottlenecks: the randomness of node energy harvesting makes it impossible to establish a clear analytical model for information age; at the same time, the SIC decoding order is a high-dimensional discrete variable, making it difficult to achieve joint optimization with transmit power, and existing technologies cannot effectively overcome this problem. Summary of the Invention

[0005] This application provides a method, apparatus, device, and storage medium for optimizing the information timeliness of communication networks. By using a two-stage model to optimize power and decoding order in stages, it effectively solves the technical problem that continuous transmission power and high-dimensional discrete decoding order are difficult to optimize together, and realizes continuous optimization of information timeliness in wireless power supply networks.

[0006] In a first aspect, this application provides a method for optimizing the information timeliness of a communication network. The communication network includes a monitoring node and multiple sensor nodes connected to the monitoring node. The method includes: acquiring the system state of each time slot, the system state including at least the information age of each sensor node at the monitoring node, the battery energy of each sensor node, and the uplink channel state; inputting the system state into a pre-trained two-stage resource optimization model, the two-stage resource optimization model including a first-stage model and a second-stage model, wherein the first-stage model is used to output the power allocation result of each sensor node, the power allocation result including the scheduling variable of each sensor node, the transmit power, and the power allocation ratio of the transmit power among each sub-message. Based on the power allocation result, the sub-message transmission power of each sub-message in each sensor node is calculated, and the total information age of the system is calculated based on the information age of each sensor node. The sub-message transmission power, the total information age of the system, the information age of each sensor node, and the uplink channel state are input into the second stage model so that the second stage model outputs the complete decoding order of all sub-messages in each sensor node. The sub-messages are obtained by the sensor nodes from the state update data packets. According to the power allocation result and the decoding order, each sensor node is controlled to send the state update data packets to the monitoring node in the current time slot to update the information age of each sensor node at the monitoring node and the battery energy of each sensor node.

[0007] In one possible implementation, the step of inputting the sub-message transmission power, the total system information age, the information age of each sensor node, and the uplink channel state into the second-stage model, so that the second-stage model outputs the complete decoding order of all sub-messages in each sensor node, specifically includes: taking all sub-messages not selected into the decoding order in the current time slot as a candidate sub-message set; using the self-attention mechanism in the second-stage model to extract features from the input sub-message transmission power, the total system information age, the information age of each sensor node, and the uplink channel state, to calculate the score vector for each candidate sub-message selected in the candidate sub-message set; selecting a target sub-message from the candidate sub-message set and adding it to the decoding order according to the score vector, and using a masking mechanism to mask the target sub-message from the candidate sub-message set to update the candidate sub-message set; repeating the above steps until all sub-messages in the candidate sub-message set are selected into the decoding order, generating the complete decoding order of all sub-messages in each sensor node.

[0008] In one possible implementation, updating the information age of each sensor node at the monitoring node specifically includes: calculating the total transmission rate of the sensor node corresponding to the state update data packet; if the total transmission rate meets the preset data packet transmission requirements, then the state update of the sensor node is determined to be successful, and the information age of the sensor node at the monitoring node is reset to the current local information age of the sensor node; if the total transmission rate does not meet the preset data packet transmission requirements, then the information age of the sensor node at the monitoring node is incremented.

[0009] In one possible implementation, updating the battery energy of each sensor node specifically includes: acquiring the battery energy of each sensor node in the current time slot and the base energy of each sensor node; if the battery energy in the current time slot is greater than the base energy, acquiring the energy power collected by each sensor node in the current time slot, the transmission power, and the maximum battery energy, and calculating the battery energy of each sensor node in the next time slot based on the battery energy, the energy power, the transmission power, the base energy, and the maximum battery energy; if the battery energy in the current time slot is not greater than the base energy, acquiring the energy power collected by each sensor node in the current time slot and the maximum battery energy, and calculating the battery energy of each sensor node in the next time slot based on the battery energy, the energy power, and the maximum battery energy; and updating the battery energy of each sensor node based on the battery energy corresponding to the next time slot.

[0010] In one possible implementation, the two-stage resource optimization model is constructed with minimizing the system information age as the optimization cost. The training process of the two-stage resource optimization model specifically includes: calculating the optimization cost sample of the current time slot; constructing a first-stage training sample set based on the optimization cost sample of the current time slot, the system state sample of the current time slot, the power allocation result sample of the current time slot, and the system state sample of the next time slot; constructing a second-stage training sample set based on the optimization cost sample of the current time slot, the decoding order of all sub-messages in the current time slot, and the second-stage input state sample of the current time slot, wherein the second-stage input state sample of the current time slot includes sub-message transmit power samples, total information age samples, information age samples of each sensor node, and uplink channel state samples; updating the model parameters of the first-stage model according to the first-stage training sample set, and updating the model parameters of the second-stage model according to the second-stage training sample set, to obtain the trained two-stage resource optimization model.

[0011] In one possible implementation, the communication network is an uplink rate-division multiple access assisted wireless power communication network, and the communication network further includes a wireless power transmitter, wherein the wireless power transmitter is used to send radio frequency energy signals to each sensor node, so that each sensor node can collect energy from the radio frequency energy signals through an energy acquisition circuit, and upload a status update data packet to the monitoring node based on the collected energy.

[0012] Secondly, this application provides an information timeliness optimization device for a communication network, comprising: a state acquisition module, a two-stage resource optimization model processing module, and an execution module; wherein, the state acquisition module is used to acquire the system state of each discrete time slot, the system state including at least the information age of each sensor node at the monitoring node, the battery energy of each sensor node, and the uplink channel state; the two-stage resource optimization model processing module is used to input the system state into a pre-trained two-stage resource optimization model, the two-stage resource optimization model including a first-stage model and a second-stage model, wherein the first-stage model is used to output the power allocation result of each sensor node, the power allocation result including the scheduling variables of each sensor node, the transmit power, and the transmit power in each sub-message. The power allocation ratio between the sensors is used to calculate the sub-message transmission power of each sub-message in each sensor node based on the power allocation result. Based on the information age of each sensor node, the total information age of the system is calculated. The sub-message transmission power, the total information age of the system, the information age of each sensor node, and the uplink channel state are input into the second-stage model so that the second-stage model outputs the complete decoding order of all sub-messages in each sensor node. The sub-messages are obtained by the sensor nodes dividing the state update data packets. The execution module is used to control each sensor node to send the state update data packet to the monitoring node in the current time slot according to the power allocation result and the decoding order, so as to update the information age of each sensor node at the monitoring node and the battery energy of each sensor node.

[0013] Thirdly, embodiments of this application also provide a computer device, which includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the above-described method.

[0014] Fourthly, embodiments of this application also provide a computer-readable storage medium storing a computer program that, when executed by a processor, can implement the above-described method.

[0015] This application provides a method, apparatus, device, and storage medium for optimizing information timeliness in communication networks, which has the following advantages compared with the prior art: This method acquires the system state for each time slot, which includes at least the information age of each sensor node at the monitoring node, the battery energy of each sensor node, and the uplink channel state. The system state is then input into a pre-trained two-stage resource optimization model, comprising a first-stage model and a second-stage model. The first-stage model outputs the power allocation results for each sensor node, and the second-stage model outputs the decoding order of all sub-messages, where the sub-messages are obtained by the sensor nodes from the state update data packets. Based on the power allocation results and the decoding order, each sensor node sends the state update data packets to the monitoring node in the current time slot to update the information of each sensor node at the monitoring node. The present application's technical solution uses a pre-trained two-stage resource optimization model to make decisions, based on the information age of each sensor node and the battery energy of each sensor node. Compared with existing technical solutions, the present application's technical solution uses a pre-trained two-stage resource optimization model to make decisions. It directly uses the real-time collected information age, battery energy, and uplink channel status as inputs. The model implicitly learns the correlation between variables, avoiding the problem that it is difficult to establish an explicit analytical model for information age due to the randomness of node energy collection. At the same time, the power allocation and decoding order optimization are split into two stages and implemented sequentially, effectively solving the problem that it is difficult to jointly optimize continuous transmission power and high-dimensional discrete decoding order. By updating the system state after making decisions in each time slot, the network can continuously optimize information timeliness under dynamic energy supply and time-varying channel conditions. Attached Figure Description

[0016] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, those skilled in the art can obtain other drawings based on these drawings without creative effort.

[0018] One or more embodiments are illustrated by way of example with reference numerals in the accompanying drawings. These illustrations do not constitute a limitation on the embodiments. Elements with the same reference numerals in the drawings are denoted as similar elements. Unless otherwise stated, the figures in the drawings are not to be limited by scale.

[0019] Figure 1 This is a flowchart illustrating an embodiment of a method for optimizing the timeliness of information in a communication network provided in this application; Figure 2 This is a schematic diagram of the structure of an embodiment of an information timeliness optimization device for a communication network provided in this application; Figure 3 This is a schematic diagram of the structure of a computer device provided in this application. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0021] The following disclosure provides numerous different embodiments or examples for implementing various structures of this application. To simplify the disclosure, specific examples of components and arrangements are described below. These are merely examples and are not intended to limit the scope of this application. Furthermore, reference numerals and / or letters may be repeated in different examples. Such repetition is for simplification and clarity and does not in itself indicate a relationship between the various embodiments and / or arrangements discussed.

[0022] It should be understood that, when used in this specification and the appended claims, the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0023] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the scope of the application. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0024] It should also be further understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0025] As used in this specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrase "if determined" or "if [described condition or event] is detected" may be interpreted, depending on the context, as "once determined," "in response to determination," "once [described condition or event] is detected," or "in response to detection of [described condition or event]."

[0026] Example 1, see Figure 1 , Figure 1 This is a flowchart illustrating an embodiment of a method for optimizing the timeliness of information in a communication network provided in this application. Figure 1 As shown, the communication network includes a monitoring node and multiple sensing nodes connected to the monitoring node. The method includes steps 101-103, as follows: Step 101: Obtain the system status for each time slot. The system status includes at least the information age of each sensor node at the monitoring node, the battery energy of each sensor node, and the uplink channel status.

[0027] In one embodiment, the communication network is an uplink rate-division multiple access assisted wireless power communication network, and the communication network further includes a wireless power transmitter, wherein the wireless power transmitter is used to send radio frequency energy signals to each sensor node, so that each sensor node can collect energy from the radio frequency energy signals through an energy acquisition circuit, and upload a status update data packet to the monitoring node based on the collected energy.

[0028] Specifically, the wireless power supply communication network includes: a wireless power transmitter, a monitoring node, and multiple distributed sensor nodes. The set of sensor nodes is represented as follows: Among them, the wireless power transmitter and each sensor node form a downlink power transmission link, and the sensor node and the monitoring node form an uplink data transmission link.

[0029] Specifically, the operating time of the wireless power supply communication network is divided into multiple continuous discrete time slots, each with a fixed duration, denoted as the time slot duration. Within each time slot, the system sequentially performs operations such as energy acquisition, status sampling, uplink data transmission, information age update, and battery energy update, forming a periodic acquisition-decision-transmission-update closed loop. The acquisition of the system status is uniformly completed in the initial stage of a single time slot.

[0030] In one embodiment, the information age of each sensor node at the monitoring node is... This data is used to characterize the freshness of data on the monitoring node side. The monitoring node is responsible for real-time statistics and storage of this status data, which is read and acquired in batches at the beginning of each time slot.

[0031] Specifically, after completing uplink data transmission in the previous time slot, the monitoring node updates the information age of the corresponding sensor node based on the data packet transmission results. Upon entering the current time slot, the monitoring node retrieves the latest information age value for each sensor node from its local storage unit, summarizing them to form the information age set of all sensor nodes in the current time slot. This data requires no additional communication interaction and is read directly from the status register of the monitoring node.

[0032] In one embodiment, battery energy It reflects the current remaining available power of the sensor node and determines whether the node can complete sampling and uplink transmission. This status is collected locally by each sensor node and reported to the monitoring node, and is summarized at the beginning of the time slot.

[0033] Specifically, each sensor node has a built-in power detection module and battery storage unit. After the previous time slot ends, the node updates its remaining battery energy by combining the energy collected by wireless power supply, basic circuit power consumption, signal transmission power consumption, and the battery capacity limit. At the start of the current time slot, the sensor node uses its built-in power sensor to detect the current remaining battery energy value in real time and sends this data to the monitoring node through the feedback channel. After receiving the battery energy data reported by all sensor nodes, the monitoring node processes the data to obtain the battery energy status of all sensor nodes in the current time slot.

[0034] In one embodiment, uplink channel state This refers to the uplink channel power gain between the sensing node and the monitoring node, which is a key environmental parameter affecting the transmission rate and decoding performance.

[0035] Specifically, the uplink channel power gain between each sensing node and the monitoring node can be calculated using standard channel estimation methods. For example, before the formal transmission of service data in each time slot, all sensing nodes send dedicated pilot signals to the monitoring node in sequence. The channel estimation unit of the monitoring node receives the pilot signals and, in combination with factors such as signal attenuation and noise interference, calculates the uplink channel power gain corresponding to each sensing node, which is used as the uplink channel state of the current time slot.

[0036] In one embodiment, the information age of each sensor node at the monitoring node, the battery energy of each sensor node, and the uplink channel status are collected and integrated into a complete system state. , .

[0037] Preferably, the system state is re-acquired and updated in each time slot to ensure that the system state on which the model decision is based can accurately reflect the latest situation of the network in terms of energy supply, information freshness and channel conditions, thereby achieving adaptive resource optimization in a dynamic and time-varying environment.

[0038] Step 102: Input the system state into a pre-trained two-stage resource optimization model. The two-stage resource optimization model includes a first-stage model and a second-stage model. The first-stage model is used to output the power allocation results of each sensor node. The power allocation results include the scheduling variables, transmit power, and power allocation ratio of the transmit power among each sub-message of each sensor node. Based on the power allocation results, the sub-message transmit power of each sub-message in each sensor node is calculated, and the total information age of the system is calculated based on the information age of each sensor node. The sub-message transmit power, the total information age of the system, the information age of each sensor node, and the uplink channel state are input into the second-stage model so that the second-stage model outputs the complete decoding order of all sub-messages in each sensor node. The sub-messages are obtained by the sensor nodes from the state update data packets.

[0039] In one embodiment, before inputting the system state into the pre-trained two-stage resource optimization model, the sensing node generates current environmental monitoring data based on the sampling results of the current time slot, packages the environmental monitoring data into a state update data packet, and divides the state update data packet into multiple sub-messages. The set of sub-messages is represented as follows: .

[0040] Specifically, in each time slot, the wireless power transmitter radiates radio frequency energy signals into space at a fixed or controllable transmission power. Each sensor node receives these radio frequency energy signals through its built-in energy harvesting circuit and converts them into DC power usable by the energy harvesting circuit. Due to the nonlinear characteristics of the energy harvesting circuit, the actual energy power harvested by each sensor node in time slot t is determined by the input power of sensor node m in time slot t, the maximum output power of the energy harvesting circuit, and the parameters of the energy parameter circuit. In this way, the sensor nodes can continuously obtain the energy required to maintain operation from the environment.

[0041] Preferably, when determining the actual energy power collected by each sensing node in time slot t, the channel power gain of the energy acquisition circuit is first obtained, and the input power of sensing node m in time slot t is calculated based on the channel power gain. Then, the input power of sensing node m in time slot t, the maximum output power of the energy acquisition circuit, and the parameters of the energy parameter circuit are substituted into the preset energy power calculation formula to calculate the actual energy power collected by each sensing node in time slot t. The specific calculation process is as follows: The input power of sensor node m in the t-th time slot is: , ,in, Let m be the input power of sensor node m in the t-th time slot. The transmission power for transmitting radio frequency energy signals to a wireless energy transmitter; Channel power gain for the transmit power energy harvesting link; For wireless power transmitters to sensor nodes The path loss factor between these two values ​​represents the path loss factor in large-scale fading. For wireless power transmitters to sensor nodes The power gain between these terms represents the power gain of small-scale fading.

[0042] The energy power collected by sensor node m in the t-th time slot for: ; in, This represents the maximum output power of the energy harvesting circuit, and a and b are the parameters of the energy parameter circuit.

[0043] Specifically, within each time slot, if the battery energy of the sensor node is sufficient to support basic energy consumption, the sensor node performs a state sampling, acquires the current environmental monitoring data, and packages the environmental monitoring data into a state update data packet.

[0044] Specifically, each sensor node divides the status update data packet into multiple sub-messages, either equally or as needed, enabling the system to implement resource allocation and interference management at a finer granularity. On one hand, each sub-message can be independently allocated power and its decoding order arranged, allowing the system to flexibly allocate resources for each sub-message based on channel quality, battery capacity, and information age, achieving differentiated transmission guarantees. On the other hand, monitoring nodes can decode each sub-message sequentially according to the optimized SIC decoding order, removing each sub-message from the superimposed signal after decoding, thereby effectively reducing interference in subsequent decoding and improving the decoding success rate of multi-node concurrent transmission. Through this flexible scheduling and interference elimination at the sub-message granularity, the system can prioritize decoding sub-messages from nodes with older information ages, achieving incremental status updates, effectively reducing the overall long-term expected total information age of the network, and improving the information timeliness performance of the wireless power supply network under dynamic power supply and time-varying channel conditions.

[0045] In one embodiment, the system state is input into the first-stage model so that the first-stage model outputs the power allocation results for each sensing node.

[0046] Specifically, the complete system state is input into the first stage of the pre-trained two-stage resource optimization model. The neural network inside the first stage model reads the three types of state features corresponding to each sensor node in the system state, and completes feature depth extraction through a multi-layer fully connected network to explore the implicit correlation between information age, battery level, uplink channel state and optimal resource allocation strategy. Unlike the traditional optimization method that relies on explicit analytical equations, this first stage model relies on the network parameters learned offline and can synchronously output discrete and continuous mixed-type action outputs based solely on the system state input in the current time slot, without the need to manually construct the mathematical relationship between information age and control variables.

[0047] Preferably, the first stage model for: ;in, This indicates the input state of the power distribution module, including the system state corresponding to the M sensor nodes. This represents the power allocation strategy function. This represents the model parameters of the first-stage model.

[0048] Specifically, the first-stage model is based on a pre-trained power allocation policy function. The system state input is processed, and the power allocation result is output. The power allocation result consists of three components: the scheduling variables of the sensor nodes. Transmit power of sensor nodes and the power allocation ratio among the sub-messages within the sensor node. Among them, the scheduling variable is a discrete decision variable used to indicate whether each sensor node is activated to perform state sampling and uplink transmission in the current time slot; the transmit power is a continuous decision variable, representing the total transmit power available for uplink transmission by each sensor node in the current time slot; the power allocation ratio is also a continuous decision variable, representing the proportion of the total transmit power allocated among sub-messages within each sensor node, and satisfying that the sum of the power allocation ratios of all sub-messages within the same sensor node is 1, i.e. .

[0049] In one embodiment, to ensure that the power allocation result is physically feasible, the system also needs to perform energy causality constraint processing on the output transmit power.

[0050] Specifically, the sensing node m requires basic energy to maintain circuit operation and state sampling: ,in, As basic energy, To maintain the circuit's operating energy, Energy is sampled for the state.

[0051] Specifically, obtain the battery energy of sensor node m at the beginning of the t-th time slot. The energy difference between the battery energy and the base energy is divided by the duration of the time slot. The first transmission power is obtained, and then compared with a preset maximum transmission power. The minimum value between the first transmission power and the preset maximum transmission power is selected as the transmission power constraint value. Based on the transmission power constraint value, the output transmission power is constrained to ensure that the output transmission power does not exceed the transmission power constraint value; that is... If the transmit power output by the first-stage model exceeds the transmit power constraint value, the system will prune it to limit it within the transmit power constraint value, thereby ensuring that the sensor node does not fail due to over-discharge. Only the power allocation result after energy causality constraint pruning can be used for subsequent uplink transmission execution.

[0052] In one embodiment, based on the power allocation result, the sub-message transmission power of each sub-message in each sensing node is calculated, and based on the information age of each sensing node, the total information age of the system is calculated.

[0053] Specifically, for sensing nodes The Each sub-message has a corresponding sub-message transmission power. The total transmit power of this sensing node Multiply by the first Power allocation ratio corresponding to each sub-message and the scheduling variables of the sensor nodes. Determined, that is Through this calculation, the total transmit power output by the first-stage model is refined to the level of each sub-message, providing the necessary sub-message power information for the subsequent second-stage model to make decoding order decisions.

[0054] Preferably, the total transmit power of sensor node m in the t-th time slot is... The sub-message transmission power of all sub-messages corresponding to this sensor node. The sum, that is .

[0055] In one embodiment, the total information age of the system The sum of the information ages of all sensing nodes at the monitoring node is, i.e. It is used to characterize the overall information obsolescence of the current time slot system.

[0056] In one embodiment, the sub-message transmission power, the total information age of the system, the information age of each sensor node, and the uplink channel state are input into the second stage model so that the second stage model outputs the complete decoding order of all sub-messages in each sensor node.

[0057] Specifically, the second-stage model is ,in, This indicates the input state of the second-stage model. This represents the decoding order strategy function. This represents the parameters of the second-stage model, whose input states include the sub-message transmission power of each sub-message. System total information age Information age of each sensor node and uplink channel status .

[0058] Specifically, all sub-messages not selected for the decoding order within the current time slot are used as a candidate sub-message set. Using the self-attention mechanism in the second-stage model, features are extracted from the input sub-message transmit power, the total system information age, the information age of each sensor node, and the uplink channel state to calculate the score vector for each candidate sub-message selected from the candidate sub-message set. Based on the score vector, a target sub-message is selected from the candidate sub-message set and added to the decoding order. A masking mechanism is then used to remove the target sub-message from the candidate sub-message set to update the candidate sub-message set. This process is repeated until all sub-messages in the candidate sub-message set are selected for the decoding order, generating a complete decoding order for all sub-messages in each sensor node.

[0059] Specifically, the second-stage model employs an autoregressive sequence generation method to successively determine each element in the decoding order. At the start of each selection round, the system first considers all sub-messages in the current time slot that have not yet been selected for the decoding order as a candidate sub-message set; assuming the current time slot has a total of There are 10 sensor nodes, and each sensor node divides the state update data packet into 10 parts. If there are 100 sub-messages, then the total number of sub-messages is 100. In the initial state of the generation process, all None of the sub-messages were selected for decoding, therefore the candidate sub-message set contains all sub-messages. After each sub-message is selected, it is removed from the candidate set, and the size of the candidate set decreases by 1 with each selection. Through this dynamic maintenance method, the system ensures that the selection object in each round is always the remaining sub-messages that have not yet been assigned a decoding order.

[0060] Specifically, in each round of selection, the second-stage model encodes the input state using the self-attention mechanism in the Transformer architecture. The core idea of ​​the self-attention mechanism is to dynamically extract context-aware feature representations by calculating the association weights between each element and other elements in the sequence. In this stage, each candidate sub-message is treated as an element in the sequence, and each candidate sub-message contains additional sub-message states. The sub-message transmission power of the sub-message is included. Age of the associated sensor node and uplink channel status Information such as...

[0061] Preferably, the self-attention mechanism can calculate the score vector of each candidate sub-message in the following way: The input features of each candidate sub-message are mapped to a query vector, key vector, and value vector through a linear transformation; the relevance score between the query vector of each candidate sub-message and the key vectors of all candidate sub-messages is calculated, and the attention weights are obtained after scaling and Softmax normalization; the attention weights are weighted and summed with the corresponding value vectors to obtain enhanced features that incorporate global context information; the enhanced features are mapped to scalar scores through a feedforward network; and the scores of all candidate sub-messages constitute the score vector. ,in, Indicates the index of the current decoding step.

[0062] Specifically, each element in the scoring vector reflects the appropriateness of the corresponding candidate sub-message to be prioritized in the current decoding step; the higher the score, the more likely the sub-message should be decoded.

[0063] Specifically, the system applies the Softmax function to the scoring vector, converting it into a probability distribution: the generation probability of the decoding order in the t-th time slot is expressed as... To prevent the same sub-message from being selected repeatedly, a masking mechanism is introduced during each step of the decoding sequence generation process. The generation probability of each decoded element is: ;in, Represents element-wise product. For the mask vector, For the current time slot The complete decoding order, For decoding order The Middle Sub-messages at each location, For the decoding order, the first Sub-messages with confirmed locations To select the next step given that the previously selected sub-messages are already known. The conditional probability as the next decoded element.

[0064] Specifically, based on the probability distribution mentioned above, the system selects a target sub-message from the candidate set by taking the maximum probability. After selecting the target sub-message, the system adds it to the end of the decoding order and updates the mask vector by setting the mask value corresponding to the sub-message from 0 to negative infinity, thereby removing it from the candidate sub-message set.

[0065] Specifically, the system repeatedly executes the steps of calculating the score, selecting the target, and updating the mask; after each round of selection, the length of the decoding sequence increases by 1, and the size of the candidate sub-message set decreases by 1. After... After rounds of iteration, all sub-messages are selected for the decoding order, and the candidate sub-message set becomes empty. At this point, the complete decoding order of all sub-messages in each sensor node is generated. .

[0066] Specifically, the complete decoding sequence It is a length of The sequence of elements corresponds to a unique identifier for each sub-message. This order indicates the decoding priority of each sub-message by the monitoring node during serial interference cancellation: the earlier the sub-message is listed, the higher its decoding priority. It is decoded first at the receiving end and removed from the superimposed signal, thereby reducing interference to subsequent decoded sub-messages.

[0067] In one embodiment, the two-stage resource optimization model is constructed with minimizing the system information age as the optimization cost.

[0068] Specifically, since information age, measured from the moment data is generated at the sensor node to the moment it is successfully received by the monitored node, comprehensively measures the total time the data has experienced, it can fully reflect the impact of various factors such as sampling interval, transmission waiting time, and retransmission delay on information freshness. Therefore, it is more suitable as a performance evaluation indicator for this type of service. However, minimizing the total information age of the system faces two core challenges: first, the randomness and causality of node energy collection make it difficult to establish an explicit analytical model between information age and decision variables; second, the strong coupling between transmit power and decoding order makes direct joint optimization difficult. This invention sets the objective as minimizing the system information age, while using a pre-trained two-stage resource optimization model to circumvent the explicit modeling problem. It also breaks down the hybrid optimization problem into two stages: power allocation and decoding order, which are implemented sequentially. This allows for continuous reduction of the network's information age under dynamic energy supply and time-varying channel conditions, achieving long-term optimization of information timeliness.

[0069] Specifically, the long-term expected total information age (ESA) of the system is defined as: ; Minimize the system information age as follows: ;

[0070]

[0071] ; In the formula, For the entire time span The power allocation sequence within, This represents the power allocation result for the t-th time slot. For the entire time span The decoding sequence within, This represents the complete decoding order of the t-th time slot.

[0072] Specifically, the joint action is performed on time slot t, i.e., the total information age of the system after power allocation and decoding order decision is used as the optimization cost of time slot t. The total information age of the system is defined as the sum of the information ages of all sensing nodes at the monitoring node. ,Right now ,in, Indicates the completion of the first step After the transmission of each time slot, the monitoring node has information about the sensing node. The information age. The smaller the value of this optimization cost, the fresher the overall information of the system, and the better the decision-making effect. The goal of the entire training process is to teach the two-stage resource optimization model to select the power allocation and decoding order combination that minimizes this cost in each time slot.

[0073] In one embodiment, the training process of the two-stage resource optimization model specifically includes: calculating the optimization cost sample of the current time slot; constructing a first-stage training sample set based on the optimization cost sample of the current time slot, the system state sample of the current time slot, the power allocation result sample of the current time slot, and the system state sample of the next time slot; constructing a second-stage training sample set based on the optimization cost sample of the current time slot, the decoding order of all sub-messages in the current time slot, and the second-stage input state sample of the current time slot, wherein the second-stage input state sample of the current time slot includes the sub-message transmit power sample, the total information age sample, the information age sample of each sensor node, and the uplink channel state sample of the current time slot; updating the model parameters of the first-stage model according to the first-stage training sample set, and updating the model parameters of the second-stage model according to the second-stage training sample set, to obtain the trained two-stage resource optimization model.

[0074] Preferably, the first-stage model is a power allocation module based on hybrid action reinforcement learning; the second-stage model is a decoding order optimization module based on sequence decision reinforcement learning.

[0075] Specifically, in each training round of the model, the complete system state of the current time slot is first collected and fed into the two-stage resource optimization model for inference, thereby obtaining the power allocation results of each sensor node and the decoding order of all sub-messages. The monitoring node controls the sensor nodes to complete uplink data packet transmission according to this set of decisions. Then, based on the success or failure of data packet transmission, the monitoring terminal information age and the battery energy of each sensor node across the entire network are updated synchronously. After completing all network interactions in this time slot, the optimization cost corresponding to the current time slot is calculated based on the updated information age of the entire network. This optimization cost uses the total information age of the system as the core metric, quantifying the information timeliness loss caused by the entire decision-making process of power allocation and decoding order in the current time slot, and serving as the reward and penalty basis for subsequent model learning.

[0076] Specifically, after completing the time slot cost calculation, multiple sets of time-series data generated in this round of interaction are collected simultaneously to build two types of independent training samples; for the first-stage model, the original system state samples of the current time slot are extracted. Sample power allocation results from the model output The time slot optimization cost sample just calculated and the system state sample generated in the next time slot after the interactive update is completed. The four elements are encapsulated into standardized quadruplet samples. The first-stage training sample set is continuously added, fully recording the network state, power actions, instantaneous costs, and state transition relationships. For the second-stage model, cost samples for the current time slot are extracted for optimization. The complete decoding order of all sub-messages output by the model and input state samples specific to the second stage. The second-stage input state sample integrates four types of data: sub-message transmission power converted from power allocation results, total system information age, monitoring terminal information age corresponding to each sensor node, and uplink channel status of the entire network. These contents are encapsulated into triplet samples. They are then collected into the second-stage training sample set.

[0077] Specifically, after the two types of sample sets accumulate to a preset size, parameter iterative updates are performed separately. The first-stage training sample set is read, and based on the state transition and slot cost information contained in the samples, a reinforcement learning algorithm adapted to a hybrid discrete-continuous action space is used to update all network parameters within the first-stage model, optimizing the output strategies for node scheduling, transmission power, and sub-message power allocation. Then, the second-stage training sample set is read, and based on the second-stage input state, decoding sequence, and slot cost within the samples, a policy gradient algorithm is used to update the second-stage model parameters, optimizing the self-attention feature extraction and sub-message ranking mask decision logic. The entire process of slot interaction, cost calculation, sample collection, and parameter update is executed iteratively, continuously optimizing the two sub-models until the models converge, ultimately resulting in a two-stage resource optimization model that has completed offline pre-training and can be deployed for online real-time resource decision-making in various slots.

[0078] Specifically, through the training process described above, the two-stage models co-evolve under the common optimization objective of minimizing the total information age of the system. The first-stage model learns to output a reasonable power allocation based on the system state, while the second-stage model learns to output the optimal complete decoding order based on the given power allocation result. After sufficient training, the two-stage models can be used for real-time decision-making in the online execution stage, enabling the output of the power allocation result and decoding order for the current time slot based on the current system state. ;in, Indicates the joint state of the system. Indicates model parameters.

[0079] Step 103: Based on the power allocation result and the decoding order, control each sensor node to send the status update data packet to the monitoring node in the current time slot to update the information age of each sensor node at the monitoring node and the battery energy of each sensor node.

[0080] In one embodiment, each sensor node performs uplink transmission based on the power allocation result output by the first-stage model.

[0081] Specifically, each sensing node follows the scheduling variables in the power allocation result. Determine whether it is activated in the current time slot: if the scheduling variable If the sensor node is activated, its transmit power will be determined according to the power allocation result. and power distribution ratio Allocate transmit power to each sub-message And the pre-divided sub-messages are sent to the monitoring node via the uplink channel at the corresponding sub-message transmission power; if If the sensor node remains silent in the current time slot and does not transmit uplink data, the total transmit power has been pruned in the previous steps to ensure that it does not exceed the maximum transmit power allowed by the battery energy in the current time slot, thus ensuring that the transmission behavior is physically feasible.

[0082] Specifically, each activated sensor node simultaneously sends its own sub-message to the monitoring node. Since all sensor nodes share the same uplink channel, the signal received by the monitoring node is the superposition of the sub-message transmission signals from all activated sensor nodes. The monitoring node then decodes the sub-messages according to the complete decoding sequence output by the second-stage model. The received superimposed signals are decoded to eliminate serial interference.

[0083] Specifically, when each activated sensor node simultaneously sends its own sub-message to the monitoring node, the uplink signal received by the monitoring node in the t-th time slot is: ,in, This represents the j-th sub-message of sensor node m. Let represent the transmit power corresponding to the j-th sub-message of sensor node m, and let n(t) represent the additive white Gaussian noise in the t-th time slot.

[0084] Specifically, the monitoring node decodes each sub-message sequentially using serial interference cancellation technology according to the complete decoding order: first, it decodes the first-ranked sub-message in the complete decoding order and removes it from the superimposed signal; then, it decodes the second-ranked sub-message in the remaining signal and removes it again; and so on, until all sub-messages are decoded; through this successive decoding and interference cancellation process, the monitoring node can recover the status update data packets uploaded by each sensor node from the multi-user superimposed signal.

[0085] In one embodiment, updating the information age of each sensor node at the monitoring node specifically includes: calculating the total transmission rate of the sensor node corresponding to the state update data packet; if the total transmission rate meets the preset data packet transmission requirements, then the state update of the sensor node is determined to be successful, and the information age of the sensor node at the monitoring node is reset to the current local information age of the sensor node; if the total transmission rate does not meet the preset data packet transmission requirements, then the information age of the sensor node at the monitoring node is incremented.

[0086] Specifically, given the power allocation and decoding order, the sub-message of time slot t... Sub-message transmission rate for: ; In the formula, For system bandwidth, The Shannon capacity formula represents the maximum spectral efficiency achievable for a given signal-to-interference-plus-noise ratio (SINR), and is the standard rate calculation formula in wireless communication theory. For sensing nodes The The sub-message in Sub-message transmission power of each time slot; For sensing nodes The The sub-message in The arrangement of each time slot in the complete decoding sequence For noise power, For sensing nodes The The sub-message in Sub-message transmission power of each time slot, For sensing nodes The The sub-message in The arrangement of each time slot in the complete decoding sequence.

[0087] Specifically, the total transmission rate of sensor node m is .

[0088] When sensor node m meets the preset data packet transmission requirements, that is When the status update data packet is successfully transmitted, it is determined that the data packet was transmitted successfully. This represents the data volume threshold of a sensor node state update data packet, which is the total number of bits contained in a state update data packet.

[0089] Specifically, define a transmission success indicator variable.

[0090] Specifically, if the status update of the sensor node is determined to be successful, the monitoring node will then update the information age of the sensor node at the monitoring node's end. Reset to the current local information age of this sensor node. This means that the status data held by the monitoring node regarding the sensor node has been updated to the latest collected data, and the information freshness has been restored.

[0091] Preferably, the local information age of the local state update data of sensor node m is defined as follows: Its evolution process is as follows: .

[0092] in, Let be the sampling indicator variable, which satisfies ,in, This represents the battery energy of sensor node m at the beginning of the t-th time slot. As basic energy, For sensing nodes In the The probability of performing state sampling in each time slot. This is the preset upper limit for the age of the information.

[0093] Preferably, the information age of sensor node m at the monitoring node is defined as... Its evolution process is as follows

[0094] Specifically, if the sensor node's state update is determined to have failed, the monitoring node will then monitor the information about the sensor node's age on the monitoring node's end. Increment by one time slot duration However, it should not exceed the preset upper limit for the age of the information. This means that the data held by the monitoring node about that sensor node is still outdated, and the information age continues to increase as time slots pass. Through the above mechanism, the monitoring node independently maintains and updates the information age for each sensor node, providing accurate information age input for system status acquisition and model decision-making in the next time slot.

[0095] In one embodiment, updating the battery energy of each sensor node specifically includes: acquiring the battery energy of each sensor node in the current time slot and the base energy of each sensor node; if the battery energy in the current time slot is greater than the base energy, acquiring the energy power collected by each sensor node in the current time slot, the transmission power, and the maximum battery energy, and calculating the battery energy of each sensor node in the next time slot based on the battery energy, the energy power, the transmission power, the base energy, and the maximum battery energy; if the battery energy in the current time slot is not greater than the base energy, acquiring the energy power collected by each sensor node in the current time slot and the maximum battery energy, and calculating the battery energy of each sensor node in the next time slot based on the battery energy, the energy power, and the maximum battery energy; and updating the battery energy of each sensor node based on the battery energy corresponding to the next time slot.

[0096] Specifically, after completing the uplink transmission of the current time slot, each sensor node updates its own battery energy status based on the energy collection and consumption of the current time slot, providing a basis for power supply in the next time slot. The battery energy update needs to be handled in two ways, depending on whether the battery energy of the current time slot is greater than the base energy.

[0097] Specifically, if the battery energy of the sensing node at the start of the current time slot... Greater than the base energy This indicates that the node has the capability to perform sampling and uplink transmission in the current time slot. In this case, the system acquires the energy power collected by the sensing node within the current time slot. Transmission power Time slot duration Basic energy and the battery's maximum energy The battery energy of the sensing node at the start of the next time slot. Calculate the energy power collected within the current time slot as follows: Subtract transmission power The difference between the time slot duration and the time slot duration Multiply them to get the first product, then calculate the first product and the battery energy. The first sum, minus the base energy. The first battery energy is obtained, and the minimum value between the first battery energy and the maximum battery energy is selected as the battery energy corresponding to each sensing node in the next time slot.

[0098] Specifically, if the battery energy of the sensing node at the start of the current time slot... Not greater than the base energy This indicates that the node does not have the capability to perform sampling and uplink transmission in the current time slot. In this case, the sensor node does not perform state sampling or uplink transmission, but only energy harvesting. The system obtains the energy power harvested by the sensor node in the current time slot. Time slot duration and the battery's maximum energy The battery energy of the sensing node at the start of the next time slot. Calculate energy power as follows: With time slot duration The second product, and calculate the second product with the battery energy. The second sum is selected as the minimum value between the second sum and the maximum battery energy, and this minimum value is taken as the battery energy corresponding to each sensing node in the next time slot.

[0099] Specifically, the evolution of battery energy into .

[0100] Through the two battery energy update mechanisms described above, the battery energy of each sensing node can accurately reflect the real-time changes in dynamic energy supply and consumption. (Updated battery energy...) As part of the system state in the next time slot, it provides energy constraints for the power allocation decision of the next round of two-stage resource optimization model, thus forming a closed-loop energy management system of energy harvesting, state awareness, model decision-making, uplink transmission, energy update, and the next time slot.

[0101] In one embodiment, the radio frequency energy provided by the wireless power transmitter determines the available energy budget of the sensor node. The available energy budget constrains the sensor node's transmit power and transmission success rate, which in turn directly affects the information age of each sensor node at the monitoring node. Based on this coupling relationship, this application uses a pre-trained two-stage resource optimization model to jointly optimize power allocation and decoding order in each time slot based on the current system state, including the information age of each sensor node, battery energy, and uplink channel state. This achieves continuous optimization of information timeliness under dynamic energy supply and time-varying channel conditions.

[0102] Example 2, see Figure 2 , Figure 2 This is a schematic diagram of an embodiment of an information timeliness optimization device for a communication network provided in this application. Corresponding to the above-described information timeliness optimization method for communication networks, this application also provides an information timeliness optimization device for communication networks. This information timeliness optimization device includes modules for executing the above-described information timeliness optimization method for communication networks, and can be configured in terminals such as desktop computers, tablet computers, and laptops. Specifically, the information timeliness optimization device includes a status acquisition module 201, a two-stage resource optimization model processing module 202, and an execution module 203.

[0103] The status acquisition module 201 is used to acquire the system status of each discrete time slot. The system status includes at least the information age of each sensor node at the monitoring node, the battery energy of each sensor node, and the uplink channel status.

[0104] The two-stage resource optimization model processing module 202 is used to input the system state into a pre-trained two-stage resource optimization model. The two-stage resource optimization model includes a first-stage model and a second-stage model. The first-stage model is used to output the power allocation results of each sensor node. The power allocation results include the scheduling variables, transmission power, and power allocation ratio of the transmission power among each sub-message of each sensor node. Based on the power allocation results, the sub-message transmission power of each sub-message in each sensor node is calculated, and the total information age of the system is calculated based on the information age of each sensor node. The sub-message transmission power, the total information age of the system, the information age of each sensor node, and the uplink channel state are input into the second-stage model so that the second-stage model outputs the complete decoding order of all sub-messages in each sensor node. The sub-messages are obtained by the sensor nodes from the state update data packets.

[0105] The execution module 203 is used to control each sensor node to send the status update data packet to the monitoring node in the current time slot according to the power allocation result and the decoding order, so as to update the information age of each sensor node at the monitoring node and the battery energy of each sensor node.

[0106] In one embodiment, the two-stage resource optimization model processing module 202 is used to input the sub-message transmission power, the total system information age, the information age of each sensor node, and the uplink channel state into the second-stage model, so that the second-stage model outputs the complete decoding order of all sub-messages in each sensor node. Specifically, this includes: taking all sub-messages not selected into the decoding order in the current time slot as a candidate sub-message set; using the self-attention mechanism in the second-stage model to extract features from the input sub-message transmission power, the total system information age, the information age of each sensor node, and the uplink channel state, to calculate the score vector for each candidate sub-message selected in the candidate sub-message set; selecting a target sub-message from the candidate sub-message set and adding it to the decoding order according to the score vector, and using a masking mechanism to mask the target sub-message from the candidate sub-message set to update the candidate sub-message set; repeating the above steps until all sub-messages in the candidate sub-message set are selected into the decoding order, generating the complete decoding order of all sub-messages in each sensor node.

[0107] In one embodiment, the execution module 203 is used to update the information age of each sensor node at the monitoring node, specifically including: calculating the total transmission rate of the sensor node corresponding to the status update data packet; if the total transmission rate meets the preset data packet transmission requirements, then the status update of the sensor node is determined to be successful, and the information age of the sensor node at the monitoring node is reset to the current local information age of the sensor node; if the total transmission rate does not meet the preset data packet transmission requirements, then the information age of the sensor node at the monitoring node is incremented.

[0108] In one embodiment, the execution module 203 is used to update the battery energy of each sensor node, specifically including: acquiring the battery energy of each sensor node in the current time slot and the base energy of each sensor node; if the battery energy in the current time slot is greater than the base energy, acquiring the energy power, transmission power, and maximum battery energy collected by each sensor node in the current time slot, and calculating the battery energy corresponding to each sensor node in the next time slot based on the battery energy, the energy power, the transmission power, the base energy, and the maximum battery energy; if the battery energy in the current time slot is not greater than the base energy, acquiring the energy power and maximum battery energy collected by each sensor node in the current time slot, and calculating the battery energy corresponding to each sensor node in the next time slot based on the battery energy, the energy power, and the maximum battery energy; and updating the battery energy of each sensor node based on the battery energy corresponding to the next time slot.

[0109] In one embodiment, the two-stage resource optimization model is constructed with minimizing the system information age as the optimization cost.

[0110] In one embodiment, the training process of the two-stage resource optimization model specifically includes: calculating the optimization cost sample of the current time slot; constructing a first-stage training sample set based on the optimization cost sample of the current time slot, the system state sample of the current time slot, the power allocation result sample of the current time slot, and the system state sample of the next time slot; constructing a second-stage training sample set based on the optimization cost sample of the current time slot, the decoding order of all sub-messages in the current time slot, and the second-stage input state sample of the current time slot, wherein the second-stage input state sample of the current time slot includes the sub-message transmit power sample, the total information age sample, the information age sample of each sensor node, and the uplink channel state sample of the current time slot; updating the model parameters of the first-stage model according to the first-stage training sample set, and updating the model parameters of the second-stage model according to the second-stage training sample set, to obtain the trained two-stage resource optimization model.

[0111] In one embodiment, the communication network is an uplink rate-division multiple access assisted wireless power communication network, and the communication network further includes a wireless power transmitter, wherein the wireless power transmitter is used to send radio frequency energy signals to each sensor node, so that each sensor node can collect energy from the radio frequency energy signals through an energy acquisition circuit, and upload a status update data packet to the monitoring node based on the collected energy.

[0112] The aforementioned communication network information timeliness optimization device can implement the communication network information timeliness optimization method of the above method embodiments. The options in the above method embodiments are also applicable to this embodiment, and will not be detailed here.

[0113] like Figure 3 As shown, Figure 3 This is a schematic diagram of the structure of a computer device provided in this application; it includes a processor 111, a communication interface 112, a memory 113 and a communication bus 114, wherein the processor 111, the communication interface 112 and the memory 113 communicate with each other through the communication bus 114, and the memory 113 is used to store computer programs.

[0114] In one embodiment of this application, the processor 111, when executing the program stored in the memory 113, implements the information timeliness optimization method for the communication network provided in any of the foregoing method embodiments.

[0115] It will be understood by those skilled in the art that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program may be stored in a storage medium, which is a computer-readable storage medium. The computer program is executed by at least one processor in the computer system to implement the process steps of the embodiments of the above methods.

[0116] Therefore, embodiments of this application also provide a computer-readable storage medium storing a computer program thereon, wherein the computer program, when executed by a processor, implements the steps of the information timeliness optimization method for communication networks provided in any of the foregoing method embodiments.

[0117] The storage medium is a physical, non-transient storage medium, such as a USB flash drive, external hard drive, read-only memory (ROM), magnetic disk, or optical disk, or any other physical storage medium capable of storing program code. The computer-readable storage medium can be non-volatile or volatile.

[0118] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this application.

[0119] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For example, the division of each unit is merely a logical functional division, and there may be other division methods in actual implementation. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed.

[0120] The steps in the methods of this application embodiment can be adjusted, merged, or deleted according to actual needs. The units in the apparatus of this application embodiment can be merged, divided, or deleted according to actual needs. Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0121] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a terminal, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application.

[0122] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0123] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Since these modifications and variations fall within the scope of the claims and their equivalents, this application also intends to include these modifications and variations.

[0124] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for optimizing the timeliness of information in a communication network, characterized in that, The communication network includes a monitoring node and multiple sensing nodes connected to the monitoring node, and the method includes: The system status of each time slot is obtained, and the system status includes at least the information age of each sensor node at the monitoring node, the battery energy of each sensor node, and the uplink channel status. The system state is input into a pre-trained two-stage resource optimization model, which includes a first-stage model and a second-stage model. The first-stage model outputs the power allocation results of each sensor node, including the scheduling variables, transmit power, and the power allocation ratio of the transmit power among the sub-messages of each sensor node. Based on the power allocation results, the sub-message transmit power of each sub-message in each sensor node is calculated, and the total information age of the system is calculated based on the information age of each sensor node. The sub-message transmit power, the total information age of the system, the information age of each sensor node, and the uplink channel state are input into the second-stage model so that the second-stage model outputs the complete decoding order of all sub-messages in each sensor node. The sub-messages are obtained by the sensor nodes from dividing the state update data packets. Based on the power allocation result and the decoding order, each sensor node is controlled to send the status update data packet to the monitoring node in the current time slot to update the information age of each sensor node at the monitoring node and the battery energy of each sensor node.

2. The method according to claim 1, characterized in that, The step of inputting the sub-message transmission power, the total system information age, the information age of each sensor node, and the uplink channel state into the second-stage model, so that the second-stage model outputs the complete decoding order of all sub-messages in each sensor node, specifically includes: All sub-messages in the current time slot that were not selected for the decoding order are taken as a set of candidate sub-messages; The self-attention mechanism in the second stage model is used to extract features from the input sub-message transmission power, the total information age of the system, the information age of each sensor node, and the uplink channel state, so as to calculate the score vector of each candidate sub-message selected in the candidate sub-message set. Based on the scoring vector, a target sub-message is selected from the candidate sub-message set and added to the decoding order. The target sub-message is then masked from the candidate sub-message set using a masking mechanism to update the candidate sub-message set. Repeat the above steps until all sub-messages in the candidate sub-message set are selected into the decoding order, generating a complete decoding order for all sub-messages in each sensing node.

3. The method according to claim 1, characterized in that, The information age of each sensor node at the updated monitoring node specifically includes: Calculate the total transmission rate of the sensor node corresponding to the state update data packet; If the total transmission rate meets the preset data packet transmission requirements, it is determined that the status update of the sensor node is successful, and the information age of the sensor node at the monitoring node is reset to the current local information age of the sensor node. If the total transmission rate does not meet the preset data packet transmission requirements, the information age of the sensor node at the monitoring node will be incremented.

4. The method according to claim 1, characterized in that, The updating of the battery energy of each sensing node specifically includes: Obtain the battery energy of each sensor node in the current time slot, as well as the base energy of each sensor node; If the battery energy in the current time slot is greater than the base energy, then the energy power, transmission power, and maximum battery energy collected by each sensing node in the current time slot are obtained. Based on the battery energy, the energy power, the transmission power, the base energy, and the maximum battery energy, the battery energy corresponding to each sensing node in the next time slot is calculated. If the battery energy in the current time slot is not greater than the base energy, then obtain the energy power collected by each sensing node in the current time slot and the maximum battery energy. Based on the battery energy, the energy power and the maximum battery energy, calculate the battery energy corresponding to each sensing node in the next time slot. The battery energy of each sensing node is updated based on the battery energy corresponding to the next time slot.

5. The method according to claim 1, characterized in that, The two-stage resource optimization model is constructed with minimizing the system information age as the optimization cost; The training process of the two-stage resource optimization model specifically includes: Calculate the optimized cost sample for the current time slot, and construct the first-stage training sample set based on the optimized cost sample for the current time slot, the system state sample for the current time slot, the power allocation result sample for the current time slot, and the system state sample for the next time slot. Based on the optimized cost sample of the current time slot, the decoding order of all sub-messages in the current time slot, and the second-stage input state sample of the current time slot, a second-stage training sample set is constructed. The second-stage input state sample of the current time slot includes the sub-message transmission power sample, the total information age sample, the information age sample of each sensor node, and the uplink channel state sample of the current time slot. The model parameters of the first-stage model are updated based on the first-stage training sample set, and the model parameters of the second-stage model are updated based on the second-stage training sample set, to obtain the trained two-stage resource optimization model.

6. The method according to claim 1, characterized in that, The communication network is an uplink rate segmented multiple access assisted wireless power supply communication network. The communication network also includes a wireless power transmitter, which is used to send radio frequency energy signals to each sensor node, so that each sensor node can collect energy from the radio frequency energy signals through an energy acquisition circuit, and upload a status update data packet to the monitoring node based on the collected energy.

7. A communication network information timeliness optimization device, characterized in that, include: The module consists of a status acquisition module, a two-stage resource optimization model processing module, and an execution module. The state acquisition module is used to acquire the system state of each discrete time slot. The system state includes at least the information age of each sensor node at the monitoring node, the battery energy of each sensor node, and the uplink channel state. The two-stage resource optimization model processing module is used to input the system state into a pre-trained two-stage resource optimization model. The two-stage resource optimization model includes a first-stage model and a second-stage model. The first-stage model is used to output the power allocation results of each sensor node. The power allocation results include the scheduling variables, transmission power, and power allocation ratio of the transmission power among each sub-message of each sensor node. Based on the power allocation results, the sub-message transmission power of each sub-message in each sensor node is calculated, and the total information age of the system is calculated based on the information age of each sensor node. The sub-message transmission power, the total information age of the system, the information age of each sensor node, and the uplink channel state are input into the second-stage model so that the second-stage model outputs the complete decoding order of all sub-messages in each sensor node. The sub-messages are obtained by the sensor nodes from the state update data packets. The execution module is configured to control each sensor node to send the status update data packet to the monitoring node in the current time slot according to the power allocation result and the decoding order, so as to update the information age of each sensor node at the monitoring node and the battery energy of each sensor node.

8. A computer device, characterized in that, The computer device includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the method as described in any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, The storage medium stores a computer program that, when executed by a processor, can implement the method as described in any one of claims 1-6.

Images