Wireless communication transmission method based on intelligent reflecting surface

Abstract

The invention discloses a wireless communication transmission method based on an intelligent reflecting surface, and relates to the technical field of wireless communication transmission, and the method comprises the steps: carrying out the time sequence modeling through employing an LSTM model, generating a channel characteristic model, generating an IRS initial reflection phase matrix based on the model, and optimizing the wireless channel configuration; dynamically adjusting a reflection phase matrix through a convex optimization algorithm, and continuously updating an IRS configuration scheme by taking maximization of a main channel gain and weakening of secondary channel interference as targets; on this basis, in combination with a channel characteristic model and an updated IRS configuration scheme, communication requirements between devices in an industrial environment are analyzed, link priorities are calculated, a low-delay communication topology is constructed, IRS resources are allocated through the priorities, and the signal quality of delay-sensitive task links is enhanced; furthermore, real-time data flow information is collected under the optimized topological structure, a task scheduling model is constructed by adopting a PPO reinforcement learning algorithm, a task transmission sequence and a resource allocation strategy are dynamically adjusted, and the resource utilization rate is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of wireless communication transmission, and particularly to a wireless communication transmission method based on an intelligent reflecting surface. Background Art

[0002] With the rapid development of industrial automation and intelligent manufacturing, wireless communication technology has become a key supporting means, providing a basis for the connection and data interaction between devices.

[0003] However, the wireless communication environment in industrial scenarios has dynamic interference sources and complex propagation conditions. Operating mechanical equipment, electromagnetic interference, and the multipath effect of metal structures will all cause signal attenuation and uncertainty in the propagation path. In this case, communication delay and unstable transmission have become common phenomena. In scenarios such as the coordinated control of industrial robots or the scheduling of automated production lines, communication delay will directly lead to task failure and even pose safety risks; to address these problems, some traditional wireless communication methods use high-frequency signals or dedicated frequency bands to reduce interference, but the spectrum resources themselves are limited and costly; or improve signal quality by increasing transmission power and receiving sensitivity, but this will also increase energy consumption and costs; some solutions deploy relay nodes to expand the signal coverage, but such solutions increase the complexity of the network and may introduce additional delays.

[0004] It can be seen that the existing solutions do not fundamentally solve this problem. Especially in industrial scenarios, most tasks are highly sensitive to delay, and the limitations of traditional methods are becoming more prominent. Therefore, how to achieve stable, fast, and reliable communication in complex environments has become an urgent problem to be solved in the current field. Summary of the Invention

[0005] In view of the above existing problems, the present invention is proposed.

[0006] The present invention provides a wireless communication transmission method based on an intelligent reflecting surface to solve the problems that in industrial automation scenarios, communication delay and unstable transmission will have a greater impact on automated tasks and even pose safety risks, while traditional solutions have problems such as insufficient utilization of spectrum resources, high energy consumption and high cost, increased network complexity, and insufficient support for delay-sensitive tasks.

[0007] To solve the above technical problems, the present invention provides the following technical solutions:

[0008] An embodiment of the present invention provides a wireless communication transmission method based on an intelligent reflecting surface, which includes,

[0009] Step S1, deploy sensors and an intelligent reflecting surface (IRS) in an industrial environment. The sensors collect channel characteristic data in real time, use a long short-term memory (LSTM) deep learning model to model the channel characteristic data to generate a channel characteristic model, and generate an IRS configuration scheme based on the channel characteristic model.

[0010] Step S2, use a convex optimization algorithm to dynamically adjust the reflection phase matrix of the IRS and continuously update the IRS configuration scheme.

[0011] Step S3, based on the updated IRS configuration scheme, analyze the communication requirements between devices in the industrial environment in combination with the channel characteristic model, adjust the priority of the communication link, and construct a communication topology.

[0012] Step S4, collect real-time data stream information under the topology of Step S3.

[0013] Use a proximal policy optimization (PPO) reinforcement learning algorithm based on a gradient strategy to construct a task scheduling model and adjust the transmission order of tasks and the resource allocation strategy.

[0014] Step S5, perform task transmission according to the adjusted scheme in Step S4, monitor the performance metrics of the transmission link in real time, and feedback the monitoring data to Step S1 and Step S2 to update the channel characteristic model and adjust the IRS configuration; the performance metrics include latency and packet loss rate.

[0015] As a preferred solution of the wireless communication transmission method based on an intelligent reflecting surface according to the present invention, wherein: the channel characteristic model is used to predict the changing trend of channel characteristics.

[0016] The installation position of the intelligent reflecting surface IRS covers the signal path of the main channel of wireless communication transmission.

[0017] The channel characteristic data includes signal attenuation, multipath effect, interference, dynamic occlusion, temperature and humidity changes, node state fluctuations, and spectrum usage changes.

[0018] The generation of the IRS configuration scheme refers to: deriving the initial reflection phase matrix of the IRS based on the channel characteristic model and generating a configuration scheme adapted to the wireless channel.

[0019] As a preferred solution of the wireless communication transmission method based on an intelligent reflecting surface according to the present invention, wherein: the steps of using an LSTM deep learning model to model the channel characteristic data to generate a channel characteristic model and generating an IRS configuration scheme based on the channel characteristic model are as follows.

[0020] Perform normalization processing on the collected channel characteristic data, and the normalization formula is:

[0021]

[0022] where d k is the k-th data point in the original channel characteristic data, k ∈ {1, 2, …, K}, K is the total number of channel characteristic data, μ d is the mean of the channel characteristic data, σ d is the standard deviation of the channel characteristic data, is the normalized channel characteristic data,

[0023] Use the LSTM network to perform time series modeling on the channel characteristic data. The LSTM network calculation includes:

[0024] f t = σ(W f x t + U f h t-1 + b f ),

[0025] i t = σ(W i x t + U i h t-1 + b i ),

[0026] o t = σ(W o x t + U o h t-1 + b o ),

[0027]

[0028] h t = o t ⊙ tanh(c t ),

[0029] where x t is the channel characteristic data input at the t-th moment, t ∈ {1, 2, …, T}, T is the total number of time steps, h t-1 is the hidden layer state at the (t - 1)-th moment, c t-1 is the memory cell value at the (t - 1)-th moment, f t , i t , o t are the activation values of the forget gate, input gate, and output gate, is the candidate memory cell value, c t is the memory cell value at the t-th moment, h t is the hidden layer state at the t-th moment, W f , W i , W o , W cis the weight matrix input to each gate, U f , U i , U o , U c is the weight matrix from the hidden state to each gate, b f , b i , b o , b c is the bias of the gate, σ is the Sigmoid activation function, tanh is the hyperbolic tangent activation function, and ⊙ is the element-wise multiplication;

[0030] For model training, the hidden layer predicts the channel characteristics through a fully connected layer, and the prediction formula is:

[0031] y t = W y h t + b y ,

[0032] where y t is the predicted value of the channel characteristics at time t, W y is the weight matrix from the hidden layer to the output layer, b y is the bias of the output layer,

[0033] According to the predicted value of the channel characteristics model, an initial reflection phase matrix is generated, and the matrix generation formula is:

[0034]

[0035] where,

[0036] Θ0 is the initial reflection phase matrix, Θ is an arbitrary reflection phase matrix, G s (Θ) is the main channel signal gain function, G i (Θ) is the secondary channel interference intensity function, and γ1 is the interference weight factor.

[0037] As a preferred scheme of the wireless communication transmission method based on intelligent reflecting surface described in the present invention, wherein: in step S2, the objective of the convex optimization algorithm is to maximize the main channel signal gain and weaken the interference intensity of the secondary channel to adapt to the real-time communication requirements.

[0038] As a preferred scheme of the wireless communication transmission method based on intelligent reflecting surface described in the present invention, wherein: the step of dynamically adjusting the reflection phase matrix of the IRS by using the convex optimization algorithm and continuously updating the IRS configuration scheme is,

[0039] Dynamically adjust the IRS reflection phase matrix, and the adjustment formula is:

[0040]

[0041] Among them, G s (Θ) is the main channel signal gain function, G i (Θ) is the secondary channel interference intensity function, Θ is the reflection phase matrix, with dimensions M×N, where M and N are the number of rows and columns of the IRS units respectively, and γ2 is the interference weight factor.

[0042] Set the constraint conditions:

[0043] Among them, θ mn is the phase value of the m-th row and n-th column in the reflection phase matrix, m ∈ {1, 2, …, M}, n ∈ {1, 2, …, N}.

[0044] Calculate the gradient to obtain the update direction, and the calculation formula is:

[0045]

[0046] Among them, is the gradient of the objective function with respect to Θ,

[0047] is the gradient of the main channel signal gain with respect to the reflection phase matrix,

[0048] is the gradient of the secondary channel interference intensity with respect to the reflection phase matrix;

[0049] Use the gradient descent method for iterative optimization, and the iterative optimization formula:

[0050]

[0051] Among them, Θ (k) is the reflection phase matrix at the k-th iteration, Θ (k+1) is the reflection phase matrix at the (k + 1)-th iteration, and η is the learning rate;

[0052] In each iteration, based on the channel characteristic model, predict and update the main channel gain and secondary channel interference, and the prediction formula is:

[0053]

[0054] Among them, U and V are the numbers of the main channel and the secondary channel respectively, w u , v v are the weights of the main channel and the secondary channel, h u , g v are the signal gains of the main channel and the secondary channel;

[0055] Through multiple iterations, obtain the optimized reflection phase matrix and generate a new IRS configuration scheme, denoted as:

[0056]

[0057] Among them,

[0058] Θ new is the finally optimized reflection phase matrix.

[0059] As a preferred solution of the wireless communication transmission method based on an intelligent reflecting surface according to the present invention, wherein: the method for adjusting the priority of the communication link is:

[0060] Assign a high priority to the latency-sensitive task link, improve the signal quality of the high-priority link through the IRS, and simultaneously construct a low-latency communication topology.

[0061] As a preferred solution of the wireless communication transmission method based on an intelligent reflecting surface according to the present invention, wherein: the steps of analyzing the communication requirements between devices in the industrial environment in combination with the channel characteristic model, adjusting the priority of the communication link, and constructing the communication topology structure are as follows

[0062] According to the updated IRS configuration scheme, analyze the communication requirements of the link, and define the communication requirements as a set D:

[0063] D = {(λ i , ρ i , π i ) | i ∈ {1, 2,..., N}},

[0064] Among them, D is the communication requirement set, λ i is the latency requirement of the i-th link, ρ i is the load of the i-th link, π i is the priority of the i-th link, and N is the total number of links.

[0065] Calculate the priority of each link according to the communication requirements, and the calculation formula is:

[0066]

[0067] Among them, π i is the priority of the i-th link, ω1, ω2 are the weight factors for priority calculation, λ i is the latency requirement of the i-th link, ρ i is the load of the i-th link;

[0068] Sort the links from high to low according to the priority to form a priority link list, expressed as:

[0069] Allocate more IRS resources to the high-priority links, and allocate the remaining resources to the low-priority links according to the communication requirements.

[0070] Construct a communication topology based on a priority link list, and the construction formula is:

[0071]

[0072] Among them, Topology is the final communication topology, is the set of all possible communication topologies,

[0073] is the weighted sum of the total link delays;

[0074] Allocate IRS resources according to the communication topology and link priorities, and the allocation formula is:

[0075]

[0076] Among them, is the IRS resource allocated to the i-th link, R total is the total resource capacity of the IRS, π i is the priority of the i-th link.

[0077] As a preferred solution of the wireless communication transmission method based on intelligent reflecting surface described in the present invention, wherein: the real-time data stream information includes task size, delay requirement, and link load.

[0078] As a preferred solution of the wireless communication transmission method based on intelligent reflecting surface described in the present invention, wherein: the step of constructing a task scheduling model by using the proximal policy optimization (PPO) reinforcement learning algorithm based on gradient strategy to adjust the transmission order of tasks and resource allocation strategy is,

[0079] Collect the real-time data stream information of the link, and define the data stream information as set F:

[0080] F = {(T q , D q , L q ) | q ∈ {1, 2,..., Q}},

[0081] Among them, F is the real-time data stream set, T q is the task size of the q-th data stream, D q is the delay requirement of the q-th data stream, L q is the link load of the q-th data stream, and Q is the total number of data streams;

[0082] Define the state space of the PPO reinforcement learning algorithm as S t :

[0083] S t = {(T q , D q , Lq , P q ) | q ∈ {1, 2, …, Q}},

[0084] Among them, S t is the system state at time step t, P q is the priority of the link corresponding to the q-th data stream;

[0085] Define the action space as the task transmission order and resource allocation strategy, denoted as A t :

[0086] A t = {a r | r ∈ {1, 2, …, Q}},

[0087] Among them, A t is the action set at time step t, a r is the scheduling and allocation decision for the r-th data stream;

[0088] The goal of the PPO algorithm is to maximize the cumulative reward J(π), and the goal formula is:

[0089]

[0090] Among them, J(π) is the cumulative reward of the policy π, T is the total number of time steps for task scheduling, E[·] is the expectation function, r t is the immediate reward at time step t, and γ is the discount factor;

[0091] Calculate the immediate reward according to the task completion situation, delay, and link load. The calculation formula is:

[0092]

[0093] Among them, r t is the immediate reward at time step t, φ is the delay reward factor, ζ is the load penalty factor, δ q is a binary indicator of whether the q-th data stream is completed, 1 means completed, 0 means not completed, D q is the delay requirement of the q-th data stream, and L q is the link load of the q-th data stream;

[0094] Update the policy using the PPO algorithm. The update formula is:

[0095]

[0096] Among them, θ t is the policy parameter at time step t, η is the learning rate, is the policy gradient;

[0097] The updated model generates a task scheduling order and a resource allocation scheme O:

[0098] O = {(a r , R r ) | r ∈ {1, 2, …, Q}},

[0099] where O is the set of optimized task scheduling results, and R r is the data stream resource allocated to the r-th task.

[0100] As a preferred solution of the wireless communication transmission method based on intelligent reflecting surface described in the present invention, wherein: the step of performing task transmission according to the adjusted solution in step S4, monitoring the performance indicators of the transmission link in real time, and feeding back the monitoring data to steps S1 and S2 to update the channel characteristic model and adjust the IRS configuration is

[0101] According to the optimized task scheduling scheme O generated in step S4, perform task transmission according to the scheduling order a r and the resource allocation R r During the task transmission process, monitor the link performance indicators, including the average delay and the packet loss rate. The performance indicators are defined as:

[0102]

[0103] where Δ is the average delay of task transmission, and P loss is the average packet loss rate of task transmission, δ q is a binary indicator of whether the q-th task is completed, D q is the delay of the q-th task, is the packet loss rate of the q-th task;

[0104] Feed back the monitored performance data to steps S1 and S2 for updating the channel characteristic model and the IRS configuration scheme. The feedback of delay data affects the update of the channel characteristic model, and the feedback of packet loss data affects the optimization of the IRS reflection phase matrix;

[0105] Use the feedback data to finely adjust the parameters of the LSTM model. The adjustment formula is:

[0106]

[0107] where θ (t) is the parameter of the channel characteristic model at time step t, η is the learning rate, is the gradient of the loss function with respect to the model parameters, y is the actual channel characteristic data, is the predicted value of the model, is the loss function;

[0108] Adjust the reflection phase matrix according to the link performance feedback, and the adjustment formula is:

[0109]

[0110] Wherein,

[0111] Θ new is the updated reflection phase matrix, Θ prev is the reflection phase matrix before update, ζ is the update step size, is the gradient of the performance metric with respect to the reflection phase matrix.

[0112] The beneficial effects of the present invention are as follows: In the present invention, an LSTM model is used for time series modeling to generate a channel characteristic model, and an initial reflection phase matrix of the IRS is generated based on this model to optimize the wireless channel configuration; the reflection phase matrix is dynamically adjusted through a convex optimization algorithm with the goal of maximizing the main channel gain and weakening the secondary channel interference, and continuously updating the IRS configuration scheme; on this basis, combining the channel characteristic model and the updated IRS configuration scheme, analyzing the communication requirements between devices in the industrial environment, calculating the link priority, constructing a low-latency communication topology, and enhancing the signal quality of the delay-sensitive task link by allocating IRS resources according to the priority; further, collecting real-time data stream information under the optimized topology structure, using the PPO reinforcement learning algorithm to construct a task scheduling model, dynamically adjusting the task transmission order and resource allocation strategy to achieve efficient resource utilization; during the task transmission process, the performance metrics of the link are monitored in real time, and the monitoring data is fed back to the channel characteristic model and the IRS configuration to update the LSTM model and the reflection phase matrix, realizing adaptive closed-loop optimization.

[0113] In summary, the present invention solves the disadvantages of the traditional solution in terms of dynamic interference, delay sensitivity, energy consumption cost, and resource utilization rate, and effectively improves the stability, reliability, and efficiency of communication in complex industrial scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0114] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0115] Figure 1 is the flowchart of the wireless communication transmission method based on the intelligent reflecting surface of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0116] In order to make the above objects, features, and advantages of the present invention more obvious and understandable, the following will make a detailed description of the specific embodiments of the present invention in conjunction with the drawings of the specification.

[0117] In the following description, many specific details are set forth in order to provide a thorough understanding of the present invention. However, the present invention may be practiced in other ways than those specifically described herein, and those skilled in the art may make similar extensions without departing from the spirit of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0118] Secondly, as used herein, an "embodiment" or "embodiments" refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The appearances of "in an embodiment" in different places in this specification do not all refer to the same embodiment, nor are they separate or alternative embodiments that exclude each other.

[0119] Embodiment 1, referring to Figure 1 , this embodiment provides a wireless communication transmission method based on an intelligent reflecting surface, including the following steps:

[0120] Step S1, deploy sensors and an intelligent reflecting surface IRS in an industrial environment. The sensors collect channel characteristic data in real time, use the deep learning model LSTM to model the channel characteristic data to generate a channel characteristic model, and generate an IRS configuration scheme based on the channel characteristic model;

[0121] The channel characteristic model is used to predict the changing trend of channel characteristics;

[0122] The installation position of the intelligent reflecting surface IRS covers the signal path of the main channel of wireless communication transmission;

[0123] The channel characteristic data includes signal attenuation, multipath effect, interference, dynamic occlusion, temperature and humidity changes, node state fluctuations, and spectrum usage changes;

[0124] Generating the IRS configuration scheme means: deriving the initial reflection phase matrix of the IRS based on the channel characteristic model and generating a configuration scheme adapted to the wireless channel;

[0125] The steps of using the deep learning model LSTM to model the channel characteristic data to generate a channel characteristic model and generating an IRS configuration scheme based on the channel characteristic model are

[0126] Normalize the collected channel characteristic data, and the normalization formula is:

[0127]

[0128] where d k is the k-th data point in the original channel characteristic data, k ∈ {1, 2,..., K}, K is the total number of channel characteristic data, μ d is the mean of the channel characteristic data, σd is the standard deviation of the channel characteristic data, is the normalized channel characteristic data,

[0129] Use the LSTM network to perform time series modeling on the channel characteristic data. The LSTM network calculation includes:

[0130] f t = σ(W f x t + U f h t-1 + b f ),

[0131] i t = σ(W i x t + U i h t-1 + b i ),

[0132] o t = σ(W o x t + U o h t-1 + b o ),

[0133]

[0134] h t = o t ⊙ tanh(c t ),

[0135] where x t is the channel characteristic data input at the t-th moment, t ∈ {1, 2,..., T}, T is the total number of time steps, h t-1 is the hidden layer state at the (t - 1)-th moment, c t-1 is the memory cell value at the (t - 1)-th moment, f t , i t , o t are the activation values of the forget gate, input gate, and output gate, is the candidate memory cell value, c t is the memory cell value at the t-th moment, h t is the hidden layer state at the t-th moment, W f , W i , W o , W c are the weight matrices input to each gate, U f , U i , U o , U c are the weight matrices from the hidden state to each gate, bf , b i , b o , b c For the bias of the gate, σ is the Sigmoid activation function, tanh is the hyperbolic tangent activation function, and ⊙ is the element-wise multiplication;

[0136] Model training is carried out, and the hidden layer predicts the channel characteristics through a fully connected layer. The prediction formula is:

[0137] y t = W y h t + b y ,

[0138] where y t is the predicted value of the channel characteristics at the t-th moment, W y is the weight matrix from the hidden layer to the output layer, and b y is the bias of the output layer.

[0139] According to the predicted value of the channel characteristics model, an initial reflection phase matrix is generated. The matrix generation formula is:

[0140]

[0141] where

[0142] Θ0 is the initial reflection phase matrix, Θ is any reflection phase matrix, G s (Θ) is the main channel signal gain function, G i (Θ) is the secondary channel interference intensity function, and γ1 is the interference weight factor;

[0143] Specifically, a channel characteristics model is generated through time series modeling of the LSTM network, and then the initial reflection phase matrix of the IRS is optimized to capture the changing law of the channel characteristics.

[0144] Step S2: Dynamically adjust the reflection phase matrix of the IRS using a convex optimization algorithm, and continuously update the IRS configuration scheme;

[0145] In step S2, the goal of the convex optimization algorithm is to maximize the main channel signal gain and weaken the interference intensity of the secondary channel to adapt to real-time communication requirements;

[0146] The steps of dynamically adjusting the reflection phase matrix of the IRS using a convex optimization algorithm and continuously updating the IRS configuration scheme are as follows.

[0147] Dynamically adjust the reflection phase matrix of the IRS. The adjustment formula is:

[0148]

[0149] where G s$(\Theta)$ is the main channel signal gain function, $G$ i $(\Theta)$ is the secondary channel interference intensity function, $\Theta$ is the reflection phase matrix with dimensions $M\times N$, where $M$ and $N$ are the number of rows and columns of the IRS units respectively, and $\gamma_2$ is the interference weight factor.

[0150] Set the constraint conditions:

[0151] where, $\theta$ mn is the phase value of the $m$-th row and $n$-th column in the reflection phase matrix, $m\in\{1,2,\ldots,M\}$, $n\in\{1,2,\ldots,N\}$.

[0152] Calculate the gradient to obtain the update direction, and the calculation formula is:

[0153]

[0154] where, is the gradient of the objective function with respect to $\Theta$,

[0155] is the gradient of the main channel signal gain with respect to the reflection phase matrix,

[0156] is the gradient of the secondary channel interference intensity with respect to the reflection phase matrix;

[0157] Use the gradient descent method for iterative optimization, and the iterative optimization formula:

[0158]

[0159] where, $\Theta$ (k) is the reflection phase matrix at the $k$-th iteration, $\Theta$ (k+1) is the reflection phase matrix at the $(k + 1)$-th iteration, and $\eta$ is the learning rate;

[0160] In each iteration, based on the channel characteristic model, predict and update the main channel gain and secondary channel interference, and the prediction formula is:

[0161]

[0162] where, $U$ and $V$ are the numbers of the main channel and the secondary channel respectively, $w$ u , $v$ v are the weights of the main channel and the secondary channel, $h$ u , $g$ v are the signal gains of the main channel and the secondary channel;

[0163] Through multiple iterations, obtain the optimized reflection phase matrix and generate a new IRS configuration scheme, denoted as:

[0164]

[0165] Among them,

[0166] Θ new is the finally optimized reflection phase matrix.

[0167] Specifically, by dynamically adjusting the reflection phase matrix through a convex optimization algorithm, the adaptation problem of dynamic channel changes is solved.

[0168] Step S3: Based on the updated IRS configuration scheme, analyze the communication requirements between devices in the industrial environment in combination with the channel characteristic model, adjust the priorities of communication links, and construct a communication topology structure;

[0169] The way to adjust the priorities of communication links is as follows:

[0170] Assign high priorities to latency-sensitive task links, enhance the signal quality of high-priority links through IRS, and simultaneously construct a low-latency communication topology;

[0171] The steps of analyzing the communication requirements between devices in the industrial environment in combination with the channel characteristic model, adjusting the priorities of communication links, and constructing a communication topology structure are as follows.

[0172] According to the updated IRS configuration scheme, analyze the communication requirements of the links, and define the communication requirements as a set D:

[0173] D = {(λ i , ρ i , π i ) | i ∈ {1, 2, …, N}},

[0174] Among them, D is the communication requirement set, λ i is the latency requirement of the i-th link, ρ i is the load of the i-th link, π i is the priority of the i-th link, and N is the total number of links.

[0175] Calculate the priority of each link according to the communication requirements. The calculation formula is:

[0176]

[0177] Among them, π i is the priority of the i-th link, ω1 and ω2 are weight factors for priority calculation, λ i is the latency requirement of the i-th link, and ρ i is the load of the i-th link;

[0178] Sort the links from high to low according to the priorities to form a priority link list, denoted as:

[0179] Allocate more IRS resources to high - priority links, and allocate the remaining resources to low - priority links according to communication requirements.

[0180] Construct a communication topology structure based on the priority link list. The construction formula is:

[0181]

[0182] where Topology is the final communication topology structure, is the set of all possible communication topology structures,

[0183] is the weighted sum of the total link delays;

[0184] Allocate IRS resources according to the communication topology and link priorities. The allocation formula is:

[0185]

[0186] where, is the IRS resources allocated to the i - th link, R total is the total resource capacity of the IRS, π i is the priority of the i - th link;

[0187] Specifically, step S3 dynamically adjusts the link priorities and optimizes the communication topology structure based on the updated IRS configuration scheme and channel characteristic model, classifies and sorts the links through the priority formula, gives priority to processing delay - sensitive tasks, and at the same time reduces the overall link delay through topology optimization; and dynamically adjusts the IRS resources according to the priorities.

[0188] Step S4, collect real - time data stream information under the topology structure of step S3;

[0189] Adopt the Proximal Policy Optimization (PPO) reinforcement learning algorithm based on gradient strategy to construct a task scheduling model and adjust the transmission order and resource allocation strategy of tasks

[0190] The real - time data stream information includes task size, delay requirement, and link load;

[0191] The steps of adopting the Proximal Policy Optimization (PPO) reinforcement learning algorithm based on gradient strategy to construct a task scheduling model and adjust the transmission order and resource allocation strategy of tasks are as follows.

[0192] Collect the real - time data stream information of the links and define the data stream information as set F:

[0193] F = {(T q , D q , L q ) | q ∈ {1, 2, …, Q}},

[0194] Among them, F is the set of real-time data streams, and T q is the task size of the q-th data stream, D q is the latency requirement of the q-th data stream, L q is the link load of the q-th data stream, and Q is the total number of data streams;

[0195] Define the state space of the PPO reinforcement learning algorithm as S t :

[0196] S t ={(T q , D q , L q , P q ) | q ∈ {1, 2, …, Q}},

[0197] where S t is the system state at time step t, and P q is the priority of the link corresponding to the q-th data stream;

[0198] Define the action space as the task transmission order and resource allocation strategy, denoted as A t :

[0199] A t ={a r | r ∈ {1, 2, …, Q}},

[0200] where A t is the action set at time step t, and a r is the scheduling and allocation decision for the r-th data stream;

[0201] The goal of the PPO algorithm is to maximize the cumulative reward J(π), and the goal formula is:

[0202]

[0203] where J(π) is the cumulative reward of policy π, T is the total number of time steps for task scheduling, E[·] is the expectation function, r t is the immediate reward at time step t, and γ is the discount factor;

[0204] Calculate the immediate reward according to the task completion situation, latency, and link load. The calculation formula is:

[0205]

[0206] where r t is the immediate reward at time step t, φ is the latency reward factor, ζ is the load penalty factor, δ q is a binary indicator of whether the q-th data stream is completed, 1 means completed, 0 means not completed, D qis the latency requirement for the q-th data stream, L q is the link load of the q-th data stream;

[0207] Use the PPO algorithm to update the policy, and the update formula is:

[0208]

[0209] where, θ t is the policy parameter at time step t, η is the learning rate, is the policy gradient;

[0210] The updated model generates the task scheduling order and resource allocation scheme O:

[0211] O = {(a r , R r ) | r ∈ {1, 2, …, Q}},

[0212] where, O is the set of optimized task scheduling results, R r is the data stream resource allocated to the r-th task;

[0213] Specifically, step S4 collects real-time data stream information, uses the PPO algorithm to optimize the task scheduling policy, efficiently processes task scheduling requirements under dynamic topologies, optimizes link resource allocation, and achieves low latency and efficient communication.

[0214] In step S5, execute task transmission according to the adjusted scheme in step S4, monitor the performance metrics of the transmission link in real time, and feedback the monitoring data to steps S1 and S2 to update the channel characteristic model and adjust the IRS configuration; the performance metrics include latency and packet loss rate;

[0215] The steps of executing task transmission according to the adjusted scheme in step S4, monitoring the performance metrics of the transmission link in real time, and feedbacking the monitoring data to steps S1 and S2 to update the channel characteristic model and adjust the IRS configuration are,

[0216] According to the optimized task scheduling scheme O generated in step S4, execute task transmission according to the scheduling order a r and resource allocation R r During the task transmission process, monitor the link performance metrics, including average latency and packet loss rate. The performance metrics are defined as:

[0217]

[0218] where, Δ is the average latency of task transmission, P loss is the average packet loss rate of task transmission, δ q is the binary indicator of whether the q-th task is completed, D q is the latency of the q-th task, is the packet loss rate of the q-th task;

[0219] Feed the monitored performance data back to steps S1 and S2 for updating the channel characteristic model and the IRS configuration scheme. Delayed data feedback affects the update of the channel characteristic model, and lost packet data feedback affects the optimization of the IRS reflection phase matrix;

[0220] Fine-tune the LSTM model parameters using the feedback data. The adjustment formula is:

[0221]

[0222] where θ (t) is the parameter of the channel characteristic model at time step t, η is the learning rate, is the gradient of the loss function with respect to the model parameters, y is the actual channel characteristic data, is the predicted value of the model, is the loss function;

[0223] Adjust the reflection phase matrix according to the link performance feedback. The adjustment formula is:

[0224]

[0225] where,

[0226] Θ new is the updated reflection phase matrix, Θ prev is the reflection phase matrix before update, ζ is the update step size, is the gradient of the performance metric with respect to the reflection phase matrix;

[0227] Specifically, step S5 performs closed-loop optimization by real-time monitoring of the link performance metrics during task transmission and feeding back the delay and packet loss rate to the channel characteristic model and the IRS configuration.

[0228] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not restrictive. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered within the scope of the claims of the present invention.

Claims

1. A wireless communication transmission method based on intelligent reflecting surface, characterized in that: including Step S1: Deploy sensors and an intelligent reflecting surface (IRS) in an industrial environment. The sensors collect channel characteristic data in real time. Use a long short-term memory (LSTM) deep learning model to model the channel characteristic data to generate a channel characteristic model, and generate an IRS configuration scheme based on the channel characteristic model. Step S2: Use a convex optimization algorithm to dynamically adjust the reflection phase matrix of the IRS and continuously update the IRS configuration scheme. Step S3: Based on the updated IRS configuration scheme, analyze the communication requirements between devices in the industrial environment in combination with the channel characteristic model, adjust the priority of the communication links, and construct a communication topology. Step S4: Collect real-time data stream information under the topology in Step S3. Use a proximal policy optimization (PPO) reinforcement learning algorithm based on gradient strategy to construct a task scheduling model and adjust the task transmission order and resource allocation strategy. Step S5: Execute task transmission according to the adjusted scheme in Step S4, monitor the performance metrics of the transmission link in real time, and feedback the monitoring data to Step S1 and Step S2 to update the channel characteristic model and adjust the IRS configuration; the performance metrics include latency and packet loss rate.

2. The wireless communication transmission method based on an intelligent reflecting surface according to claim 1, characterized in that: The channel characteristic model is used to predict the changing trend of channel characteristics. The installation location of the intelligent reflecting surface (IRS) covers the signal path of the main channel for wireless communication transmission. The channel characteristic data includes signal attenuation, multipath effect, interference, dynamic occlusion, temperature and humidity changes, node status fluctuations, and spectrum usage changes. Generating the IRS configuration scheme means: Deriving the initial reflection phase matrix of the IRS based on the channel characteristic model and generating a configuration scheme adapted to the wireless channel.

3. The wireless communication transmission method based on an intelligent reflecting surface according to claim 2, characterized in that: The steps of using the LSTM deep learning model to model the channel characteristic data to generate a channel characteristic model and generating an IRS configuration scheme based on the channel characteristic model are as follows. Perform normalization processing on the collected channel characteristic data. The normalization formula is: where d k is the k-th data point in the original channel characteristic data, k ∈ {1, 2, …, K}, K is the total number of channel characteristic data, μ d is the mean of the channel characteristic data, σ d is the standard deviation of the channel characteristic data, is the normalized channel characteristic data, Use an LSTM network to perform time series modeling on the channel characteristic data. The LSTM network calculations include: f t = σ(W f x t + U f h t-1 + b f ), i t = σ(W i x t + U i h t-1 + b i ), o t = σ(W o x t + U o h t-1 + b o ), h t = o t ⊙tanh(c t ) where x t is the channel characteristic data input at the t-th moment, t ∈ {1, 2, …, T}, T is the total number of time steps, h t-1 is the hidden layer state at the (t - 1)-th moment, c t-1 is the memory cell value at the (t - 1)-th moment, f t , i t , o t are the activation values of the forget gate, input gate, and output gate, is the candidate memory cell value, c t is the memory cell value at the t-th moment, h t is the hidden layer state at the t-th moment, W f , W i , W o , W c are the weight matrices input to each gate, U f , U i , U o , U c are the weight matrices from the hidden state to each gate, b f , b i , b o , b c are the biases of the gates, σ is the Sigmoid activation function, tanh is the hyperbolic tangent activation function, and ⊙ is the element-wise multiplication; Perform model training. The hidden layer predicts the channel characteristics through a fully connected layer. The prediction formula is: y t = W y h t + b y , where y t is the predicted value of the channel characteristics at the t-th moment, W y is the weight matrix from the hidden layer to the output layer, b y is the output layer bias, Generate an initial reflection phase matrix according to the predicted value of the channel characteristic model. The matrix generation formula is: where Θ0 is the initial reflection phase matrix, Θ is an arbitrary reflection phase matrix, G s (Θ) is the main channel signal gain function, G i (Θ) is the secondary channel interference intensity function, and γ1 is the interference weight factor.

4. The wireless communication transmission method based on an intelligent reflecting surface according to claim 3, wherein: In Step S2, the goal of the convex optimization algorithm is to maximize the main channel signal gain and weaken the interference intensity of the secondary channel to adapt to real-time communication requirements.

5. The wireless communication transmission method based on an intelligent reflecting surface according to claim 4, wherein: The steps of using the convex optimization algorithm to dynamically adjust the reflection phase matrix of the IRS and continuously update the IRS configuration scheme are as follows. Dynamically adjust the IRS reflection phase matrix. The adjustment formula is: Among them, G s (Θ) is the main channel signal gain function, G i (Θ) is the secondary channel interference intensity function, Θ is the reflection phase matrix, with dimensions of M×N, where M and N are the number of rows and columns of the IRS units respectively, and γ2 is the interference weight factor. Set constraints: where θ mn is the phase value of the m-th row and n-th column in the reflection phase matrix, m ∈ {1, 2, …, M}, n ∈ {1, 2, …, N} Calculate the gradient to obtain the update direction. The calculation formula is: Among them, is the gradient of the objective function with respect to Θ, is the gradient of the main-channel signal gain with respect to the reflection phase matrix, is the gradient of the reflection phase matrix with respect to the secondary channel interference intensity; Use the gradient descent method for iterative optimization. The iterative optimization formula: where, Θ (k) is the reflection phase matrix of the k-th iteration, Θ (k+1) is the reflection phase matrix of the (k + 1)-th iteration, and η is the learning rate; In each iteration, predict and update the main channel gain and secondary channel interference based on the channel characteristic model. The prediction formula is: Among them, U and V are the numbers of the primary channel and the secondary channel respectively, and w u , v v are the weights of the primary channel and the secondary channel, and h u , g v are the signal gains of the primary channel and the secondary channel; Through multiple iterations, obtain the optimized reflection phase matrix and generate a new IRS configuration scheme, expressed as: where Θ new is the finally optimized reflection phase matrix.

6. The wireless communication transmission method based on an intelligent reflecting surface according to claim 5, characterized in that: The method of adjusting the priority of the communication link is: Assign a high priority to the latency-sensitive task link, improve the signal quality of the high-priority link through the IRS, and simultaneously construct a low-latency communication topology.

7. The wireless communication transmission method based on intelligent reflecting surface according to claim 6, wherein: The steps of analyzing the communication requirements between devices in the industrial environment in combination with the channel characteristic model, adjusting the priority of the communication link, and constructing the communication topology are as follows: According to the updated IRS configuration scheme, analyze the communication requirements of the link and define the communication requirements as set D: D = {(λ i , ρ i , π i ) | i ∈ {1, 2, …, N}}, where D is the set of communication requirements, λ i is the latency requirement of the i-th link, ρ i is the load of the i-th link, π i is the priority of the i-th link, and N is the total number of links. Calculate the priority of each link according to the communication requirements, and the calculation formula is: where, π i is the priority of the i-th link, ω1, ω2 are the weight factors for priority calculation, λ i is the delay requirement of the i-th link, ρ i is the load of the i-th link; Sort the links from high to low according to the priority to form a priority link list, which is expressed as: Allocate more IRS resources to high-priority links, and allocate the remaining resources to low-priority links according to communication requirements. Construct the communication topology based on the priority link list, and the construction formula is: Among them, Topology is the final communication topology structure, is the set of all possible communication topology structures, is the weighted sum of the total link delays; According to the communication topology and link priority, allocate IRS resources, and the allocation formula is: Among them, is the IRS resource allocated to the i-th link, and R total is the total resource capacity of the IRS, and π i is the priority of the i-th link.

8. The wireless communication transmission method based on intelligent reflecting surface according to claim 7, characterized in that: The real-time data stream information includes task size, latency requirement, and link load.

9. The wireless communication transmission method based on an intelligent reflecting surface according to claim 8, characterized in that: The steps of using the reinforcement learning algorithm PPO based on the gradient strategy to construct a task scheduling model and adjust the transmission order and resource allocation strategy of the task are as follows: Collect the real-time data stream information of the link and define the data stream information as set F: F = {(T q , D q , L q ) | q ∈ {1, 2, …, Q}}, Among them, F is the set of real-time data streams, T q is the task size of the q-th data stream, D q is the latency requirement of the q-th data stream, L q is the link load of the q-th data stream, and Q is the total number of data streams; Define the state space of the PPO reinforcement learning algorithm as S t : S t = {(T q , D q , L q , P q ) | q ∈ {1, 2, …, Q}}, where S t is the system state at time step t, and P q is the priority of the link corresponding to the q-th data stream; Define the action space as the task transmission order and resource allocation strategy, denoted as A t : A t = {a r | r ∈ {1, 2, …, Q}}, Among them, A t is the set of actions at time step t, and a r is the scheduling and allocation decision for the r-th data stream; The goal of the PPO algorithm is to maximize the cumulative reward J(π), and the goal formula is: Among them, J(π) is the cumulative reward of policy π, T is the total number of time steps for task scheduling, E[·] is the expectation function, and r t is the immediate reward at time step t, and γ is the discount factor; Calculate the immediate reward according to the task completion situation, latency, and link load, and the calculation formula is: where, r t is the immediate reward at time step t, φ is the delayed reward factor, is the load penalty factor, δ q is a binary indicator of whether the q-th data stream is completed, 1 means completed, 0 means not completed, D q is the latency requirement of the q-th data stream, L q is the link load of the q-th data stream; Use the PPO algorithm to update the policy, and the update formula is: Among them, θ t is the policy parameter at time step t, η is the learning rate, is the policy gradient; The updated model generates a task scheduling order and a resource allocation scheme O: O = {(a r , R r ) | r ∈ {1, 2, …, Q}}, Among them, O is the set of optimized task scheduling results, and R r is the data flow resource allocated to the r-th task.

10. A wireless communication transmission method based on an intelligent reflecting surface according to claim 9, characterized in that: The steps of performing task transmission according to the adjusted scheme in step S4, real-time monitoring the performance indicators of the transmission link, and feeding back the monitoring data to step S1 and step S2 to update the channel characteristic model and adjust the IRS configuration are as follows: According to the optimized task scheduling scheme O generated in step S4, in accordance with the scheduling order a r and resource allocation R r Execute task transmission, and monitor the link performance metrics during the task transmission, including average latency and packet loss rate. The performance metrics are defined as: Among them, Δ is the average delay of task transmission, P loss is the average packet loss rate of task transmission, δ q is the binary indicator of whether the q-th task is completed, D q is the delay of the q-th task, is the packet loss rate of the q-th task; Feed back the monitored performance data to step S1 and step S2 for updating the channel characteristic model and the IRS configuration scheme. The feedback of latency data affects the update of the channel characteristic model, and the feedback of packet loss data affects the optimization of the IRS reflection phase matrix; Fine-tune the LSTM model parameters using the feedback data, and the adjustment formula is: where θ (t) is the parameter of the channel characteristic model at time step t, η is the learning rate, is the gradient of the loss function with respect to the model parameters, y is the actual channel characteristic data, is the predicted value of the model, is the loss function; Adjust the reflection phase matrix according to the link performance feedback, and the adjustment formula is: Among them, Θ new is the updated reflection phase matrix, Θ prev is the reflection phase matrix before update, is the update step size, is the gradient of the performance metric with respect to the reflection phase matrix.

Images