Resource scheduling method based on perception, communication and calculation integrated network

Abstract

The invention relates to the technical field of communication, and discloses a resource scheduling method based on a perception, communication and calculation integrated network, and the method comprises the steps: obtaining a radar perception signal-to-noise ratio, an uplink communication calculation rate and a user local calculation rate through a channel model which is a model constructed based on a reconfigurable intelligent surface, according to a resource scheduling algorithm based on block coordinate descent and a radar sensing signal-to-noise ratio, an uplink communication calculation rate and a user local calculation rate are processed to obtain a plurality of initialized variable blocks, and the plurality of initialized variable blocks are alternately optimized and subjected to variable updating to obtain an optimal resource scheduling result. Through the channel model constructed based on the reconfigurable intelligent surface, mutual interference between radar perception and communication calculation is effectively reduced, a block coordinate descent framework is adopted, cooperative scheduling of cross-layer resources is realized through alternate optimization, and the resource scheduling efficiency is improved.

Description

Technical Field

[0001] This application relates to the field of communication technologies, and particularly to a resource scheduling method based on a sensing, communication, and computing integrated network. Background Art

[0002] With the continuous increase in communication requirements, future networks not only need to achieve reliable data transmission but also support high-precision sensing and low-latency computing, which has prompted extensive research on sensing, communication, and computing integrated networks. In the resource scheduling of sensing, communication, and computing integrated networks, when simultaneously implementing radar sensing, communication transmission, and edge computing, the existing technologies face two core problems: channel interference and cross-layer resource scheduling, resulting in low resource scheduling efficiency. Summary of the Invention

[0003] The main purpose of this application is to provide a resource scheduling method based on a sensing, communication, and computing integrated network, aiming to solve the technical problem that the existing technology has low resource scheduling efficiency due to channel interference and cross-layer resource scheduling obstacles.

[0004] To achieve the above purpose, this application proposes a resource scheduling method based on a sensing, communication, and computing integrated network, and the method includes:

[0005] Obtain the radar sensing signal-to-noise ratio, uplink communication computing rate, and user local computing rate through a channel model, where the channel model is a model constructed based on a reconfigurable intelligent surface;

[0006] Process the uplink communication computing rate and the user local computing rate according to a resource scheduling algorithm based on block coordinate descent and the radar sensing signal-to-noise ratio to obtain multiple initial variable blocks;

[0007] Perform alternating optimization and variable update on the multiple initial variable blocks to obtain an optimal resource scheduling result.

[0008] In one embodiment, the resource scheduling algorithm includes a signal-to-interference-plus-noise ratio quadratic transformation formula, a Lagrangian dual transformation formula, and a successive convex approximation formula. The step of performing alternating optimization and variable update on the multiple initial variable blocks to obtain an optimal resource scheduling result includes:

[0009] Solve the energy allocation parameter of the user terminal in the multiple initial variable blocks through the signal-to-interference-plus-noise ratio quadratic transformation formula to determine the optimized energy allocation parameter;

[0010] Solve the received beamforming vector of the multiple initial variable blocks according to a convex optimization algorithm to determine the optimized received beamforming vector of the base station;

[0011] Solve the reflection beamforming matrix of multiple initialized variable blocks through the Lagrangian dual transformation formula and the successive convex approximation formula to determine the optimized reflection beamforming matrix of the reconfigurable intelligent surface;

[0012] Based on the optimized energy allocation parameter, the optimized receive beamforming vector, and the optimized reflection beamforming matrix, obtain the objective function minimized after one iteration;

[0013] If the current iteration number reaches the maximum iteration number and the improvement amplitude of the objective function is less than the convergence threshold, stop the iteration and output the optimal resource scheduling result.

[0014] In one embodiment, the step of solving the energy allocation parameter of the user terminal in multiple initialized variable blocks through the signal-to-interference-plus-noise ratio quadratic transformation formula to determine the optimized energy allocation parameter includes:

[0015] Fix the receive beamforming vector and the reflection beamforming matrix of the reconfigurable intelligent surface to obtain fixed variables;

[0016] Based on the current channel state and the fixed variables, determine the initial auxiliary variables;

[0017] Solve the energy allocation parameter of the user terminal in multiple initialized variable blocks through the signal-to-interference-plus-noise ratio quadratic transformation formula to obtain the solved energy allocation parameter;

[0018] Update the initial auxiliary variables based on the solved energy allocation parameter to obtain the updated auxiliary variables;

[0019] Optimize and iterate the solved energy allocation parameter according to the updated auxiliary variables to obtain the optimized energy allocation parameter.

[0020] In one embodiment, the step of solving the reflection beamforming matrix of multiple initialized variable blocks through the Lagrangian dual transformation formula and the successive convex approximation formula to determine the optimized reflection beamforming matrix of the reconfigurable intelligent surface includes:

[0021] Fix the optimized energy allocation parameter and the optimized receive beamforming vector;

[0022] Introduce auxiliary variables through the Lagrangian dual transformation formula and update the auxiliary variables to obtain the updated auxiliary variables;

[0023] Optimize the reflection beamforming matrix of multiple initialized variable blocks based on the successive convex approximation algorithm to determine the optimized reflection beamforming matrix of the reconfigurable intelligent surface.

[0024] In one embodiment, the steps of obtaining the radar sensing signal-to-noise ratio, the uplink communication computing rate, and the user's local computing rate through the channel model include:

[0025] The base station transmits a radar signal and, based on the radar signal, receives an echo signal from a radar target;

[0026] Adjust the reflection beamforming matrix according to the reconfigurable intelligent surface to obtain an optimized channel condition;

[0027] Based on the optimized channel condition, determine the radar sensing signal-to-noise ratio according to the ratio of the intensity of the echo signal to the noise power;

[0028] The user terminal sends an uplink communication signal to the base station to obtain the user's transmit power, the base station's receive beamforming vector, and channel interference information;

[0029] Integrate the user's transmit power, the receive beamforming vector, and the channel interference information to determine the uplink communication signal-to-interference-plus-noise ratio;

[0030] Determine the uplink communication computing rate according to the uplink communication signal-to-interference-plus-noise ratio;

[0031] Based on the energy allocation parameter of the user terminal, obtain the user's local computing rate.

[0032] In one embodiment, the steps of processing the uplink communication computing rate and the user's local computing rate according to the resource scheduling algorithm based on block coordinate descent and the radar sensing signal-to-noise ratio to obtain multiple initial variable blocks include:

[0033] Based on the resource scheduling algorithm based on block coordinate descent, under the constraint that the radar sensing signal-to-noise ratio is not lower than a preset threshold, perform a weighted sum of the computing rates to obtain an optimization objective of the total computing rate;

[0034] Based on the optimization objective, determine the energy allocation parameter according to the energy allocation ratio of each user for local computing and communication offloading;

[0035] Determine the receive beamforming vector according to the beamforming weights for signal reception of each user at the base station;

[0036] Determine the reflection beamforming matrix according to the phase adjustment parameters of each reflection unit of the reconfigurable intelligent surface;

[0037] Initialize the energy allocation parameter, the receive beamforming vector, and the reflection beamforming matrix to obtain multiple initial variable blocks.

[0038] In addition, to achieve the above object, the present application also proposes a resource scheduling device based on a perception, communication, and computing integrated network, where the resource scheduling device based on the perception, communication, and computing integrated network includes:

[0039] An information acquisition module, configured to obtain the radar sensing signal-to-noise ratio, the uplink communication computing rate, and the user local computing rate through a channel model, where the channel model is a model constructed based on a reconfigurable intelligent surface;

[0040] A variable partitioning module, configured to process the uplink communication computing rate and the user local computing rate according to a resource scheduling algorithm based on block coordinate descent and the radar sensing signal-to-noise ratio to obtain a plurality of initial variable blocks;

[0041] A scheduling optimization module, configured to perform alternating optimization and variable update on the plurality of initial variable blocks to obtain an optimal resource scheduling result.

[0042] In addition, to achieve the above object, the present application also proposes a resource scheduling device based on a perception, communication, and computing integrated network, where the device includes: a memory, a processor, and a computer program stored on the memory and executable on the processor, and the computer program is configured to implement the steps of the resource scheduling method based on the perception, communication, and computing integrated network as described above.

[0043] In addition, to achieve the above object, the present application also proposes a storage medium, where the storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium, and when the computer program is executed by a processor, it implements the steps of the resource scheduling method based on the perception, communication, and computing integrated network as described above.

[0044] In addition, to achieve the above object, the present application also provides a computer program product, where the computer program product includes a computer program, and when the computer program is executed by a processor, it implements the steps of the resource scheduling method based on the perception, communication, and computing integrated network as described above.

[0045] The technical solution proposed in this application obtains the radar sensing signal-to-noise ratio, the uplink communication computing rate, and the user's local computing rate through a channel model. Among them, the channel model is a model constructed based on a reconfigurable intelligent surface. According to the resource scheduling algorithm based on block coordinate descent and the radar sensing signal-to-noise ratio, the uplink communication computing rate and the user's local computing rate are processed to obtain multiple initial variable blocks, and the multiple initial variable blocks are alternately optimized and variable updated to obtain the optimal resource scheduling result. This application effectively reduces the mutual interference between radar sensing and communication computing through the channel model constructed based on the reconfigurable intelligent surface, adopts the block coordinate descent framework, and realizes the collaborative scheduling of cross-layer resources through alternating optimization, improving the resource scheduling efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] The accompanying drawings herein are incorporated into the specification and constitute a part of this specification, showing embodiments consistent with this application, and are used together with the specification to explain the principles of this application.

[0047] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the following will briefly introduce the accompanying drawings required for use in the description of the embodiments or the prior art. Obviously, for those of ordinary skill in the art, other drawings can also be obtained based on these drawings without creative efforts.

[0048] Figure 1 FIG. is a schematic flowchart provided for Embodiment 1 of the resource scheduling method based on the integrated network of sensing, communication, and computing of this application;

[0049] Figure 2 FIG. is a schematic flowchart provided for Embodiment 2 of the resource scheduling method based on the integrated network of sensing, communication, and computing of this application;

[0050] Figure 3 FIG. is a schematic flowchart of the resource scheduling algorithm based on the block coordinate descent optimization framework;

[0051] Figure 4 FIG. is a schematic flowchart provided for Embodiment 3 of the resource scheduling method based on the integrated network of sensing, communication, and computing of this application;

[0052] Figure 5 FIG. is a schematic diagram of the integrated network of radar sensing and edge computing;

[0053] Figure 6 FIG. is a schematic diagram of the module structure of the resource scheduling device based on the integrated network of sensing, communication, and computing in the embodiments of this application;

[0054] Figure 7 FIG. is a schematic diagram of the device structure of the hardware operating environment involved in the resource scheduling method based on the integrated network of sensing, communication, and computing in the embodiments of this application.

[0055] The implementation, functional features, and advantages of this application will be further described in conjunction with the embodiments and with reference to the accompanying drawings. Detailed implementation manners

[0056] It should be understood that the specific embodiments described herein are only used to explain the technical solutions of this application and are not used to limit this application.

[0057] To better understand the technical solutions of this application, the following will be described in detail in conjunction with the accompanying drawings of the specification and specific implementation manners.

[0058] In the resource scheduling of the integrated sensing, communication, and computing network, when simultaneously implementing radar sensing, communication transmission, and edge computing, the prior art faces two core problems: channel interference and cross-layer resource scheduling, resulting in low resource scheduling efficiency.

[0059] Therefore, to overcome the above defects, this application provides a solution. By using a channel model constructed based on reconfigurable intelligent surfaces, the mutual interference between radar sensing and communication computing is effectively reduced. The block coordinate descent framework is adopted, and the cross-layer resource collaborative scheduling is realized through alternating optimization, improving the resource scheduling efficiency.

[0060] It should be noted that the execution subject of each embodiment of this application can be a computing service system with data processing, network communication, and program running functions, such as an electronic system capable of implementing the above functions, a resource scheduling system based on the integrated sensing, communication, and computing network, etc. The following takes the resource scheduling system based on the integrated sensing, communication, and computing network (hereinafter referred to as the "system") as an example to describe the following embodiments.

[0061] Based on this, the embodiments of this application provide a resource scheduling method based on the integrated sensing, communication, and computing network, with reference to Figure 1 , Figure 1 is the flowchart of the first embodiment of the resource scheduling method based on the integrated sensing, communication, and computing network of this application.

[0062] In this embodiment, the resource scheduling method based on the integrated sensing, communication, and computing network includes steps S10 to S30:

[0063] Step S10, obtaining the radar sensing signal-to-noise ratio, the uplink communication computing rate, and the user local computing rate through the channel model, where the channel model is a model constructed based on reconfigurable intelligent surfaces.

[0064] In a network that integrates sensing, communication, and computing, to simultaneously achieve radar sensing, communication transmission, and edge computing, it is necessary to address the problem of low resource scheduling efficiency caused by channel interference and cross-layer resource scheduling obstacles. To solve these problems, this application proposes a resource scheduling method for a sensing, communication, and computing integrated network. The Reconfigurable Intelligent Surface (RIS) technology is used to mitigate the channel interference between radar sensing and computing offloading. For the cross-layer resource scheduling problem, a resource scheduling algorithm based on the block coordinate descent framework is designed to improve the overall performance of the sensing, communication, and computing integrated network.

[0065] In this step, it should be noted that the channel model is a mathematical model of the wireless propagation environment constructed based on RIS technology. It is a model integrating an integrated base station, user terminals, radar targets, and reconfigurable intelligent surfaces. Among them, the integrated base station includes radar sensing: target detection is achieved through multi-beam radar signal transmission and echo reception; wireless communication: uplink data transmission and radar sensing signal interaction with user terminals; edge computing: resource scheduling for user local computing tasks and tasks offloaded to the base station to optimize the overall system computing rate. The radar sensing signal-to-noise ratio is an index to measure the performance of radar target detection, representing the power ratio of the echo signal to the noise; the uplink communication computing rate refers to the data transmission rate at which users offload tasks to the base station through the wireless link; the user local computing rate is the ability of users to process tasks using their own computing resources, which is usually related to the frequency and energy allocation of the Central Processing Unit (CPU).

[0066] It can be understood that the channel model covers the signal superposition effect of the direct link and the reflected link. Through this model, the radar sensing signal-to-noise ratio (reflecting target detection performance), the uplink communication computing rate (reflecting the quality of the communication link), and the user local computing rate (reflecting local processing ability) can be calculated respectively. The core of this step is to utilize the channel regulation ability of RIS to provide accurate input parameters for subsequent resource scheduling.

[0067] Step S20: Process the uplink communication computing rate and the user local computing rate according to the resource scheduling algorithm based on block coordinate descent and the radar sensing signal-to-noise ratio to obtain multiple initial variable blocks.

[0068] It should be noted that block coordinate descent is an optimization algorithm that approximates the optimal solution by decomposing complex problems into multiple sub-problems and solving them alternately; initializing variable blocks can include energy allocation parameters, the receive beamforming vector of the base station, and the RIS reflection beamforming matrix, corresponding to the allocation of communication and computing resources, signal reception optimization, and channel condition regulation respectively; the radar sensing signal-to-noise ratio is used as one of the constraints to ensure that the sensing performance is not sacrificed during the resource scheduling process.

[0069] Based on the requirements of the radar sensing signal-to-noise ratio, the resource scheduling algorithm based on block coordinate descent divides the optimization problem into three variable blocks: the energy allocation parameter determines the allocation of user energy between local computing and task offloading; the receive beamforming vector is used to optimize the reception quality of the user signal by the base station; the RIS reflection beamforming matrix is used to adjust the channel conditions to reduce interference. During initialization, reasonable initial values are assigned to the above three variables.

[0070] Step S30, alternately optimize and update the variables of the multiple initialized variable blocks to obtain the optimal resource scheduling result.

[0071] It can be understood that alternating optimization is an iterative method that fixes some variables and optimizes the remaining variables each time, gradually approaching the global optimal solution; variable update refers to adjusting the variable values according to the current optimization results, for example, solving sub-problems through fractional programming or convex approximation methods; the optimal resource scheduling result is the final combination of energy allocation, beamforming, and RIS configuration obtained after the algorithm converges.

[0072] In the specific implementation, first optimize the energy allocation to improve communication energy efficiency, then optimize the receive beamforming to enhance signal quality, and finally optimize the RIS reflection matrix to improve channel conditions. After each iteration, check whether the objective function converges. If it does not converge, repeat the optimization process. Through this alternating update strategy, the algorithm finally outputs the optimal resource scheduling scheme that meets the sensing and communication requirements.

[0073] In this embodiment, the radar sensing signal-to-noise ratio, the uplink communication computing rate, and the user local computing rate are obtained through the channel model. Among them, the channel model is a model constructed based on the reconfigurable intelligent surface. According to the resource scheduling algorithm based on block coordinate descent and the radar sensing signal-to-noise ratio, the uplink communication computing rate and the user local computing rate are processed to obtain multiple initialized variable blocks, and the multiple initialized variable blocks are alternately optimized and variable updated to obtain the optimal resource scheduling result. Through the channel model constructed based on the reconfigurable intelligent surface, the mutual interference between radar sensing and communication computing is effectively reduced. Using the block coordinate descent framework, cross-layer resource collaborative scheduling is achieved through alternating optimization, improving the resource scheduling efficiency.

[0074] Based on the first embodiment of this application, in the second embodiment of this application, the same or similar content as that in the above-mentioned first embodiment can be referred to the above introduction and will not be elaborated hereinafter. On this basis, the resource scheduling algorithm includes a signal-to-interference-plus-noise ratio (SINR) quadratic transformation formula, a Lagrangian dual transformation formula, and a successive convex approximation formula. Referring to Figure 2 , step S30 may include steps S301 to S305:

[0075] Step S301, solve the energy allocation parameters of the user terminal in multiple initialization variable blocks through the SINR quadratic transformation formula to determine the optimized energy allocation parameters.

[0076] It should be noted that the SINR quadratic transformation formula is a mathematical transformation method used to convert a non-convex fractional optimization problem into a solvable form, which is applicable to optimizing the energy allocation parameters of the user terminal. By introducing auxiliary variables, this formula transforms the original problem into a more tractable convex optimization problem, thereby efficiently solving the optimal energy allocation.

[0077] As an implementation manner, step S301 in this embodiment may include: fixing the receive beamforming vector and the reflection beamforming matrix of the reconfigurable intelligent surface to obtain fixed variables; determining initial auxiliary variables based on the current channel state and the fixed variables; solving the energy allocation parameters of the user terminal in multiple initialization variable blocks through the SINR quadratic transformation formula to obtain the solved energy allocation parameters; updating the initial auxiliary variables based on the solved energy allocation parameters to obtain the updated auxiliary variables; and optimizing and iterating the solved energy allocation parameters according to the updated auxiliary variables to obtain the optimized energy allocation parameters.

[0078] In specific implementation, first fix the current values of the receive beamforming vector and the reflection beamforming matrix and use them as known conditions. The purpose is to decompose the multi-variable joint optimization problem into independent sub-problems, thereby simplifying the solution process of the energy allocation parameters and enabling focus on the local optimization of the energy allocation parameters to ensure the feasibility and efficiency of each iteration step.

[0079] The channel state reflects the transmission characteristics of the wireless link, including path loss, interference intensity, etc.; the initial auxiliary variable is an intermediate variable introduced by the SINR quadratic transformation formula, which is used to convert the fractional-form optimization objective into a solvable convex problem. In addition, the initial value of the auxiliary variable is usually calculated based on the current configuration of the channel state and the fixed variables. The core of calculating the initial auxiliary variable is to convert the non-linear fractional term in the original problem into a linear expression through mathematical transformation to reasonably select the initial auxiliary variable to accelerate the convergence speed of the algorithm.

[0080] The solved energy allocation parameter is the user energy allocation ratio obtained through the current iteration, which is used to balance the resource allocation between local computing and task offloading. Under the condition of fixing other variables and auxiliary variables, the optimization problem of the energy allocation parameter is transformed into a convex optimization problem and solved. That is, through quadratic transformation, the non-linear characteristics of the objective function are eliminated, enabling the problem to be efficiently solved by standard convex optimization methods. The solution result is the energy allocation parameter under the current iteration. Then, the obtained energy allocation parameter is substituted into the update formula of the auxiliary variable to recalculate its value. By dynamically adjusting the auxiliary variable, the approximate form of the objective function is made closer to the original problem, thereby gradually approaching the global optimal solution. The updated auxiliary variable will be used for the optimization of the energy allocation parameter in the next round.

[0081] It should be noted that the optimization iteration refers to the process of gradually improving the accuracy of the solution by repeatedly performing the processes of updating the auxiliary variable and solving the energy allocation parameter. The system uses the updated auxiliary variable to solve the optimization problem of the energy allocation parameter again. Through multiple iterations, the auxiliary variable and the energy allocation parameter promote each other until the change amplitude of the objective function is lower than the preset threshold, and finally the optimized energy allocation parameter is output.

[0082] Step S302: Solve the receive beamforming vectors of multiple initial variable blocks according to the convex optimization algorithm to determine the optimized receive beamforming vector of the base station.

[0083] The convex optimization algorithm is used to solve the receive beamforming vector, which determines how the base station combines multi-antenna received signals to maximize the signal-to-interference-plus-noise ratio. Its optimization problem can usually be transformed into a generalized eigenvalue problem, and its optimal solution corresponds to the maximum eigenvector of a specific matrix to effectively suppress interference and enhance the target signal.

[0084] In specific implementation, under the condition of fixing the energy allocation parameter and the reflection beamforming matrix, the optimization problem of the receive beamforming vector is modeled as a convex optimization problem. By solving this problem, the algorithm determines the optimal receive beamforming vector, enabling the base station to decode user signals with the highest efficiency.

[0085] Step S303: Solve the reflection beamforming matrices of multiple initial variable blocks through the Lagrangian dual transformation formula and the successive convex approximation formula to determine the optimized reflection beamforming matrix of the reconfigurable intelligent surface.

[0086] It should be noted that the Lagrangian dual transformation formula is used to process the logarithmic term in the objective function and convert it into a more easily optimized form; the successive convex approximation formula approximates the non-convex constraint as a convex constraint through an iterative linearization method, thereby gradually approaching the optimal solution.

[0087] As an implementation manner, in this embodiment, step S303 may include: fixing the optimized energy allocation parameter and the optimized received beamforming vector; introducing an auxiliary variable through the Lagrangian dual transformation formula, and updating the auxiliary variable to obtain the updated auxiliary variable; optimizing the reflection beamforming matrix of multiple initialized variable blocks based on the successive convex approximation algorithm, and determining the optimized reflection beamforming matrix of the reconfigurable intelligent surface.

[0088] In specific implementation, first, fix the energy allocation parameter and the received beamforming vector that have been optimized through the previous steps. Then, use the Lagrangian dual transformation formula to process the logarithmic term in the objective function, converting the logarithmic barrier term in the original objective function into a linear form. This conversion is achieved by introducing an auxiliary variable, and then update these auxiliary variables according to the value of the current reflection beamforming matrix. Further, the successive convex approximation algorithm constructs a convex approximation of the objective function near the current point, solves this convex problem to obtain a new reflection beamforming matrix. Through multiple iterations, each time reconstruct a convex approximation near the new solution and solve it, gradually approaching the optimal solution of the original non-convex problem, ensuring to find the phase configuration scheme that optimizes the system performance under the premise of satisfying the modulus constraint of the RIS unit, and finally output the optimized reflection beamforming matrix.

[0089] Step S304, based on the optimized energy allocation parameter, the optimized received beamforming vector, and the optimized reflection beamforming matrix, obtain the objective function minimized after one iteration.

[0090] It can be understood that the objective function is the core index of the resource scheduling problem, usually defined as the total system computing rate, including a comprehensive measure of communication and computing performance. After each iteration, recalculate the objective function value according to the updated variables to evaluate the optimization degree of the current solution, that is, after completing the optimization of energy allocation, received beamforming, and reflection beamforming matrix, substitute the three into the objective function formula to calculate the system performance of the current iteration and quantify the optimization effect.

[0091] Step S305, if the current iteration number reaches the maximum iteration number and the improvement amplitude of the objective function is less than the convergence threshold, stop the iteration and output the optimal resource scheduling result.

[0092] It should be noted that the maximum iteration number is one of the preset algorithm termination conditions to prevent infinite loops; the convergence threshold is used to determine whether the optimization of the objective function tends to be stable. When the improvement amplitude of the objective function is lower than this threshold, it indicates that the space for further optimization is limited and the algorithm can terminate.

[0093] In a specific implementation, it is checked whether the stopping condition is satisfied after each iteration. If the maximum number of iterations is reached or the change in the objective function is small enough, the optimization process is terminated, and the current energy allocation parameters, receive beamforming vector, and reflection beamforming matrix are output as the final solution.

[0094] For ease of understanding, reference is made to Figure 3 for illustration, but it does not limit the resource scheduling method of the present application based on the integrated network of sensing, communication, and computing. For the resource scheduling management problem P0 of maximizing the total system computing rate, a resource scheduling algorithm based on the block coordinate descent optimization framework is introduced. The schematic diagram of the resource scheduling algorithm based on the block coordinate descent optimization framework is as Figure 3 shown. The implementation steps are as follows:

[0095] A. Divide variable blocks: According to the characteristics and constraints of the problem, it is divided into 3 variable blocks, including variable block 1: energy allocation parameter α; variable block 2: receive beamforming vector w at the base station; variable block 3: reflection beamforming matrix Φ of the reconfigurable intelligent surface. By dividing the variable blocks, the original joint optimization problem is split into more tractable sub-problems, facilitating the use of different methods to solve them separately.

[0096] B. Initialize variables: Initialize the initial values of all variable blocks, initialize the variable values of α (0) , w (0) and Φ (0) ; Set the maximum number of iterations and the convergence tolerance. By initializing the variables, a reasonable initial point is provided to start the alternating optimization process, enhancing the feasibility and convergence speed of the algorithm.

[0097] C. Alternately optimize and update variables: In each iteration, select a block, fix the values of other blocks, and then optimize the selected block to minimize the objective function. Obtain the current updated variables according to the optimal value of the problem. Solve each sub-problem (problem P1, problem P2, and problem P3) by calling the signal-to-interference-plus-noise ratio quadratic transformation formula, the Lagrangian dual transformation formula, and the successive convex approximation formula. Optimize one variable block according to the following strategy in each iteration, and the remaining blocks remain fixed. Specifically: Given w (1) and Φ (1) , use the signal-to-interference-plus-noise ratio quadratic transformation formula to solve problem P1 and obtain the energy allocation parameter α (l+1) . By optimizing the allocation of user power, the data transmission rate and energy efficiency of the communication link are improved; Given α (l+1) and Φ (l) , solve problem P2 to obtain the receive beamforming vector w (l+1) of the base station to improve the signal-to-interference-plus-noise ratio at the receiving end, enhance the quality of the received signal, and reduce interference; Given w (l+1) and α (l+1), the Lagrange dual transformation formula and the successive convex approximation formula are used to jointly solve problem P3 to obtain the reflection beamforming matrix Φ of the reconfigurable intelligent surface (l+1) , to adjust the phase of the reconfigurable intelligent surface to enhance the channel gain, thereby improving the overall communication channel quality.

[0098] A complex non-convex optimization problem is decomposed into multiple smaller and relatively easy-to-solve sub-problems. By gradually optimizing each part of the variables, optimizing the current variable block while fixing other variables, and gradually approaching the local optimal solution or KKT point (Karush-Kuhn-Tucker conditions) of the overall problem, the system performance is improved. Among them, the KKT point refers to the point that satisfies the KKT conditions. The KKT conditions (Karush-Kuhn-Tucker conditions) are a set of necessary conditions for optimization problems in mathematics, used to describe the optimality conditions of constrained optimization problems.

[0099] D. Check for convergence: After each iteration, check whether the stopping condition is met. If the improvement amplitude of the objective function is less than the set threshold or the maximum number of iterations is reached, terminate the optimization to avoid redundant calculations.

[0100] E. Repeat iteration: If the stopping condition is not met, continue with step C for the next iteration until the stopping condition is satisfied. The number of iterations l = l + 1.

[0101] It should be noted that the process steps of the resource scheduling algorithm based on block coordinate descent are as follows:

[0102] 1. Input: Initialize the iteration index l = 0, the feasible value α (0) , w (0) and Φ (0) , the convergence threshold ξ;

[0103] 2. Repeat

[0104] 3. Given w (1) and Φ (1) , apply the signal-to-interference-plus-noise ratio quadratic transformation formula to solve problem P1 to obtain α (l+1) ;

[0105] 4. Given α (l+1) and Φ (l) , solve problem P2 to obtain w (l+1) ;

[0106] 5. Given w (l+1) and α (l+1) , apply the Lagrange dual transformation formula and the successive convex approximation formula

[0107] 6. Solve problem P3 using the formula and obtain Φ (l+1) ;

[0108] 7. Until the increase in the P0 target value is lower than the threshold ξ;

[0109] 8. Output: Optimal variable α * , w * and Φ * and the optimal value.

[0110] Among them, Repeat includes the following three parts:

[0111] ① The process steps of applying the signal-to-interference-plus-noise ratio (SINR) quadratic transformation formula are as follows:

[0112] 1. Initialization, obtain the network state, and set the maximum number of iterations I MAX and the convergence threshold δ; assign the initial energy parameter α to each user k k ;

[0113] 2. Main iteration loop (for loop: for t = 0, 1,..., I MAX do), iteratively optimize the objective function successively;

[0114] 3. Fix α and optimize the auxiliary variable y k * ;

[0115] 4. Fix y and solve the convex optimization problem;

[0116] 5. Judge the convergence. If satisfied, terminate the iteration;

[0117] 6. Otherwise, continue the for loop iteration and update the iteration count t.

[0118] ② The process steps of applying the Lagrange dual transformation formula are as follows:

[0119] 1. Input: Initial iteration n = 0, set the iteration starting point V (0) , step size γ, error threshold ∈, δ, and the maximum number of iterations N max ;

[0120] 2. Main loop (While n < N max do), execute the successive convex approximation iteration process;

[0121] 3. Calculate the values and gradients of the current objective function f q (V) and the constraint term f incq (V) respectively;

[0122] 4. Solve the linearized convex sub-problem to obtain

[0123] 5. Judge the convergence. If satisfied, terminate the iteration; otherwise, continue the iteration;

[0124] 6.Continue the iteration of the while loop (step 2) and update t; update V = V (n) +γ(V new -V (n) );

[0125] Step 7: Finally output the optimal value V opt = V.

[0126] ③The process steps of applying the successive convex approximation formula are as follows:

[0127] 1. Input: the initial values of parameters η, y, and V;

[0128] 2. Outer iteration of the main loop (while the stop condition is not satisfied do), update η by formula (20);

[0129] 3. Nested while loop, update y and V simultaneously in the inner loop, optimize the auxiliary variable y; call the Lagrangian dual transformation formula to optimize the variable V, and if the stop condition is satisfied, jump out of the nested while loop.

[0130] 4. If the stop condition is satisfied, jump out of the outer while loop;

[0131] 5. Finally output the optimal solution V * .

[0132] In this embodiment, through the collaborative application of three mathematical transformation formulas, the energy allocation parameter, the receive beamforming vector, and the reflection beamforming matrix are optimized respectively, so that each module maximizes the calculation rate while ensuring the sensing performance. Moreover, the non-convex problem is transformed into a solvable convex optimization problem by using the signal-to-interference-plus-noise ratio quadratic transformation, and the energy allocation parameter is ensured to converge quickly through the iterative update of the auxiliary variable. Combining the dual optimization of Lagrangian duality transformation and successive convex approximation, the fast convergence of the reflection matrix is achieved under the condition of satisfying the unit modulus constraint. Compared with the traditional method, it improves the optimization speed of the reflection beamforming and can adapt to the channel change.

[0133] Based on the first embodiment of the present application, in the third embodiment of the present application, the content that is the same as or similar to the above-mentioned embodiment 1 can be referred to the above introduction and will not be elaborated hereinafter. On this basis, please refer to Figure 4 , the step S10 may include steps S101 to S107:

[0134] Step S101, transmit a radar signal through the base station, and based on the radar signal, receive an echo signal from the radar target.

[0135] It should be noted that the radar signal is a dedicated electromagnetic wave signal transmitted by the base station to detect targets, with specific waveforms and modulation methods; the echo signal refers to the signal reflected back after the radar signal encounters a target, and its characteristics include information such as the position and speed of the target. In specific implementation, the base station first generates and transmits the designed radar signal waveform. These signals propagate in space, are reflected when encountering radar targets, and form echo signals. The base station captures these echo signals through the receiving antenna array.

[0136] Step S102, adjust the reflection beamforming matrix according to the reconfigurable intelligent surface to obtain the optimized channel condition.

[0137] It can be understood that according to the current communication and sensing requirements, by dynamically adjusting the reflection beamforming matrix of the RIS and optimizing the phase of each reflection unit in real time, the RIS can enhance the energy in the direction of the desired signal and suppress the signals in the interference direction, thereby improving the overall channel condition.

[0138] Step S103, based on the optimized channel condition, determine the radar sensing signal-to-noise ratio according to the ratio of the intensity of the echo signal to the noise power.

[0139] In this step, first measure the intensity (power) of the echo signal optimized by the RIS, then calculate the noise power in the receiving system, and obtain the real-time radar sensing signal-to-noise ratio through the ratio of the two. Among them, the noise power includes the influence of environmental noise and system internal noise.

[0140] Step S104, send an uplink communication signal from the user terminal to the base station to obtain the user transmit power, the receiving beamforming vector of the base station, and the channel interference information.

[0141] It should be noted that the uplink communication signal is a wireless signal for the user terminal to transmit data to the base station; the user transmit power determines the transmission distance and quality of the signal; the channel interference information includes the interference conditions caused by other user signals and radar signals.

[0142] Each user terminal sends an uplink communication signal to the base station according to the allocated resources. The base station records the transmit power of each user and calculates the optimal receiving beamforming vector to enhance the target signal. At the same time, the system measures the interference power from other users and the radar system.

[0143] Step S105, comprehensively determine the uplink communication signal-to-interference-plus-noise ratio based on the user transmit power, the receiving beamforming vector, and the channel interference information.

[0144] In a specific implementation, the user transmit power, the base station receive beamforming vector, and the channel interference information are substituted into the signal-to-interference-plus-noise ratio calculation formula. By accurately calculating the enhancement effect of the desired signal and the suppression effect of the interference signal, the uplink communication signal-to-interference-plus-noise ratio of each user is obtained.

[0145] Step S106: Determine the uplink communication calculation rate according to the uplink communication signal-to-interference-plus-noise ratio.

[0146] The uplink communication calculation rate represents the data transmission ability of the user to offload tasks to the base station through the wireless link, usually in bits per second. The calculated signal-to-interference-plus-noise ratio can be substituted into the Shannon capacity formula. Considering factors such as the system bandwidth, the theoretical maximum uplink communication rate of each user can be calculated. This rate index reflects the ability of the user to successfully transmit the calculation task data to the base station under the current resource configuration.

[0147] Step S107: Obtain the user local calculation rate based on the energy allocation parameter of the user terminal.

[0148] It can be understood that according to the proportion of energy allocated by the user for local calculation and combined with the calculation efficiency parameter of the device, the maximum calculation rate supported by the local CPU is calculated. This index reflects the ability of the user to complete the calculation task independently without relying on the base station. The user local calculation rate and the uplink communication calculation rate together constitute the evaluation of the user's total calculation ability.

[0149] As an implementation manner, step S20 may include: based on the resource scheduling algorithm of block coordinate descent, under the constraint condition that the radar sensing signal-to-noise ratio is not lower than a preset threshold, perform a weighted sum on the calculation rate to obtain an optimization target of the total calculation rate; based on the optimization target, determine the energy allocation parameter according to the energy allocation ratio of each user for local calculation and communication offloading; determine the receive beamforming vector according to the beamforming weights of the base station for receiving signals of each user; determine the reflection beamforming matrix according to the phase adjustment parameters of each reflection unit of the reconfigurable intelligent surface; initialize the energy allocation parameter, the receive beamforming vector, and the reflection beamforming matrix to obtain a plurality of initialized variable blocks.

[0150] It should be noted that the optimization target of the total calculation rate is a system performance index that comprehensively considers the uplink communication calculation rate and the user local calculation rate.

[0151] In a specific implementation, an optimization problem is first established with the goal of maximizing the total computing rate, where the total computing rate is the weighted sum of the uplink communication computing rate and the user's local computing rate. At the same time, this optimization problem needs to satisfy the constraint that the radar sensing signal-to-noise ratio is not lower than a preset threshold to ensure that the sensing performance is not affected. Then, according to the established optimization goal, the energy allocation parameter is taken as the first variable block to be optimized. By adjusting this ratio, while ensuring the basic communication requirements, the overall computing ability of the system can be optimized. Then, the receiving beamforming vector of the base station and the reflection beamforming matrix of the RIS are optimized one by one.

[0152] Reasonable initial values are set for the energy allocation parameter, the receiving beamforming vector, and the reflection beamforming matrix respectively to form three initialization variable blocks. These initial values can be determined by methods such as uniform distribution, random generation, or predicted values based on historical data.

[0153] In this embodiment, the sensing signal-to-noise ratio and communication parameters are obtained synchronously through the channel model, realizing the cooperative processing of radar and communication signals, improving the spectrum utilization rate, while ensuring that the sensing accuracy does not decrease. Moreover, the total computing rate target is decomposed into three independently adjustable dimensions of energy, beam, and phase, and fast convergence is achieved through variable block initialization, reducing the number of algorithm convergence iterations.

[0154] For ease of understanding, reference is made to Figure 5 for illustration, but it does not limit the resource scheduling method of the present application for the integrated network of sensing, communication, and computing. Figure 5 Fig. 13 is a schematic diagram of an integrated network of radar sensing and edge computing, considering a reconfigurable intelligent surface-assisted integrated network of radar sensing and edge computing. The network consists of a communication and sensing integrated base station, K communication users represented by the set K = {1......K}, a radar target, and a reconfigurable intelligent surface composed of M reflection units.

[0155] To improve the utilization efficiency of spectrum and time, the radar sensing and edge offloading system share the same frequency band and time resources. In particular, the base station is equipped with N t transmitting antennas and N r receiving antennas, which are arranged in a uniform linear array form to communicate with all single-antenna users in the uplink and detect nearby radar targets. For simplicity of analysis, it is assumed that N t = N r = N. Since the communication and sensing integrated base station integrates an edge server, it can perform the dual functions of radar sensing and receiving user offloading simultaneously on the same spectrum channel. The specific steps are as follows:

[0156] 1. Base station transmits signals

[0157] To achieve the above functions, the base station broadcasts a dedicated radar signal s0 in the form of multiple beams. In particular, assume that where the covariance matrix The signal s0 can be further decomposed by the eigenvalue decomposition of R0 to obtain the following form:

[0158]

[0159] All sensing waveforms are statistically independent of each other and follow s 0，i ~CN(0,1). w 0，i denotes the transmit beamforming vector of s 0，i and is uniquely determined by the eigenvalue decomposition of R0.

[0160] 2. Model of Reconfigurable Intelligent Surface

[0161] The reconfigurable intelligent surface consists of M reflecting elements, and the phase shift of each reflecting element is represented by θ m ∈[0,2π), m ∈ M. The diagonal elements of the reflection beamforming matrix are composed of and the amplitude of all reflecting elements is 1. Therefore, the mathematical model of Φ can be expressed as The channel between the reconfigurable intelligent surface and the base station is represented as The channel between user k and the reconfigurable intelligent surface is represented as while the channel between the radar target and the reconfigurable intelligent surface is represented as

[0162] 3. Models of Communication and Computing

[0163] Each user has a limited energy budget but needs to complete intensive computing tasks. The available computing energy of user k is denoted as E k . This example adopts a partial task offloading mode, and partial offloading aims to process computationally flexible tasks, enabling the user's local computing and edge computing on the base station to run in parallel.

[0164] For uplink transmission, denotes the information symbol of the task offloaded by user k, and p k denotes the transmit power of user k. Among them, all users with offloading requirements send signals to the base station simultaneously, and the base station will receive the radar echo signal at the same time. Therefore, the corresponding received signal at the base station can be expressed as:

[0165]

[0166] where The equivalent baseband channel from user k to the base station; g t The equivalent baseband channel from the radar target to the base station; n is additive white Gaussian noise that is independent and identically distributed with a mean of 0 and a variance of σ 2 , that is

[0167] By deploying a reconfigurable intelligent surface, the equivalent baseband channel from user k to the base station includes the direct link and the reflected link from the reconfigurable intelligent surface. Therefore, can be modeled as:

[0168]

[0169] The equivalent baseband channel from the radar target to the base station includes the direct link and the reflected link from the reconfigurable intelligent surface. Therefore, g t can be modeled as:

[0170] g t = h dr + G H Φg tr (4)

[0171] For the offloading task, the energy partitioning parameter α k ∈[0,1], α k E k represents the energy used for computing offloading. Accordingly, the transmit power p k used by user k for computing offloading is:

[0172]

[0173] where T is the length of the running time slot.

[0174] Considering the linear beamforming strategy and denoting as the receive beamforming vector of the base station for decoding . Based on Equation (2), the signal received by user k at the base station is represented by and is given by:

[0175]

[0176] Therefore, the signal-to-interference-plus-noise ratio of the uplink signal of user k observed at the base station is given by:

[0177]

[0178] Then the computing rate generated by user k due to offloading is:

[0179]

[0180] where B is the system bandwidth.

[0181] For local computing, the computing rate is expressed as f k / C k , where C k is the number of CPU cycles required for the user to compute each bit of input data, and f k is the user's CPU frequency. By adopting dynamic voltage and frequency scaling technology, the user's computing energy efficiency is improved by adaptively controlling the CPU frequency of local computing. Specifically, the user's energy consumption can be calculated as where is the effective capacitance coefficient of user k. In addition, since (1 - a k )E k represents the energy of local computing, so there is Therefore, the local computing rate of user k is:

[0182]

[0183] 4. Model of Radar Sensing

[0184] After the receiver decodes the communication signal, the communication signal is removed from the observed waveform using successive interference cancellation, and an interference-free radar echo signal is obtained. Therefore, with the assistance of the reconfigurable intelligent surface, the echo signal collected from the base station can be expressed as:

[0185] y rad = g t (g t ) H s0 + n (10)

[0186] The sensing signal-to-noise ratio is given by:

[0187]

[0188] The goal of this example is to maximize the task computing rates (including offloading and local computing) of all users on the premise of ensuring that the sensing signal-to-noise ratio is large enough for reliable target detection within a given duration T. Therefore, there is the following optimization problem P0:

[0189]

[0190] The optimization problem considers the receive beamforming vector w at the base station to enhance the reception quality of the offloading signal and introduces the energy allocation parameter α to balance the resources allocated for local computing and offloading computing. It can be seen from Eqs. (7) and (11) that the reflection beamforming matrix Φ of the reconfigurable intelligent surface acts jointly in the dual functions, where an effective trade-off must be made between conflicting sensing and communication metrics. Therefore, {α, Φ, w} are jointly designed to maximize the task computing rate of all users. Constraint 1 (a k ∈[0,1], k ∈ K) restricts the energy allocation ratio to be between 0 and 1. Constraint 2 (|Φ mm | = 1, m ∈ M) ensures that the amplitude of the reflection unit of the reconfigurable intelligent surface is 1. Constraint 3 (r ad (Φ) ≥ γ min ) ensures that the sensing signal-to-noise ratio is not lower than a specific sensing threshold.

[0191] 5. Resource Scheduling Method for Sensing, Communication, and Computing Integrated Network of This Application

[0192] To solve the complex non-convex problem P0, the resource scheduling method for the sensing, communication, and computing integrated network of this application is introduced. The specific implementation steps are as follows: First, according to the characteristics and constraints of the problem, it is divided into the energy allocation parameter α, the reflection beamforming matrix Φ of the reconfigurable intelligent surface, and the receive beamforming vector w at the base station. Then, the variable values of α (0) , w (0) , and Φ (0) are initialized. Subsequently, the variables are alternately optimized and updated. In each iteration, one block is selected, the values of other blocks are fixed, and then the selected block is optimized to minimize the objective function. The current updated variables are obtained according to the optimal value of the problem. Finally, after each iteration, it is checked whether the stopping condition is satisfied. If the condition is satisfied, the iteration is directly exited. If the stopping condition is not reached, the next iteration is performed until the stopping condition is satisfied. Among them, the specific derivation processes for solving each sub-problem (Problem P1, Problem P2, and Problem P3) are as follows:

[0193] (1) Update the energy allocation parameter α (solve Problem P1)

[0194] With other variables kept fixed, the update of the energy allocation parameter α can be achieved by solving sub-problem P1. The objective of sub-problem P1 is to optimize the energy allocation to maximize the system performance under the current conditions. The sub-problem P1 for updating the energy allocation parameter α is:

[0195]

[0196] This problem retains the terms coupled with a k . Due to the channel interference terms, the energy allocation parameter a of each devicek are mutually coupled. This problem is a complex non-convex optimization problem.

[0197] Since the main component of the objective function has a fractional form and each signal-to-interference-plus-noise ratio (SINR) term is inside a logarithmic function, the function is non-decreasing and concave. The numerator and denominator of the SINR terms are concave and convex functions respectively, while the constraint conditions form a convex set in standard form. Therefore, the main form of P1 belongs to the multi-ratio concave-convex fractional programming problem. Such problems can be solved by applying a quadratic transformation and ensuring convergence to a stationary point through an iterative method. For this purpose, a fractional programming transformation method based on quadratic transformation is developed, enabling the effective solution of this complex non-convex problem.

[0198] First, to prove the legitimacy of rewriting problem P1 by applying the SINR quadratic transformation formula, the following properties are given.

[0199] Proposition 1: The energy allocation maximization problem P1 is equivalent to problem P q , that is

[0200]

[0201] where y represents the set of auxiliary variables y k ; y k is introduced by the quadratic transformation of the uplink of each user k.

[0202] Based on Proposition 1, the variables α and y are iteratively updated. When α is fixed, the function f q (α, y) is concave with respect to y. Therefore, setting to 0 gives the globally optimal y * , that is:

[0203]

[0204] Finding the optimal α with y fixed * is a convex problem, and problem P q can converge to a stationary point. The iterative method based on quadratic transformation is as shown in the SINR quadratic transformation formula. This algorithm requires solving a convex problem in each iteration and can be initialized using simple and reasonable heuristics. In the simulation, a random starting point is used to ensure fairness in comparison, and the average value is calculated from the collected results. In addition, a small constant δ > 0 is set, and the convergence criterion where t is the iteration index.

[0205] (2) Update the receive beamforming vector w (solve problem P2)

[0206] When other variables are fixed, the sub-problem P2 for updating the receive beamforming vector w is as follows:

[0207]

[0208] Since the uplink signal-to-interference-plus-noise ratio term of user k only depends on its own receive beamforming vector w k , each w can be optimized in parallel k , and the objective function is the signal-to-interference-plus-noise ratio term in Equation (7). Here, the form of the k-th sub-problem is:

[0209]

[0210] Equation (17) is equivalent to

[0211] where and This is a generalized eigenvector problem, and its optimal solution is the eigenvector corresponding to the largest eigenvalue of the matrix (B k ) -1 A k .

[0212] (3) Update the reflection beamforming matrix Φ of the reconfigurable intelligent surface (solve problem P3)

[0213] When other variables are kept fixed, the update of the reflection beamforming matrix Φ of the reconfigurable intelligent surface can be achieved by solving the sub-problem P3. The sub-problem P3 is

[0214]

[0215] where:

[0216]

[0217] is the constraint condition, which ensures the target sensing performance.

[0218] Due to the high-order terms and coupled constraints in the denominator of the objective function, it is difficult to solve the sub-problem P3. To overcome this difficulty, a Lagrangian dual reconstruction based on problem P3 is first proposed, and then a quadratic transformation method is applied for problem equivalent transformation. Finally, the successive convex approximation algorithm is used to handle this equivalent problem to find an effective solution to this problem.

[0219] For the convenience of solution, the variable Φ is transformed into the variable First, the Lagrangian dual transformation is applied. By introducing auxiliary variables, the logarithmic function outside the fraction is removed, and a new fraction without the logarithmic function will appear after the transformation. Problem P3 is a fractional programming problem with multiple ratios, which can be equivalent to the following problem PL :

[0220]

[0221] where η = {η1, η2 ……, η k} and the objective function is At this time, the variables V and η need to be iteratively updated to solve problem P L . When the variable V is fixed, the function f(V, η) is concave with respect to η. Therefore, making can obtain the globally optimal It can be obtained that:

[0222]

[0223] When the variable η is fixed, the optimization problem with respect to the variable V is still difficult to solve. Because new fractional terms have been generated in the log function. Therefore, the fractional terms of the objective function need to be quadratically transformed. Further, the target problem P L is changed to problem P q , that is:

[0224]

[0225] where the auxiliary variable y = {y1, y2,......, y k}. The objective function is According to the iterative update rule of the quadratic transformation, when the variable V is fixed, the optimal is directly given as:

[0226]

[0227] Finally, when the variable y is fixed, the optimization problem of the variable V needs to be solved. The objective function is still complex and non-convex with respect to y. Therefore, the successive convex approximation algorithm is used to handle problem P V ,

[0228]

[0229] where

[0230] Since the non-convex optimization problem can be approximately transformed into a solvable convex form, the successive convex approximation-based method has been widely used in the literature. The following is the derivation process of using the successive convex approximation algorithm to handle problem P V .

[0231] First, relax the equality constraint, that is, |V m|≤1. Then, perform a first-order Taylor expansion on the objective function f q (V) to obtain an approximate function:

[0232]

[0233] where is the gradient at the current iteration point V (n) at the n-th iteration.

[0234] For the non-convex constraint is defined as Perform a - order Taylor expansion on the constraint function f ineq (V) to obtain an approximate constraint function, that is:

[0235]

[0236] where is the gradient at the current iteration point V (n) at the n-th iteration. Since the objective function at this time is a convex function and the constraint is a linear convex constraint. Therefore, at the n-th iteration, the following convex optimization sub-problem is formed:

[0237]

[0238] Through such a processing method, the non-convex problem can be approximated as a series of convex optimization sub-problems, enabling a relatively simple convex problem to be solved in each iteration, thus gradually approaching the solution of the original non-convex problem. The algorithm flow is shown in the Lagrangian dual transformation formula. This algorithm is a successive convex approximation solution method for the sub-problem P V , aiming to convert the originally complex non-convex optimization problem into a series of relatively easy-to-solve convex optimization sub-problems through successive approximation. The algorithm starts iterating from the initial value V (0) . In each iteration, first calculate the current objective function f q (V) and its gradient while calculating the constraint function f ineq (V) and its gradient. Then, through convexification, the non-convex sub-problem is transformed into a convex sub-problem, and the updated solution V new is obtained by solving this sub-problem (Equation (26)). Subsequently, the algorithm checks the convergence condition. If the change in the objective function value or the norm of the variable update is less than the preset thresholds ∈ and δ, the iteration stops. If the convergence condition is not reached, the algorithm performs a linear update using the step size parameter γ, advancing in the direction of V new , and continues the next iteration. Through this step-by-step approximation method, the successive convex approximation algorithm can gradually simplify the non-convex problem into multiple convex optimization problems until the optimal solution V opt of the problem is found.This method is applicable to solving non-convex problems with high complexity and can ensure convergence and improve the computational efficiency to a certain extent.

[0239] In summary, the algorithmic process for solving problem P3 is the successive convex approximation formula. The successive convex approximation formula is the overall solution for solving sub-problem P3, which combines Lagrangian dual transformation and quadratic transformation, and optimizes variables η, y, and V through multi-level iterative optimization. First, variable η is updated according to equation (20), which is related to the user signal-to-interference-plus-noise ratio term and is used to guide the optimization process. Then, variable y is updated and precisely optimized according to equation (22). In each outer iteration, the Lagrangian dual transformation formula based on successive convex approximation is used to update variable V. This method transforms non-convex problems into multiple solvable convex sub-problems and gradually approaches the optimal solution by alternately optimizing variables, thus effectively solving complex coupled non-convex constraint problems and improving the solution efficiency and convergence.

[0240] This application discloses a resource scheduling method based on a sensing, communication, and computing integrated network. Aiming at the problem of low resource scheduling efficiency caused by channel interference problems and cross-layer resource scheduling obstacles in the radar sensing and computing offloading process, an innovative technical solution is proposed, that is, combining reconfigurable intelligent surface technology and a resource scheduling algorithm based on the block coordinate descent framework. The technical protection points are as follows:

[0241] 1. Application of RIS technology in the sensing, communication, and computing integrated network: As a transformative technology capable of dynamically regulating the wireless propagation environment, RIS technology optimizes the channel conditions by intelligently reflecting signals, effectively alleviating the problems of wireless path loss and multi-link channel interference. Its flexible signal regulation ability provides a new solution for the coexistence of radar sensing and computing offloading in the sensing, communication, and computing integrated network, significantly reducing the mutual interference between the two.

[0242] 2. Problem formulation for maximizing the total system computing rate: First, a system model of the joint radar sensing and edge computing network is proposed, a channel model introducing reconfigurable intelligent surface technology is given, and the radar sensing signal-to-noise ratio, user local computing rate, and uplink communication computing rate are derived. Then, the optimization problem P0 of maximizing the task computing rate of all users within a given duration while ensuring that the sensing signal-to-noise ratio is large enough for reliable target detection is described.

[0243] 3. For the resource management problem P0 of maximizing the total system computing rate, an overall iterative algorithm based on the block coordinate descent optimization framework is designed, that is, the resource scheduling algorithm based on the block coordinate descent framework.

[0244] It should be noted that the above examples are only for understanding this application and do not constitute a limitation on the resource scheduling method of this application based on the integrated network of sensing, communication, and computing. Any simple transformation in more forms based on this technical concept is within the protection scope of this application.

[0245] This application also provides a resource scheduling device based on the integrated network of sensing, communication, and computing. Please refer to Figure 6 , the resource scheduling device based on the integrated network of sensing, communication, and computing includes:

[0246] An information acquisition module 10, configured to obtain the radar sensing signal-to-noise ratio, the uplink communication computing rate, and the user local computing rate through a channel model, where the channel model is a model constructed based on a reconfigurable intelligent surface;

[0247] A variable partitioning module 20, configured to process the uplink communication computing rate and the user local computing rate according to a resource scheduling algorithm based on block coordinate descent and the radar sensing signal-to-noise ratio to obtain multiple initial variable blocks;

[0248] A scheduling optimization module 30, configured to alternately optimize and update variables for the multiple initial variable blocks to obtain an optimal resource scheduling result.

[0249] This application provides a resource scheduling device based on the integrated network of sensing, communication, and computing. The resource scheduling device based on the integrated network of sensing, communication, and computing includes: at least one processor; and a memory communicatively connected to at least one processor.

[0250] Next, refer to Figure 7 , which shows a schematic structural diagram of a resource scheduling device based on the integrated network of sensing, communication, and computing suitable for implementing the embodiments of this application. The resource scheduling device based on the integrated network of sensing, communication, and computing may include a processing device 1001 (such as a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory 1002 or a program loaded from a storage device 1003 into a random access memory 1004. In the random access memory 1004, various programs and data required for the operation of the resource scheduling device based on the integrated network of sensing, communication, and computing are also stored. The processing device 1001, the read-only memory 1002, and the random access memory 1004 are connected to each other through a bus 1005. An input / output interface 1006 is also connected to the bus.

[0251] As described above, this is only the specific implementation manner of the present application. However, the protection scope of the present application is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present application can easily think of changes or substitutions, which should all be covered within the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

[0252] The present application provides a computer-readable storage medium having computer-readable program instructions (i.e., computer programs) stored thereon, and the computer-readable program instructions are used to execute the resource scheduling method based on the integrated network of perception, communication, and computing in the above embodiments.

[0253] The present application further provides a computer program product, including a computer program, and when the computer program is executed by a processor, the steps of the resource scheduling method based on the integrated network of perception, communication, and computing as described above are implemented.

[0254] The above description is only partial embodiments of the present application and does not limit the patent scope of the present application. Any equivalent structural transformation made by using the content of the specification and drawings of the present application under the technical concept of the present application, or any direct / indirect application in other related technical fields is included within the patent protection scope of the present application.

Claims

1. A resource scheduling method based on a perception, communication, and computing integrated network, characterized in that The method includes the following steps: Obtain the radar sensing signal-to-noise ratio, the uplink communication computing rate, and the user's local computing rate through a channel model, where the channel model is a model constructed based on a reconfigurable intelligent surface; Process the uplink communication computing rate and the user's local computing rate according to a resource scheduling algorithm based on block coordinate descent and the radar sensing signal-to-noise ratio to obtain multiple initial variable blocks; Perform alternating optimization and variable update on the multiple initial variable blocks to obtain an optimal resource scheduling result.

2. The resource scheduling method based on the integrated network of sensing, communication, and computing according to claim 1, wherein The resource scheduling algorithm includes a signal-to-interference-plus-noise ratio quadratic transformation formula, a Lagrangian dual transformation formula, and a successive convex approximation formula. The step of performing alternating optimization and variable update on the multiple initial variable blocks to obtain an optimal resource scheduling result includes: Solve the energy allocation parameter of the user terminal in the multiple initial variable blocks through the signal-to-interference-plus-noise ratio quadratic transformation formula to determine the optimized energy allocation parameter; Solve the received beamforming vector of the multiple initial variable blocks according to a convex optimization algorithm to determine the optimized received beamforming vector of the base station; Solve the reflection beamforming matrix of the multiple initial variable blocks through the Lagrangian dual transformation formula and the successive convex approximation formula to determine the optimized reflection beamforming matrix of the reconfigurable intelligent surface; Obtain the objective function minimized after one iteration based on the optimized energy allocation parameter, the optimized received beamforming vector, and the optimized reflection beamforming matrix; If the current iteration number reaches the maximum iteration number and the improvement amplitude of the objective function is less than the convergence threshold, stop the iteration and output the optimal resource scheduling result.

3. The resource scheduling method based on the integrated network of sensing, communication, and computing according to claim 2, wherein The step of solving the energy allocation parameter of the user terminal in the multiple initial variable blocks through the signal-to-interference-plus-noise ratio quadratic transformation formula to determine the optimized energy allocation parameter includes: Fix the received beamforming vector and the reflection beamforming matrix of the reconfigurable intelligent surface to obtain fixed variables; Determine an initial auxiliary variable based on the current channel state and the fixed variables; Solve the energy allocation parameter of the user terminal in the multiple initial variable blocks through the signal-to-interference-plus-noise ratio quadratic transformation formula to obtain the solved energy allocation parameter; Update the initial auxiliary variable based on the solved energy allocation parameter to obtain the updated auxiliary variable; Optimize and iterate the solved energy allocation parameter according to the updated auxiliary variable to obtain the optimized energy allocation parameter.

4. The resource scheduling method based on an integrated perception, communication, and computing network according to claim 2, wherein The step of solving the reflection beamforming matrix of the multiple initial variable blocks through the Lagrangian dual transformation formula and the successive convex approximation formula to determine the optimized reflection beamforming matrix of the reconfigurable intelligent surface includes: Fix the optimized energy allocation parameter and the optimized received beamforming vector; Introduce an auxiliary variable through the Lagrangian dual transformation formula and update the auxiliary variable to obtain the updated auxiliary variable; Optimize the reflection beamforming matrix of the multiple initial variable blocks based on the successive convex approximation algorithm to determine the optimized reflection beamforming matrix of the reconfigurable intelligent surface.

5. The resource scheduling method based on the integrated network of perception, communication and computing according to any one of claims 1 to 4, characterized in that The steps of obtaining the radar sensing signal-to-noise ratio, the uplink communication computing rate, and the user local computing rate through the channel model include: The base station transmits radar signals and receives echo signals from radar targets based on the radar signals. Adjust the reflection beamforming matrix according to the reconfigurable intelligent surface to obtain the optimized channel condition. Based on the optimized channel condition, determine the radar sensing signal-to-noise ratio according to the ratio of the intensity of the echo signal to the noise power. The user terminal sends uplink communication signals to the base station to obtain the user transmission power, the receiving beamforming vector of the base station, and the channel interference information. Integrate the user transmission power, the receiving beamforming vector, and the channel interference information to determine the uplink communication signal-to-interference-plus-noise ratio. Determine the uplink communication computing rate according to the uplink communication signal-to-interference-plus-noise ratio. Obtain the user local computing rate based on the energy allocation parameter of the user terminal.

6. The resource scheduling method based on the integrated network of sensing, communication and computing according to any one of claims 1 to 4, characterized in that, The steps of processing the uplink communication computing rate and the user local computing rate according to the resource scheduling algorithm based on block coordinate descent and the radar sensing signal-to-noise ratio to obtain multiple initial variable blocks include: Based on the resource scheduling algorithm based on block coordinate descent, under the constraint that the radar sensing signal-to-noise ratio is not lower than a preset threshold, perform weighted summation on the computing rates to obtain the optimization objective of the total computing rate. Based on the optimization objective, determine the energy allocation parameter according to the energy allocation ratio of each user for local computing and communication offloading. Determine the receiving beamforming vector according to the beamforming weights for receiving signals of each user at the base station side. Determine the reflection beamforming matrix according to the phase adjustment parameters of each reflection unit of the reconfigurable intelligent surface. Initialize the energy allocation parameter, the receiving beamforming vector, and the reflection beamforming matrix to obtain multiple initial variable blocks.

7. A resource scheduling device based on a perception, communication, and computing integrated network, characterized in that, The resource scheduling device based on the integrated sensing, communication, and computing network includes: An information acquisition module, configured to obtain the radar sensing signal-to-noise ratio, the uplink communication computing rate, and the user local computing rate through a channel model, where the channel model is a model constructed based on a reconfigurable intelligent surface. A variable partitioning module, configured to process the uplink communication computing rate and the user local computing rate according to the resource scheduling algorithm based on block coordinate descent and the radar sensing signal-to-noise ratio to obtain multiple initial variable blocks. A scheduling optimization module, configured to perform alternating optimization and variable update on the multiple initial variable blocks to obtain the optimal resource scheduling result.

8. A resource scheduling device based on a perception, communication, and computing integrated network, characterized in that, The resource scheduling device based on the integrated sensing, communication, and computing network includes: a memory, a processor, and a resource scheduling program based on the integrated sensing, communication, and computing network stored on the memory and executable on the processor. When the resource scheduling program based on the integrated sensing, communication, and computing network is executed by the processor, it implements the resource scheduling method based on the integrated sensing, communication, and computing network according to any one of claims 1 to 6.

9. A storage medium, characterized in that, A resource scheduling program based on a perception, communication, and computing integrated network is stored on the storage medium, and when the resource scheduling program based on the perception, communication, and computing integrated network is executed by a processor, it implements the resource scheduling method based on the perception, communication, and computing integrated network according to any one of claims 1 to 6.

10. A computer program product, characterized in that, The computer program product includes a computer program, and when the computer program is executed by a processor, it implements the steps of the resource scheduling method based on the perception, communication, and computing integrated network according to any one of claims 1 to 6.

Images