A distributed high-efficiency power control method in a device pass-through network

Abstract

This invention discloses a distributed high-efficiency power control method in device-to-device (D2D) communication, belonging to the field of resource allocation technology. The controlled object of this invention is a device-to-device network, comprising N D2D user pairs and K resource blocks, where each device has independent transmit and receive functions and is connected to adjacent individual devices for data communication via D2D links. Based on this distributed high-efficiency power control method in device-to-device networks, this invention establishes a device-to-device network model and constructs an optimization problem aimed at maximizing network energy efficiency. It employs the Continuous Convex Approximation (SCA) algorithm to obtain the optimal power control scheme, effectively improving the energy efficiency of the D2D network and reducing computational complexity. By combining the Alternating Direction Multiplier Method (ADMM), the computational load is distributed among the devices, reducing base station load and computational overhead, and allowing users to allocate resources more flexibly, thus improving communication performance.

Description

Technical Field

[0001] This invention belongs to the field of resource allocation technology in device-to-device (D2D) communication, and relates to a distributed high-efficiency power control method in a device-to-device network. Background Technology

[0002] In recent years, with the rapid development of technology, the widespread adoption of smartphones and other electronic devices has spurred an explosive growth in high-speed wireless multimedia services. Users' demands for service volume, variety, and quality are constantly increasing, posing greater challenges to wireless communication systems in terms of data transmission rates, service coverage, and data transmission methods. Device-to-Device (D2D) communication, as a communication technology that can effectively support these demands, is considered one of the key technologies in 3GPP LTE-Advanced. In D2D communication mode, because users can achieve short-range direct communication, the channel quality is relatively higher, and data transmission loss is less. Therefore, D2D communication can achieve higher data transmission rates, lower power consumption, and lower latency. Furthermore, base stations can further improve coverage by controlling widely distributed user terminals, achieving efficient utilization of spectrum resources, supporting more flexible network architectures and connection methods, and improving link flexibility and network reliability.

[0003] However, when using D2D communication technology, sharing the same resources between two types of users inevitably leads to interference and performance loss. Therefore, reasonable power control of D2D networks is necessary to fully improve network energy efficiency. Traditional centralized power control methods typically rely on a central control node for calculations. The central base station collects channel information from each user and allocates resources among them. This algorithm and computational performance place high demands on the network infrastructure, resulting in significant overhead for base station operation and control. Furthermore, the computational cost increases substantially with larger network sizes. Summary of the Invention

[0004] The main objective of this invention is to provide a distributed high-efficiency power control method in device-to-device (D2D) networks. Addressing the interference problem between user communications in D2D networks, this method establishes a D2D network model and constructs an optimization problem aimed at maximizing network energy efficiency. The optimal power control scheme is obtained using the Continuous Convex Approximation (SCA) algorithm, effectively improving the energy efficiency of D2D networks and reducing computational complexity. Furthermore, the alternating direction multiplier method (ADMM) distributes computation across various devices, reducing base station load and computational overhead. Users can also allocate resources more flexibly, improving communication performance.

[0005] The objective of this invention is achieved through the following technical solution:

[0006] This invention discloses a distributed high-efficiency power control method in a device pass-through network, comprising the following steps:

[0007] Step 1: Construct a device pass-through network model. The distributed high-efficiency power control object is the device pass-through network, which includes N D2D user pairs and K resource blocks. Each device has independent transmit and receive functions and is connected to adjacent individual devices through D2D links for data communication, reducing energy loss. Construct a device pass-through network power loss model, considering the power loss of the RF power amplifier and the static circuit power loss caused by signal processing and active circuit blocks. Construct a device pass-through network communication power control model and constrain the transmit power and throughput to meet channel bandwidth limitations and ensure user quality of service requirements.

[0008] Step 1.1: Construct a device pass-through network model. In this model, communication spectrum resources are provided in the form of resource blocks. The D2D user pair set and the resource block set are respectively represented as follows: and Let p n,k h is the transmit power of the nth D2D user when it occupies the kth resource block. n,n,k and h n,n′,k Let represent the channel coefficients between the receiver and transmitter on the nth link and the transmitter on the n′ link, respectively, when using the kth resource block. Then, when the user equipment transmits data using the kth resource block on the nth link, the signal-to-interference-plus-noise ratio (SIR) γ at the receiver is... n,k Represented as:

[0009]

[0010] Where, σ 2 The variance is zero-mean additive white Gaussian noise. The maximum data transmission amount for the nth user is:

[0011]

[0012] Where B is the bandwidth of a resource block. Then, calculate the total network throughput R of the device according to formula (3). tot :

[0013]

[0014] Step 1.2: Construct a power loss model for the device pass-through network, considering the power loss of the RF power amplifier and the static circuit power loss caused by signal processing and active circuit blocks. The power loss of the RF power amplifier is closely related to the amplifier's efficiency. The power loss P of the RF power amplifier is calculated using formula (4), which is the reciprocal of the drain efficiency of the power amplifier. m :

[0015]

[0016] Static circuit power loss is caused by signal processing and active circuit blocks, which require a certain current to maintain normal operation. The average circuit power consumption is represented by the static variable P. c If P represents the total circuit power consumption, then Ptotal is the power consumption of the circuit. s Represented as:

[0017] P s =NP c (5)

[0018] Therefore, the power loss model of the device pass-through network is constructed as shown in formula (6).

[0019] P tot =P m +P s (6)

[0020] Step 1.3: Construct a device pass-through network communication power control model, based on the total throughput R of the device pass-through network obtained in Steps 1.2 and 1.3. tot and power loss P tot The energy efficiency η of the device's direct-to-network connection is calculated using formula (7):

[0021]

[0022] Using the device pass-through network communication power control model shown in Equation (8) as the optimization objective and the lower limit of channel bandwidth and throughput requirements as constraints, a distributed high-efficiency power control optimization problem for device pass-through networks is constructed.

[0023]

[0024] The transmit power vector is represented by p = [p1, p2, ..., p]. N ] T It means that p n =[p n,1 ,p n,2 ,…,p n,K ]; and These represent the maximum allowed total transmit power and minimum throughput for the nth user, respectively, to ensure the communication service quality requirements of each user.

[0025] Step 2: Based on the device pass-through network communication power control model constructed in Step 1, the device pass-through network communication power control optimization problem is reconstructed into a difference form of two convex functions using the properties of logarithmic functions. The continuous convex approximation algorithm is then used to approximate the reconstructed device pass-through network communication power control optimization problem as a continuous and differentiable convex function, thus transforming the optimization problem into solving a continuous convex function. Compared with directly solving the non-convex optimization problem of device pass-through network communication power control, this significantly reduces the computational difficulty.

[0026] Step 2.1: To make the expression for the device pass-through network communication power control optimization problem more concise, let Using the properties of the logarithmic function in formula (2), the maximum data transmission amount R of the nth user can be expressed as follows: n Expressed in the following subtraction form:

[0027] R n =g n (p)-h n (p) (9)

[0028] in,

[0029] The total network throughput of the device is R tot Reconstructed as:

[0030]

[0031] Step 2.2: Introduce the variable w = 1 / P tot and θ n,k =wp n,k , and θ={θ n,k The device pass-through network communication power control model constructed in step one is transformed into a device pass-through network communication power control model with convexity as shown in formula (11):

[0032]

[0033] Due to g n (p), h n Since (p), g(p), and h(p) are all concave functions, the maximum data transmission amount R for the nth user is... n Total network throughput R of devices directly connected to the network tot Both are in the difference form of two concave functions, thus proving the transformed device pass-through network communication power control model. The objective function is a convex function, and all its constraints satisfy the convexity constraint. Therefore, the transformed device pass-through network communication power control model... It is a convexity programming problem, and it is related to the power control model of the device's direct network communication. They have the same optimal solution. The reconstructed device pass-through network communication power control optimization problem is approximated as a continuous, differentiable convex function using the continuous convex approximation SCA algorithm. This transforms the optimization problem into solving a continuous convex function to obtain a solution that satisfies the Karush-Kuhn-Tucker (KKT) conditions of the device pass-through network communication power control model. In the t-th iteration, the approximate expression for the convex function subproblem is:

[0034]

[0035] Where, θ (t) , and w (t) Let θ and θ represent the optimal solution to the t-th subproblem, respectively. n,k and w.

[0036] Step 3: In the t-th iteration, the objective function and constraints of the approximate convex subproblem are inseparable. Therefore, the Alternating Direction Multiplier Method (ADMM) cannot be directly applied to solve it. It is necessary to transform the approximate convex subproblem and then construct the Lagrangian function of the approximate convex subproblem to facilitate further solution.

[0037] Step 3.1: Introduce a new variable I n,n′,k =G n,n′,k θ n′,k , Define auxiliary variable w n , and Where w n , and w and I respectively n,n′,k and I n′,n,k In the local copy of device n, it is specified that The feasible set of the local variable for the iterative subproblem in round t is constructed as follows:

[0038]

[0039] in, and The optimal solution of the SCA subproblem in round t and w (t) Obtained through calculation. Definition. l={l n}, ξ={ξ n}, I n Include The global variables corresponding to the elements in the middle will approximate the convex function subproblem. Perform the following equivalent transformation:

[0040]

[0041] Step 3.2: Approximate convex function subproblems obtained after equivalent transformation The augmented Lagrangian function is shown in equation (15):

[0042]

[0043] Where α={α n}, β and λ={λ n Let} be the Lagrange multiplier vector, and ρ>0 is called the penalty coefficient, which determines the size of the update step. and ξ n From the properties, we can obtain that λ n Includes Lagrange multipliers and Corresponding variables and w n .

[0044] Step 4: The Alternating Direction Multiplier Method (ADMM) is employed to solve optimization problems in a distributed environment by sequentially updating local variables, global variables, and Lagrange multipliers. A dynamically updated penalty coefficient is used, adaptively adjusting its value based on the results of each iteration to further improve the convergence performance of the ADMM algorithm. Local variable updates are independent of other nodes and are computed in parallel locally; global variables are used for information exchange and coordination between nodes; and Lagrange multipliers are used to handle constraints. Through this alternating update method, the distributed high-efficiency power control algorithm can gradually approach the optimal solution and achieve efficient optimization in a distributed environment.

[0045] Step 4.1: Initialize variables Initialize the SCA subproblem iteration round t in step two;

[0046] Step 4.2: Initialize ADMM iteration round j;

[0047] Step 4.3: Update local variables Local variables are variables localized to each distributed node. Each node is responsible for solving its own sub-problems and performing local optimization based on the characteristics of the problem. Updates to local variables are performed without considering information from other nodes; each node independently solves its own optimal solution, and computations are performed locally in parallel. Local variables are updated according to formula (16):

[0048]

[0049] Step 4.4: Update the global variable {ξ} (j+1) ,l (j+1)Global variables are variables shared among all nodes, used to store information exchange and coordination between nodes. In the ADMM algorithm, global variables are typically used to transmit shared information about constraints and the problem among nodes, and updates to global variables are achieved through communication and information exchange between nodes. Using the local variables obtained in step 4.3, the global variables are updated sequentially as follows:

[0050]

[0051]

[0052] Since formula (17) is an unconstrained quadratic optimization problem, the global variable ξ is calculated according to formulas (19) and (20). (j+1) The closed-form solution is:

[0053]

[0054]

[0055] To derive l as shown in formula (18) (j+1) Closed-form solution, using global variable l (j+1) Transform it into the form of square terms as shown in formula (21):

[0056]

[0057] in,

[0058] By setting the first derivative of formula (21) to zero, we can write the global variable l. (j+1) Closed-form solution:

[0059]

[0060] Step 4.5: Update the Lagrange multipliers Lagrange multipliers are used to introduce constraints into the optimization problem and gradually adjust their values ​​through alternating updates. Each node has a corresponding Lagrange multiplier, which is used to handle the constraints associated with that node. Through the update of the Lagrange multipliers, the ADMM algorithm exchanges and coordinates information between the global and local levels to obtain the optimal result. The update rules of the Lagrange multipliers are shown in formulas (23)-(25):

[0061]

[0062]

[0063]

[0064] Step 4.6: Update the penalty coefficient according to formula (26):

[0065]

[0066] Where μ>0, τ incr >1 and τ decr >1 By adaptively adjusting the penalty coefficient based on the results of the current iteration, the selection can be made in combination with different problem structures and complexities.

[0067] Step 4.7: Convergence condition check. If the ADMM convergence condition is met (∈ is a very small positive number): If so, continue with step 4.8; otherwise, let j = j + 1 and jump to step 4.3 to continue the update iteration.

[0068] Step 4.8: Determine if |θ is satisfied (t+1) -θ (t) If |≤ε or t≥T, where ε is the convergence error and T is the maximum number of iterations, then the algorithm terminates; otherwise, continue to step 4.2.

[0069] The above distributed high-efficiency power control algorithm steps ultimately yield the converted device direct-to-network communication power control model. The local KKT point, i.e., {θ * ,w *}

[0070] Step 5: Based on the optimal solution {θ} obtained in Step 4 * ,w *}, by calculating p * =θ * / w * Ultimately, the power control result that optimizes the network energy efficiency (EE) is obtained.

[0071] It also includes step six: Based on the device pass-through network distributed high-efficiency power control optimization results obtained in step five, the optimal communication power allocation in the D2D network is realized, which can reduce the energy consumption of devices during communication, improve the communication quality between devices, enhance the signal coverage, thereby reducing resource consumption, improving network capacity and performance, and providing users with a more reliable communication experience.

[0072] Beneficial effects:

[0073] 1. The present invention discloses a distributed high-efficiency power control method in a device-to-device (D2D) network. It constructs a communication power control model for a D2D network, considers the maximum allowable total transmit power and minimum throughput constraints, and improves network energy efficiency while meeting user service quality requirements by selecting the optimal power control scheme.

[0074] 2. The present invention discloses a distributed high-efficiency power control method in a device pass-through network. The method uses the continuous convex approximation (SCA) algorithm to iteratively approximate the communication power control problem of the device pass-through network by using subproblems with standard convex optimization forms, thereby reducing the difficulty and complexity of the calculation and making the calculation of high-efficiency power control schemes in device pass-through networks more efficient.

[0075] 3. This invention discloses a distributed high-efficiency power control method in a device pass-through network. By using the ADMM algorithm, tasks or data are distributed across multiple nodes for execution. Each node only needs to handle its local sub-problem, and computation can be performed in parallel, effectively improving computational efficiency and reducing the computational burden on the central control unit. Even if one node fails or malfunctions, the entire system can continue to operate. In contrast, centralized algorithms impose an excessive computational burden on the central node and have poor handling capabilities for sudden errors; a single node failure can paralyze the entire system. The distributed high-efficiency power control method in a device pass-through network disclosed in this invention has higher reliability, making it more robust to hardware failures, network interruptions, or other abnormal situations.

[0076] 4. In traditional ADMM algorithms, the penalty coefficient is usually fixed and does not change with the iteration process. This may lead to slow convergence or failure to converge in some cases. This is because when the constraints of the primal problem are strong, a fixed penalty coefficient may cause inaccurate solutions to the dual problem, thus affecting the convergence of the entire algorithm. In contrast, the distributed high-efficiency power control method in device pass-through networks disclosed in this invention adopts dynamically updated penalty coefficients, achieving convergence with fewer iterations. Dynamically selecting the value adaptively adjusts the size of the penalty coefficient based on the results of the current iteration, better balancing the relationship between the primal and dual problems. It can adapt to different problem structures and complexities, thereby enabling the distributed algorithm to converge faster and improving the convergence performance of the model compared to a fixed value. Attached Figure Description

[0077] To more clearly illustrate the technical solutions in the specific embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below.

[0078] Figure 1 This is a flowchart of a distributed high-efficiency power control method in a device pass-through network according to the present invention.

[0079] Figure 2 This is a schematic diagram of a device-to-network communication system model;

[0080] Figure 3This is a comparison of the simulation convergence curves of the ADMM algorithm with dynamically updated penalty coefficients and fixed-value penalty coefficients, in conjunction with Embodiment 1 of the present invention, "A Distributed High-Efficiency Power Control Method in a Device Straight-Through Network".

[0081] Figure 4 This is a schematic diagram comparing the results of the distributed method and the centralized power control method of the present invention in Example 1. Detailed Implementation

[0082] To make the objectives and technical solutions of the present invention clearer, the embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

[0083] Example 1:

[0084] This invention provides a distributed high-efficiency power control method in a device pass-through network. Considering, for example... Figure 1 The illustrated D2D communication scenario shows each device connecting to adjacent individual devices via D2D links for data communication. Assume D2D pairs are uniformly distributed throughout the network, with a D2D user pair distance of 10m, a base station coverage radius of 25m, a path loss exponent of 3.8, a carrier frequency of 2GHz, a reference distance of 100m, a subcarrier bandwidth of 180kHz, a noise power spectral density of -174dBm / Hz, a shadowing variance of 8dB, and shadowing fading following a log-normal distribution. Fixed value ρ and dynamic initial value ρ (0) All are set to 1.1. The drain efficiency and average circuit power consumption P of the power amplifier are... c The values ​​are 0.38 and 10W respectively. To address the problems of excessive load on the central unit and high cost caused by centralized algorithms, this invention proposes a distributed algorithm in the context of D2D communication that aims to maximize network EE, based on a combination of ADMM and SCA. The specific method includes the following steps:

[0085] Step 1: Construct a device pass-through network model. The distributed high-efficiency power control object is the device pass-through network, which includes N D2D user pairs and K resource blocks. Each device has independent transmit and receive functions and is connected to adjacent individual devices through D2D links for data communication, reducing energy loss. Construct a device pass-through network power loss model, considering the power loss of the RF power amplifier and the static circuit power loss caused by signal processing and active circuit blocks. Construct a device pass-through network communication power control model and constrain the transmit power and throughput to meet channel bandwidth limitations and ensure user quality of service requirements.

[0086] Step 1.1: Construct a device pass-through network model. In this model, communication spectrum resources are provided in the form of resource blocks. The D2D user pair set and the resource block set are respectively represented as follows: and Let p n,k h is the transmit power of the nth D2D user when it occupies the kth resource block. n,n,k and h n,n′,k Let represent the channel coefficients between the receiver and transmitter on the nth link and the transmitter on the n′ link, respectively, when using the kth resource block. Then, when the user equipment transmits data using the kth resource block on the nth link, the signal-to-interference-plus-noise ratio (SIR) γ at the receiver is... n,k Represented as:

[0087]

[0088] Where, σ 2 The variance is zero-mean additive white Gaussian noise. The maximum data transmission amount for the nth user is:

[0089]

[0090] Where B is the bandwidth of a resource block. Then, calculate the total network throughput R of the device according to formula (3). tot :

[0091]

[0092] Step 1.2: Construct a power loss model for the device pass-through network, considering the power loss of the RF power amplifier and the static circuit power loss caused by signal processing and active circuit blocks. The power loss of the RF power amplifier is closely related to the amplifier's efficiency. The power loss P of the RF power amplifier is calculated using formula (4), which is the reciprocal of the drain efficiency of the power amplifier. m :

[0093]

[0094] Static circuit power loss is caused by signal processing and active circuit blocks, which require a certain current to maintain normal operation. The average circuit power consumption is represented by the static variable P. c If P represents the total circuit power consumption, then Ptotal is the power consumption of the circuit. s Represented as:

[0095] P s =NP c (5)

[0096] Therefore, the power loss model of the device pass-through network is constructed as shown in formula (6).

[0097] P tot =P m +P s (6)

[0098] Step 1.3: Construct a device pass-through network communication power control model, based on the total throughput R of the device pass-through network obtained in Steps 1.2 and 1.3. tot and power loss P tot The energy efficiency η of the device's direct-to-network connection is calculated using formula (7):

[0099]

[0100] Using the device pass-through network communication power control model shown in Equation (8) as the optimization objective and the lower limit of channel bandwidth and throughput requirements as constraints, a distributed high-efficiency power control optimization problem for device pass-through networks is constructed.

[0101]

[0102] The transmit power vector is represented by p = [p1, p2, ..., p]. N ] T It means that p n =[p n,1 ,p n,2 ,…,p n,K ]; and These represent the maximum allowed total transmit power and minimum throughput for the nth user, respectively, to ensure the communication service quality requirements of each user.

[0103] Step 2: Based on the device pass-through network communication power control model constructed in Step 1, the device pass-through network communication power control optimization problem is reconstructed into a difference form of two convex functions by utilizing the properties of logarithmic functions. The continuous convex approximation algorithm is then used to approximate the reconstructed device pass-through network communication power control optimization problem as a continuous and differentiable convex function, thereby transforming the optimization problem into solving a continuous convex function. Compared with directly solving the non-convex optimization problem of device pass-through network communication power control, this significantly reduces the computational difficulty.

[0104] Step 2.1: To make the expression for the device pass-through network communication power control optimization problem more concise, let Using the properties of the logarithmic function in formula (2), the maximum data transmission amount R of the nth user can be expressed as follows: n Expressed in the following subtraction form:

[0105] R n =g n (p)-h n (p) (9)

[0106] in,

[0107] The total network throughput of the device is R tot Reconstructed as:

[0108]

[0109] Step 2.2: Introduce the variable w = 1 / P tot and θ n,k =wp n,k , and θ={θ n,k The device pass-through network communication power control model constructed in step one is transformed into a device pass-through network communication power control model with convexity as shown in formula (11):

[0110]

[0111] Due to g n (p), h n Since (p), g(p), and h(p) are all concave functions, the maximum data transmission amount R for the nth user is... n Total network throughput R of devices directly connected to the network tot Both are in the difference form of two concave functions, thus proving the transformed device pass-through network communication power control model. The objective function is a convex function, and all its constraints satisfy the convexity constraint. Therefore, the transformed device pass-through network communication power control model... It is a convexity programming problem, and it is related to the power control model of the device's direct network communication. They have the same optimal solution. The reconstructed device pass-through network communication power control optimization problem is approximated as a continuous, differentiable convex function using the continuous convex approximation SCA algorithm. This transforms the optimization problem into solving a continuous convex function to obtain a solution that satisfies the Karush-Kuhn-Tucker (KKT) conditions of the device pass-through network communication power control model. In the t-th iteration, the approximate expression for the convex function subproblem is:

[0112]

[0113] Where, θ (t) , and w (t) Let θ and θ represent the optimal solution to the t-th subproblem, respectively. n,k and w.

[0114] Step 3: The convex function subproblem approximated in the t-th iteration The objective function and constraints are inseparable, therefore the Alternating Direction Multiplier Method (ADMM) cannot be directly applied to solve it. Appropriate mathematical transformations are needed to approximate the convex subproblem before constructing the approximate convex subproblem. The Lagrange function is used to facilitate further solutions.

[0115] Step 3.1: Introduce a new variable I n,n′,k =G n,n′,k θ n′,k , Define auxiliary variable w n , and Where w n , and w and I respectively n,n′,k and I n′,n,k In the local copy of device n, it is specified that The feasible set of local variables in the t-th round iterative subproblem is constructed as follows:

[0116]

[0117] in, and The optimal solution to the SCA subproblem in round t can be obtained from... and w (t) Obtained through calculation. Definition. l={l n}, ξ={ξ n}, I n Include The global variables corresponding to the elements in the middle will approximate the convex function subproblem. Perform the following equivalent transformation:

[0118]

[0119] Step 3.2: Approximate convex function subproblems obtained after equivalent transformation The augmented Lagrangian function is shown in equation (15):

[0120]

[0121] Where α={α n}, β and λ={λ n Let} be the Lagrange multiplier vector, and ρ>0 is called the penalty coefficient, which determines the size of the update step. and ξ n From the properties, we can obtain that λ n Includes Lagrange multipliers and Corresponding variables and w n .

[0122] Step 4: The Alternating Direction Multiplier Method (ADMM) is employed to solve optimization problems in a distributed environment by sequentially updating local variables, global variables, and Lagrange multipliers. A dynamically updated penalty coefficient is used, adaptively adjusting its value based on the results of each iteration to further improve the convergence performance of the ADMM algorithm. Local variable updates are independent of other nodes and are computed in parallel locally; global variables are used for information exchange and coordination between nodes; and Lagrange multipliers are used to handle constraints. Through this alternating update method, the distributed high-efficiency power control algorithm can gradually approach the optimal solution and achieve efficient optimization in a distributed environment.

[0123] Step 4.1: Initialize variables Initialize the SCA subproblem iteration round t in step two;

[0124] Step 4.2: Initialize ADMM iteration round j;

[0125] Step 4.3: Update local variables Local variables are variables localized to each distributed node. Each node is responsible for solving its own sub-problems and performing local optimization based on the characteristics of the problem. Updates to local variables are performed without considering information from other nodes; each node independently solves its own optimal solution, and computations are performed locally in parallel. Local variables are updated according to formula (16):

[0126]

[0127] Step 4.4: Update the global variable {ξ} (j+1) ,l (j+1) Global variables are variables shared among all nodes, used to store information exchange and coordination between nodes. In the ADMM algorithm, global variables are typically used to transmit shared information about constraints and the problem among nodes, and updates to global variables are achieved through communication and information exchange between nodes. Using the local variables obtained in step 4.3, the global variables are updated sequentially as follows:

[0128]

[0129]

[0130] Since formula (17) is an unconstrained quadratic optimization problem, the global variable ξ is calculated according to formulas (19) and (20). (j+1) The closed-form solution is:

[0131]

[0132]

[0133] To derive l as shown in formula (18) (j+1) Closed-form solution, using global variable l (j+1) Transform it into the form of square terms as shown in formula (21):

[0134]

[0135] in,

[0136] By setting the first derivative of formula (21) to zero, we can write the global variable l. (j+1) Closed-form solution:

[0137]

[0138] Step 4.5: Update the Lagrange multipliers Lagrange multipliers are used to introduce constraints into the optimization problem and gradually adjust their values ​​through alternating updates. Each node has a corresponding Lagrange multiplier, which is used to handle the constraints associated with that node. Through the update of the Lagrange multipliers, the ADMM algorithm exchanges and coordinates information between the global and local levels to obtain the optimal result. The update rules of the Lagrange multipliers are shown in formulas (23)-(25):

[0139]

[0140]

[0141]

[0142] Step 4.6: Update the penalty coefficient according to formula (26):

[0143]

[0144] Where μ>0, τ incr >1 and τ decr >1 By adaptively adjusting the penalty coefficient based on the results of the current iteration, the selection can be made in combination with different problem structures and complexities.

[0145] Step 4.7: Convergence condition check. If the ADMM convergence condition is met (∈ is a very small positive number): If so, continue with step 4.8; otherwise, let j = j + 1 and jump to step 4.3 to continue the update iteration.

[0146] Step 4.8: Determine if |θ is satisfied(t+1) -θ (t) If |≤ε or t≥T, where ε is the convergence error and T is the maximum number of iterations, then the algorithm terminates; otherwise, continue to step 4.2.

[0147] The above distributed high-efficiency power control algorithm steps ultimately yield the converted device direct-to-network communication power control model. The local KKT point, i.e., {θ * ,w *}

[0148] Step 5: Based on the optimal solution {θ} obtained in Step 4 * ,w *}, by calculating p * =θ * / w * Ultimately, the power control result that optimizes the network energy efficiency (EE) is obtained.

[0149] In this example, Figure 3 To demonstrate the convergence curve of the ADMM algorithm for solving the SCA convex optimization subproblem, compared to the results obtained by traditional centralized convex optimization algorithms, the ADMM algorithm can obtain the same optimal solution to the SCA convex optimization subproblem in a distributed manner. Furthermore, in the traditional ADMM algorithm, a fixed penalty coefficient may lead to inaccurate solutions to the dual problem, thus affecting the convergence of the entire algorithm. In contrast, the method of dynamically updating the penalty coefficient achieves convergence with fewer iterations, improving the algorithm's convergence performance.

[0150] A comparison of the results of the distributed method and the centralized power control method of the present invention is as follows: Figure 4 As shown, the proposed distributed algorithm, after final convergence, obtains an optimal solution similar to that of the centralized algorithm. In practical applications, the distributed algorithm distributes the computational load across various devices, with each node only needing to handle its local subproblem. Furthermore, the computation can be performed in parallel, effectively improving computational efficiency, reducing the computational load on the central node, and achieving network energy efficiency similar to the centralized approach.

[0151] The above detailed description further illustrates the purpose, technical solution, and beneficial effects of the invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A distributed high-efficiency power control method in a device direct-connection network, characterized in that: Includes the following steps, Step 1: Construct a device pass-through network model. The distributed high-efficiency power control object is the device pass-through network, which includes N D2D user pairs and K resource blocks. Each device has independent transmit and receive functions and is connected to adjacent individual devices for data communication via D2D links. Construct a device pass-through network power loss model, considering the power loss of the RF power amplifier and the static circuit power loss caused by signal processing and active circuit blocks. Based on the device pass-through network power loss model and the total throughput of the device pass-through network, construct a device pass-through network communication power control model, and constrain the transmit power and throughput to meet channel bandwidth limitations and ensure user service quality requirements. Step 2: Based on the device pass-through network communication power control model constructed in Step 1, the device pass-through network communication power control optimization problem is reconstructed into the difference form of two convex functions by utilizing the properties of logarithmic functions. The reconstructed device pass-through network communication power control optimization problem is then approximated as a continuous and differentiable convex function using a continuous convex approximation algorithm, thereby transforming the optimization problem into a problem of solving a continuous convex function. Step 3: In the t-th iteration, the objective function and constraints of the convex subproblem obtained by approximation are inseparable. Therefore, the Alternating Direction Multiplier Method (ADMM) cannot be directly applied to solve it. It is necessary to transform the approximated convex subproblem and then construct the Lagrangian function of the approximated convex subproblem to facilitate further solution. Step 4: The Alternating Direction Multiplier Method (ADMM) is employed to solve optimization problems in a distributed environment by sequentially updating local variables, global variables, and Lagrange multipliers. A dynamically updated penalty coefficient is used, adaptively adjusting its value based on the results of each iteration to further improve the convergence performance of the ADMM algorithm. Local variable updates are independent of other nodes and are computed in parallel locally. Global variables are used for information exchange and coordination between nodes. Lagrange multipliers are used to handle constraints. Through alternating updates, the distributed high-efficiency power control algorithm can gradually approach the optimal solution and achieve efficient optimization in a distributed environment. Step 5: Based on the optimal solution {θ} obtained in Step 4 * ,w * }, by calculating p * =θ * / w * Ultimately, the power control result that optimizes the network energy efficiency (EE) is obtained.

2. The distributed high-efficiency power control method in a device pass-through network as described in claim 1, characterized in that: It also includes step six, which, based on the device pass-through network distributed high-efficiency power control optimization results obtained in step five, achieves optimal communication power allocation in the D2D network. This can reduce the energy consumption of devices during communication, improve the communication quality between devices, enhance signal coverage, thereby reducing resource consumption and improving network capacity and performance.

3. A distributed high-efficiency power control method in a device pass-through network as described in claim 1 or 2, characterized in that: The implementation method for step one is as follows: Step 1.1: Construct a device pass-through network model. In this model, communication spectrum resources are provided in the form of resource blocks. The D2D user pair set and the resource block set are respectively represented as follows: and Let p n,k h is the transmit power of the nth D2D user when it occupies the kth resource block. n,n,k and h n,n′,k Let represent the channel coefficients between the receiver and transmitter on the nth link and the transmitter on the n′ link, respectively, when using the kth resource block. Then, when the user equipment transmits data using the kth resource block on the nth link, the signal-to-interference-plus-noise ratio (SIR) γ at the receiver is... n,k Represented as: Where, σ 2 The variance of the zero-mean additive white Gaussian noise is: The maximum data transmission amount for the nth user is: Where B is the bandwidth of a resource block; then, the total throughput R of the device's direct network is calculated according to formula (3). tot : Step 1.2: Construct a power loss model for the device pass-through network, considering the power loss of the RF power amplifier and the static circuit power loss caused by signal processing and active circuit blocks; the power loss of the RF power amplifier is closely related to the amplifier's efficiency, given... The power loss P of the RF power amplifier is calculated using formula (4), which is the reciprocal of the drain efficiency of the power amplifier. m : Static circuit power loss is caused by signal processing and active circuit blocks, which require a certain current to maintain normal operation; the average circuit power consumption is represented by the static variable P. c If P represents the total circuit power consumption, then Ptotal is the power consumption of the circuit. s Represented as: P s NP c (5) Therefore, the power loss model of the device pass-through network is constructed as shown in formula (6); P tot =P m +P s (6) Step 1.3: Based on Step 1.2 and Step 1.3, calculate the total network throughput R of the device through-network. tot and power loss P tot The energy efficiency η of the device's direct-to-network connection is calculated using formula (7): Using the device pass-through network communication power control model as shown in formula (8) as the optimization objective and the lower limit of channel bandwidth and throughput requirements as constraints, a distributed high-efficiency power control optimization problem for the device pass-through network is constructed. The transmit power vector is represented by p = [p1, p2, ..., p]. N ] T It means that p n =[p n,1 ,p n,2 ,...,p n,K ]; and These represent the maximum allowed total transmit power and minimum throughput for the nth user, respectively, to ensure the communication service quality requirements of each user.

4. The distributed high-efficiency power control method in a device pass-through network as described in claim 3, characterized in that: The second step is implemented as follows: Step 2.1: To make the expression for the device pass-through network communication power control optimization problem more concise, let Using the properties of the logarithmic function in formula (2), the maximum data transmission amount R of the nth user can be expressed as follows: n Expressed in the following subtraction form: R n =g n (p)-h n (p) (9) in, The total network throughput R of the device is directly connected to the network. tot Reconstructed as: Step 2.2: Introduce the variable w = 1 / P tot and θ n,k =wp n,k , and θ={θ n,k The device pass-through network communication power control model constructed in step one is transformed into a device pass-through network communication power control model with convexity as shown in formula (11): Due to g n (p), h n (p) and Both are concave functions, therefore the maximum data transmission amount R for the nth user is... n Total network throughput R of devices directly connected to the network tot Both are in the difference form of two concave functions, thus proving the transformed device pass-through network communication power control model. The objective function is a convex function, and all its constraints satisfy the convexity constraint. Therefore, the transformed device pass-through network communication power control model... It is a convexity programming problem, and it is related to the power control model of the device's direct network communication. Having the same optimal solution; the reconstructed device pass-through network communication power control optimization problem is approximated as a continuous, differentiable convex function using the continuous convex approximation SCA algorithm, thus transforming the optimization problem into solving a continuous convex function to obtain a solution that satisfies the Karush-Kuhn-Tucker (KKT) conditions of the device pass-through network communication power control model; in the t-th iteration, the approximate expression of the convex function subproblem is: Where, θ (t) , and w (t) Let θ and θ represent the optimal solution to the t-th subproblem, respectively. n,k and w.

5. The distributed high-efficiency power control method in a device pass-through network as described in claim 4, characterized in that: The method for implementing step three is as follows: Step 3.1: Introduce a new variable I n,n′,k =G n,n′,k θ n′,k , Define auxiliary variable w n , and Where w n , and w and I respectively n,n′,k and I n′,n,k In the local copy of device n, it is specified that The feasible set of the local variable for the iterative subproblem in round t is constructed as follows: in, and The optimal solution of the SCA subproblem in round t and w (t) Calculated; defined l={l n }, ξ={ξ n }, I n Include The global variables corresponding to the elements in the middle will approximate the convex function subproblem. Perform the following equivalent transformation: Step 3.2: Approximate convex function subproblems obtained after equivalent transformation The augmented Lagrangian function is shown in equation (15): Where α={α n }, β and λ={λ n } represents the Lagrange multiplier vector, ρ > 0 is called the penalty coefficient, which determines the size of the update step; by and ξ n From the properties, we can obtain that λ n Includes Lagrange multipliers and Corresponding variables and w n .

6. The distributed high-efficiency power control method in a device pass-through network as described in claim 5, characterized in that: Step four is implemented as follows: Step 4.1: Initialize variables Initialize the SCA subproblem iteration round t in step two; Step 4.2: Initialize ADMM iteration round j; Step 4.3: Update local variables Local variables are variables localized to each distributed node; each node is responsible for solving its own sub-problems and performing local optimization based on the characteristics of the problem; local variables are updated without considering information from other nodes, each node independently solves its own optimal solution, and the computation is performed in parallel locally; local variables are updated according to formula (16): Step 4.4: Update the global variable {ξ} (j+1) ,l (j+1) Global variables are variables shared among all nodes, used to store information exchange and coordination between nodes. In the ADMM algorithm, global variables are typically used to transmit shared information about constraints and problems between nodes. Updates to global variables are achieved through communication and information exchange between nodes. Using the local variables obtained in step 4.3, the global variables are updated sequentially as follows: Since formula (17) is an unconstrained quadratic optimization problem, the global variable ξ is calculated according to formulas (19) and (20). (j+1) The closed-form solution is: To derive l as shown in formula (18) (j+1) Closed-form solution, using global variable l (j+1) Transform it into the form of square terms as shown in formula (21): in, By setting the first derivative of formula (21) to zero, we can write the global variable l. (j+1) Closed-form solution: Step 4.5: Update the Lagrange multipliers Lagrange multipliers are used to introduce constraints into the optimization problem and gradually adjust their values ​​through alternating updates; each node has a corresponding Lagrange multiplier, which is used to handle the constraints associated with that node; through the update of the Lagrange multipliers, the ADMM algorithm exchanges and coordinates information between the global and local systems to obtain the optimal result; the update rules of the Lagrange multipliers are shown in formulas (23)-(25): Step 4.6: Update the penalty coefficient according to formula (26): Where μ>0, τ incr >1 and τ decr >1 The penalty coefficient can be adaptively adjusted based on the result of the current iteration, and can be selected according to different problem structures and complexities; Step 4.7: Convergence condition judgment. If the ADMM convergence condition is met: If so, continue with step 4.8; otherwise, let j = j + 1 and jump to step 4.3 to continue the update iteration. Step 4.8: Determine if θ is satisfied (t+1) -θ (t) If |≤ε or t≥T, where ε is the convergence error and T is the maximum number of iterations, then the algorithm terminates; otherwise, continue to step 4.

2. The above distributed high-efficiency power control algorithm steps ultimately yield the converted device direct-to-network communication power control model. The local KKT point, i.e., {θ * ,w * }

Images