Energy efficiency optimization method for distributed intelligent reflector assisted cellular network removal

Abstract

The invention provides an energy efficiency optimization method for a distributed intelligent reflector-assisted cellular network, and the method specifically comprises the steps: firstly constructing a system model of the intelligent reflector-assisted cellular network and a controllable energy consumption model of the on-off state of an intelligent reflector; constructing a distributed intelligent reflecting surface to assist in eliminating the system energy efficiency maximization problem of the cellular network, and jointly designing transmitting beam forming, an intelligent reflecting surface phase and a switch state; auxiliary variables are introduced by using a Dinkelbach method, Lagrange dual transformation and a quadratic programming method to decouple problems, and optimization methods of auxiliary vectors, transmitted beam forming, phase coefficients of intelligent reflecting surfaces and switch states are given; and iteratively and alternately optimizing the auxiliary variable, the transmitting beam forming, the intelligent reflecting surface phase and the switch state until convergence, and finally obtaining an intelligent reflecting surface control state and a base station beam vector, and obtaining a configuration scheme of energy efficiency optimization.

Description

Technical Field

[0001] The present invention relates to the field of wireless communication technology, and in particular to an energy efficiency optimization method of a distributed intelligent reflecting surface (IRS) in a de-cellular network in future sixth generation mobile communications (6th Generation, 6G). Background Art

[0002] Decellularized network technology is developed from massive multiple-input multiple-output (MIMO) technology and distributed MIMO in the fifth generation of wireless communications. Under this architecture, a large number of distributed access points (APs) form a distributed massive MIMO array to provide services to multiple users in the same time-frequency resources. Since the concept of cell boundaries no longer exists and signal processing effectively handles interference between users, decellularized networks can fully utilize the advantages of cooperative transmission to significantly increase system capacity. However, in order to cope with the thousand-fold increase in network capacity in the future, this method of achieving a significant increase in capacity by deploying a large number of AP base stations is no longer feasible, because this method will bring serious deployment costs and energy consumption problems, which will pose a major problem in the future pursuit of green and energy-saving 6G communications.

[0003] The emerging smart reflector technology stands out recently with its passive, energy-saving and high-gain characteristics, which can solve the high energy consumption problem of the above-mentioned decellularized network system to a certain extent. The smart reflector is composed of passive programmable reflective units, which can realize intelligent control of signal beams through controllers, thereby achieving signal enhancement, interference elimination and other effects. Smart reflectors do not require high-energy-consuming hardware such as radio frequency links, and their energy consumption is low, while at the same time achieving system capacity improvement. Therefore, introducing smart reflectors into decellularized networks can ensure system performance at a lower cost and in a more energy-efficient way compared to traditional decellularized systems.

[0004] However, considering that the large-scale deployment of smart reflective surfaces will still bring considerable energy consumption and the capacity improvement is not obvious, in order to further improve the energy efficiency of the system and support the concept of green communication, the on or off state of the smart reflective surface can be further reasonably designed. Most of the existing research on energy efficiency only considers the scenario where the smart reflective surface remains on all the time. Therefore, how to reasonably design the AP beamforming, the phase of the smart reflective surface, and the on / off state of the smart reflective surface to improve the energy efficiency of the system is an issue worthy of attention. Summary of the invention

[0005] The present invention proposes a distributed intelligent reflector-assisted de-cellularized network energy efficiency optimization method. Specifically, firstly, a system model of the de-cellularized network assisted by the intelligent reflector and an energy consumption model with controllable switch states of the intelligent reflector are constructed; the system energy efficiency maximization problem of the distributed intelligent reflector-assisted de-cellularized network is constructed, and the transmit beamforming, the phase of the intelligent reflector and the switch state are jointly designed to meet the user's minimum rate requirement, the base station transmit power limit and the phase constraint of the intelligent reflector; the Dinkelbach method, the Lagrange dual transformation and the quadratic programming method are used to introduce auxiliary variables to decouple the problem, and the auxiliary variables and the base station beamforming are optimized under the given phase and switch state of the intelligent reflector; under the given transmit beamforming and auxiliary vector, the phase coefficient and the switch state of the intelligent reflector are optimized; the auxiliary variables, the transmit beamforming, the phase of the intelligent reflector and the switch state are iteratively and alternately optimized until convergence, and finally the intelligent reflector control state and the base station beam vector are obtained, and the energy efficiency optimization configuration scheme is obtained.

[0006] The energy efficiency optimization method in a distributed intelligent reflector-assisted decellularized network proposed in the present invention comprises the following steps:

[0007] (1) Construct a system model of a de-cellular network assisted by an intelligent reflective surface and an energy consumption model with controllable switching states of the intelligent reflective surface.

[0008] Specifically, a de-cellularized network scenario is constructed, which includes B APs, L smart reflective surfaces, and K single-antenna users, in which all APs and all smart reflective surfaces are uniformly controlled and signal processed by the central processing unit (CPU). In this de-cellularized scenario assisted by smart reflective surfaces, all APs and smart reflective surfaces work together to serve all users, and the switching state of the smart reflective surfaces can be optimized and controlled.

[0009] In the energy efficiency optimization method proposed in the present invention, in order to further save energy consumption and avoid ineffective deployment of smart reflective surfaces, it is proposed to introduce a binary variable x for the on / off state of the smart reflective surface. l ∈{0,1} representation, added to the cascade channel part of the channel model, that is, when x l = 1, the intelligent reflective surface is in the on state, and the phase is optimized and adjusted; when x l When =0, the intelligent reflective surface is turned off and no energy is consumed.

[0010] Assume that the CPU knows all the channel status information through AP feedback, let represents the channel matrix from the bth AP to the kth user, represents the equivalent channel from the lth smart reflection surface to the kth user, represents the equivalent channel from the bth AP to the lth smart reflector, where M t Indicates the number of transmit antennas of the AP, and N indicates the number of reflective units of each smart reflective surface. is the diagonal matrix of the reflection phase of the lth smart reflection surface, Multi-hop reflection is not considered here.

[0011] Assume m represents the symbol sent by the target to user m, represents the beamforming matrix designed by APb for user m, represents the Gaussian noise at user k, then in the multi-intelligent reflector assisted scenario, the received signal model at user k can be written as

[0012]

[0013] By defining the vector And the merge vector Then the equivalent combined channel from user k to all APs is expressed as

[0014]

[0015] in, Φ=diag(Φ1,...,Φ l ,...,Φ L ).

[0016] In this system, the total power consumption includes the transmit power of APs, the hardware static power consumption of each AP, and the The hardware static power consumption P of each reflective unit of each smart reflective surface RIS , and the circuit power consumption of each user k Expressed as

[0017]

[0018] in, represents the efficiency of the transmit power amplifier,

[0019] (2) Construct a distributed intelligent reflector-assisted decellularized network to maximize the system energy efficiency, and jointly design the transmit beamforming, intelligent reflector phase, and switch state to meet the user's minimum rate requirements, base station transmit power limitations, and the phase constraints of the intelligent reflector.

[0020] Specifically, according to the channel model constructed in step (1), the signal to interference noise ratio of user k can be obtained, which is expressed as

[0021]

[0022] in, is the beamforming combination vector for each user k.

[0023] So, define and Respectively represent the set of intelligent reflective surface and each reflective unit of the reflective surface, and construct the total energy constraint condition P b,max The optimization problem of maximizing energy efficiency under the condition of satisfying the user's minimum rate requirement R min and the phase constraint of the smart reflector, the problem of jointly optimizing the transmit beamforming, the phase of the smart reflector, and the switch state can be described as:

[0024]

[0025] log2(1+γ k )≥R min ,

[0026]

[0027] (3) Auxiliary variables are introduced to solve the problem by using the Dinkelbach method, Lagrange dual transformation and quadratic programming method. Given the phase and switch state of the smart reflector, the auxiliary variables and base station beamforming are optimized.

[0028] Specifically, by using the Dinkelbach method, Lagrange dual transformation and quadratic programming method, the auxiliary variable y is introduced. The objective function can be further reformulated as

[0029]

[0030] in g(W b ) is a function representing the total power consumption The optimal values ​​of the auxiliary variables are updated as ε k ＊ =γ k ,

[0031] Therefore, after the above transformation, the original optimization problem is equivalently transformed into

[0032]

[0033]

[0034] log2(1+γ k )≥R min ,

[0035]

[0036]

[0037] When the introduced variables y, ε, u, the reflection coefficient of the smart reflector, and the switch vector x are fixed, the optimization problem of AP transmit beamforming can be restated as

[0038]

[0039] The objective expression of the above problem is a concave function with respect to W, while the second constraint is non-convex. Using the successive convex approximation (SCA) method, the second constraint, the minimum achievable rate, can be further expressed as

[0040]

[0041] In the above formula, (t) indicates that the vector is the value of the last iteration. Therefore, the problem becomes a convex quadratic constrained quadratic programming (QCQP), which can be solved by convex optimization tools (such as CVX solver).

[0042] (4) Given the transmit beamforming and auxiliary vector, optimize the phase coefficient and switching state of the smart reflector.

[0043] Specifically, given the variables y, ε, u, x and the transmit beamforming vector And ignore the constant, about the reflection coefficient of the smart reflector The sub-problem can be reformulated as

[0044]

[0045] in,

[0046]

[0047]

[0048] The above problem also involves the minimum achievable rate constraint like the previous problem. The SCA method is also used to further transform the constraints into:

[0049]

[0050] in, C k,m =|h k w m | 2 ,φ (t)represents the optimal value of the previous iteration, so the problem also becomes a convex QCQP, which can be effectively solved by the existing convex optimization solver CVX.

[0051] Then update the switch vector. Given the phase matrix Φ and the beamforming matrix W, the original problem is a nonlinear integer optimization problem about the smart reflector switch vector x. Since nonlinear integer optimization problems are generally NP-hard, it is difficult to obtain a global optimal solution with polynomial complexity. To solve this problem, we plan to use a greedy algorithm to solve the smart reflector switch optimization problem.

[0052] The basic idea is that if the energy efficiency target value can be improved and the new solution is feasible, we try to close one smart reflective surface each time. Step 1: Set the set is the set of active intelligent reflective surfaces s serving users. In step 2, the initial target value of energy efficiency E0 is calculated. Then when the set When not empty, for each If the smart reflective surface l is turned off and the resulting switch vector is feasible, the corresponding energy efficiency value is calculated and recorded as E l Otherwise, if the new switch vector is not feasible, the target value E l Recorded as zero.

[0053] Next, we compare the energy efficiency values ​​of these new switching schemes with the initial solution and obtain If k≠0, the index value k indicates that turning off a smart reflective surface can improve energy efficiency. Turning off smart reflective surface k can achieve the highest energy efficiency than turning off any other smart reflective surface l. In this case, turning off smart reflective surface k is energy-saving. Then we select Subtract index k from And update the initial energy efficiency value E0=E for the next iteration k If k = 0, it means that turning off any smart reflective surface will lead to energy inefficiency. In this case, turning off any smart reflective surface will not improve energy efficiency, which means we need to terminate the loop and output the final switch vector.

[0054] Since the target value is incremented in each iteration during this update process and the target value always has a finite upper bound, the algorithm must converge.

[0055] (5) Iteratively optimize the auxiliary variables, transmit beamforming, smart reflector phase, and switch state until convergence, and finally obtain the smart reflector control state and base station beam vector to obtain an energy-efficient configuration solution.

[0056] Specifically, the block coordinate descent method is used to iteratively optimize the auxiliary variables y, ε and u, AP transmit beamforming, smart reflector reflection coefficient Φ and smart reflector working state vector x introduced in the above steps, and find a conditional optimal solution in each iteration until the target value converges. Finally, the optimized smart reflector phase control, switch state and base station beam vector are obtained, and an energy-efficiency optimized configuration scheme is obtained.

[0057] Beneficial Effects

[0058] The present invention proposes a method for optimizing the energy efficiency of a distributed intelligent reflector-assisted de-cellularized network, establishes a system model of a de-cellularized network assisted by an intelligent reflector, and an energy consumption model with controllable switching states of the intelligent reflector; under the conditions of user minimum rate requirements, base station transmission power limitations, and intelligent reflector phase constraints, a system energy efficiency maximization problem of a distributed intelligent reflector-assisted de-cellularized network that jointly designs transmission beamforming, intelligent reflector phase, and switching states is constructed; an iterative optimization method for the proposed energy efficiency optimization problem is given using the Dinkelbach method, Lagrange dual transformation, quadratic programming method, and greedy algorithm, and the optimized intelligent reflector control state and base station beam vector are given respectively; the proposed method effectively improves the system energy efficiency through the reasonable deployment and design of intelligent reflectors, effectively improves the energy efficiency compared to de-cellularized systems without the assistance of intelligent reflectors, and provides certain guiding conclusions for the configuration of scheme parameters. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0060] Figure 1 It is a schematic diagram of an example scenario of the present invention;

[0061] Figure 2 is a flow chart of the implementation of the method in the example of the present invention;

[0062] Figure 3 It is a schematic diagram of the convergence performance of the proposed energy efficiency optimization method;

[0063] Figure 4 It is the performance trend chart of system energy efficiency under different base station transmission powers; DETAILED DESCRIPTION

[0064] The present invention proposes a distributed intelligent reflector-assisted decellularized network energy efficiency optimization method. Figure 1 The algorithm flow of this case is shown in the attached Figure 2 As shown, the specific implementation steps are:

[0065] Step 200: construct a system model of a de-cellular network assisted by an intelligent reflective surface and an energy consumption model of an intelligent reflective surface with controllable switching states.

[0066] The system model of the de-cellularized network assisted by the intelligent reflective surface includes B APs, L intelligent reflective surfaces, and K single-antenna users, wherein all APs and all intelligent reflective surfaces are uniformly controlled and signal processed by the central processing unit CPU. In this de-cellularized scenario assisted by the intelligent reflective surface, all APs and intelligent reflective surfaces work together to serve all users, and the switching state of the intelligent reflective surface can be optimized and controlled.

[0067] In the energy efficiency optimization method proposed in the present invention, in order to further save energy consumption and avoid ineffective deployment of smart reflective surfaces, it is proposed to introduce a binary variable x for the on / off state of the smart reflective surface. l ∈{0,1} representation, added to the cascade channel part of the channel model, that is, when x l = 1, the intelligent reflective surface is in the on state, and the phase is optimized and adjusted; when x l When =0, the intelligent reflective surface is turned off and no energy is consumed.

[0068] Assume that the CPU knows all the channel status information through AP feedback, let represents the channel matrix from the bth AP to the kth user, represents the equivalent channel from the lth smart reflection surface to the kth user, represents the equivalent channel from the bth AP to the lth smart reflector, where M t Indicates the number of transmit antennas of the AP, and N indicates the number of reflective units of each smart reflective surface. is the diagonal matrix of the reflection phase of the lth smart reflection surface, Multi-hop reflection is not considered here.

[0069] Assume m represents the symbol sent by the target to user m, represents the beamforming matrix designed by APb for user m, represents the Gaussian noise at user k, then in the multi-intelligent reflector assisted scenario, the received signal model at user k can be written as

[0070]

[0071] By defining the vector And the merge vector Then the equivalent combined channel from user k to all APs can be expressed as

[0072]

[0073] in, Φ=diag(Φ1,...,Φ l ,...,Φ L ).

[0074] Step 210, construct the system energy efficiency maximization problem of the distributed intelligent reflector-assisted decellularized network, and jointly design the transmit beamforming, the phase of the intelligent reflector, and the switch state to meet the user's minimum rate requirement, the base station transmit power limit, and the phase constraint of the intelligent reflector.

[0075] According to the channel model constructed in step 200, the signal to interference noise ratio of user k can be obtained, which can be expressed as

[0076]

[0077] in, is the beamforming combination vector for each user k. Then the sum rate of all K users can be expressed as

[0078]

[0079] In this system, the total power consumption includes the transmit power of APs, the hardware static power consumption of each AP, and the The hardware static power consumption P of each reflective unit of each smart reflective surface RIS , and the circuit power consumption of each user k Expressed as

[0080]

[0081] in, So, define and Respectively represent the set of intelligent reflective surface and each reflective unit of the reflective surface, and construct the total energy constraint condition P b,max The optimization problem of maximizing energy efficiency under the condition of satisfying the user's minimum rate requirement R min and the phase constraint of the smart reflector, the problem of jointly optimizing the transmit beamforming, the phase of the smart reflector, and the switch state can be described as:

[0082]

[0083]

[0084] log2(1+γ k )≥R min ,

[0085]

[0086]

[0087] Step 220, using the Dinkelbach method, Lagrange dual transformation and quadratic programming method to introduce auxiliary variables to solve the problem, and optimizing the auxiliary variables and base station beamforming under the given phase and switch state of the smart reflector.

[0088] It can be observed that the optimization objective function is a single ratio form of the logarithm and function of the signal-to-interference-noise ratio, which makes the energy efficiency optimization problem difficult to handle. This form is processed by the Dinkelbach method in the fractional programming method, and then the Lagrange dual transformation and quadratic transformation methods are used to decouple and simplify the optimization variables.

[0089] First, by introducing an auxiliary variable y, since the target value remains unchanged when the auxiliary variable y takes the optimal value, the target expression can be equivalent to

[0090]

[0091] To maximize f′(W,Φ,x,y), the optimal value of y can be updated as

[0092]

[0093] Among them, g(W b ) is a function representing the total power consumption,

[0094] However, the new objective function contains the sum of ratios, so the problem is still intractable and non-convex for variables W and φ. Therefore, we further adopt the Lagrangian dual transformation and introduce a slack variable Then f′(W,Φ,x,y) can be equivalently replaced by

[0095]

[0096] Among them, the optimal ε k It can be obtained by taking the first-order partial derivative of f″(W,Φ,x,y,ε), so that So we can update ε k ＊ =γ k .

[0097] However, γ k It is still a non-convex fractional form. Through the quadratic transformation method, an auxiliary variable u is introduced for each ratio k , for the fraction γ in the objective function k The objective function can be further reformulated as

[0098]

[0099] in And, k The optimal value of can be given by To determine,

[0100]

[0101] Therefore, after the above transformation, the auxiliary variables y, ε and u are introduced, and the original optimization problem is equivalently transformed into

[0102]

[0103]

[0104] log2(1+γ k )≥R min ,

[0105]

[0106]

[0107] When the introduced variables y, ε, u, the reflection coefficient of the smart reflector, and the switch vector x are fixed, the optimization problem of AP transmit beamforming can be restated as

[0108]

[0109] The objective expression of the above problem is a concave function with respect to W, while the second constraint is non-convex. The second constraint, the minimum achievable rate, can be further expressed as

[0110]

[0111] Using the continuous convex approximation method, the left side of the inequality can be bounded by its first-order Taylor expansion, so the above formula can be approximated as

[0112]

[0113] In the above formula, (t) indicates that the vector is the value of the last iteration. Therefore, the problem becomes a convex quadratic constrained quadratic programming (QCQP), which can be solved by convex optimization tools (such as CVX solver).

[0114] Step 230: Optimize the phase coefficient and switch state of the smart reflective surface under given transmit beamforming and auxiliary vectors.

[0115] Given variables y, ε, u, x and the transmit beamforming vector And ignore the constant, about the reflection coefficient of the smart reflector The sub-problem can be reformulated as

[0116]

[0117] in,

[0118]

[0119]

[0120] The above problem also involves the minimum achievable rate constraint like the previous problem. Then, we first expand the constraints to

[0121]

[0122] Then it is further transformed into:

[0123]

[0124] in, C k,m =|h k w m | 2 . The SCA method is also used, so the constraint can be further approximated as

[0125]

[0126] where φ (t) represents the optimal value of the previous iteration, so the problem also becomes a convex QCQP, which can be effectively solved by the existing convex optimization solver CVX.

[0127] Then update the switch vector. Given the phase matrix Φ and the beamforming matrix W, the original problem is a nonlinear integer optimization problem about the smart reflector switch vector x. Since nonlinear integer optimization problems are generally NP-hard, it is difficult to obtain a global optimal solution with polynomial complexity. To solve this problem, we plan to use a greedy algorithm to solve the smart reflector switch optimization problem.

[0128] The basic idea is that if the energy efficiency target value can be improved and the new solution is feasible, we try to close one smart reflective surface each time. Step 1: Set the set is the set of active intelligent reflective surfaces s serving users. In step 2, the initial target value of energy efficiency E0 is calculated. Then when the set When not empty, for each If the smart reflective surface l is turned off and the resulting switch vector is feasible, the corresponding energy efficiency value is calculated and recorded as El Otherwise, if the new switch vector is not feasible, the target value E l Recorded as zero.

[0129] Next, we compare the energy efficiency values ​​of these new switching schemes with the initial solution and obtain If k≠0, the index value k indicates that turning off a smart reflective surface can improve energy efficiency. Turning off smart reflective surface k can achieve the highest energy efficiency than turning off any other smart reflective surface l. In this case, turning off smart reflective surface k is energy-saving. Then we select Subtract index k from And update the initial energy efficiency value E0=E for the next iteration k If k = 0, it means that turning off any smart reflective surface will lead to energy inefficiency. In this case, turning off any smart reflective surface will not improve energy efficiency, which means we need to terminate the loop and output the final switch vector.

[0130] Since the target value is incremented in each iteration during this update process and the target value always has a finite upper bound, the algorithm must converge.

[0131] Step 240, iteratively and alternately optimize the auxiliary variables, transmit beamforming, smart reflector phase, and switch state until convergence, and finally obtain the smart reflector control state and base station beam vector, and obtain an energy efficiency optimized configuration solution.

[0132] The block coordinate descent method is used to iteratively optimize the auxiliary variables y, ε and u, AP transmit beamforming, smart reflector reflection coefficient Φ and smart reflector working state vector x introduced in the above steps, and find a conditional optimal solution in each iteration until the target value converges. Finally, the optimized smart reflector phase control, switch state and base station beam vector are obtained, and the energy efficiency optimized configuration scheme is obtained.

[0133] The simulation results are shown in the attached Figure 3 And attached Figure 4 As shown. In the simulation scenario, we consider 4 APs and 8 smart reflectors to serve 4 users. Consider a 200m×200m map. The 4 APs are located at (0,0), (0,200m), (200m,0), (200m,200m), and the users are randomly distributed in a circle with a radius of 20m centered at (100,100). The smart reflectors are evenly distributed at all angles centered at (100,100) and randomly distributed in radius. Considering that the APs in cellular networks are usually small in size, the number of AP antennas is set to M. t =2, where in the simulation parameters, it is assumed that all users have the same minimum QoS requirements, i.e., R min=1bps / Hz, and the static power consumption of each AP hardware is P b B =9dBW, the hardware static power consumption of each reflective unit of each smart reflective surface is P RIS =10dBm, and the circuit power consumption of each user k represents the efficiency of the transmit power amplifier, τ b =1.2, noise power σ 2 =-90dBm. In addition, unless otherwise specified, the number of reflection units of each smart reflection surface is N=50.

[0134] Attached Figure 3 The convergence of the proposed method is verified. As can be seen from the figure, the method converges in about 7 iterations, and the convergence speed is fast. In addition, the figure analyzes the impact of different numbers of smart reflector units and different AP transmit powers on its convergence performance. When the number of units N increases from 50 to 100, the energy efficiency performance can be improved to a certain extent due to the rich scattering channel gain brought by more smart reflector units, while the convergence speed is slightly slowed down due to the increase in the dimension of the calculated matrix. Similarly, as the AP transmit power increases from 10dBm to 20dBm, the energy efficiency performance is greatly improved, while the convergence speed is slightly slowed down.

[0135] Attached Figure 4 The system energy efficiency performance trend with the increase of AP transmit power is given. As can be seen from the figure, the system energy efficiency of the proposed architecture first increases with the increase of AP transmit power, and then decreases after exceeding a certain threshold. The proposed smart reflector-assisted decellularization architecture is significantly better than the smart reflector random phase, decellularization without smart reflector assistance, and traditional decellularization, where the traditional decellularization architecture refers to the case where all smart reflectors are replaced by APs. It is observed that deploying smart reflectors in the decellularization system can effectively improve the energy efficiency of the multi-user communication system, and when the AP transmit power is small, the energy efficiency gain brought by the smart reflector is more obvious, because the smart reflector can significantly improve the system capacity while introducing less energy consumption in this case. At the same time, it is observed that the traditional decellularization system provides more transmit power through more APs to achieve a higher total rate. Its energy efficiency needs to be based on a lower AP total transmit power constraint. Therefore, when the transmit power is low, its energy efficiency is more considerable. When the AP transmit power is greater than a certain value, the hardware static power consumption becomes the main factor affecting the energy efficiency. Therefore, the proposed smart reflector-assisted system can obtain better energy efficiency performance than the traditional non-cellular system.

Claims

1. A distributed intelligent reflector-assisted decellularized network energy efficiency optimization method, characterized in that: The following steps are involved: Firstly, a system model of a de-cellularized network assisted by an intelligent reflector and an energy consumption model with controllable switching states of the intelligent reflector are constructed; the system energy efficiency maximization problem of a distributed intelligent reflector-assisted de-cellularized network is constructed, and the transmit beamforming, the phase of the intelligent reflector and the switching state are jointly designed to meet the user's minimum rate requirement, the base station transmit power limit and the phase constraint of the intelligent reflector; the Dinkelbach method, Lagrange dual transformation and quadratic programming method are used to introduce auxiliary variables to decouple the problem, and the auxiliary variables and the base station beamforming are optimized given the phase and switching state of the intelligent reflector; the phase coefficient and the switching state of the intelligent reflector are optimized given the transmit beamforming and the auxiliary vector; the auxiliary variables, transmit beamforming, the phase of the intelligent reflector and the switching state are iteratively optimized until convergence, and finally the control state of the intelligent reflector and the base station beam vector are obtained, and the configuration scheme with energy efficiency optimization is obtained.

2. The system model for constructing a decellularized network assisted by an intelligent reflective surface according to claim 1, characterized in that: Construct a decellularized network scenario with B APs, L smart reflectors, and K single-antenna users. All APs and all smart reflectors are uniformly controlled and signal processed by the central processing unit (CPU). In this decellularized scenario assisted by smart reflectors, all APs and smart reflectors work together to serve all users, and the on / off state of the smart reflectors can be optimized and controlled. A binary variable x is introduced. l ∈{0,1} represents the switch state, that is, when x l = 1, the intelligent reflective surface is in the on state, and the phase is optimized and adjusted; when x l = 0, the intelligent reflector is closed and does not consume energy. Further define the vector Construct an equivalent combined channel from user k to all APs Expressed as in, represents the equivalent combined channel from all base stations to the kth user, represents the equivalent combined channel from all smart reflectors to the kth user, G represents the equivalent combined channel from all APs to all smart reflectors, Φ=diag(Φ1,...,Φ l ,...,Φ L ) represents the diagonal matrix combination of the reflection phases of L smart reflection surfaces.

3. According to claim 1, the energy consumption model of the smart reflective surface with controllable switch state for building a smart reflective surface-assisted decellularized network is characterized in that: In this system, the total power consumption includes the transmit power of APs, the hardware static power consumption of each AP, and the The hardware static power consumption P of each reflective unit of each smart reflective surface RIS , and the circuit power consumption of each user k Expressed as Among them, x l represents the switch state of the lth smart reflective surface, N represents the number of reflective units of each smart reflective surface, Indicates the efficiency of the transmit power amplifier, W b =[w b,1 ,...,w b,K ],w b,k represents the beamforming matrix designed by APb for user k.

4. According to the problem of maximizing the system energy efficiency of building a distributed intelligent reflector-assisted decellularized network as described in claim 1, it is characterized in that: Construct an energy efficiency maximization problem and jointly design the transmit beamforming, smart reflector phase, and switch state to meet the user's minimum rate requirement R min , Base station transmission power limit P b,max And the phase constraint of the smart reflector Among them, R sum is the system user and rate, γ k is the signal-to-interference-noise ratio of user k, is the phase of the nth unit of the lth smart reflector, and They respectively represent the set of smart reflective surface and each reflective unit of the reflective surface.

5. According to claim 1, the method of using the Dinkelbach method, Lagrange dual transformation and quadratic programming method to introduce auxiliary variables to decouple the problem, optimizing the auxiliary variables and base station beamforming under the given phase and switch state of the smart reflector, is characterized in that: By using the Dinkelbach method, Lagrange dual transformation and quadratic programming method, auxiliary variables are introduced The objective function can be further reformulated as in g(W b ) is a function representing the total power consumption is the beamforming combination vector for each user k, and the optimal values ​​of the auxiliary variables are updated as Given the introduced variables y, ε, u and the reflection coefficient of the smart reflector and the switch vector x, solve the optimization problem of AP transmit beamforming The continuous convex approximation method is used to deal with the second constraint approximation: In the above formula, (t) indicates that the vector is the value of the last iteration, making the problem convexly solvable.

6. The method for optimizing the phase coefficient and switch state of the smart reflector under given transmit beamforming and auxiliary vector according to claim 1, characterized in that: Given variables y, ε, u, x and the transmit beamforming vector Ignore the constant and solve for the reflection coefficient of the smart reflector Sub-problem in, And the SCA method is also used to process the first constraint, which is further transformed into: in, C k,m =|h k w m | 2 ,φ (t) represents the optimal value of the previous iteration, so the problem also becomes a convex QCQP, which can be effectively solved by the existing convex optimization solver CVX. Then, the greedy algorithm is used to solve the smart reflector switch optimization problem.

Images