Active-passive hybrid intelligent metasurface-assisted low-power-consumption beam forming method

Abstract

The invention provides an active-passive hybrid intelligent metasurface-assisted low-power-consumption beam forming method, which belongs to the technical field of wireless communication, and comprises the following steps: designing an active-passive hybrid intelligent metasurface, and constructing an intelligent metasurface-assisted communication system model; analyzing the reflection mode, amplitude and phase physical constraint conditions of the intelligent metasurface, the power constraint conditions of the base station and the intelligent metasurface, and the minimum rate constraint condition of each user in the communication system model, and constructing a first optimization problem of total power consumption minimization according to beam forming design requirements; and converting the first optimization problem to obtain a second optimization problem, carrying out continuous relaxation processing on the second optimization problem to obtain a continuous variable optimization problem, and carrying out iterative updating on coupling variables involved in the continuous variable optimization problem by adopting an alternate optimization framework to obtain a final joint beam forming optimization result. On the premise of meeting the required communication quality, the power consumption overhead of the system can be reduced to the maximum extent.

Description

Technical Field

[0001] The present invention belongs to the technical field of wireless communication, and particularly relates to a low-power beamforming method assisted by an active-passive hybrid intelligent metasurface. Background Art

[0002] With the development of the new generation of wireless communication networks, technologies such as ultra-massive multiple-input multiple-output, ultra-dense networking, and millimeter-wave communication have been widely studied to meet the future communication requirements of high data rate, wide coverage, and low latency. However, these technologies still follow the traditional idea of trading scale for gain and face problems such as high hardware cost, high energy consumption overhead, and sensitivity to signal blockage. Currently, the intelligent metasurface technology based on sub-wavelength metamaterials is considered to be one of the key enabling technologies for future networks, which can flexibly control parameters such as the phase, amplitude, and polarization of spatial electromagnetic waves, and has the advantages of low cost, low energy consumption, and easy deployment, and can provide more spatial degrees of freedom to meet the growing communication requirements of future networks.

[0003] Traditional intelligent metasurfaces include two types: all-passive and all-active based on a fully connected structure. The all-passive intelligent metasurface integrates a large number of low-cost passive reflection elements and dynamically adjusts the phase of the incident signal to adapt to the changing transmission environment. Affected by the multiplicative fading effect of the cascaded reflection path, the performance gain of the all-passive metasurface-assisted communication system is relatively limited. Although deploying more reflection elements can alleviate this problem, due to the lack of signal processing capabilities of the all-passive intelligent metasurface, the large-scale metasurface antenna array configuration depends on the system channel state information feedback, which will lead to unaffordable communication overhead. The all-active intelligent metasurface can amplify the signal amplitude by integrating a reflective amplifier for each element to effectively overcome the influence of path loss. However, the large number of integrated reflective amplifiers will result in high hardware cost and energy consumption overhead, which is not suitable for large-scale deployment in future networks.

[0004] Most of the existing research focuses on the beamforming design of communication systems assisted by fully connected intelligent metasurfaces, facing high hardware cost and control overhead. Some research has proposed new all-passive and all-active intelligent metasurface structures based on sub-array division, sacrificing some performance in exchange for the reduction of system overhead. However, the beamforming design freedom of the intelligent metasurface based on a single active / passive reflection type is still relatively limited. Summary of the Invention

[0005] The purpose of the present invention is to propose a low-power beamforming method assisted by an active-passive hybrid intelligent metasurface, which can minimize the system power consumption overhead on the premise of meeting the required communication quality.

[0006] The present invention is realized through the following technical solutions: A low-power beamforming method assisted by an active-passive hybrid intelligent metasurface, comprising the following steps: Step S1, design an active-passive hybrid intelligent metasurface based on subarray division, obtain the reflection coefficient matrix of the intelligent metasurface, construct a communication system model assisted by the intelligent metasurface, and determine the achievable total rate and total power consumption expressions of the communication system model; Step S2, analyze the reflection mode, amplitude and phase physical constraint conditions of the intelligent metasurface, the power constraint conditions of the base station and the intelligent metasurface, and the minimum rate constraint conditions of each user in the communication system model. According to the beamforming design requirements, construct a first optimization problem for minimizing the total power consumption, which involves the joint design of the base station transmit precoding vector and the reflection beamforming matrix of the intelligent metasurface; Step S3, transform the first optimization problem constructed in Step S2 to obtain a second optimization problem, perform continuous relaxation processing on the second optimization problem to obtain a continuous variable optimization problem, and use an alternating optimization framework to iteratively update the coupled variables involved in the continuous variable optimization problem to obtain the final joint beamforming optimization result.

[0007] Further, in Step S1, the intelligent metasurface includes N reflection elements divided into M subarrays, each subarray can be switched between an active reflection operating mode and a passive reflection operating mode, and the reflection matrix of the intelligent metasurface is expressed as , where represents the overall phase shift matrix of the intelligent metasurface when all M subarrays are active subarrays, represents the overall phase shift matrix of the intelligent metasurface when all M subarrays are passive subarrays, , represents the identity matrix, is a vector of all 1s with dimension , is the reflection amplitude matrix when all M subarrays are active subarrays, is the mode switching matrix, represents the transpose operation, represents the Kronecker product operation.

[0008] Further, in Step S1, the communication system model includes a base station with N t antennas deployed, an intelligent metasurface with N reflection elements, and K single-antenna users. The achievable total rate expression of the communication system model is , the total power consumption expression of the communication system model is , where is the set of single-antenna users, represents the beamforming vector of the base station for transmitting information symbols to user k , represents the effective channel between the base station and user k , represents the channel state information from the base station to the intelligent metasurface, represents the channel state information from the base station to user k , represents the channel state information from the intelligent metasurface to user k , represents the variance of the thermal noise at the active sub-array of the intelligent metasurface, represents user k the variance of the additive white Gaussian noise at, represents the reflection matrix corresponding to the active sub-array part of the intelligent metasurface, is the power consumption expression at the base station side, is the power consumption expression at the intelligent metasurface side, is the reciprocal of the amplification efficiency of the transmit power amplifier at the base station, Tr(.) is the trace function, P a is the power consumption of a single antenna of the base station, P s is the static power consumption of the baseband processing unit of the base station, is the reciprocal of the amplification efficiency of the reflection power amplifier of the intelligent metasurface, P c represents the power consumption of the phase shifter circuit of the intelligent metasurface, P DC represents the DC operating power consumption of the reflection power amplifier of the intelligent metasurface, represents the conjugate transpose operation.

[0009] Further, in step S2, the first optimization problem of minimizing the total power consumption is , where and are auxiliary parameters, P B is the maximum power load of the base station, P R is the maximum power load of the intelligent metasurface side, γ min is the minimum achievable communication rate of the user, ρ min and ρ maxThey represent the lower and upper limits of the range of the gain factor of the reflective power amplifier, respectively. For the matrix No. n diagonal elements, This is a modulo operation.

[0010] Furthermore, in step S3, the second optimization problem is expressed as ,in, , is the introduced semi-positive definite diagonal matrix, which represents the signal amplification power of the active sub-array, for M The connection relationship between the elements and the reflective power amplifier when all sub-arrays are active sub-arrays. is an auxiliary variable, which is defined as , is an auxiliary parameter, Represents the conjugate operation.

[0011] Furthermore, in step S3, the continuous variable optimization problem is expressed as ,in, is the introduced exponential term.

[0012] Furthermore, in step S3, the coupling variable includes the base station end transmitting precoding vector , Joint reflection amplitude and mode switching matrix of smart metasurface , phase shift matrix , using the alternating optimization framework to , and Iterative updates are performed, and when solving a certain coupling variable, the values ​​of the other coupling variables are kept unchanged. Each iteration selects the result of the previous iteration as the initial point, and the variable iterative update process is cyclic until the objective function value involved in the continuous variable optimization problem converges.

[0013] Furthermore, in step S3, the combined reflection amplitude and mode switching matrix of the fixed intelligent metasurface and phase shift matrix , the continuous variable optimization problem is about the base station end transmitting precoding vector The sub-problem is expressed as , using continuous convex approximation technology, in t +1 iteration outputs results based on the previous iteration , constrain the non-convex minimum achievable rate Approximately , the converted base station transmit precoding vector Solve the sub-problem of to obtain the transmit precoding vector The optimization result, where is an auxiliary parameter, , is the operation of taking the real part.

[0014] Furthermore, in the step S3, the joint reflection amplitude and mode switching matrix of the intelligent metasurface and the base station transmit precoding vector are fixed. The sub-problem of the continuous variable optimization problem with respect to the intelligent metasurface phase shift matrix is expressed as , where is the stacked optimization vector, , , vecd(.) is the operation of extracting the diagonal elements of the matrix and arranging them into a column vector. The positive semi-definite matrix is the coupling parameter of the conjugate quadratic form variable in the power constraint condition. The positive semi-definite matrices , and are the coupling parameters of the conjugate quadratic form with respect to the variables , and in the user minimum achievable rate constraint, respectively. , is the coupling parameter obtained based on the active subarray structure simplification. is the linear parameter with respect to the linear form variable . , is the -th element of the vector n . In the t +1-th iteration, based on the output result of the previous iteration , combined with the constant modulus property of the phase shift variable and using Taylor expansion, the upper bound linear convex surrogate functions of the objective function and the constraint conditions of the sub-problem with respect to the intelligent metasurface phase shift vector are and , respectively. And the approximate linear programming problem is solved to obtain the optimization result of the original intelligent metasurface phase shift matrix , where represents the operation of taking the maximum eigenvalue. , and are all constant terms independent of the optimization variables obtained in the Taylor expansion approximation process.

[0015] Furthermore, in the step S3, the base station transmit precoding vector and the of the intelligent metasurface are fixed.The continuous variable optimization problem with respect to the joint reflection amplitude and mode switching matrix of the intelligent reflecting surface The sub-problem is expressed as where is the introduced auxiliary variable. Using the continuous convex approximation technique, the sub-problem is approximated as a convex problem at the t -th iteration . By solving this convex problem and continuously iterating, the original joint reflection amplitude and mode switching matrix of the intelligent reflecting surface can be obtained. Among them, the semi-positive definite matrix is the coupling parameter of the source sub-array signal power amplification term in the objective function, and the semi-positive definite matrices are all coupling parameters of the conjugate quadratic form optimization variables in the user minimum achievable rate constraint.

[0016] The present invention has the following beneficial effects: 1. The present invention first designs an active-passive hybrid intelligent reflecting surface based on sub-array division, constructs a communication system model assisted by this intelligent reflecting surface, and then analyzes the physical constraint conditions of the reflection mode, amplitude and phase of the intelligent reflecting surface, the power constraint conditions of the base station and the intelligent reflecting surface, and the minimum rate constraint conditions of each user in the communication system model. According to the beamforming design requirements, a first optimization problem with minimized total power consumption is constructed. Finally, the first optimization problem is transformed into a second optimization problem, and the second optimization problem is continuously relaxed to obtain a continuous variable optimization problem. The coupling variables involved in the continuous variable optimization problem are iteratively updated using an alternating optimization framework to obtain the final joint beamforming optimization result. It can be applied to the downlink communication scenario where the base station serves multiple users simultaneously. Each sub-array can switch between active and passive operating modes according to the actual transmission environment, thereby minimizing the system power consumption overhead as much as possible while meeting the required communication quality, and can support future large-scale intelligent reflecting surface assisted communication scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The present invention will be further described in detail below with reference to the drawings.

[0018] Figure 1 is the flowchart of the present invention.

[0019] Figure 2 is the detailed flowchart of the present invention.

[0020] Figure 3 is the schematic diagram of the system scenario of the present invention.

[0021] Figure 4 is the structure diagram of the intelligent reflecting surface of the present invention.

[0022] Figure 5This is a comparison curve graph of the total power consumption obtained by the present invention with respect to the change in the minimum achievable communication rate of users and a comparison scheme.

[0023] Figure 6 This is a comparison curve graph of the total power consumption obtained by the present invention with respect to the change in the number of users and a comparison scheme.

[0024] Figure 7 This is a comparison curve graph of the total power consumption obtained by the present invention with respect to the change in the distance between the base station and users and a comparison scheme. Detailed implementation manners

[0025] As Figure 1 and Figure 2 shown, the active-passive hybrid intelligent metasurface-assisted low-power beamforming method includes the following steps: Step S1: Design an active-passive hybrid intelligent metasurface based on subarray division, obtain the reflection coefficient matrix of the intelligent metasurface, construct a communication system model assisted by the intelligent metasurface, and determine the expressions of the achievable total rate and total power consumption of the communication system model; Specifically, as Figure 3 shown, the downlink communication system model constructed in this embodiment includes a communication base station, an active-passive subarray hybrid intelligent metasurface, and K single-antenna users. The set of single-antenna users is denoted by , the base station antenna is a uniform linear array including N t antennas, and the number of reflection units (i.e., reflection elements) of the intelligent metasurface is denoted as N .

[0026] As Figure 4 shown, the constructed active-passive subarray hybrid intelligent metasurface evenly divides the reflection elements into subarrays, each subarray contains elements, and each subarray can flexibly switch between the active reflection mode and the passive reflection mode to adapt to the complex and changeable communication environment. In the active reflection mode, all elements within the subarray are respectively connected to different phase shifter circuits, which can independently regulate the phase of the incident signal. All phase shifter circuits are connected to a common active reflection amplifier, which can uniformly regulate the amplitude of the incident signal, further reducing the hardware cost and power consumption overhead of the active subarray; in the passive reflection mode, all elements within the subarray are connected to a common multi-phase shifter, and the incident signal of the subarray shares the same phase regulation parameter, which can significantly reduce the beamforming design complexity, and the subarray cost overhead is also the lowest. The hardware module and control method for realizing the switching of the subarray between the passive reflection mode and the active reflection mode are prior arts.

[0027] To further explore the reflection signal model of the active-passive subarray hybrid intelligent metasurface, define are the incident signal vector and the reflected signal vector of the sub-array m respectively. Considering the sub-array m operating in the passive reflection mode, since the reflection units of the sub-array share the same phase shifter parameters , the mapping relationship between the incident and reflected signals of this sub-array is . On the other hand, considering the sub-array m operating in the active reflection mode, the incident signal is first independently phase-regulated by Q phase shifters, and then combined into a single signal and input into a reflection power amplifier for amplitude regulation. The amplified signal can be expressed as , where represents the amplification amplitude coefficient of the m -th active sub-array, and represents the phase shift coefficient vector of the phase shifters of the sub-array Q ; without loss of generality, the amplified signal is equally divided into Q paths of signals, and each path of signal is again regulated by the phase shifter circuit within the sub-array to complete signal reflection. This process can be expressed as . Introduce a diagonal mode switching matrix to characterize the active-passive hybrid reflection characteristics of this intelligent metasurface. Its diagonal elements satisfy , and represent that the m -th sub-array operates in the passive and active modes respectively. Accordingly, the reflection matrix of the m -th sub-array can be expressed as , and the overall reflection matrix of the intelligent metasurface can be expressed as , where represents the reflection matrix corresponding to the active sub-array part of the intelligent metasurface, represents the reflection matrix corresponding to the passive sub-array part of the intelligent metasurface, represents the phase shift matrix when all M sub-arrays are active sub-arrays, represents the phase shift matrix when all M sub-arrays are passive sub-arrays, , represents the identity matrix, is a vector of all 1s with dimension Q , , is the reflection amplitude matrix when all M sub-arrays are active sub-arrays. represents the transpose operation, represents the Kronecker product operation.

[0028] Define , , are the channel state information between the base station-intelligent reflecting surface, the base station-user k , and the intelligent reflecting surface-user k respectively. And represent the information symbol transmitted from the base station to the user k and its corresponding beamforming vector respectively. Assuming that all channels are quasi-static flat fading channels over the entire transmission block and the global channel state information is known, the received signal at the user k is , where and represent the thermal noise when all M sub-arrays of the intelligent reflecting surface act as active sub-arrays and the additive white Gaussian noise at the user respectively. According to the above intelligent reflecting surface model and received signal model, the achievable total rate expression of the communication system model is , where represents the effective channel between the base station and the user k , represents the variance of the thermal noise at the active sub-array of the intelligent reflecting surface, represents the variance of the additive white Gaussian noise at the user k , and represents the conjugate transpose operation.

[0029] The power consumption expression at the base station side is , where is the reciprocal of the amplification efficiency of the transmit power amplifier of the base station considering the actual power conversion loss, P a is the power consumption of a single antenna of the base station, P s is the static power consumption of the baseband processing unit of the base station, Tr(X) represents the trace function of matrix X, and the power consumption expression at the intelligent reflecting surface side is , where is the reciprocal of the amplification efficiency of the reflective power amplifier, P c represents the power consumption of the phase shifter circuit of the intelligent reflecting surface, P DC represents the DC operating power consumption of the reflective power amplifier of the intelligent reflecting surface. Therefore, the total power consumption index expression of the communication system model is . Under the given conditions of the base station side transmit power constraint, the signal amplification power constraint and the dynamic structure physical constraint of the intelligent reflecting surface, as well as the user minimum rate constraint, considering the beamforming parameter design of the base station and the intelligent reflecting surface comprehensively, the system total power consumption minimization problem of the joint beamforming design can be established.

[0030] Step S2: Analyze the reflection pattern, amplitude, and phase physical constraint conditions of the intelligent metasurface, the power constraint conditions of the base station and the intelligent metasurface, and the minimum rate constraint conditions of each user in the communication system model. According to the beamforming design requirements, construct a first optimization problem for minimizing the total power consumption, which involves the joint design of the base station transmission precoding vector and the reflection beamforming matrix of the intelligent metasurface; The present invention aims to minimize the total power consumption and jointly designs the set of transmission precoding vectors at the base station end and the reflection beamforming matrix of the hybrid active and passive subarray intelligent metasurface , where is jointly determined by the reflection amplitude and mode switching matrix , the phase shift matrices of the active and passive subarrays . Considering the system constraint conditions, the set of transmission precoding vectors needs to satisfy the maximum power constraint conditions of the base station and the intelligent metasurface; the active-passive reflection mode switching of the intelligent metasurface subarray is modeled as an integer discrete variable constraint, and the phase shift coefficients of the phase shifters need to satisfy the unit constant modulus constraint; each user has a minimum achievable communication rate constraint to ensure its communication quality. Therefore, the first optimization problem for minimizing the total power consumption is , where , define the auxiliary parameters and , represents the signal amplification power consumption of all active subarrays of the intelligent metasurface, is a parameter value independent of the optimization variables, constraints (1b), (1c) represent the power constraint conditions of the base station and the intelligent metasurface, P B is the maximum power load of the base station, P R is the maximum power load of the intelligent metasurface end, represents the number of active subarrays, (1d) is the constraint condition on the minimum communication rate γ min of each user, (1e)-(1g) respectively represent the mode switching constraint, reflection amplitude constraint, and constant modulus phase constraint conditions of the intelligent metasurface, is the -th diagonal element of the matrix n , is the modulus operation, ρ min and ρ maxThey respectively represent the lower bound and the upper bound of the value range of the amplification factor of the reflective power amplifier. This problem is a mixed-integer programming problem, including discrete dynamic subarray pattern switching constraints (1e), non-convex user minimum achievable rate constraints (1d), and constant modulus phase constraints (1g). The variables are strongly coupled and difficult to solve directly. Therefore, the present invention aims to solve the above mixed-integer programming problem based on the idea of continuous relaxation. First, the problem is transformed to obtain an easier-to-process problem form, and then the discrete integer variables are relaxed to continuous variables within the interval, and an exponential term penalty factor is introduced to ensure that the optimization result of the continuous variables approaches the discrete values. Then, based on the alternating optimization framework, the joint beamforming variable result is obtained.

[0031] Step S3: Transform the first optimization problem constructed in step S2 to obtain a second optimization problem, perform continuous relaxation processing on the second optimization problem to obtain a continuous variable optimization problem, and use the alternating optimization framework to iteratively update the coupled variables involved in the continuous variable optimization problem to obtain the final joint beamforming optimization result.

[0032] To effectively solve the constructed non-convex mixed-integer programming problem (the first optimization problem), first introduce a semi-definite diagonal matrix representing the signal amplification power of the active subarray. Combining the binary discrete characteristics of the diagonal elements of the mode switching matrix , the following amplification power equality relationship can be obtained , and minimizing the objective function can be further transformed into , where is an auxiliary parameter related to the matrix . Combining the minimization characteristics of the objective function, the above equality relationship about can be equivalently scaled to an inequality relationship . Define the auxiliary variable , and the first optimization problem can be transformed into the following second optimization problem: , where is the connection relationship between the elements and the reflective power amplifier when all M subarrays are active subarrays, is an auxiliary parameter, and represents the conjugate operation.

[0033] The above second optimization problem is still a mixed-integer programming problem containing non-convex discrete variables. Directly solving this problem will result in exponential computational complexity. Accordingly, the present invention performs continuous relaxation processing on the discrete mode switching variables, approximates as continuous variables within the interval , at the same time, to ensure that the optimization result of continuous relaxation is close to the feasible discrete solution set, an exponential term is introduced for the continuous variable and obtain to ensure the equivalence of the continuous relaxation of discrete variables. On the one hand, the exponential term ensures that its output result is sufficiently close to the binary discrete value through penalty . As increases, the optimal objective function value of the optimized problem after continuous relaxation is closer to the objective value of the original mixed-integer optimization problem. On the other hand, a larger will make the feasible region of the problem-solving smaller, and the optimization result is likely to fall into a local optimal point with poor performance. Based on the above processing, the second optimization problem can be transformed into the following continuous variable optimization problem . In practice, it is necessary to select a suitable value for according to the application scenario to ensure that a high-quality sub-optimal solution can be obtained by solving this optimization problem.

[0034] To effectively solve the above continuous variable optimization problem, the present invention adopts an alternating optimization algorithm framework, and sequentially updates three types of coupled variables: the transmit precoding vector at the base station side, the joint reflection amplitude and mode switching matrix of the intelligent surface, and the phase shift matrix . When solving a certain type of variable, the values of the remaining variables are kept unchanged. Each iteration selects the result of the previous iteration as the initial point, and the loop variable iteration update process continues until the objective function value involved in the continuous variable optimization problem converges, so as to obtain the final joint beamforming optimization result.

[0035] Specifically, for the transmit precoding vector at the base station side, fixing the joint reflection amplitude and mode switching matrix and the phase shift matrix of the intelligent surface, the sub-problem of the continuous variable optimization problem with respect to the transmit precoding vector at the base station side is expressed as , where is an auxiliary parameter, , is the real part operation. This problem includes the transmit power constraint at the base station side, the signal amplification power constraint at the intelligent surface side, and the non-convex minimum achievable rate constraint for each user, which is difficult to solve directly. The present invention uses the continuous convex approximation technology to perform a first-order Taylor expansion on the right side of the minimum achievable rate constraint t in the +1th iteration based on the output result of the previous iteration to obtain the corresponding linear lower bound approximation expression , then the above sub-problem is transformed into​ , the optimized result of the transmit precoding vector at the base station side can be obtained by solving the transformed problem using a numerical solution toolbox such as CVX.

[0036] For the intelligent reflecting surface phase shift matrix , fixing the combined reflection amplitude and mode switching matrix of the intelligent reflecting surface and the transmit precoding vector at the base station side, the sub-problem of the continuous variable optimization problem with respect to the intelligent reflecting surface phase shift matrix is expressed as , where , is the -th element of the vector n , is the modulus operation, and represent the vectors formed by the diagonal elements of the intelligent reflecting surface phase shift matrix according to the position order, is the introduced stacked vector, vecd(.) is the operation of extracting the diagonal elements of the matrix and arranging them into a column vector, the positive semi-definite matrix is the coupling parameter for the conjugate quadratic form variable in the power constraint condition, the positive semi-definite matrices , and are the coupling parameters for the conjugate quadratic forms of the variables , and in the user minimum achievable rate constraint respectively, , is the coupling parameter obtained based on the active sub-array structure simplification, is the linear parameter with respect to the linear form variable . This problem includes non-convex minimum achievable rate constraints for each user and the constant modulus property constraint of the phase shift matrix, and the in the objective function and constraints is in the fourth order form for both, making it difficult to solve directly. In the t +1-th iteration, based on the output result of the previous iteration, the upper bound linear convex surrogate functions of the objective function and constraints are derived using Taylor expansion: , and , then an additional stacked vector is defined, and the above sub-problem of the continuous variable optimization problem with respect to the intelligent reflecting surface phase shift vector is approximated as a linear programming sub-problem form: , where the auxiliary function is defined as , , the parameter , and are all constant terms independent of the optimization variables in the Taylor expansion approximation process. Since the linear programming problem with non-convex constant modulus constraints has strong duality, it can be solved by the corresponding Lagrangian dual problem. Accordingly, the Lagrangian dual function can be expressed as , where is the Lagrangian dual variable corresponding to the inequality constraint. Using the constant modulus property of the phase shift vector and the Cauchy-Schwarz inequality, the optimal phase shift vector expression can be derived as . Based on the solved dual variable , the optimization result of the phase shift vector t in the -th iteration can be obtained. Repeating the above process until the objective function converges, the optimization result of the final phase shift vector can be obtained, and the phase shift matrix result of the intelligent metasurface and can be obtained.

[0037] For the reflection amplitude and mode switching matrix of the intelligent metasurface, fix the phase shift matrix of the intelligent metasurface and the transmit precoding vector at the base station side. For the convenience of problem solving, define the stacked variable for variable substitution, and the reflection matrix of the intelligent metasurface can be restated as . The sub-problem of the continuous variable optimization problem with respect to the joint reflection amplitude and mode switching matrix of the intelligent metasurface is expressed as , where the positive semi-definite matrix is the coupling parameter of the active subarray signal power amplification term in the objective function, and the positive semi-definite matrices are all coupling parameters of the conjugate quadratic form optimization variables in the minimum achievable rate constraint of each user. This sub-problem includes the signal amplification power constraint at the intelligent metasurface side, the reflection amplitude constraints of the dynamic structure and subarrays, and the minimum achievable rate constraints of each user. Using the continuous convex approximation technique to transform this non-convex optimization sub-problem, specifically, the amplification power inequality constraint can be restated as . In the t -th iteration of the sub-problem, the right convex function of this term is approximated by the first-order linear lower bound as , and this non-convex constraint can be approximated as the following convex constraint . Similarly, the non-convex minimum achievable rate constraint of each user can be approximated as . Based on the above processing, the above sub-problem can be approximated as a convex problem: , the convex problem can be solved by using numerical optimization tools such as CVX. By continuously iterating the above process until convergence, the reflection amplitude and mode switching matrix of the intelligent metasurface can be obtained based on the results of .

[0038] The above is the specific update process for the three types of optimization variables, namely the transmit precoding vector at the base station end , the reflection amplitude and mode switching matrix of the intelligent metasurface , and the phase shift matrix . By continuously alternating the solution of the optimization variables in the iteration until the calculation result converges, and then discretizing the relaxed continuous variables , and further updating the remaining variables, a feasible beamforming result can be finally obtained.

[0039] More specifically, the performance of the present invention is verified by using Matlab numerical simulation software.

[0040] Simulation parameters: The total number of elements of the intelligent metasurface (referred to as RIS in the figure) is , and it is set that the total number of sub-arrays included in the RIS of the present invention is , and each sub-array contains 4 elements; the power consumption of the phase shifter circuit is , and the DC power consumption of the reflective power amplifier circuit is .

[0041] The comparison schemes are as follows: (a) Fully connected passive RIS: The intelligent metasurface adopts a fully connected structure, and each reflection element is equipped with a phase shifter circuit separately. The RIS reflection matrix is a diagonal matrix, and the modulus values of the diagonal elements are all 1, which can adjust the phase of the incident signal; (b) Fully connected active RIS: The intelligent metasurface adopts a fully connected structure, and each reflection element is equipped with an independent phase shifter and a reflective power amplifier circuit separately. The RIS reflection matrix is a diagonal matrix, which can adjust the amplitude and phase of the incident signal; (c) Fixed active and passive sub-array hybrid RIS: The intelligent metasurface adopts the sub-array connection structure proposed in the present invention, and it is set that the number of active and passive sub-arrays is equal, and the reflection mode of each sub-array is fixed.

[0042] Figures 5 to 7 In

[0043] , the dynamic active and passive sub-array hybrid RIS is the present invention. Figure 5 As shown in , the curve of the total system power consumption varying with the minimum achievable communication rate of the user obtained by the present invention in the user number scenario. It can be observed from the figure that when ), the fully-connected active RIS exhibits lower system power consumption performance compared to the fully-connected passive RIS. When takes a smaller value ( ), the performance of the fully-connected passive RIS is better. This is because when the communication rate requirement is high, the phase adjustment ability of the fully-connected passive RIS alone cannot meet this requirement, so a higher base station transmission power is needed. At this time, the fully-connected active RIS can significantly reduce the base station transmission power by amplifying the signal amplitude. As the communication rate requirement decreases, the high hardware cost of the fully-connected active RIS leads to higher power consumption overhead. In addition, the present invention exhibits the lowest power consumption overhead and is significantly better than the fixed scheme without subarray dynamic allocation.

[0044] As Figure 6 shown, the curve of the total system power consumption varying with the number of users obtained by the present invention under different scenarios of the minimum achievable communication rate of users . It can be observed from the figure that as the number of users increases, the system power consumption required by each scheme increases to meet the growing system communication requirements and fully suppress the interference between users. When the minimum user rate is a relatively high , the fully-connected passive RIS scheme has the highest system power consumption overhead. As the system communication requirements increase due to the increase in the number of users, the fully-connected active RIS with more active power amplifiers has lower power consumption overhead compared to the fixed active-passive subarray hybrid RIS. The present invention has the lowest system power consumption overhead compared to other schemes, because the active-passive subarray hybrid RIS structure has lower hardware cost, and the dynamic subarray allocation can achieve the optimal trade-off between the active power amplification gain and the passive beamforming gain, and can better adapt to the changing communication environment.

[0045] As Figure 7 shown, the curve of the total system power consumption varying with the distance between the base station and the users obtained by the present invention under the scenario of the minimum achievable communication rate of users and the number of users . It can be observed from the figure that as the distance between the base station and the users increases, the path loss becomes more serious, and each scheme requires more total system power consumption to meet the user communication requirements. In particular, when the distance between the base station and the users is large, the system power consumption of the fully-connected passive RIS scheme is significantly higher than that of other schemes, because it lacks signal amplification ability and relies on the base station to increase the transmission energy to meet the communication requirements. With the sufficient signal amplification ability, the power consumption performance of the fully-connected active RIS scheme is better than that of the fixed active-passive subarray hybrid RIS. For any distance between the base station and the users, the present invention shows the lowest system power consumption overhead advantage compared to other schemes.

[0046] The above are only the preferred embodiments of the present invention, and thus cannot limit the scope of implementation of the present invention. That is, equivalent changes and modifications made according to the scope of the patent application of the present invention and the content of the specification should still fall within the scope covered by the patent of the present invention.

Claims

1. Active-passive hybrid intelligent metasurface assisted low-power beamforming method, characterized by: The steps include: Step S1, designing an active-passive hybrid smart metasurface based on subarray division, obtaining a reflection coefficient matrix of the smart metasurface to construct a communication system model assisted by the smart metasurface, and determining expressions of the achievable total rate and total power consumption of the communication system model; Step S2, analyzing the reflection mode, amplitude and phase physical constraints of the smart metasurface, the power constraints of the base station and the smart metasurface, and the minimum rate constraints of each user in the communication system model, and constructing a first optimization problem of minimizing the total power consumption according to the beamforming design requirements, the first optimization problem involves the joint design of the base station transmit precoding vector and the smart metasurface reflection beamforming matrix; Step S3, transforming the first optimization problem constructed in step S2 into a second optimization problem, performing continuous relaxation processing on the second optimization problem to obtain a continuous variable optimization problem, and using an alternating optimization framework to iteratively update the coupling variables involved in the continuous variable optimization problem to obtain the final joint beamforming optimization result.

2. The active-passive hybrid intelligent metasurface assisted low-power beamforming method according to claim 1, characterized in that: In step S1, the smart super surface includes N is divided into M The reflection elements of the sub-arrays are each switchable between an active reflection working mode and a passive reflection working mode. The reflection matrix of the smart metasurface is expressed as ,in, express M The overall phase shift matrix of the smart metasurface when all sub-arrays are used as active sub-arrays. express M When all sub-arrays are used as passive sub-arrays, the overall phase shift matrix of the smart metasurface is: , represents the identity matrix, The dimension is A vector of all 1s, for M The reflection amplitude matrix when all sub-arrays are used as active sub-arrays, is the mode switching matrix, represents the transpose operation, Represents the Kronecker product operation.

3. The active-passive hybrid intelligent metasurface assisted low-power beamforming method according to claim 2, characterized in that: In step S1, the communication system model includes a deployment N t A base station with antennas, a N Smart metasurfaces with 3 reflective elements and K The achievable total rate expression of the communication system model is , the total power consumption expression of the communication system model is: ,in, is a single-antenna user set, Indicates the base station transmits to the user k The beamforming vector of the information symbol, Indicates base station and user k The effective channel between represents the channel state information between the base station and the smart metasurface, Indicates base station to user k The channel state information between Representing smart metasurfaces to users k The channel state information between represents the variance of thermal noise at the active sub-array of the smart metasurface, Indicates user k The variance of the additive white Gaussian noise at , represents the reflection matrix corresponding to the active sub-array part of the smart metasurface, is the power consumption expression of the base station, is the power consumption expression of the smart metasurface, is the inverse of the amplification efficiency of the base station's transmitting power amplifier, Tr(.) is the trace function, P a is the power consumption of a single antenna of the base station, P s is the static power consumption of the baseband processing unit of the base station, is the inverse of the amplification efficiency of the reflective power amplifier of the smart metasurface, P c represents the power consumption of the phase shifter circuit of the smart metasurface, P DC represents the DC operating power consumption of the reflective power amplifier of the smart metasurface, Represents the conjugate transpose operation.

4. The active-passive hybrid intelligent metasurface assisted low-power beamforming method according to claim 3, characterized in that: In step S2, the first optimization problem of minimizing the total power consumption is: ,in, and is an auxiliary parameter, P B is the maximum power load of the base station, P R is the maximum power load of the smart metasurface, γ min is the minimum achievable communication rate for the user, ρ min and ρ max They represent the lower and upper limits of the range of the gain factor of the reflective power amplifier, respectively. For the matrix No. n diagonal elements, This is a modulo operation.

5. The active-passive hybrid intelligent metasurface assisted low-power beamforming method according to claim 4, characterized in that: In step S3, the second optimization problem is expressed as ,in, , is the introduced semi-positive definite diagonal matrix, which represents the signal amplification power of the active sub-array, for M The connection relationship between the elements and the reflective power amplifier when all sub-arrays are active sub-arrays. is an auxiliary variable, which is defined as , is an auxiliary parameter, Represents the conjugate operation.

6. The active-passive hybrid intelligent metasurface assisted low-power beamforming method according to claim 5, characterized in that: In step S3, the continuous variable optimization problem is expressed as ,in, is the introduced exponential term.

7. The active-passive hybrid intelligent metasurface assisted low-power beamforming method according to claim 6, characterized in that: In step S3, the coupling variable includes the base station end transmitting precoding vector , Joint reflection amplitude and mode switching matrix of smart metasurface , phase shift matrix , using the alternating optimization framework to , and Iterative updates are performed, and when solving a certain coupling variable, the values ​​of the other coupling variables are kept unchanged. Each iteration selects the result of the previous iteration as the initial point, and the variable iterative update process is cyclic until the objective function value involved in the continuous variable optimization problem converges.

8. The active-passive hybrid intelligent metasurface assisted low-power beamforming method according to claim 7, characterized in that: In step S3, the combined reflection amplitude and mode switching matrix of the smart metasurface are fixed. and phase shift matrix , the continuous variable optimization problem is about the base station end transmitting precoding vector The sub-problem is expressed as , using continuous convex approximation technology, in t +1 iteration outputs results based on the previous iteration , constrain the non-convex minimum achievable rate Approximately , the converted base station transmit precoding vector Solve the sub-problem of to obtain the transmit precoding vector The optimization result of is an auxiliary parameter, , This is the real part operation.

9. The active-passive hybrid intelligent metasurface assisted low-power beamforming method according to claim 8, characterized in that: In step S3, the combined reflection amplitude and mode switching matrix of the smart metasurface are fixed. and the base station transmits the precoding vector , the continuous variable optimization problem is about the smart metasurface phase shift matrix The sub-problem is expressed as ,in Optimize vectors for stacking, , , vecd(.) is an operation to extract the diagonal elements of a matrix and organize them into column vectors, and the semi-positive matrix is the coupling parameter of the power constraint condition with respect to the conjugate quadratic variable, the semi-positive definite matrix , and They are the variables in the user's minimum achievable rate constraint. , and Coupling parameters of the conjugate quadratic form, , is the coupling parameter obtained based on the simplified active subarray structure, For a linear variable The linear parameters of , For vector No. n The element in t +1 iteration, output results based on the previous iteration , combined with the constant mode characteristics of the phase shift variable and using Taylor expansion, the phase shift vector of the smart metasurface is obtained The objective function and upper bound linear convex proxy functions of the subproblem are and , and solve the approximate linear programming problem to obtain the original intelligent metasurface phase shift matrix The optimization result of represents the maximum eigenvalue operation, , and They are all constant terms independent of the optimization variables obtained in the Taylor expansion approximation process.

10. The active-passive hybrid intelligent metasurface assisted low-power beamforming method according to claim 9, characterized in that: In step S3, the fixed base station transmits the precoding vector and smart metasurfaces , the continuous variable optimization problem is about the joint reflection amplitude and mode switching matrix of the smart metasurface The sub-problem is expressed as ,in, As the auxiliary variable introduced, the continuous convex approximation technique is used in the t The subproblem is approximated as a convex problem in the iteration , the convex problem is solved and iterated continuously to obtain the original joint reflection amplitude and mode switching matrix of the smart metasurface The optimization result of , where the semi-positive definite matrix is the coupling parameter of the active subarray signal power amplification term in the objective function, and the semi-positive definite matrix They are coupling parameters of the conjugate quadratic optimization variables in the user's minimum achievable rate constraint.

Images