Active and passive beamforming collaborative design method for non-orthogonal multiple access secure transmission based on smart reflector

Abstract

The present invention provides a collaborative design method for active and passive beamforming in non-orthogonal multiple access secure transmission based on intelligent reflecting surfaces, which belongs to the security field of non-orthogonal multiple access technology. In a downlink multi-input single-output non-orthogonal multiple access network, internal untrusted users are interested in the transmission information of secure users and are potential internal eavesdroppers of the system. The characteristics of the wireless channel reconstruction by intelligent reflecting surfaces are used to suppress the eavesdropping of secure user information by internal eavesdroppers, thereby ensuring the maximization of the security rate. Based on this model, the present invention provides a design method for jointly optimizing the active beam transmitted by the base station and the passive beam reflected by the reflecting surface. The method collaboratively optimizes the active and passive beams according to the different channel gains of the two users, while satisfying the service quality requirements of the internal eavesdropping users and the base station transmission power requirements and other constraints, ultimately achieving the maximization of the system security rate and improving the physical layer security of the non-orthogonal multiple access network.

Description

Technical Field

[0001] The present invention belongs to the security field of non-orthogonal multiple access (NOMA) technology, and more specifically, to a design method based on an intelligent reflecting surface (IRS)-assisted non-orthogonal multiple access network to enhance communication security through active and passive beamforming collaborative optimization. Background Art

[0002] With the increasing popularity of smart terminals and the continuous growth of demand for new mobile communication services, the demand for wireless transmission rate has exploded. In the face of a series of higher standards such as lower latency, higher speed, massive device interconnection, green and secure communication in the fifth-generation mobile communication network, it is crucial to improve the spectrum efficiency of existing available frequency bands. In recent years, non-orthogonal multiple access technology has received extensive research attention due to its efficient spectrum utilization. In the mainstream NOMA scheme, the transmitter adopts superposition coding and realizes power domain multiplexing by assigning different power levels to users. Generally speaking, in order to help the receiver eliminate multiple access interference between users, users with good channel quality will be allocated less power, while users with weak channel gain will be allocated more power. At the receiving end, multiple access interference is eliminated by serial interference cancellation technology. The receiver demodulates the signals of different users in sequence according to the power size of the received signal until the information of the user is demodulated. Therefore, by adopting power domain multiplexing and serial interference cancellation technology, NOMA can realize resource sharing in the time domain or frequency domain, thereby improving spectrum efficiency.

[0003] In actual communication systems, it is often difficult for NOMA transmitters to obtain perfect user channel state information, which poses a challenge to the power allocation of different users and may cause the user's multi-access interference to be unable to be completely eliminated, thereby affecting the user's signal detection and hindering the receiver's signal demodulation. In addition, users with strong channel gain within the NOMA system can easily obtain information about the paired users with weak channel gain, thereby threatening the confidential communications of users with weak channel gain. Although the use of NOMA can significantly improve the spectrum efficiency of wireless communications, the reliability and security of NOMA networks need to be further strengthened. The reasonable power allocation of different users at the transmitter and the effective suppression of interference and eavesdropping at the receiver are currently urgent issues to be solved.

[0004] Recently, intelligent reflective surfaces have attracted widespread attention in academia and industry due to their ability to reconstruct wireless transmission environments. IRS is an artificial metasurface composed of several low-cost passive reflective elements. It does not use active transmission modules, such as power amplifiers, and only reflects the received signal as a passive array. By adjusting the reflection coefficient of its surface reflective elements through software control, the received signal of the receiver can be purposefully enhanced, weakened, or even eliminated. Therefore, by properly deploying IRS in the NOMA network, it is expected to enhance or suppress the reception of base station transmission signals by certain users, enhance the transmission rate of confidential information, and thus improve the security of wireless communication networks.

[0005] To address the problem of user interference and eavesdropping in existing NOMA networks, the present invention designs an IRS-assisted downlink multi-input and single-output dual-user NOMA network. By utilizing the characteristics of the intelligent reflection surface that can reconstruct the wireless transmission channel environment, a method is proposed to jointly optimize the active beam transmitted by the base station and the passive beam reflected by the IRS, strengthen the demodulation of secure user information, suppress the interference and eavesdropping of internal eavesdroppers, thereby effectively reducing information leakage and maximizing the system security rate. Summary of the invention

[0006] The purpose of this invention is to solve the internal eavesdropping interference problem of NOMA network by using intelligent reflective surface. For the designed downlink multi-antenna transmission single antenna reception NOMA network, a method is proposed to maximize the security rate of secure users by using IRS assistance and active and passive beam joint optimization. The specific scheme is as follows Figure 1 As shown in the figure. By introducing IRS in the NOMA network, resetting the demodulation order of the user receiver, and jointly optimizing the active beam transmitted by the base station and the passive beam reflected by the IRS, the security rate is maximized, while meeting the Quality of Service (QoS) requirements of ordinary users to enhance the security of the NOMA network.

[0007] In order to solve the above technical problems, the technical solution adopted by the present invention is as follows:

[0008] A collaborative design method for active and passive beamforming in non-orthogonal multiple access secure transmission based on intelligent reflective surface, comprising the following steps:

[0009] The first step is to build a system model:

[0010] 1) In a NOMA network with multiple inputs and single outputs in the downlink, a base station (BS) with a fixed position and equipped with M antennas provides services to two full-duplex users, namely, ordinary user U1 and secure user U2. Among them, ordinary user U1 is interested in the information of secure user U2 and attempts to steal its information, and is regarded as an internal eavesdropper of the system. In order to suppress the internal eavesdropping of secure users by ordinary users and maximize the security rate of secure users, an IRS carrying N adjustable reflection elements is introduced into the system to improve the transmission security of the communication system. At this time, there is a direct link between the base station and the user (BS→U i , i∈{1,2}) and IRS reflection (BS→IRS→U i ) two communication links. Assume that BS to U i The transmission channel of the direct link follows the Rayleigh distribution, from BS to IRS, and from IRS to U i The transmission channel follows the Rice distribution.

[0011] In this model, the distance between BS and IRS is set to d sr , BS to U i The distance is expressed as IRS to U i The distance is expressed as For BS to U i The channel vector of the direct link is expressed as (Note: represents an M×1-dimensional complex vector). For the IRS reflected communication link, the channel vector from BS to IRS is expressed as IRS to U i The channel vector is expressed as Where L0 represents the path loss constant per unit distance (1m), α sr and Represent the corresponding path attenuation exponents respectively. It obeys a complex Gaussian distribution with a mean of 0 and a covariance matrix that is the unit matrix. and Respectively represent the line-of-sight (LoS) and non-line-of-sight (NLoS) transmission components. Similarly, Also includes LoS component 1 and NLoS components β sr and They represent the Ricean factors of the corresponding channels respectively.

[0012] 2) The BS transmission signal is set to x=w1s1+w2s2, where s1 and s2 both obey a normal distribution with a mean of 0 and a variance of 1, representing the transmission signals to the common user U1 and the secure user U2, respectively. is the corresponding precoding vector.

[0013] So, U i The received signal can be expressed as

[0014]

[0015] in, Represents a vector The transpose of , the same below. represents the IRS reflection coefficient matrix (Note: diag(.) represents the diagonalization operation of the vector.), a n ∈[0,1] represents the amplitude of the reflection coefficient of the nth element on the reflection surface. In order to simplify the model, this paper does not consider the influence of the reflection surface on the amplitude of the incident signal, that is, a n =1. θ n ∈[0,2π) represents the phase shift of the reflection coefficient of the nth element on the reflection surface, n∈{1,2,…,N}. i represents additive Gaussian white noise, with a mean of 0 and a variance of

[0016] Let v = [v1, v2, ..., v n ] H ,v n =Φ n,n ,make so So, BS to U i The equivalent communication link between i express

[0017]

[0018] Therefore, a (n+1)×1-dimensional IRS reflection vector is redefined

[0019] In previous NOMA, base stations usually allocate transmission power to users based on their channel gain. Generally, users with weaker channels are often allocated more power. At the receiving end, NOMA users decode using serial interference cancellation technology. For users with stronger channels, it is usually necessary to demodulate the weak user's signal first and then decode their own information. Although this helps eliminate multiple access interference, it also creates the risk of users with strong channel gains eavesdropping on weak user information. In order to suppress internal eavesdropping in the NOMA system, this method uses IRS to assist in configuring the channel and redesigns the user's demodulation order so that the receivers of both users demodulate U1's information first. Therefore, the signals received at the two user receivers must meet the following constraints

[0020]

[0021] For U1, the signal to interference plus noise ratio (SINR) at its receiver can be expressed as

[0022]

[0023] in, For U2, the receiver needs to first identify and eliminate the information of U1 in the superimposed signal before decoding its own signal. Therefore, the SINR of the U1 signal at the U2 receiver is expressed as

[0024]

[0025] After eliminating the information from BS to U1, the signal-noise ratio (SNR) of U2 at the receiver is expressed as

[0026]

[0027] In addition, since U1 is interested in U2’s information, its SNR for eavesdropping on the secure user is

[0028]

[0029] According to information theory, the transmission rate of U1 is expressed as

[0030]

[0031] The transmission rate of U2 is expressed as

[0032]

[0033] The eavesdropping rate of U1 on U2 is

[0034]

[0035] Therefore, the safe rate of U2 is expressed as

[0036]

[0037] in,

[0038] The second step is to simplify the objective function and list the optimization problem:

[0039] According to the system settings, the maximum security rate optimization problem of the IRS-assisted NOMA system against internal eavesdropping is proposed. The specific optimization problem can be expressed as

[0040]

[0041] Among them, C1 represents the QoS requirement constraint of U1, is its QoS requirement value, C2 represents the power constraint of the transmitted signal, P max is the maximum transmission power of the BS, C3 ensures that the reflection coefficient modulus corresponding to each reflection element on the IRS is a constant 1, and C4 is the power constraint of the signal on the two user receivers, so that both user receivers demodulate the information of U1 first. Mutual coupling, objective function C s As well as the non-convex constraints, problem (12) is difficult to solve directly. Therefore, an alternating optimization algorithm is used to solve it.

[0042] The third step is to solve the optimization problem:

[0043] In order to solve the coupling between variables, the original optimization problem is divided into two sub-problems. Using the relevant theories of convex optimization, the two sub-problems are transformed into convex optimization problems for solution.

[0044] 1) Fixed IRS reflection vector Optimize the active beam vectors w1 and w2 transmitted by the BS.

[0045] Let the matrix matrix and matrix Where W1 and W2 satisfy the rank of 1. The corresponding transformation is as follows

[0046]

[0047] Among them, Tr() means finding the trace of the matrix, then the objective function is expressed as

[0048]

[0049] Therefore, the first sub-problem of problem (12) can be summarized as

[0050]

[0051] For the objective function (14), it can be equivalent to Arrangement available

[0052]

[0053] It is known that the convex function f(t) = -tx + lnt + 1, x is a constant and greater than zero, when t = 1 / x, Similarly, we introduce a non-negative variable t b and a convex function

[0054]

[0055] in It can be deduced that when t b =1 / x1, function (17) has a maximum value

[0056]

[0057] Ignoring the constant ln2, (18) can be equivalent to the objective function C s For non-convex constraints C1 and C2, they can be organized into convex expressions as follows:

[0058]

[0059] For C5 of (15), the rank 1 constraint is non-convex and can be removed using the semi-definite relaxation (SDR) algorithm. Therefore, problem (15) can be finally transformed into

[0060]

[0061] At this time, the optimization problem (20) is b are all convex problems. Given the initial W2, t b The optimal solution is

[0062]

[0063] fixed Solve the optimization problem (22).

[0064]

[0065] Use a convex optimization toolbox, such as MATLAB's CVX convex tool, to solve (22). If the objective function (14) does not converge, substitute the currently solved W2 into (21) to update Continue to optimize and solve (22) until convergence. At this time, the obtained W1 and W2 are the optimal active beam matrices. If rank(W1) = rank(W2) = 1, the corresponding optimal active beam vectors w1 and w2 can be directly obtained through eigenvalue decomposition. Otherwise, the approximate value of the active beam vector is obtained by Gaussian randomization.

[0066] 2) Fix the active beam vector and optimize the phase shift vector of IRS reflection

[0067] Let the matrix matrix and matrix So the objective function is expressed as

[0068]

[0069] The second sub-problem of the optimization problem (12) can be expressed as

[0070]

[0071] Similar to the first sub-problem, for the objective function, a convex function ψ(V,z b ) and the new variable z b >0,

[0072]

[0073] in It can be deduced that when z b =1 / x2,

[0074]

[0075] Ignoring the constant ln2, we can use the convex function ψ(V,z b ) to approximate the non-convex C s The QoS constraint on U1 in (24) can be organized into the following convex form:

[0076]

[0077] Using SDR again, we remove the rank 1 constraint rank(V)=1, and transform problem (24) into a convex optimization problem.

[0078] For the function ψ(V,z b ), given an initial variable V, z b The corresponding optimal solution can be expressed as Use the CVX convex toolbox to solve the following optimization problem

[0079]

[0080] If the objective function (23) has not converged, the current solution V is updated Continue to solve (28) until convergence. Finally, if the solution satisfies rank(V) = 1, the passive beam vector of IRS reflection can be obtained by eigenvalue decomposition: Otherwise, the passive beam vector is approximated with the aid of Gaussian randomization.

[0081] At this time, problem (12) is divided into two sub-problems and solved separately, but the original problem may not converge. By fixing the active beam vector and the passive beam vector respectively, the two sub-problems are optimized alternately until their objective functions converge and the safety rate is maximized. The corresponding active and passive beam vectors are the optimal solutions.

[0082] 4) The present invention proposes a method for joint optimization of active and passive beams to maximize the safety rate of the NOMA system through alternating iterations. For the proposed optimization problem, in order to solve the mutual coupling between the active and passive beam variables, the original problem is split into two sub-problems and solved separately, which greatly reduces the complexity of solving the problem. When solving each sub-optimization problem, auxiliary variables and a convex function are introduced to convexly approximate the objective function, and the SDR is used to relax the rank 1 constraint, so that the sub-problem is converted into a convex optimization problem, which can be directly solved using the convex optimization CVX tool. By solving the two sub-problems multiple times through alternating iterations until the objective function of the original optimization problem converges, the maximum safety rate can be guaranteed, and the corresponding optimal active and passive beam vector values ​​can be obtained.

[0083] The beneficial effects of the present invention are: by providing the locations of specific base stations, internal eavesdropping users, secure users and IRS, the security rate of the NOMA communication system is maximized through the joint optimization design of active and passive beamforming. The present invention provides a reference method for how to use IRS to assist the NOMA network to effectively suppress internal eavesdropping and improve the security of wireless communication transmission. BRIEF DESCRIPTION OF THE DRAWINGS

[0084] Figure 1 It is a downlink multiple-input single-output wireless communication NOMA network system diagram;

[0085] Figure 2 It is the location setting diagram of each network node;

[0086] Figure 3 is a comparison chart of the safe rate of user 2 as N changes under different conditions;

[0087] Figure 4 The transmission rate of user 1 and the security rate of user 2 vary under different conditions. Comparison of changes; Figure (a) represents user 1, and Figure (b) represents user 2;

[0088] Figure 5 is the safe rate of user 2 under different communication links with P max Comparative graph of changes;

[0089] Figure 6 This is a diagram showing the impact of two deployment positions of the user relative to the base station on the security rate of user 2. DETAILED DESCRIPTION

[0090] The present invention proposes a collaborative design method for active and passive beamforming in non-orthogonal multiple access secure transmission based on an intelligent reflective surface. In order to better understand the above technical solution, the experimental results are described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0091] An implementation of a secure communication method for non-orthogonal multiple access networks with active and passive beamforming collaborative optimization is proposed. Consider a dual-user NOMA communication network with multiple inputs and single outputs in the downlink, as shown in the attached figure. Figure 1 shown.

[0092] The first step is to build a system model: given the location of each network node, the specific settings are as shown in the attached Figure 2 Assume that the number of base station transmitting antennas is M = 4, and its position coordinates are (0,0,5), (distance unit: meter, the same below), the position coordinates of ordinary user U1 are (5,18,0), and the coordinates of secure user U2 are (0,40,0). Figure 2 Here, we consider two cases where the IRS is deployed near the BS and near U2, which are located at (0,10,2) and (0,35,2) respectively. The maximum transmit power at the base station is set to P max =30dBm, and the QoS requirement of U1 is set to Other fixed parameters are set as follows: L0 = -30dB, Under the condition of satisfying the transmission power limit, the base station randomly generates the superposition signal x=w1s1+w2s2 transmitted to the two users, and initializes the reflection coefficient vector of the IRS reflection element but According to the given parameters and the characteristics of Rayleigh channel and Rice channel, BS to IRS, U i , and IRS to U i The transmission channel of the two users can be expressed as formula (1). At the receiving end, in order to suppress the internal eavesdropping of the NOMA system, the IRS is used to assist in configuring the channel so that the receivers of the two users demodulate the information of U1 first.

[0093] The second step is to list the optimization problem: According to the specific parameters and constraints of the designed NOMA system, list the optimization problem (12).

[0094] The third step is to solve the optimization problem: fix the active beam vectors w1, w2 and the passive beam vectors Split problem (12) into two sub-problems (15) and (24) and solve them separately. For problem (15), introduce the convex function and auxiliary variable t b The objective function is transformed into a convex optimization problem (22) through SDR. For problem (24), the convex function ψ(V,z b ) and auxiliary variable z b , which is transformed into a convex optimization problem (28) through SDR. Using the CVX convex optimization tool, we solve problems (22) and (28) respectively and alternately optimize the active beam vectors w1, w2 and the passive beam vectors The security rate is maximized until convergence, and the corresponding active and passive beam vectors are finally obtained to complete the NOMA network security design. On this basis, the following studies the changes in the security rate of the NOMA system with the increase in the number of IRS reflection elements and the QoS requirements of ordinary users, and compares the security rate under the same conditions without IRS assistance or without optimizing the IRS reflection vector, thereby verifying the feasibility of the IRS-assisted NOMA security transmission scheme and obtaining a better IRS deployment scheme.

[0095] The solution proposed in this invention to maximize the NOMA security performance varies with the reflection element. Figure 3 As shown in the figure, the security rate results under the random IRS reflection vector scheme are also given as a comparison. From the experimental results, it can be seen that the security rate of the optimized IRS scheme is significantly higher than that of the random IRS scheme, and as the reflection element N of the IRS increases from 20 to 60, the security rate of the optimized IRS scheme gradually increases, while the results of the random IRS scheme remain basically unchanged. It is worth noting that when the IRS is deployed near U2, its security rate is much higher than when the IRS is deployed near the BS.

[0096] Figure 4 The results of comparing the U1 transmission rate and U2 security rate as the QoS requirement value of U1 increases from 1 bit / s / Hz to 5 bit / s / Hz when the number of IRS reflection elements is 20, 40, and 60 respectively. Figure 4 It can be seen that when the number of IRS reflection elements N = 20, 40 and 60, the transmission rate of U1 under the two IRS position deployments is the same and equal to its QoS requirement value. When the QoS requirement increases, the transmission rate of U1 increases synchronously, while the security rate of U2 decreases. This is because the increase in U1's QoS requirement will increase its transmission power demand. When the total BS transmission power is constant, the power allocated to the security user U2 will decrease. In addition, Figure 4 The results again show that the system can achieve higher security rates by deploying the IRS near the security users rather than near the BS.

[0097] To further determine the security performance of the proposed scheme, we compare the security rates under three different transmission link benchmarks, namely, “direct link only”, “reflection link only”, and “direct + reflection link”. Figure 5The relationship between the safety rate of U2 and the total BS transmission power under different communication link transmission is shown. The safety rate of all benchmark experiments increases with the increase of BS transmission power. The safety rate of the proposed "direct + reflection link" scheme is better than the other two schemes, and even the safety performance of the "direct + reflection link" scheme when N = 30 is better than the "reflection link only" scheme when N = 50. For the "reflection link only" scheme, under the same transmission power, the safety rate of U2 increases with the increase of the number of reflection elements, but this scheme is not necessarily better than the "direct link only" scheme when the IRS has fewer reflection elements. This shows that active and passive beam collaborative optimization has a positive significance for improving the security of the system, and the IRS with large-scale reflection elements has a very significant auxiliary effect on wireless communication transmission when only the reflection link is transmitted.

[0098] Figure 6 The impact of IRS on system security performance when two users are at different distances from the base station is compared. The position of the BS remains unchanged, and the position of the IRS is reset to (2.5, 25, 1). Below: The coordinates of U1 are set to (5,20,0) and the coordinates of U2 are set to (0,40,0), which means that U1 is closer to BS. Bottom: The coordinates of U1 are set to (5,40,0) and the coordinates of U2 are set to (0,20,0), which means that U2 is closer to BS. Figure 6 is the security rate when the number of reflection elements increases from 20 to 60 under the two position relationships, and the security rate results of the scheme without deploying IRS are compared. As the number of reflection elements increases, the security rates of the two position schemes with IRS deployment gradually increase and are significantly better than the scheme without IRS deployment. In the case of more challenging security users with worse channel conditions than internal eavesdroppers, that is, The scheme poses a greater challenge to the secure transmission of the system. However, by deploying IRS with a large number of reflective elements, the NOMA network can still effectively suppress internal eavesdropping and ensure communication security.

Claims

1. A collaborative design method for active and passive beamforming in non-orthogonal multiple access secure transmission based on smart reflective surface, characterized in that: The following steps are involved: The first step is to build a system model: 1) In a NOMA network with multiple inputs and single outputs downlink, a fixed-position base station BS equipped with M antennas provides services for two full-duplex users, namely, ordinary user U1 and secure user U2. Among them, ordinary user U1 is interested in the information of secure user U2 and attempts to steal its information, and is regarded as an internal eavesdropper of the system. In order to suppress the internal eavesdropping of secure users by ordinary users and maximize the security rate of secure users, an IRS carrying N adjustable reflection elements is introduced into the system to improve the transmission security of the communication system. At this time, there is a direct link BS→U between the base station and the user. i , i∈{1,2} and IRS reflects BS→IRS→U i Two communication links; assuming BS to U i The transmission channel of the direct link follows the Rayleigh distribution, from BS to IRS, and from IRS to U i The transmission channel follows the Rice distribution; In this model, the distance between BS and IRS is set to d sr , BS to U i The distance is expressed as IRS to U i The distance is expressed as For BS to U i The channel vector of the direct link is expressed as For the IRS reflected communication link, the channel vector from BS to IRS is expressed as IRS to U i The channel vector is expressed as Where L0 represents the path loss constant per unit distance, α sr and They represent the corresponding path attenuation index respectively; It obeys complex Gaussian distribution, with mean 0 and covariance matrix is ​​the unit matrix; and Respectively represent the line-of-sight (LoS) and non-line-of-sight (NLoS) transmission components; similarly, Also includes LoS component and NLoS components β sr and They represent the Ricean factors of the corresponding channels respectively; 2) The BS transmission signal is set to x=w1s1+w2s2, where s1 and s2 both obey a normal distribution with a mean of 0 and a variance of 1, representing the transmission signals to the common user U1 and the secure user U2, respectively. is the corresponding precoding vector; U i The received signal is represented as: in, Represents a vector The transpose of , the same below; represents the IRS reflection coefficient matrix, a n ∈[0,1] represents the amplitude of the reflection coefficient of the nth element on the reflection surface, without considering the influence of the reflection surface on the amplitude of the incident signal; θ n ∈[0,2π) represents the phase shift of the reflection coefficient of the nth element on the reflection surface, n∈{1,2,…,N}; n i represents additive Gaussian white noise, with a mean of 0 and a variance of Let v = [v1, v2, ..., v n ] H ,v n =Φ n,n ,make so So, BS to U i The equivalent communication link between i express: Therefore, we redefine an (n+1)×1 dimensional IRS reflection vector Using IRS to assist in channel configuration, the demodulation order of users is redesigned so that the receivers of both users demodulate the information of U1 first. Therefore, the signals received by the receivers of the two users must meet the following constraints: For U1, the signal-to-noise-interference ratio SINR at its receiver is expressed as: in, For U2, the receiver needs to first identify and eliminate the information of U1 in the superimposed signal before decoding its own signal; therefore, the SINR of the U1 signal at the U2 receiver is expressed as: After eliminating the information from BS to U1, the signal-to-noise ratio (SNR) of U2 at the receiver is expressed as: In addition, since U1 is interested in U2's information, its SNR for eavesdropping on the secure user is: According to information theory, the transmission rate of U1 is expressed as: The transmission rate of U2 is expressed as: The eavesdropping rate of U1 on U2 is: Therefore, the safe rate of U2 is expressed as: in, The second step is to simplify the objective function and list the optimization problem: According to the system settings, the maximum security rate optimization problem of the IRS-assisted NOMA system against internal eavesdropping is proposed. The specific optimization problem is expressed as: Among them, C1 represents the QoS requirement constraint of U1, is its QoS requirement value, C2 represents the power constraint of the transmitted signal, P max is the maximum transmit power of the BS, C3 ensures that the reflection coefficient modulus corresponding to each reflection element on the IRS is a constant 1, and C4 is the power constraint of the signal on the two user receivers, so that both user receivers demodulate the information of U1 first; The third step is to use the alternating optimization algorithm to solve the optimization problem: The original optimization problem is divided into two sub-problems. Using the relevant theories of convex optimization, the two sub-problems are transformed into convex optimization problems for solution. 1) Fixed IRS reflection vector Optimize the active beam vectors w1 and w2 transmitted by the BS; Let the matrix matrix and matrix Where W1 and W2 satisfy the rank of 1; the corresponding transformation is as follows: Among them, Tr() means finding the trace of the matrix, then the objective function is expressed as: Therefore, the first sub-question of question (12) can be summarized as follows: For the objective function (14), it can be equivalent to After finishing, we can get: Similarly, we introduce a non-negative variable t b and a convex function but: For non-convex constraints C1 and C2, they can be organized into convex expressions as follows: For C5 of (15), the rank 1 constraint is non-convex and can be removed by using the semidefinite relaxation SDR algorithm; therefore, problem (15) is finally transformed into: At this time, the optimization problem (20) is b All of them are convex problems; given the initial W2, t b The optimal solution is: fixed Solve the optimization problem (22); Use the convex optimization toolbox to solve formula (22); if the objective function (14) does not converge, substitute the currently solved W2 into (21) to update Continue to optimize and solve (22) until convergence; at this time, the solved W1 and W2 are the optimal active beam matrices; if rank(W1)=rank(W2)=1, the corresponding optimal active beam vectors w1 and w2 can be directly obtained by eigenvalue decomposition, otherwise, the approximate value of the active beam vector is obtained by Gaussian randomization; 2) Fix the active beam vector and optimize the phase shift vector of IRS reflection Let the matrix matrix and matrix So the objective function is expressed as: The second sub-problem of the optimization problem (12) can be expressed as: Similar to the first sub-problem, for the objective function, a convex function ψ(V,z b ) and the new variable z b >0, then: in The QoS constraint on U1 in (24) can be organized into the following convex form: Using SDR again, we remove the rank 1 constraint rank(V)=1, and transform problem (24) into a convex optimization problem; For the function ψ(V,z b ), given an initial variable V, z b The corresponding optimal solution can be expressed as Use the CVX convex toolbox to solve the following optimization problem: If the objective function (23) has not converged, the current solution V is updated Continue to solve (28) until convergence; finally, if the solution satisfies rank(V) = 1, the passive beam vector of IRS reflection can be obtained by eigenvalue decomposition: Otherwise, the passive beam vector is approximated with the help of Gaussian randomization; At this time, problem (12) is divided into two sub-problems and solved separately, but the original problem may not converge. By fixing the active beam vector and the passive beam vector respectively, the two sub-problems are optimized alternately until their objective functions converge and the safety rate is maximized. The corresponding active and passive beam vectors are the optimal solutions.

Images