Receiving end rate optimization method of RIS auxiliary relay cooperative communication system

Abstract

A receiving end rate optimization method of an RIS auxiliary relay cooperative communication system comprises the following steps: establishing an RIS auxiliary relay cooperative downlink multiple-input multiple-output communication system model which comprises a base station, an intelligent reflector RIS, a relay and a receiving end, and establishing a maximum receiving end rate optimization problem under the system model; an original problem is converted into two sub-problems to be solved by adopting an alternating optimization method, namely, an active emission covariance matrix and passive beam forming, in an active beam forming optimization problem, an optimal solution is solved by adopting an SDR and a water injection algorithm, a reflection coefficient of an RIS is optimized by adopting a particle swarm algorithm, the condition that the position of a receiving end is changeable is fully considered, and the optimal solution is obtained. And large-area receiving equipment and users can be covered. According to the alternative optimization method, the constructed problem is effectively solved in an iteration mode, the advantages of the RIS technology and the relay technology are combined, and the rate of the receiving end is remarkably improved.

Description

Technical Field

[0001] The present invention relates to the fields of unmanned aerial vehicle communication and deep reinforcement learning, and in particular to a receiving end rate optimization method for a RIS-assisted relay cooperative communication system. Background Art

[0002] Reflective intelligent surface (RIS) is a passive, low-cost, easy-to-deploy communication technology, which consists of multiple low-cost, low-power reflective units. By adjusting the reflection amplitude, phase and other parameters of the incident signal, RIS can change the unfavorable wireless propagation environment and improve the system's high-efficiency transmission.

[0003] Relay-assisted intelligent reflector-assisted wireless communication is a novel communication technology, and its research is gradually deepening. The RIS-assisted relay cooperative communication model combines the advantages of relay technology and RIS technology, further improving the coverage, transmission efficiency and reliability of the wireless communication system, and providing target users with more efficient, stable and reliable communication services. Relays and RIS collaborate effectively to optimize the transmission path and enhance signal quality, maximize the performance of the communication system, and show better performance in terms of energy efficiency. However, most studies have considered systems with IRS or relays, and focused on single-antenna single-user or single-antenna multi-user systems, with insufficient receiving end rate. Summary of the invention

[0004] The purpose of the present invention is to provide a receiving end rate optimization method for a RIS-assisted relay cooperative communication system, which enhances the transmission rate and signal strength of the receiving end by jointly optimizing the transmission covariance matrix of the base station and the relay and the reflection phase shift of the RIS element.

[0005] In order to achieve the above object, the technical solution adopted by the present invention is: a receiving end rate optimization method of a RIS-assisted relay cooperative communication system, comprising the following steps: S1. Establish a downlink multiple-input multiple-output communication system model with RIS-assisted relay coordination, including a base station, an intelligent reflective surface RIS, a relay and a receiving end, wherein the base station has M antennas, the RIS has N reflective elements, the relay has L antennas, and the receiving end has K antennas; S2. Assume that the channel matrix from the base station to the receiving end is The channel matrix from the relay to the receiver is The channel coefficient matrix from the base station to the RIS is The channel coefficient matrix from RIS to the receiver is It is expressed as: In formula (1) and (2), β 1 , β 2 is the Rice fading coefficient, For line-of-sight links, It is a non-line-of-sight link; The reflection coefficient of S3 and RIS at the receiving end is: In formula (3), represents the phase shift of the reflected signal of the nth element of the RIS, β n ∈[0,1] represents the amplitude reflection coefficient; The transmission signal vectors of S4, base station and relay are The covariance matrices of the transmitted signals of the base station and the relay are defined as And Q>0, F>0; The signal vectors received by the receiving end from the base station and the relay are: y 1 =(H 2 ΦH 1 +H dir )x 1 +n 1 =Hx 1 +n 1 (4); y 2 =hx 2 +n 2 (5); In formula (4) and (5), They respectively indicate that the mean value received by the user is zero and the variance is Additive white Gaussian noise (AWGN); The total signal received by the receiver is: y=(H 2 ΦH 1 +H dir )x 1 +hx 2 +n (6); In formula (6), n = n 1 +n 2 , is the noise vector, Φ represents the reflected beamforming of the signal by RIS; The data transmission rate at the receiving end is: S4. By jointly optimizing Q, F and Φ, the transmission rate of the receiving end is maximized. The solution to the maximum rate problem is optimized as follows: In formula (8), P s , P r are the average transmission power constraints of the base station and the relay, respectively, Ε{||x 1 || 2}≤P s , Ε{||x 2 || 2}≤P r ; Solve through the iterative algorithm of alternating optimization. First, the optimal solution of Q and F is obtained by giving Φ, and then the optimal solution of Q and F is used as the given Q and F to obtain the optimal solution of Φ, and the cycle is repeated alternately. After each round of iteration, the transmission rate after the current round of iteration is compared with the transmission rate after the previous round of iteration using the fitness function. When the comparison result reaches the preset threshold, the iteration is stopped, and the maximum rate of the receiving end is obtained. S41. Given Φ, find the optimal solution of Q and F Define Q = qq H , F = ff H , The optimization problem is transformed into: Definition H H =H H H,h h =h H h, optimize the objective function R u is equivalent to: Combining equations (9) and (10), the optimization problem is transformed into: The optimization problem is then simplified to: The semi-positive definite programming (SDR) algorithm is used for optimization, and then the eigenvalue decomposition (EVD) is used to obtain the optimal solution under given phase conditions. The process of solving the optimal Q is as follows: H = U 1 Λ 1 V 1 H is the truncated singular value decomposition of H, where D = rank (H) ≤ min (M, K) represents the maximum number of data streams that H can transmit; then the power is allocated according to the water injection algorithm to obtain the optimal power amount p for each data stream. 1i , and then get the optimal Q; The power p of each data stream 1i According to the Lagrange multiplier method, the optimization problem is converted to: In formula (13), λ is the Lagrange multiplier; P 1i The derivative is make get: In formula (14), The optimal Q is: Q * =V 1 diag{p 11 ,...,p 1D}V 1 H (15); The optimal F solution process is as follows: as the truncated singular value decomposition of h, where Dd=rank(d)≤min(L,K) represents the maximum number of data streams that h can transmit; then the power is allocated according to the water injection algorithm to obtain the optimal power amount p for each data stream 2i , and then the optimal F is: S42. Given Q and F, find the optimal solution of Φ The optimization problem is transformed into: The particle position is set to a set of solutions of the phase shift vector of RIS, the particle swarm is initialized, and the initial velocity v and position x are randomly generated, which can be expressed as: v p (i+1)=wv p (i)+c 1 r 1 (pbest p (i)-x p (i))+c 2 r 2 (gbest p (i)-x p (i)) (18); x p (i+1)=x p (i)+v p (i+1) (19); In formulas (18) and (19), w represents the weight, c 1 、c 2 represents the learning factor, r 1 、r 2represents a random number that follows a uniform distribution between [0,1], and the particle position x is the phase shift vector, constrained between [0,2π); Then, the objective function is used to calculate the fitness value of each particle, find the individual optimal pbest and the global optimal gbest, and adjust the next flight speed and position. During the iteration process, the fitness function is used to evaluate each particle. When the maximum number of iterations is reached, the optimal fitness value, that is, the maximum rate Φ, is obtained.

[0006] Preferably, in s42, a nonlinear function is constructed to adjust the weight w, and a random number is added to the particle velocity update method, so as to improve the dynamic search capability of the algorithm by dynamically adjusting the motion trajectory of the particles. The adjusted weight w and particle velocity v are expressed as: In formula (20) and (21), w max 、w max is the maximum and minimum value of the weight, Iter is the maximum number of iterations, and rand() represents a random number that follows a uniform distribution between [0,1]. is the adjustment threshold with a value of [0,1], It is expressed as: In formula (22), are the maximum adjustment threshold and the minimum adjustment threshold.

[0007] Compared with the prior art, the present invention has the following beneficial effects: The RIS and relays of the present invention cooperate to assist in wireless communication transmission, fully considering the variable position of the receiving end, and can cover a large area of ​​receiving devices and users. Due to the complexity of the original optimization problem, the original problem is decoupled into two sub-problems for solution, and the alternating optimization method is adopted to effectively solve the constructed problem in an iterative manner. The SDR and water injection algorithm are used to process the transmission covariance matrix, and the PSO algorithm is used to obtain the optimal solution of the RIS phase shift, which combines the advantages of RIS technology and relay technology and significantly improves the rate of the receiving end. DETAILED DESCRIPTION

[0008] The embodiment provides a receiving end rate optimization method of a RIS-assisted relay cooperative communication system, comprising the following steps: S1. A downlink MIMO communication system model with RIS-assisted relay collaboration is established, including a base station, an intelligent reflecting surface RIS, a relay and a receiving end, wherein the base station has M antennas, the RIS has N reflecting elements, the relay has L antennas, and the receiving end has K antennas.

[0009] S2. Assume that the channel matrix from the base station to the receiving end is The channel matrix from the relay to the receiver is The channel coefficient matrix from the base station to the RIS is The channel coefficient matrix from RIS to the receiver is It is expressed as: In formula (1) and (2), β 1 , β 2 is the Rice fading coefficient, For line-of-sight links, It is a non-line-of-sight link.

[0010] The reflection coefficient of S3 and RIS at the receiving end is: In formula (3), represents the phase shift of the reflected signal of the nth element of the RIS, β n ∈[0,1] represents the amplitude reflection coefficient.

[0011] The transmission signal vectors of S4, base station and relay are The covariance matrices of the transmitted signals of the base station and the relay are defined as And Q>0, F>0.

[0012] The signal vectors received by the receiving end from the base station and the relay are: y 1 =(H 2 ΦH 1 +H dir )x 1 +n 1 =Hx 1 +n 1 (4); y 2 =hx 2 +n 2 (5); In formula (4) and (5), They respectively indicate that the mean value received by the user is zero and the variance is Additive White Gaussian Noise (AWGN).

[0013] The total signal received by the receiver is: y=(H 2 ΦH 1 +H dir )x 1 +hx 2 +n (6); In formula (6), n = n 1 +n 2 , is the noise vector, and Φ represents the reflected beamforming of the signal by RIS.

[0014] According to Shannon's formula, the data transmission rate at the receiving end is:

[0015] S4. By jointly optimizing Q, F and Φ, the transmission rate of the receiving end is maximized. The solution to the maximum rate problem is optimized as follows: In formula (8), P s , P r are the average transmission power constraints of the base station and the relay, respectively, Ε{||x 1 || 2}≤P s , Ε{||x 2 || 2}≤P r .

[0016] Since the constraints of the optimization problem are highly coupled and present a non-convex problem, an iterative algorithm based on alternating optimization is considered to effectively optimize one of the variables while keeping the other variables fixed.

[0017] First, Φ is given to obtain the optimal solution of Q and F. Then, the optimal solution of Q and F is used as the given Q and F to obtain the optimal solution of Φ, and the cycle is repeated alternately.

[0018] After each round of iteration, the transmission rate after this round of iteration is compared with the transmission rate after the previous round of iteration using the fitness function. When the comparison result reaches a preset threshold, the iteration is stopped, and the maximum rate of the receiving end is obtained.

[0019] S41. Given Φ, the process of finding the optimal solution of Q and F is: Define Q = qq H , F = ff H , The optimization problem is transformed into:

[0020] Definition H H =H H H,h h =h H h, optimize the objective function R u is equivalent to:

[0021] Combining equations (9) and (10), the optimization problem is transformed into:

[0022] Since log2 is a monotonic function, the optimization problem is simplified to:

[0023] This problem is a convex semidefinite programming problem (SDP). We use the existing convex optimization solver, namely, the semidefinite programming SDR algorithm, to optimize it. Then, we consider the rank-one problem just ignored and use the eigenvalue decomposition (EVD) to obtain the optimal solution Q under the given phase condition. * , F * .

[0024] The water injection algorithm to optimize Q and F under given Φ conditions is as follows: Let H = U 1 Λ 1 V 1 H is the truncated singular value decomposition of H, where D = rank (H) ≤ min (M, K) represents the maximum number of data streams that H can transmit; then the power is allocated according to the water injection algorithm to obtain the optimal power amount p for each data stream. 1i , and then get the optimal Q.

[0025] The power p of each data stream 1i According to the Lagrange multiplier method, the optimization problem is converted to: In formula (13), λ is the Lagrange multiplier.

[0026] P 1i The derivative is make get: In formula (14),

[0027] The optimal Q is: Q * =V 1 diag{p 11 ,...,p 1D}V 1 H (15).

[0028] The optimization of the relay transmission covariance matrix F is similar to the optimization of Q above. as the truncated singular value decomposition of h, where Dd=rank(d)≤min(L,K) represents the maximum number of data streams that h can transmit; then the power is allocated according to the water injection algorithm to obtain the optimal power amount p for each data stream 2i, and then the optimal F is:

[0029] S42. Given Q and F, the process of finding the optimal solution of Φ is: The goal is to optimize the reflection coefficient of the RIS phase shift given the transmission covariance matrix Q and F of the BS and relay. The PSO algorithm is used to solve the optimization problem, which is transformed into:

[0030] The particle position is set to a set of solutions of the phase shift vector of RIS, the particle swarm is initialized, and the initial velocity v and position x are randomly generated, which can be expressed as: v p (i+1)=wv p (i)+c 1 r 1 (pbest p (i)-x p (i))+c 2 r 2 (gbest p (i)-x p (i)) (18); x p (i+1)=x p (i)+v p (i+1) (19); In formulas (18) and (19), w represents the weight, c 1 、c 2 represents the learning factor, r 1 、r 2 represents a random number that follows a uniform distribution between [0,1], and the particle's position x is a phase shift vector constrained between [0,2π).

[0031] Then, the objective function is used to calculate the fitness value of each particle, find the individual optimal pbest and the global optimal gbest, and adjust the next flight speed and position. During the iteration process, the fitness function is used to evaluate each particle. When the maximum number of iterations is reached, the optimal fitness value, that is, the maximum rate Φ, is obtained.

[0032] Since the particles’ ability to utilize the surrounding environment becomes weaker and weaker during the evolution process, that is, the search ability of the PSO algorithm is gradually weakened, and it is easy to fall into the local optimum. Therefore, a nonlinear function is constructed to adjust the weight w, and random numbers are added to the particle speed update method. By dynamically adjusting the particle’s motion trajectory, the dynamic search ability of the algorithm is improved. The adjusted weight w and particle speed v are expressed as: In formula (20) and (21), w max 、wmax is the maximum and minimum value of the weight, Iter is the maximum number of iterations, and rand() represents a random number that follows a uniform distribution between [0,1]. is the adjustment threshold with a value of [0,1], It is expressed as: In formula (22), are the maximum adjustment threshold and the minimum adjustment threshold.

[0033] The performance of the method is verified by numerical simulation results. The parameters are initialized as M=4, N=40, L=4, K=2, P s =1,P r =1, noise power sigma = 1, iterative convergence accuracy is ε = 10 -3 , the spatially correlated Rice fading model is used to model the channel, the Rice factor is set to k = 1, the population size is par = 50, the maximum number of iterations Iter = 100, c 1 =c 2 =2,w min =0.4, w max =0.95,

[0034] Two benchmark schemes are set up for comparison with this embodiment: (1) RIS with random phase; (2) without RIS and relay. At the same time, the alternating optimization algorithm based on SDR and PSO and the alternating optimization algorithm based on PSO with water injection algorithm are simulated respectively.

[0035] Under the alternating optimization algorithm based on SDR and PSO, as the number of RIS elements continues to increase, the alternating optimization algorithm of this embodiment has better performance than other solutions, and the RIS and relay jointly assist system model can effectively improve the maximum rate. s =5, when the number of BSs increases from 4 to 6, the achievable rate also increases. The more antennas there are, the greater the channel gain. When the number of BSs remains unchanged, P s When increasing from 5 to 10, the achievable rate increases significantly.

[0037] Under the alternating optimization algorithm based on the water injection algorithm and PSO, as the number of RIS elements continues to increase, the alternating optimization algorithm with relays has better performance than other schemes, and the RIS and relay jointly assist the system model to effectively improve the maximum rate. As the number of relay antennas or relay transmission power increases, the achievable rate also increases. The more antennas there are, the greater the transmission power, the greater the channel gain, and the stronger the signal strength received by the receiving end, so the achievable rate is greater.

[0038] In summary, the present invention adopts RIS and relay nodes to assist transmission at the same time, and jointly optimizes the active transmission covariance matrix of the base station and relay node and the passive beamforming matrix at the RIS to maximize the rate of the receiving end. Due to the coupling and non-convexity of the optimization problem, the original problem is decoupled into two sub-problems, namely the active transmission covariance matrix and passive beamforming. In the active beamforming optimization problem, SDR and water injection algorithm are used to solve the optimal solution, and the reflection coefficient of RIS is optimized by particle swarm algorithm, which improves the rate of the system and has a faster convergence speed while ensuring the quality of service.

Claims

1. A method for optimizing the receiving end rate of a RIS-assisted relay cooperative communication system, characterized in that: The following steps are involved: S1. Establish a downlink multiple-input multiple-output communication system model with RIS-assisted relay coordination, including a base station, an intelligent reflective surface RIS, a relay and a receiving end, wherein the base station has M antennas, the RIS has N reflective elements, the relay has L antennas, and the receiving end has K antennas; S2. Assume that the channel matrix from the base station to the receiving end is The channel matrix from the relay to the receiver is The channel coefficient matrix from the base station to the RIS is The channel coefficient matrix from RIS to the receiver is It is expressed as: In formula (1) and (2), β1 and β2 are Rice fading coefficients. For line-of-sight links, It is a non-line-of-sight link; The reflection coefficient of S3 and RIS at the receiving end is: In formula (3), represents the phase shift of the reflected signal of the nth element of the RIS, β n ∈[0,1] represents the amplitude reflection coefficient; The transmission signal vectors of S4, base station and relay are The covariance matrices of the transmitted signals of the base station and the relay are defined as And Q>0, F>0; The signal vectors received by the receiving end from the base station and the relay are: y1=(H2ΦH1+H dir )x1+n1=Hx1+n1 (4); y2=hx2+n2 (5); In formula (4) and (5), They respectively indicate that the mean value received by the user is zero and the variance is Additive white Gaussian noise (AWGN); The total signal received by the receiver is: y=(H2ΦH1+H dir )x1+hx2+n (6); In formula (6), n = n1 + n2, is the noise vector, Φ represents the reflected beamforming of the signal by RIS; The data transmission rate at the receiving end is: S4. By jointly optimizing Q, F and Φ, the transmission rate of the receiving end is maximized. The solution to the maximum rate problem is optimized as follows: In formula (8), P s , P r are the average transmission power constraints of the base station and the relay, Ε{‖x1‖ 2 }≤P s , Ε{||x2|| 2 }≤P r ; Solve through the iterative algorithm of alternating optimization. First, the optimal solution of Q and F is obtained by giving Φ, and then the optimal solution of Q and F is used as the given Q and F to obtain the optimal solution of Φ, and the cycle is repeated alternately. After each round of iteration, the transmission rate after the current round of iteration is compared with the transmission rate after the previous round of iteration using the fitness function. When the comparison result reaches the preset threshold, the iteration is stopped, and the maximum rate of the receiving end is obtained. S41. Given Φ, find the optimal solution of Q and F Define Q = qq H , F = ff H , The optimization problem is transformed into: Definition H H =H H H,h h =h H h, optimize the objective function R u is equivalent to: Combining equations (9) and (10), the optimization problem is transformed into: The optimization problem is then simplified to: The semi-positive definite programming (SDR) algorithm is used for optimization, and then the eigenvalue decomposition (EVD) is used to obtain the optimal solution under given phase conditions. The optimal Q solution process is: H = U1Λ1V1 H is the truncated singular value decomposition of H, where D = rank (H) ≤ min (M, K) represents the maximum number of data streams that H can transmit; then the power is allocated according to the water injection algorithm to obtain the optimal power amount p for each data stream. 1i , and then get the optimal Q; The power p of each data stream 1i According to the Lagrange multiplier method, the optimization problem is converted to: In formula (13), λ is the Lagrange multiplier; P 1i The derivative is make get: In formula (14), The optimal Q is: Q * =V1diag{p 11 ,...,p 1D }V1 H (15); The optimal F solution process is as follows: as the truncated singular value decomposition of h, where Dd=rank(d)≤min(L,K) represents the maximum number of data streams that h can transmit; then the power is allocated according to the water injection algorithm to obtain the optimal power amount p for each data stream 2i , and then the optimal F is: S42. Given Q and F, find the optimal solution of Φ The optimization problem is transformed into: The particle position is set to a set of solutions of the phase shift vector of RIS, the particle swarm is initialized, and the initial velocity v and position x are randomly generated, which can be expressed as: v p (i+1)=wv p (i)+c1r1(pbest p (i)-x p (i))+c2r2(gbest p (i)-x p (i)) (18); x p (i+1)=x p (i)+v p (i+1) (19); In equations (18) and (19), w represents the weight, c1 and c2 represent learning factors, r1 and r2 represent random numbers that obey uniform distribution between [0, 1], and the particle position x is the phase shift vector constrained between [0, 2π); Then, the objective function is used to calculate the fitness value of each particle, find the individual optimal pbest and the global optimal gbest, and adjust the next flight speed and position. During the iteration process, the fitness function is used to evaluate each particle. When the maximum number of iterations is reached, the optimal fitness value, that is, the maximum rate Φ, is obtained.

2. The method for optimizing the receiving end rate of a RIS-assisted relay cooperative communication system according to claim 1, characterized in that: In s42, a nonlinear function is constructed to adjust the weight w, and random numbers are added to the particle velocity update method. The dynamic search ability of the algorithm is improved by dynamically adjusting the particle's motion trajectory. The adjusted weight w and particle velocity v are expressed as: In formula (20) and (21), w max 、w max is the maximum and minimum value of the weight, Iter is the maximum number of iterations, rand() represents a random number that follows a uniform distribution between [0,1], and θ is the adjustment threshold whose value is [0,1]. θ is expressed as: In formula (22), θ max ,θ min are the maximum adjustment threshold and the minimum adjustment threshold.

Images