A robust optimization method and system for intelligent reflecting surface assisted wireless relay NOMA communication system

Abstract

The application relates to the technical field of communication technology, and discloses a robust optimization method and system for an intelligent reflecting surface (IRS) assisted wireless relay NOMA communication system, which comprises the following specific steps: S1, according to the IRS assisted wireless relay NOMA communication system, a system model is constructed, which comprises joint optimization of a base station end, a relay end, power distribution coefficients, time slot distribution coefficients and IRS reflection coefficients; S2, an optimization problem of the system model is constructed according to the system model; S3, a relaxation variable is introduced, the optimization problem is solved, and the robust optimization of the IRS assisted wireless relay NOMA communication system is completed. The application solves the problem that the prior art lacks a robust optimization method for a NOMA communication system under imperfect channel state information, and has the characteristics of effectively improving the performance of a communication system.

Description

Technical Field

[0001] This invention relates to the field of communication technology, and more specifically, to a robust optimization method and system for an intelligent reflector-assisted wireless relay NOMA communication system. Background Technology

[0002] Intelligent reflecting surfaces (IRS) are a revolutionary new technology that effectively improves resource utilization and expands coverage in wireless communication systems. They consist of numerous low-cost passive reflective elements, each capable of independently adjusting its reflection coefficient to alter the amplitude and phase of the incident signal. By densely deploying IRSs in a wireless network and coordinating their reflection coefficients, signals can be coherently superimposed in desired directions, while interference can be eliminated in certain directions. This enables intelligent control of the propagation environment, transforming wireless communication from passively adapting to the environment to actively changing it, thus improving system performance. Another key feature of IRS is its low power consumption. Compared to traditional repeater technologies, IRS does not require a radio frequency link; it passively reflects signals, requiring only low power to maintain the reflective elements and avoiding the introduction of noise signals seen in repeaters. Furthermore, IRS can be flexibly deployed in various environments and is compatible with all wireless networks, offering lower hardware and deployment costs.

[0003] Non-orthogonal multiple access (NOMA) is considered an effective solution for improving spectral efficiency and achieving massive connectivity in future wireless networks, and is expected to meet the high spectral efficiency and massive connectivity requirements of 5G. Traditional orthogonal multiple access (OMA) allocates orthogonal resources to different users. While this avoids mutual interference, it falls short in meeting the demands of massive connectivity and high spectral efficiency in future networks. Unlike traditional OMA, NOMA allows multiple users to access the same resource block, significantly improving spectral efficiency. At the transmitter, signals from different users use different power levels based on differences in channel gain, and are then superimposed in the power domain. The receiver can distinguish the information from different users based on the power of the received signal, and uses serial interference cancellation (SIC) to eliminate multiple access interference and recover the desired information.

[0004] However, NOMA only achieves better spectral efficiency than OMA when the channel gain differences among users are significant. Because IRS has the ability to reconfigure the wireless environment, adjusting the IRS reflection coefficient can increase the channel gain difference between users, thus leveraging the advantages of NOMA. Therefore, NOMA is often used in conjunction with IRS. Furthermore, NOMA can effectively ensure fairness among users. Most current research on IRS and NOMA assumes perfect channel state information. However, due to the passive nature of IRS and factors such as channel estimation and quantization errors, the actual obtained channel state information is often imperfect. Especially in harsh propagation environments such as oceans and deserts, it is almost impossible for base stations to obtain perfect channel state information. Therefore, optimizing wireless resources under imperfect channel state information is of greater practical significance.

[0005] In recent years, wireless relay has been considered an important technology for achieving wide coverage. By deploying relays in a system, base stations can provide communication services to remote users through multi-hop communication, thereby expanding the coverage area of ​​the base station. However, traditional relay schemes suffer from severe path loss due to the long propagation path from the base station to the relay, resulting in low system communication rates. We urgently need a solution to improve the communication performance of relay systems.

[0006] Referring to four papers on IRS-assisted wireless communication systems under channel uncertainty, and specifically IRS-assisted wireless communication in marine environments, the paper "G. Zhou, C. Pan, H. Ren, K. Wang, MDRenzo and A. Nallanathan, 'Robust Beamforming Design for Intelligent Reflecting Surface Aided MISO Communication Systems,'" in IEEE Wireless Communications Letters, vol. 9, no. 10, pp. 1658-1662, Oct. 2020, doi:10.1109 / LWC.2020.3000490," proposes a design method for IRS-assisted wireless communication in terrestrial scenarios with channel uncertainty. This method jointly optimizes the beamforming vectors of the base station and the IRS to minimize the base station's transmit power under imperfect channel state information. However, this paper does not consider NOMA technology or IRS-assisted relaying. The paper “Q.Sun, P.Qian, W.Duan, J.Zhang, J.Wang and K.-K.Wong, Ergodic Rate Analysis and IRS Configuration for Multi-IRS Dual-Hop DFRelaying Systems,” in IEEE Communications Letters, vol.25, no.10, pp.3224-3228, Oct.2021, considers a communication system model with multiple IRS-assisted DF relays and proposes a method for maximizing ergodic rate and an expression for the upper bound of ergodic rate. However, this paper considers a single-user system, does not use NOMA technology, and focuses on resource allocation optimization under perfect channel state information, while not considering resource allocation optimization under imperfect channel state information.The paper "Z. Zhou, N. Ge, W. Liu and Z. Wang, RIS-Aided Offshore Communications with Adaptive Beamforming and Service Time Allocation, ICC2020-2020 IEEE International Conference on Communications (ICC), 2020, pp. 1-6, doi:10.1109 / ICC40277.2020.9148833." studies IRS-assisted offshore systems and employs reconfigurable reflective arrays (RRAs) at the base station to maximize the effective communication rate by jointly optimizing beamforming vectors and time allocation. However, this paper does not consider the case of IRS-assisted radio relay, nor does it consider non-orthogonal multiple access technologies.

[0007] Because IRS has high beamforming gain and can intelligently reconstruct the propagation environment, IRS-assisted wireless relay can effectively improve the performance of the communication system. In addition, since NOMA technology has high spectral efficiency, its use can increase the user's communication rate. Furthermore, since we often cannot obtain accurate channel state information in real-world scenarios, existing patents lack robust optimization methods for NOMA communication systems with imperfect channel state information. Therefore, how to invent a robust optimization method for a NOMA communication system based on robust beamforming and intelligent reflector-assisted wireless relay is a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0008] To address the lack of robust optimization methods for NOMA communication systems with imperfect channel state information in existing technologies, this invention provides a robust optimization method for intelligent reflector-assisted wireless relay NOMA communication systems, which effectively improves the performance of the communication system.

[0009] To achieve the above-mentioned objectives of this invention, the technical solution adopted is as follows:

[0010] A robust optimization method for a smart reflector-assisted wireless relay NOMA communication system includes the following specific steps:

[0011] S1. Based on the intelligent reflector-assisted wireless relay NOMA communication system, construct a system model that includes joint optimization of the power allocation coefficient, time slot allocation coefficient, and IRS reflection coefficient of the base station and relay ends.

[0012] S2. Optimization problem of constructing the system model based on the system model;

[0013] S3. Introduce slack variables, solve the optimization problem, and complete the robust optimization of the intelligent reflector-assisted wireless relay NOMA communication system.

[0014] Preferably, the intelligent reflector-assisted wireless relay (NOMA) communication system includes a shore-based base station, an IRS, and k+1 users; the IRS includes M reflector units.

[0015] Furthermore, in step S1, based on the intelligent reflector-assisted wireless relay NOMA communication system, a system model is constructed that includes jointly optimizing the power allocation coefficient, time slot allocation coefficient, and IRS reflection coefficient of the base station and relay ends, specifically as follows:

[0016] S101. In the first time slot phase of the intelligent reflector-assisted wireless relay (NOMA) communication system, the channel coefficient of the base station to IRS communication link is defined as follows: The channel coefficients for the base station to user 0 and IRS to user 0 communication links are respectively and In the second time slot phase of a smart reflector-assisted wireless relay (NOMA) communication system, the channel coefficient of the relay-to-IRS link is defined as follows: The channel coefficients of the relay to user i and the IRS to user i communication links are respectively and in

[0017] i∈{1,2…K};

[0018] In the first time slot phase, the reflection phase shift matrix of the IRS is defined as φ1 = diag(θ1), where θ1 represents the reflection phase.

[0019] Position coefficient; in the second time slot phase, the IRS reflection coefficient matrix is ​​defined as φ2=diag(θ2), where θ2 represents the reflection coefficient.

[0020] Phase coefficient;

[0021] S102. Establish a channel model, where large-scale fading is represented by a two-path fading model, and small-scale fading by a [missing information - likely a specific model or model].

[0022] The stag fading model; the channel model expression is as follows:

[0023]

[0024]

[0025]

[0026]

[0027] in,

[0028] h BI This represents the channel model from the base station to the IRS, h. d This represents the channel model from the base station to the user, h. i This represents the channel model from relay to user, PL(d) B0 ), PL(d i ), PL(d ri The following expressions represent the channel power gain caused by large-scale fading in the links from base station to IRS, base station to user 0, relay to user i, and IRS to user i, respectively, and are expressed as follows:

[0029] PL(d)=(λ c / (4πd)) 2 *4(sin(2πH t H r / (λ c d))) 2

[0030] H t H r , λ c d and d represent the transmit antenna height, receive antenna height, center subcarrier wavelength, and wavelength, respectively.

[0031] Transmission distance; for h BI Small-scale fading for h d Small-scale fading for h i Small-scale fading

[0032] for Small-scale fading is modeled as a Ricean fading model; the levels from the base station to the IRS and from the IRS to the user are defined.

[0033] The connection channel is Define the cascaded channel from relay to IRS and from IRS to user i as follows: The equation is:

[0034]

[0035]

[0036] Since it is almost impossible for a base station to obtain perfect channel state information, let h be the direct link from the base station to the relay. d ,

[0037] g0, the direct link h from the relay to user i i and the cascaded link g from the relay to the IRS to the user i iThere is a channel error.

[0038] The actual channel is modeled as follows:

[0039]

[0040]

[0041] Where h d g0, h i and g i For the actual channel coefficients, and The channel coefficients estimated for the base station and relay, and Δh d , Δg0, Δh i and Δg i The channel estimation error is represented by a bounded CSI error model, which is used to describe the channel error as follows:

[0042] ||Δh d ||≤ε d ||Δg0||≤ε g0

[0043] ||Δh i ||≤ε i ||Δg i ||≤ε gi i∈{0,…,K}

[0044] Where ||a|| denotes the 2-norm operation of vector a, ε d ε g0 ε i and ε gi The radius representing the uncertainty range of the corresponding channel error;

[0045] S103. Construct transmission and reception signal models; in the first time slot phase, represent the baseband signal transmitted by the base station to the relay as follows: Where α m P is the power allocation factor for user m. t s represents the maximum transmission power of the base station. m Let be the data symbol transmitted by the base station to user m, and s m The signal follows an independent, identically distributed, circularly symmetric complex Gaussian distribution with mean 0 and variance 1. After receiving all user information, the relay decodes each user's information sequentially using serial interference cancellation technology, and re-encodes and transmits the information of user i in the second time slot. In the second time slot, the baseband signal transmitted by the relay is represented as... Where β m P is the power factor allocated to user m by the relay. rThis is the maximum transmit power of the relay;

[0046] In the first time slot phase, the signal received by the relay is represented as:

[0047]

[0048] Where n0 represents a mean of 0 and a variance of 0. Gaussian white noise signal;

[0049] In the second time slot phase, the signal received by user i is represented as:

[0050]

[0051] Where n i This indicates that the mean and variance received at user i are 0. Gaussian white noise signal; without loss of generality, assume Furthermore, we stipulate that the relay decodes user information in descending order, that is, first decodes the information of user K, then decodes the information of user K-1, and so on, finally decoding the information of user 0. According to the NOMA decoding rules, when decoding the information of user i, the information of all users t (t = i+1, ..., K) will be decoded first, then the information of user t will be removed from the received signal, and then the information of user i will be decoded. In the first time slot, the received signal after removing the information of user i is represented as follows:

[0052]

[0053] In the first time slot phase, the communication rate of user i is:

[0054]

[0055] Where t1 is the transmission time of the first time slot, α i Let the power allocation coefficient be for user i, and specify... During the second time slot phase, when decoding the information of user i at user j (1≤j≤i≤K), the residual received signal at user j after user j completes SIC is represented as follows:

[0056]

[0057] The feasible rate for user i to decode at user j is:

[0058]

[0059] Where t2 is the transmission time of the second time slot, σ j 2 β represents the power of the noise signal received at user j.i Let the power factor allocated to user i by the relay be... According to the rules of NOMA-based serial interference cancellation technology, the signal of user i needs to be decoded by all users j, where j∈{1,...,i}. The achievable rate of user k in the second time slot is expressed as:

[0060]

[0061] After transmission through two time slots, the communication rate for user i will be:

[0062] R0 = R0 1

[0063]

[0064] R0 is the feasible rate for user 0 decoding. 1 Define set Λ i ={(Δh) i , Δg i )|||Δh i ||≤ε i ,||Δg i ||≤ε gi}, i∈{1,...,K}, define set Λ0={(Δh d ,Δg0)|||Δh d ||≤ε d , ||Δg0||≤ε g0 Considering the worst-case user communication rate, the worst-case communication rate of user k is expressed as:

[0065]

[0066] The worst-case communication rate for user 0 is expressed as:

[0067]

[0068] Furthermore, in step S2, the optimization problem of the system model specifically includes:

[0069] P1:

[0070]

[0071]

[0072] |θ 1，ii |=1,|θ 2，ii |=1,i=1,…M (1d)

[0073] t1+t2≤1, t1≥0, t2≥0 (1e)

[0074] α i ≥0, i=0, …K (1f)

[0075] β i ≥0, i=1, …K (1g)

[0076] Where P1 is the first optimization problem, Let α represent the user's communication rate, and let (1a) represent the objective function. The optimization objective is to maximize the communication rate of the minimum number of users in the system under the worst-case scenario. i β represents the power allocation factor at the base station. i θ represents the power distribution factor at the relay end. 1，ii The unit mode constraint for the IRS reflection coefficient of the first time slot, θ 2，ii The unit mode constraint represents the IRS reflection coefficient of the second time slot, t1 is the allocation coefficient of the first time slot, and t2 is the allocation coefficient of the second time slot.

[0077] Furthermore, in step S3, slack variables are introduced to solve the optimization problem through non-convex optimization. The specific steps are as follows:

[0078] S301. Introduce the slack variable λ and rewrite the first optimization problem as the second optimization problem P2:

[0079]

[0080]

[0081]

[0082] λ≥0 (2d)

[0083] (1b)-(1g) (2e)

[0084] S302, Introduce slack variable μ i i = 0, ..., K and μ ij , 1≤j≤i, i=1,…K, rewrite the second optimization problem as the third optimization problem P3:

[0085]

[0086] st t1log2(1+-μ i )≥λ i=0…K (3b)

[0087] t2log2(1-μ ij )≥λ 1≤j≤i,i=1…K (3c)

[0088]

[0089]

[0090] μ i ≥0, i=0…K (3f)

[0091] μ ij ≥0, 1≤j≤i, i=1…K (3g)

[0092] (1b)-(1g), (2d) (3h)

[0093] Relax constraints (3d) and (3e):

[0094]

[0095]

[0096] S303, Transform the third optimization problem P3 into the fourth optimization problem P4:

[0097]

[0098] st t1log2(1+μ i )≥λ i=0…K (4b)

[0099] t2log2(1+μ ij )≥λ 1≤j≤i,i=1…K (4c)

[0100] (1b)-(1g), (2d), (3f), (3g), (3i), (3j) (4d)

[0101] S304. Introduce slack variables {ρ} i},{ρ ij}, {η i}, {η ij The fourth optimization problem, P4, is further transformed into the fifth optimization problem, P5:

[0102]

[0103] st t1log2(1+μ i )≥λ i=0…K (5b)

[0104] t2log2(1+μ ij )≥λ 1≤j≤i,i=1…K (5c)

[0105]

[0106]

[0107]

[0108]

[0109]

[0110]

[0111]

[0112] (1b)-(1g), (2d), (3f), (3g) (5j)

[0113] ρ i ≥0, η i ≥σ0 2 , i = 0…K (5k)

[0114] ρ ij ≥0, η ij ≥σ j 2 , 1≤j≤i, i=1…K (51)

[0115] S305. Solve the fifth optimization problem P5 using the alternating optimization algorithm to obtain the optimal power allocation coefficient at the base station. Optimal power distribution factor at the relay end Optimal first time slot allocation coefficient Optimal Second Time Slot Allocation Coefficient Optimal IRS reflection phase coefficient in the first time slot Second time slot phase IRS reflection phase coefficient

[0116] Furthermore, in step S305, the fifth optimization problem P5 is solved using an alternating optimization algorithm to obtain the optimal power allocation coefficient at the base station. Optimal power distribution factor at the relay end Optimal first time slot allocation coefficient Optimal Second Time Slot Allocation Coefficient Optimal IRS reflection phase coefficient in the first time slot Second time slot phase IRS reflection phase coefficient The specific steps are as follows:

[0117] A1. Initialization η ij (0) The number of iterations n = 0, and the threshold ε0 is set.

[0118] A2, obtained Given Solving the optimization problem yields

[0119] A3, Given Solving the optimization problem yields

[0120] A3. Let n = n + 1;

[0121] A4, when When the time is up, the solution process ends and the optimal solution is obtained.

[0122] Otherwise, return to step A2 for iteration.

[0123] Furthermore, in step A2, the following is obtained: Given Solving the optimization problem yields The specific steps are as follows:

[0124] B1. According to 3i, μ i along with The value increases as it increases, when hour, To obtain the maximum value, thus the optimal solution is obtained.

[0125] B2. Update optimization problem P5 (the 5th optimization problem) to optimization problem P8 (the 8th optimization problem);

[0126]

[0127]

[0128]

[0129]

[0130] (1d), (2d) (8e)

[0131] ρ ij ≥0, η ij ≥σ j 2 , 1≤j≤i, i=1…K (8f)

[0132] B3. Define matrix V2 = [θ2] T 1] H [θ2 T 1], through the equivalent transformation formula:

[0133]

[0134] in,

[0135]

[0136] By performing an equivalent transformation on optimization problem 8, we obtain optimization problem 9, P9:

[0137]

[0138] stt2log2(η ij +ρ ij )-t2log2(η ij )≥λ 1≤j≤I,i=1…K (9b)

[0139]

[0140]

[0141] (2d),(8f)(9e)

[0142] V 2, =1,i=1,…M+1(9f)

[0143] Rank(V2)=1,(9g)

[0144] B4. Apply a first-order Taylor expansion to the convex function terms in the 9th optimization problem. Perform linearization:

[0145]

[0146] B5. Relax constraint (9c) into an inequality constraint:

[0147]

[0148] B6. Similarly, for the convex function term -log2(η) in constraints (9b) and (9c) ij ) and concave function terms Perform a first-order Taylor expansion:

[0149]

[0150]

[0151] in This is the expansion point of the first-order Taylor expansion;

[0152] B7. Using a continuous convex approximation algorithm, through m1 iterations, the 9th optimization problem is optimized into the 10th optimization problem P10:

[0153]

[0154]

[0155]

[0156] (2d),(8f),(9h)(10d)

[0157] V 2, =1, i=1,…M+1(10e)

[0158] Rank(V2) = 1, (10f)

[0159] in, η ij (m1-1) This is the optimal value in the (m1-1)th iteration;

[0160] B8. Solve the 10th optimization problem P10 iteratively using the CVX optimization tool to find the optimal solution. when When the rank is 1, for The optimal IRS reflectance coefficient is obtained by performing eigenvalue decomposition. like When the rank is not equal to 1, the suboptimal solution is obtained by using the Gaussian randomization method.

[0161] Furthermore, in step A3, given Solving the optimization problem yields Specifically:

[0162] C1. Fix θ1 and θ2, and rewrite the fifth optimization problem P5 as P6:

[0163]

[0164]

[0165]

[0166] ρ i ≥η i μ i , i = 0…K (6d)

[0167] ρ ij ≥η ij μ ij , 1≤j≤i, i=1…K (6e)

[0168] (1b), (1c), (1e)-(1g), (2d), (3f), (3g), (5d)-(5g), (5k), (5l) (6j)

[0169] C2. Relax constraints (6b)-(6e):

[0170]

[0171]

[0172]

[0173] Among them, C 1i C 2ij C 3i and C 4ij The point is a constant.

[0174] C3. After relaxing constraints (6b)-(6e), the sixth optimization problem is rewritten as the seventh optimization problem P7:

[0175] st

[0176]

[0177]

[0178]

[0179] (1b), (1c), (1e)-(1g), (2d), (3f), (3g), (5d)-(5g), (5k), (51) (7j)

[0180] C4. Solve the seventh optimization problem P7 using the CVX optimization solver tool for m² iterations. Furthermore, in the m²th iteration, C... 1i C 2ij C 3i and C 4ij The update expression is:

[0181]

[0182]

[0183] A robust optimization system for a smart reflector-assisted wireless relay NOMA communication system includes a model building module, an optimization problem module, and a problem solving module.

[0184] The model building module is used to build a system model based on the intelligent reflector-assisted wireless relay NOMA communication system, including the power allocation coefficient, time slot allocation coefficient, and IRS reflection coefficient of the base station and relay.

[0185] The optimization problem module is used to construct the optimization problem of the system model;

[0186] The problem-solving module is used to introduce slack variables, solve optimization problems, and complete robust optimization of the intelligent reflector-assisted wireless relay NOMA communication system.

[0187] The beneficial effects of this invention are as follows:

[0188] This patent considers a communication system model of an IRS-assisted wireless relay and NOMA network, optimizing resource allocation under imperfect channel state information. To ensure fairness among users, under imperfect channel state information, a system model is constructed that jointly optimizes the power allocation coefficients, time slot allocation coefficients, and IRS reflection coefficients at the base station and relay ends to maximize the rate of the minimum user in the system. An optimization problem is constructed based on the system model, and finally, slack variables are introduced to solve the optimization problem, achieving robust optimization of the intelligent reflector-assisted wireless relay NOMA communication system. This invention addresses the problem of the lack of robust optimization methods for NOMA communication systems with imperfect channel state information in existing technologies and has the characteristic of effectively improving the performance of communication systems. Attached Figure Description

[0189] Figure 1 This is a flowchart illustrating the robust optimization method for this intelligent reflector-assisted wireless relay NOMA communication system.

[0190] Figure 2 This is a system schematic diagram of the intelligent reflector-assisted wireless relay NOMA communication system.

[0191] Figure 3 This diagram illustrates the variation of the minimum user rate as a function of channel uncertainty in a robust optimization method for a smart reflector-assisted wireless relay NOMA communication system.

[0192] Figure 4 This is a graph showing the variation of the minimum user rate as a function of transmit power for a robust optimization method employed in this intelligent reflector-assisted wireless relay NOMA communication system.

[0193] Figure 5 This is a graph showing the minimum user rate of the system as a function of the number of IRS reflective units, based on the robust optimization method of this intelligent reflector-assisted wireless relay NOMA communication system. Detailed Implementation

[0194] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments.

[0195] Example 1

[0196] like Figure 1 As shown, a robust optimization method for a smart reflector-assisted wireless relay NOMA communication system includes the following specific steps:

[0197] S1. Based on the intelligent reflector-assisted wireless relay NOMA communication system, construct a system model that includes joint optimization of the power allocation coefficient, time slot allocation coefficient, and IRS reflection coefficient of the base station and relay ends.

[0198] S2, the optimization problem of constructing a system model;

[0199] S3. Introduce slack variables, solve the optimization problem, and complete the robust optimization of the intelligent reflector-assisted wireless relay NOMA communication system.

[0200] Example 2

[0201] More specifically, such as Figure 2 As shown, the intelligent reflector-assisted wireless relay NOMA communication system includes a shore base station, an IRS, and k+1 users; the IRS includes M reflector units.

[0202] In one specific embodiment, step S1 involves constructing a system model based on the intelligent reflector-assisted wireless relay NOMA communication system, including jointly optimized power allocation coefficients, time slot allocation coefficients, and IRS reflection coefficients for the base station and relay ends. Specifically:

[0203] S101. In the first time slot phase of the intelligent reflector-assisted wireless relay (NOMA) communication system, the channel coefficient of the base station to IRS communication link is defined as follows: The channel coefficients for the base station to user 0 and IRS to user 0 communication links are respectively and In the second time slot phase of a smart reflector-assisted wireless relay (NOMA) communication system, the channel coefficient of the relay-to-IRS link is defined as follows: The channel coefficients of the relay to user i and the IRS to user i communication links are respectively and Where i∈{1,2…K};

[0204] In this embodiment, the distance from the base station to user i (i∈{1,2…K}) is relatively long. Due to the severe path loss during long-distance propagation, we do not consider the communication link between the base station and user i (i∈{1,2…K}).

[0205] In the first time slot phase, the reflection phase shift matrix of the IRS is defined as φ1=diag(θ1), where θ1 represents the reflection phase coefficient; in the second time slot phase, the reflection coefficient matrix of the IRS is defined as φ2=diag(θ2), where θ2 represents the reflection phase coefficient.

[0206] In this embodiment, θ1 = [q1, q2, ..., q M ] H (X H (represents the conjugate transpose of matrix X) where Here β m and θ m Let represent the reflection amplitude and reflection phase of the m-th reflecting unit, respectively, where m ∈ {1, ..., M}. Similarly, in the second time slot phase, the IRS reflection coefficient matrix can be expressed as φ2 = diag(θ2), and similarly, e m Satisfy | e m |=1.

[0207] S102. Establish the channel model, where large-scale fading is represented by a two-path fading model and small-scale fading by a Ricean fading model; the expression for the channel model is as follows:

[0208]

[0209]

[0210]

[0211]

[0212] Among them, h BI This represents the channel model from the base station to the IRS, h. d This represents the channel model from the base station to the user, h. i This represents the channel model from relay to user, PL(d) B0 ), PL(d i ), PL(d ri The following expressions represent the channel power gain caused by large-scale fading in the links from base station to IRS, base station to user 0, relay to user i, and IRS to user i, respectively, and are expressed as follows:

[0213] PL(d)=(λ c / (4πd)) 2 *4(sin(2πH t H r / (λ c d))) 2

[0214] H t H r , λc d and d represent the transmitting antenna height, receiving antenna height, center subcarrier wavelength, and propagation distance, respectively; for h BI Small-scale fading for h d Small-scale fading for h i Small-scale fading for Small-scale fading is modeled as a Ricean fading model; the cascaded channels from the base station to the IRS and from the IRS to the user are defined as follows: Define the cascaded channel from relay to IRS and from IRS to user i as follows: The equation is:

[0215]

[0216]

[0217] In this embodiment, since it is almost impossible for the base station to obtain perfect channel state information, let h be the direct link h from the base station to the relay. d g0, direct link h from relay to user i i and the cascaded link g from the relay to the IRS to the user i i There is a channel error; the actual channel is modeled as follows:

[0218]

[0219]

[0220] Where h d g0, h i and g i For the actual channel coefficients, and The channel coefficients estimated for the base station and relay, and Δh d , Δg0, Δh i and Δg i The channel estimation error is represented by a bounded CSI error model, which is used to describe the channel error as follows:

[0221] ||Δh d ||≤ε d ||Δg0||≤ε g0

[0222] ||Δh i ||≤ε i ||Δg i ||≤εg i i∈{0,…,K}

[0223] Where ||a|| denotes the 2-norm operation of vector a, ε d ε g0 ε i and ε gi The radius representing the uncertainty range of the corresponding channel error;

[0224] S103. Construct transmission and reception signal models; in the first time slot phase, represent the baseband signal transmitted by the base station to the relay as follows: Where α m P is the power allocation factor for user m. t s represents the maximum transmission power of the base station. m Let be the data symbol transmitted by the base station to user m, and s m The signal follows an independent, identically distributed, circularly symmetric complex Gaussian distribution with mean 0 and variance 1. After receiving all user information, the relay decodes each user's information sequentially using serial interference cancellation technology, and re-encodes and transmits the information of user i in the second time slot. In the second time slot, the baseband signal transmitted by the relay is represented as... Where β m P is the power factor allocated to user m by the relay. r This is the maximum transmit power of the relay;

[0225] In the first time slot phase, the signal received by the relay is represented as:

[0226]

[0227] Where n0 represents a mean of 0 and a variance of 0. Gaussian white noise signal;

[0228] In the second time slot phase, the signal received by user i is represented as:

[0229]

[0230] Where ni represents the mean of the data received at user i is 0 and the variance is 0. Gaussian white noise signal; without loss of generality, assume Furthermore, we stipulate that the relay decodes user information in descending order, that is, first decodes the information of user K, then decodes the information of user K-1, and so on, finally decoding the information of user 0. According to the NOMA decoding rules, when decoding the information of user i, the information of all users t (t = i+1, ..., K) will be decoded first, then the information of user t will be removed from the received signal, and then the information of user i will be decoded. In the first time slot, the received signal after removing the information of user i is represented as follows:

[0231]

[0232] In the first time slot phase, the communication rate of user i is:

[0233]

[0234] Where t1 is the transmission time of the first time slot, α k Let k be the power allocation factor for user k, and specify... During the second time slot phase, when decoding the information of user i at user j (1≤j≤i≤K), the residual received signal at user j after user j completes SIC is represented as follows:

[0235]

[0236] The feasible rate for user i to decode at user j is:

[0237]

[0238] Where t2 is the transmission time of the second time slot, σ j 2 β represents the power of the noise signal received at user j. i Let the power factor allocated to user i by the relay be... According to the rules of NOMA-based serial interference cancellation technology, the signal of user i needs to be decoded by all users j, where j∈{1,...,i}. The achievable rate of user k in the second time slot is expressed as:

[0239]

[0240] After transmission through two time slots, the communication rate for user i will be:

[0241] R0 = R0 1

[0242]

[0243] R0 is the feasible rate for user 0 decoding. 1 Define set Λ i ={(Δh) i , Δg i )|||Δh i ||≤ε i ,||Δg i ||≤ε gi}, i∈{1,...,K}, define set Λ0={(Δh d ,Δg0)|||Δh d ||≤ε d , ||Δg0||≤εg0 Considering the worst-case user communication rate, the worst-case communication rate of user k is expressed as:

[0244]

[0245] The worst-case communication rate for user 0 is expressed as:

[0246]

[0247] In one specific embodiment, to ensure fairness for maritime users in the presence of channel errors, the system maximizes the worst-case communication rate for the minimum number of users by jointly optimizing the power allocation coefficients, time slot allocation coefficients, and IRS reflection coefficients at the base station and relay ends. The optimization problem of the system model in step S2 specifically involves:

[0248] P1:

[0249]

[0250]

[0251] |θ 1，ii |=1,|θ 2，ii |=1,i=1,…M (1d)

[0252] t1+t2≤1, t1≥0, t2≥0 (1e)

[0253] α i ≥0, i=0, …K (1f)

[0254] β i ≥0, i=1, …K (1g)

[0255] Where P1 is the first optimization problem, Let α represent the user's communication rate, and let (1a) represent the objective function. The optimization objective is to maximize the communication rate of the minimum number of users in the system under the worst-case scenario. i β represents the power allocation factor at the base station. i θ represents the power distribution factor at the relay end. 1，ii The unit mode constraint for the IRS reflection coefficient of the first time slot, θ 2，ii The unit mode constraint represents the IRS reflection coefficient of the second time slot, t1 is the allocation coefficient of the first time slot, and t2 is the allocation coefficient of the second time slot.

[0256] In one specific embodiment, step S3 involves introducing slack variables to solve the optimization problem using non-convex optimization. The specific steps are as follows:

[0257] S301. In this embodiment, the first optimization problem P1 is a difficult non-convex problem. By introducing a relaxation variable λ, the first optimization problem is rewritten as the second optimization problem P2:

[0258]

[0259]

[0260]

[0261] λ≥0 (2d)

[0262] (1b)-(1g) (2e)

[0263] S302. In this embodiment, we continue to introduce slack variables to transform the second optimization problem into an easier-to-handle optimization problem. We introduce slack variables μ. i i = 0, ..., K and μ ij , 1≤j≤i, i=1,…K, rewrite the second optimization problem as the third optimization problem P3:

[0264]

[0265] st t1log2(1+μ i )≥λ i=0…K (3b)

[0266] t2log2(1+μ ij )≥λ 1≤j≤i,i=1…K (3c)

[0267]

[0268]

[0269] μ i ≥≥0, i=0…K (3f)

[0270] μ ij ≥0, 1≤j≤i, i=1…K (3g)

[0271] (1b)-(1g), (2d) (3h)

[0272] In this embodiment, the third optimization problem remains a difficult non-convex problem, the most challenging aspect being the infinite number of inequality constraints (3d) and (3e). To transform constraints (3d) and (3e) into more manageable ones, the following deduction is required:

[0273]

[0274] Therefore, in set Λ0, |hd +θ1 T The following inequalities hold for g0:

[0275]

[0276] when When, the inequality on the left side holds the equality sign, when When, the inequality on the right side of the above equation takes the equality sign, where θ1 * This indicates the conjugate operation on vector θ1. Similarly, we have:

[0277]

[0278] Where j = 1, ..., K.

[0279] Based on the two inequalities above, relax constraints (3d) and (3e):

[0280]

[0281]

[0282] S303, Transform the third optimization problem P3 into the fourth optimization problem P4:

[0283]

[0284] st t1log2(1+μ i )≥λi=0…K (4b)

[0285] t2log2(1+μ ij )≥λ 1≤j≤i,i=1…K (4c)

[0286] (1b)-(1g), (2d), (3f), (3g), (3i), (3j) (4d)

[0287] S304. In this embodiment, equations (4b), (4c), (3i), and (3j) are still non-convex constraints. In order to transform the optimization problem into a solvable optimization problem, a slack variable {ρ} is introduced. i}, {p ij}, {η i}, {η ij The fourth optimization problem, P4, is further transformed into the fifth optimization problem, P5:

[0288]

[0289] st t1log2(1+μ i )≥λ i=0…K (5b)

[0290] t2log2(1+μ ij )≥λ 1≤j≤i,i=1…K (5c)

[0291]

[0292]

[0293]

[0294]

[0295]

[0296]

[0297] (1b)-(1g), (2d), (3f), (3g) (5j)

[0298] ρ i ≥0, η i ≥σ0 2 , i = 0…K (5k)

[0299] ρi j ≥0, η ij ≥σ j 2 , 1≤j≤i, i=1…K (51)

[0300] S305. Solve the fifth optimization problem P5 using the alternating optimization algorithm to obtain the optimal power allocation coefficient at the base station. Optimal power distribution factor at the relay end Optimal first time slot allocation coefficient Optimal Second Time Slot Allocation Coefficient Optimal IRS reflection phase coefficient in the first time slot Second time slot phase IRS reflection phase coefficient

[0301] In this embodiment, since the power allocation coefficient and the IRS reflection phase are coupled together, the fifth optimization problem is still a non-convex problem. In order to solve the fifth optimization problem, we propose an efficient alternating optimization algorithm.

[0302] In one specific embodiment, step S305 involves solving the fifth optimization problem P5 using an alternating optimization algorithm to obtain the optimal power allocation coefficient at the base station. Optimal power distribution factor at the relay end Optimal first time slot allocation coefficient Optimal Second Time Slot Allocation Coefficient Optimal IRS reflection phase coefficient in the first time slot Second time slot phase IRS reflection phase coefficient The specific steps are as follows:

[0303] A1. Initialization η ij (0) The number of iterations n = 0, and the threshold ε0 is set.

[0304] A2, obtained Given Solving the optimization problem yields

[0305] A3, Given Solving the optimization problem yields

[0306] A3. Let n = n + 1;

[0307] A4, when When the time is up, the solution process ends and the optimal solution is obtained.

[0308] Otherwise, return to step A2 for iteration.

[0309] In one specific embodiment, in step A2, the following is obtained: Given Solving the optimization problem yields The specific steps are as follows:

[0310] B1. According to 3i, μ i along with The value increases as it increases, when hour, To obtain the maximum value, thus the optimal solution is obtained.

[0311] B2. Update optimization problem P5 (the 5th optimization problem) to optimization problem P8 (the 8th optimization problem);

[0312]

[0313]

[0314]

[0315]

[0316] (1d), (2d) (8e)

[0317] ρ ij≥0, η ij ≥σ j 2 , 1≤j≤i, i=1…K (8f)

[0318] B3. In this embodiment, because constraints (8c), (8d), and (1d) are non-convex constraints with respect to θ2, and constraint (8b) is non-convex with respect to ρ... ij and η ij Since the functions are not jointly concave, the 8th optimization problem is a non-convex problem that is difficult to solve directly. We will now use the SDR algorithm to solve the 8th optimization problem. To use the SDR algorithm, the following equivalent transformation is required: Define matrix V2 = [θ2] T 1] H [θ2 T 1], through the equivalent transformation formula:

[0319]

[0320] in,

[0321]

[0322] By performing an equivalent transformation on optimization problem 8, we obtain optimization problem 9, P9:

[0323]

[0324] stt2log2(η ij +ρ ij )-t2log2(η ij )≥λ 1≤j≤i,i=1…K (9b)

[0325]

[0326]

[0327] (2d), (8f) (9e)

[0328] V 2，ii =1, i=1, ..., M+1 (9f)

[0329] Rank(V2) = 1, (9g)

[0330] B4. In this embodiment, due to the presence of convex function terms... Therefore, constraint (9c) is not a convex constraint, and constraints (9d) and (9b) are related to the optimization variables V2 and η. ijSince it is not a convex constraint, the optimization problem (P9) remains a difficult nonconvex problem. Next, we employ the Continuous Convex Approximation (SCA) algorithm to solve the optimization problem (P9) iteratively. To use the SCA algorithm, we first need to perform a first-order Taylor expansion on the convex function terms. Linearization: First-order Taylor expansion is used to transform the convex function terms in the 9th optimization problem. Perform linearization:

[0331]

[0332] B5. The right side of the inequality above is a convex function term. exist The first-order Taylor expansion of a point, used to replace the convex function term. Constraint (9c) can be relaxed to the following inequality constraint:

[0333]

[0334] The above inequality relates to the optimization variables V2 and ρ. ij Convex constraints;

[0335] B6. Similarly, for the convex function term -log2(η) in constraints (9b) and (9c) ij ) and concave function terms Perform a first-order Taylor expansion:

[0336]

[0337]

[0338] Where η ii (m-1) , This is the expansion point of the first-order Taylor expansion;

[0339] B7. Using a continuous convex approximation algorithm, through m1 iterations, the 9th optimization problem is optimized into the 10th optimization problem P10:

[0340]

[0341]

[0342]

[0343] (2d), (8f), (9h) (10d)

[0344] V 2，ii =1, i=1, ..., M+1 (10e)

[0345] Rank(V2) = 1, (10f)

[0346] in, This is the optimal value in the (m1-1)th iteration;

[0347] B8. In this embodiment, the 10th optimization problem P10 is solved iteratively using the CVX optimization tool. When the constraint (10f) is ignored, the 10th optimization problem is a convex optimization problem, and the optimal solution can be obtained iteratively using the CVX optimization tool. when When the rank is 1, for The optimal IRS reflectance coefficient is obtained by performing eigenvalue decomposition. like When the rank is not equal to 1, the suboptimal solution is obtained by using the Gaussian randomization method.

[0348] In one specific embodiment, in step A3, given Solving the optimization problem yields Specifically:

[0349] C1. Fix θ1 and θ2, and rewrite the fifth optimization problem P5 as P6:

[0350]

[0351]

[0352]

[0353] ρ i ≥η i μ i , i = 0…K (6d)

[0354] ρ ij ≥η ij μ ij , 1≤j≤i, i=1…K (6e)

[0355] (1b), (1c), (1e)-(1g), (2d), (3f), (3g), (5d)-(5g), (5k), (5l) (6j)

[0356] C2. In this embodiment, in the sixth optimization problem, constraints (6b)-(6e) are non-convex constraints, inspired by the inequality x. 2 +y 2 ≥2xy, x≥0, y≥0. Relax constraints (6b)-(6e):

[0357] Take constraint (6b) as an example:

[0358]

[0359] in for The upper bound of C is a convex function with respect to λ and t1. 1i Let be a constant fixed point to ensure that the inequality on the left side of the above equation is tight. Similarly, relaxing the constraint (6c)-(6e), we get:

[0360]

[0361]

[0362]

[0363] Among them, C 1i C 2ij C 3i and C 4ij The point is a constant.

[0364] C3. After relaxing constraints (6b)-(6e), the sixth optimization problem is rewritten as the seventh optimization problem P7:

[0365]

[0366]

[0367]

[0368]

[0369] (1b), (1c), (1e)-(1g), (2d), (3f), (3g), (5d)-(5g), (5k), (51) (7j)

[0370] C4. In this embodiment, all constraints in the 7th optimization problem are convex constraints, and the optimization objective function is also a convex function. Therefore, the 7th optimization problem is a convex optimization problem, which can be solved iteratively using the CVX optimization solver tool. The 7th optimization problem P7 is solved by m2 iterations using the CVX optimization solver tool.

[0371] In one specific embodiment, during the m2th iteration, C 1i C 2ij C ai and C 4ij The update expression is:

[0372]

[0373]

[0374] The robust optimization method for this intelligent reflector-assisted wireless relay NOMA communication system has the following advantages:

[0375] 1) Due to the use of wireless relay technology, which allows users close to the base station to act as relays, the coverage of the base station can be expanded through multi-hops.

[0376] 2) Because IRS technology is used in wireless relay systems, the strength of the received signal can be greatly enhanced by adjusting the reflection coefficient of the IRS, thereby improving the system's communication performance.

[0377] 3) Because the channel estimation error of the system is taken into account, the communication performance of the system is more robust, which is more practical in real-world environments.

[0378] Experimental results of the robust optimization method for this intelligent reflector-assisted wireless relay NOMA communication system are as follows: Figure 3 , Figure 4 As shown: Figure 3 , Figure 4 The English-Chinese comparison of the simulation schemes used is shown in Table 1:

[0379]

[0380] Table 1

[0381] In the experiment, the boundary of channel error was defined as Where δ represents the uncertainty of the channel, and the larger its value, the more uncertain the channel is. Figure 3The figure shows the variation of the minimum user rate of the system with channel uncertainty, where the number of IRS reflection units M = 64. From the figure, we can draw the following conclusions: (1) The optimization method adopted in this patent is basically the same as the optimal algorithm of SDR (theoretical upper limit), which verifies the effectiveness and superiority of the algorithm we adopted. (2) The minimum user rate of all schemes decreases with the increase of channel uncertainty δ. This is because as δ increases, the channel becomes more uncertain, and the user's communication rate decreases in the worst case. (3) The communication rate of the optimization method adopted in this patent is greater than that of other schemes within the range of δ variation, which proves the superiority of the proposed algorithm. (4) The performance of the optimization method adopted in this patent is better than that of the "Same-phase" scheme. This is because the proposed scheme has more optimization degrees of freedom. (5) From the performance comparison between the optimization method adopted in this patent and the "t1 = t2 = 0.5" scheme, it can be seen that optimizing the time slot allocation coefficient can effectively improve the system performance. (6) The minimum user communication rate of the system using the optimization method adopted in this patent is higher than that of the "Without-IRS" and "rand" schemes, proving the effectiveness of deploying IRS and optimizing the IRS reflection coefficient in improving the system communication rate. (7) As δ increases, the minimum user rate of the system using the optimization method adopted in this patent decreases faster than that of the "Without-IRS" scheme. This is because adding IRS introduces channel error, so the channel error of the optimization method adopted in this patent is larger than that of the "Without-IRS" scheme. Therefore, the proposed scheme is more sensitive to channel error, and the minimum user rate of the system will decrease rapidly as δ increases.

[0382] When the number of IRS reflector elements M=64, the minimum user rate of the system varies with transmit power in each scheme as follows: Figure 4 As shown; by Figure 4 It can be seen that: (1) When δ = 0 and there is no channel error, all schemes increase linearly with the increase of transmit power. This is because increasing transmit power will enhance the strength of the received signal, thereby improving the communication rate of the system. (2) When δ = 0.01, the communication rate of all schemes first increases and then remains unchanged with the increase of transmit power. This is because the error in SIC decoding leaves interference terms in the received signal. The value of the interference terms also increases linearly with the increase of transmit power. As a result, after the transmit power exceeds a certain value, the user's communication rate stabilizes at a certain value and no longer increases with the increase of transmit power. This shows the importance of accurate channel state information for improving system performance.

[0383] like Figure 5As shown: (1) When δ = 0.001, except for the “Without-IRS” scheme, the minimum user rate of the system of all other schemes increases with the increase of the number of IRS reflection units. This is because with the increase of the number of IRS reflection units, the IRS can provide a larger beamforming gain, which can more effectively enhance the strength of the received signal and reduce the signal interference between users, thereby improving the user's communication rate. (2) When δ = 0.001, the performance gain of the proposed scheme relative to the “Without-IRS” scheme increases with the increase of the number of IRS reflection units, proving the effectiveness of the IRS beamforming gain in improving the system communication rate. (3) When δ = 0.02, the minimum user communication rate of the system of the proposed algorithm hardly changes with the increase of the number of IRS reflection units. This is because deploying the IRS in the system will introduce channel error. Although increasing the number of IRS reflection units will increase the beamforming gain of the IRS, it will also increase the channel error. When channel uncertainty is high, the disadvantages of increasing the number of IRS reflection units can outweigh the benefits, resulting in a situation where the minimum user rate of the system remains almost unchanged even when the number of IRS reflection units is increased.

[0384] This patent considers a communication system model for an IRS-assisted wireless relay and NOMA network, optimizing resource allocation under imperfect channel state information. To ensure fairness among users, under imperfect channel state information, a system model is constructed that jointly optimizes the power allocation coefficients, time slot allocation coefficients, and IRS reflection coefficients at the base station and relay ends to maximize the rate of the minimum number of users. An optimization problem is constructed based on the system model, and finally, slack variables are introduced to solve the optimization problem, achieving robust optimization of the intelligent reflector-assisted wireless relay NOMA communication system. This invention addresses the lack of robust optimization methods for NOMA communication systems with imperfect channel state information in existing technologies and effectively improves the performance of the communication system.

[0385] Example 3

[0386] A robust optimization system for a smart reflector-assisted wireless relay NOMA communication system includes a model building module, an optimization problem module, and a problem solving module.

[0387] The model building module is used to build a system model based on the intelligent reflector-assisted wireless relay NOMA communication system, including the power allocation coefficient, time slot allocation coefficient, and IRS reflection coefficient of the base station and relay.

[0388] The optimization problem module is used to construct the optimization problem of the system model;

[0389] The problem-solving module is used to introduce slack variables, solve optimization problems, and complete robust optimization of the intelligent reflector-assisted wireless relay NOMA communication system.

[0390] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the implementation of the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the claims of the present invention.

Claims

1. A robust optimization method for a smart reflector-assisted wireless relay NOMA communication system, characterized in that: The specific steps include the following: S1. Based on the intelligent reflector-assisted wireless relay NOMA communication system, construct a system model that includes joint optimization of the power allocation coefficient, time slot allocation coefficient, and IRS reflection coefficient of the base station and relay ends, specifically: In the first time slot phase of a smart reflector-assisted wireless relay (NOMA) communication system, the reflection phase shift matrix of the IRS is defined as follows: , This represents the reflection phase coefficient of the first time slot; In the second time slot phase, the IRS reflection coefficient matrix is ​​defined as follows: , This represents the reflection phase coefficient of the second time slot; Assume a direct link from the base station to the relay. , Relay to user direct link and relay to IRS to user Cascaded links There is a channel error; the actual channel is modeled as follows: in For the actual channel coefficients, , , and These are the channel coefficients estimated for base stations and relays, while , , and The channel estimation error is represented by a bounded CSI error model, which is used to describe the channel error as follows: in Representing vectors L2 norm operations, , , and The radius representing the uncertainty range of the corresponding channel error; After transmission through two time slots, the communication rate of user i is expressed as: The feasible rate for user-0 decoding. To define a set Define a set , Communication rate; Considering the worst-case user communication rate, the worst-case communication rate of user i is expressed as: in, Let i be the feasible rate at which user i decodes at user j; The worst-case communication rate for user 0 is expressed as: S2. The optimization problem of constructing the system model, specifically: Where P1 is the first optimization problem, Let represent the user's communication rate. The objective function (1a) indicates that the optimization objective is to maximize the communication rate of the minimum number of users in the system under the worst-case scenario. This represents the power allocation coefficient at the base station. This represents the power distribution factor at the relay end. The unit mode constraint represents the IRS reflection coefficient of the first time slot. The unit mode constraint represents the IRS reflection coefficient of the second time slot. The allocation coefficient for the first time slot. The allocation coefficient for the second time slot; S3. Introduce slack variables, solve the optimization problem, and complete the robust optimization of the intelligent reflector-assisted wireless relay NOMA communication system.

2. The robust optimization method for the intelligent reflector-assisted wireless relay NOMA communication system according to claim 1, characterized in that: The intelligent reflector-assisted wireless relay NOMA communication system includes a shore base station, an IRS, and k+1 users; the IRS includes M reflector units.

3. The robust optimization method for the intelligent reflector-assisted wireless relay NOMA communication system according to claim 2, characterized in that: In step S1, based on the intelligent reflector-assisted wireless relay NOMA communication system, a system model is constructed that includes jointly optimizing the power allocation coefficient, time slot allocation coefficient, and IRS reflection coefficient of the base station and relay ends. Specifically: S101. In the first time slot phase of the intelligent reflector-assisted wireless relay (NOMA) communication system, the channel coefficient of the base station to IRS communication link is defined as follows: The channel coefficients of the communication links from base station to user 0 and from IRS to user 0 are respectively and In the second time slot phase of a smart reflector-assisted wireless relay (NOMA) communication system, the channel coefficient of the relay-to-IRS link is defined as follows: relay to user and IRS to users The channel coefficients of the communication link are respectively and ,in ; In the first time slot phase, the reflection phase shift matrix of the IRS is defined as follows: , Indicates the reflection phase coefficient; In the second time slot phase, the IRS reflection coefficient matrix is ​​defined as follows: , Indicates the reflection phase coefficient; S102. Establish the channel model, where large-scale fading is represented by a two-path fading model and small-scale fading by a Ricean fading model; the expression for the channel model is as follows: in, This represents the channel model from the base station to the IRS. This represents the channel model from the base station to the user. This represents the channel model for relaying data to a user. This represents the channel model for relaying to user i; in, , , , These represent the connections from the base station to the IRS, from the base station to user 0, and from the relay to the user, respectively. IRS to user The channel power gain caused by large-scale fading in the link is expressed as follows: , These represent the transmitting antenna height, receiving antenna height, center subcarrier wavelength, and propagation distance, respectively. for Small-scale fading for , for Small-scale fading for Small-scale fading is modeled as a Ricean fading model; the cascaded channels from the base station to the IRS and from the IRS to the user are defined as follows: Define the relay to the IRS, and the IRS to the user. The cascaded channel is represented as We obtain the following equation: Assume a direct link from the base station to the relay. , Relay to user direct link and relay to IRS to user Cascaded links There is a channel error; the actual channel is modeled as follows: in For the actual channel coefficients, , , and These are the channel coefficients estimated for base stations and relays, while , , and The channel estimation error is represented by a bounded CSI error model, which is used to describe the channel error as follows: in Representing vectors L2 norm operations, , , and The radius representing the uncertainty range of the corresponding channel error; S103. Construct transmission and reception signal models; in the first time slot phase, represent the baseband signal transmitted from the base station to the relay as follows: ,in For users The power allocation factor, This is the maximum transmission power of the base station. Let m be the data symbol transmitted by the base station to user m, and The signal follows an independent, identically distributed, circularly symmetric complex Gaussian distribution with mean 0 and variance 1. After receiving all user information, the relay decodes each user's information sequentially using serial interference cancellation technology, and re-encodes and transmits the information of user i in the second time slot. In the second time slot, the baseband signal transmitted by the relay is represented as... ,in The power factor allocated to user m by the relay. This is the maximum transmit power of the relay; In the first time slot phase, the signal received by the relay is represented as: in, This indicates that the mean is 0 and the variance is... Gaussian white noise signal; In the second time slot phase, the user The received signal is represented as: in This indicates that the mean and variance received at user i are 0. Gaussian white noise signal; without loss of generality, assume Furthermore, we stipulate that the relay decodes user information in descending order, that is, first decodes the information of user K, then decodes the information of user K-1, and so on, finally decoding the information of user 0; according to the NOMA decoding rules, when decoding user... i When processing information, all users will be decoded first. The information, and then the user from the received signal. t All the information is eliminated, and then the user's information is decoded. i The information; in the first time slot phase, the received signal after removing the information of user i is represented as: In the first time slot phase, the user i The communication rate is: in, The transmission time of the first time slot, Let the power allocation coefficient be for user i, and specify... In the second time slot phase, during the user j Decoding User i Information ,user j After SIC is executed, in the user j The residual received signal at that location is represented as: The feasible rate for user i to decode at user j is: in, For the transmission time of the second time slot, This represents the power of the noise signal received at user j. Let the power factor allocated to user i by the relay be... According to the rules of NOMA-based serial interference cancellation technology, user i's signal needs to be detected by all users. Decoding, in which The achievable rate of user k in the second time slot phase is expressed as: After transmission through two time slots, the user i The communication rate is: The feasible rate for user-0 decoding. To define a set Define a set Considering the worst-case user communication rate, the worst-case communication rate of user k is expressed as: The worst-case communication rate for user 0 is expressed as: 。 4. The robust optimization method for the intelligent reflector-assisted wireless relay NOMA communication system according to claim 3, characterized in that: In step S3, slack variables are introduced to solve the optimization problem through non-convex optimization. The specific steps are as follows: S301. Introduce slack variables Rewrite the first optimization problem as the second optimization problem P2: （2a） s.t. (2b) （2c） （2d） （2e） S302. Introduce slack variables and The second optimization problem is rewritten as the third optimization problem P3: （3a） s.t. (3b) （3c） （3d） （3e） （3f） （3g） （3h） Relax constraints (3d) and (3e): （3i） （3j）； S303, Transform the third optimization problem P3 into the fourth optimization problem P4: （4a） st (4b) （4c） （4d） S304. Introducing slack variables The fourth optimization problem, P4, is further transformed into the fifth optimization problem, P5: （5a） s.t. (5b) （5c） （5d） （5e） （5f） （5g） （5h） （5i） （5j） （5k） （5l） S305. Solve the fifth optimization problem P5 using the alternating optimization algorithm to obtain the optimal power allocation coefficient at the base station. Optimal IRS reflection phase coefficient in the first time slot Second time slot phase IRS reflection phase coefficient .

5. The robust optimization method for the intelligent reflector-assisted wireless relay NOMA communication system according to claim 4, characterized in that: In step S305, the fifth optimization problem P5 is solved using an alternating optimization algorithm to obtain the optimal power allocation coefficient at the base station. Optimal IRS reflection phase coefficient in the first time slot Second time slot phase IRS reflection phase coefficient The specific steps are as follows: A1. Initialization The number of iterations n=0, and a threshold is set. ; A2, obtained Given Solving the optimization problem yields ; A3, Given Solving the optimization problem yields ; A4. Let n = n + 1; A5, when When the time is up, the solution process ends and the optimal solution is obtained. Otherwise, return to step A2 for iteration.

6. The robust optimization method for the intelligent reflector-assisted wireless relay NOMA communication system according to claim 5, characterized in that: In step A2, the following is obtained: Given Solving the optimization problem yields The specific steps are as follows; B1. According to 3i, along with The value increases as it increases, when hour, To obtain the maximum value, thus ; B2. Update optimization problem P5 (the 5th optimization problem) to optimization problem P8 (the 8th optimization problem); （8a） s.t. (8b) （8c） （8d） （8e） （8f） B3. Defining a matrix Using the equivalent transformation formula: in, =tr( ) By performing an equivalent transformation on optimization problem 8, we obtain optimization problem 9, P9: （9a） st (9b) （9e） （9f） Rank =1.(9g) B4. Apply a first-order Taylor expansion to the convex function terms in the 9th optimization problem. Perform linearization: B5. Relax constraint (9c) into an inequality constraint: （9h） B6. Similarly, for the convex function terms in constraints (9b) and (9c) and concave function terms Perform a first-order Taylor expansion: in , This is the expansion point of the first-order Taylor expansion; B7. Using a continuous convex approximation algorithm, through m1 iterations, the 9th optimization problem is optimized into the 10th optimization problem P10: （10a） s.t （10b） （10c） （10d） （10e） Rank =1,(10f) in, This is the optimal value in the (m1-1)th iteration; B8. Solve the 10th optimization problem P10 iteratively using the CVX optimization tool to find the optimal solution. ,when When the rank is 1, for The optimal IRS reflectance coefficient is obtained by performing eigenvalue decomposition. ;like When the rank is not equal to 1, the suboptimal solution is obtained by using the Gaussian randomization method.

7. The robust optimization method for the intelligent reflector-assisted wireless relay NOMA communication system according to claim 6, characterized in that: In step A3, given Solving the optimization problem yields Specifically: C1, Fixed Rewrite optimization problem P5 as P6: （6a） s.t. (6b) （6c） （6d） （6e） （6j） C2. Constraints Perform relaxation: in, , , The point is a constant. C3. After considering the constraints After relaxation, the 6th optimization problem is rewritten as the 7th optimization problem P7: (7a) st (7b) （7c） （7d） （7e） （7j） C4. Solve the 7th optimization problem P7 using the CVX optimization solution tool through m2 iterations.

8. The robust optimization method for the intelligent reflector-assisted wireless relay NOMA communication system according to claim 7, characterized in that: No. In the next iteration , , The update expression is: 。 9. A robust optimization system for an intelligent reflector-assisted wireless relay NOMA communication system, characterized in that: The method for implementing the method as described in any one of claims 1 to 8 includes a model building module, an optimization problem module, and a problem solving module; The model building module is used to build a system model based on the intelligent reflector-assisted wireless relay NOMA communication system, including the power allocation coefficient, time slot allocation coefficient, and IRS reflection coefficient of the base station and relay. The optimization problem module is used to construct the optimization problem of the system model; The problem-solving module is used to introduce slack variables, solve optimization problems, and complete robust optimization of the intelligent reflector-assisted wireless relay NOMA communication system.

Images