Pilot period setting method for intelligent metasurface-assisted networks under time-varying channels

Abstract

The present invention provides a method for setting the pilot period for an intelligent metasurface-assisted network in a time-varying channel. The method comprises: obtaining channel state information through pilot signals, constructing a time-varying channel model within the pilot period for the IRS-user channel; solving the robust beamforming matrix and the IRS phase matrix using a method to eliminate the multiple random terms introduced by the model, thereby obtaining optimized capacity performance within the period; and setting a reasonable pilot period to update the IRS-user channel state information based on the performance changes over time. The proposed method can maximize the weighted sum rate of each transmission time slot and extend the pilot period under certain capacity loss, thereby avoiding the large-scale signaling overhead caused by real-time IRS channel estimation.

Description

Technical Field

[0001] The present invention relates to the field of wireless communication technology, and in particular to a method for setting a pilot period of an intelligent metasurface-assisted network under a time-varying channel in future sixth-generation mobile communications (6th Generation, 6G). Background Art

[0002] With the growing demand for high capacity in wireless networks, achieving capacity growth through technologies like base station densification and antenna density is difficult to meet and leads to high costs and energy consumption. However, intelligent metasurfaces, also known as intelligent reflective surfaces (IRS), as passive reflective devices, offer the advantages of low cost, low energy consumption, high reliability, and high capacity, making them a popular technology. IRSs intelligently construct radio propagation environments by actively modifying channels, allowing for flexible programming to improve network performance.

[0003] The system performance and user service quality of IRS-assisted networks depend heavily on coordinated signal processing between the base station and the IRS, specifically joint base station beamforming and IRS phase design. To fully exploit the potential of collaborative optimization, real-time, perfect channel state information is required.

[0004] However, the IRS's multiple independent reflector units introduce numerous and dimensional channel state information matrices between the IRS and users, and between the base station and IRS, in the IRS cascade link. Actual cellular networks periodically transmit pilot signals for channel estimation. Given that smart metasurfaces are typically deployed in a fixed manner, the base station-to-IRS channel can be acquired to some extent. However, the IRS-to-user channel is often time-varying due to user mobility, making CSI acquisition challenging.

[0005] To facilitate phase optimization in the IRS reflection unit, there is also a way to configure a sensing device on the smart metasurface to receive and transmit pilot signals, thereby measuring the channel state information between the IRS and the user and between the base station and the IRS, respectively.

[0006] Given the large number and dimensionality of channel state information matrices, channel estimation in IRS-assisted networks will introduce significant signaling overhead. Pilot transmission resources are limited in real-world network environments, and to avoid this significant signaling overhead, the pilot period needs to be lengthened. However, in high-speed mobile scenarios, user movement can increase the Doppler shift of the point-to-point link between the IRS and the user, leading to significant channel time-variability and severe channel aging within the pilot period. The channel state information estimated from the pilot signal is non-ideal, and the beamforming matrix and phase matrix designed based on this estimated channel state information are mismatched with the actual channel, reducing system throughput.

[0007] The impact of channel time-varying can be modeled by establishing an autoregressive model. This model consists of two components: previously estimated channel samples and a random error matrix. The random error matrix can be modeled as a combination of channel statistics and a Gaussian random matrix. Due to the presence of multiple random terms in this channel time-varying model, this error matrix is often treated as interference, which renders the non-robust joint beamforming matrix and phase design matrix ineffective in traditional IRS-assisted networks. Given the slow variation of channel statistics and their high estimation accuracy, how to exploit potential active and passive beamforming gains by utilizing static statistics as useful signals rather than interference is a critical issue.

[0008] To deal with the pain points and difficulties of random terms brought about by applying the statistical channel model established by static statistical information to the modeling of useful signals, it is necessary to solve the closed-form expression of the IRS cascade channel capacity. This is one of the reasons why the existing robustness design based on non-ideal CSI has not been comprehensively studied in terms of maximizing the weighted rate and. The current solution is to use a method for solving approximate closed-form expressions of various MIMO channel capacities - deterministic equivalence, so that as the matrix dimension increases, an asymptotically equivalent analytical expression with higher approximate accuracy is obtained. The present invention applies this method to the IRS auxiliary network for robust joint beamforming and phase design to obtain a closed-form target expression to guide the setting of a reasonable pilot period. Summary of the Invention

[0009] The present invention provides a method for setting a pilot period of an intelligent metasurface-assisted network under a time-varying channel, comprising: obtaining channel state information through a pilot, and constructing a time-varying channel model within a pilot period for an IRS-user channel; solving a robust beamforming matrix and an IRS phase matrix using a method for eliminating random terms introduced by the model to obtain optimized capacity performance within the period; and setting a reasonable pilot period to update the IRS-user channel state information according to changes in performance over time.

[0010] The pilot period setting method of the present invention provides the following technical solutions, including:

[0011] Step 200: Acquire channel state information through pilot signals, and construct a time-varying channel model within a pilot period for the IRS-user channel.

[0012] Consider a flat-fading MIMO system where a single base station and multiple users periodically transmit channel estimation pilot signals for channel estimation. In order to facilitate the design of the intelligent metasurface phase, sensors are configured on the IRS to receive and transmit pilot signals for channel estimation between the base station and IRS and between the IRS and users, respectively.

[0013] Based on the above definition, the following system model is established: Define the t s The channel state information between the base station and the intelligent metasurface in the transmission time slot is: Since the base station and IRS are located at fixed positions, for simplicity, the channel between the base station and the smart metasurface during the SRS period is considered to be a quasi-static channel, i.e. Define t s The channel state information between the intelligent metasurface and user k in a transmission time slot is: Assume that the channel between the base station and user k is blocked, the channel state information

[0014] To model the pilot period T ue In order to solve the time-varying channel problem in the intelligent metasurface, an autoregressive model containing past known channel samples and a random error matrix is introduced. The random error matrix attempts to apply statistical channel information modeling, thereby calculating the channel between the intelligent metasurface IRS and the user k-channel. Modeling is performed to determine the matrix With random matrices The form of the sum, setting Updated once per transmission time slot, the following time-varying channel model can be obtained

[0015]

[0016] where ρ r,k As the time domain autocorrelation coefficient of the channel, it is modeled by the Jakes channel model, that is, ρ r,k =J0(2πf c Δt), Δt represents the transmission time interval, i.e., a single transmission time slot, J0 is the zero-order Bessel function, f c is the Doppler frequency shift, i.e. f c =vf d / c, where v is the user rate, f d is the carrier frequency, c is the speed of light. s ) represents a stable random process. The present invention establishes a characteristic beam model Perform statistical channel modeling, where U, Ω, and V can be expressed by −Tue ,-2T ue And the previous channel estimation sample value is determined, The elements in are Gaussian random variables with zero mean and unit variance.

[0017] Step 210: For the multiple random terms introduced by the model, a method of eliminating random terms is used to solve the robust beamforming matrix and the IRS phase matrix to obtain the optimized capacity performance within the cycle.

[0018] Let the transmission time slot t s The phases of the N reflection units of the smart metasurface are in Transmission time slot t s The beamforming matrix of the base station to user k is expressed as and transmit data vector Then user k receives the signal It can be expressed as

[0019]

[0020] in is the noise vector at user k, satisfying Represents the noise variance. The total interference noise of each user is regarded as Gaussian noise Its covariance matrix is defined as

[0021] The expected rate of user k can be expressed as

[0022]

[0023] The design goal of the algorithm of the present invention is to optimize the total network throughput of each transmission time slot, and the optimization variable is the beamforming matrix of each transmission time slot. With the phase matrix The optimization problem is established as follows

[0024]

[0025]

[0026]

[0027] The objective function of this optimization problem is α k Characterizing user fairness, the optimization objective function is the whole network throughput formula, and all users are fair, that is, α k are all equal to 1, and the constraint condition P is the total power of the base station, that is, the total power transmitted to the user needs to be lower than the total power of the base station.

[0028] Consider the transmission time slot t sThe received signal at user k is passed through a linear receive filter The estimated signal vector is

[0029]

[0030] By transforming the target problem into the form of minimum mean square error, the MMSE optimal linear receiving filter is solved

[0031] Further expressing the user rate expression after optimizing the receiving matrix, we observe that R k It is about the mean square error matrix Convex function, applying the first-order condition of convex function, we can establish the equivalent problem of the original optimization problem, as shown below

[0032]

[0033]

[0034]

[0035] The present invention then solves this optimization problem by iteratively achieving convergence using an alternating optimization algorithm. To address limited pilot resources, reduce the signaling overhead of channel measurement, and ensure minimal capacity loss for users over longer pilot periods, the optimization problem is decomposed into two subproblems. Based on deterministic equivalence theory, a robust cooperative beamforming design is proposed, solving the robust beamforming weight design and robust beamforming phase design, respectively.

[0036] The phase matrix is fixed, and the beamforming matrix is optimized. The MM algorithm, which is commonly used in the design of convex optimization algorithms, is used. For the random terms in the optimization expression coefficients, the deterministic equivalence theory is applied to obtain the asymptotically equivalent analytical expression of the user rate, which is further solved. Then, the optimized beamforming matrix is fixed and the IRS phase is optimized. Noting that the problem expression contains multiple random term coefficients, the deterministic equivalence method is also used to solve its closed-form expression. Then, the Riemann conjugate gradient descent method is used to solve the phase optimization matrix under the unit module constraint in the manifold space.

[0037] Step 220: According to the change of performance over time, set a reasonable pilot period to update the IRS-user channel status information.

[0038] Based on the performance results within a period derived from the robustness algorithm, the capacity performance is analyzed over time to determine the most appropriate pilot period. If performance degrades at a given moment, the pilot period can be reduced, trading pilot overhead for improved performance. If performance is good at that moment, with no significant degradation, the pilot period can be extended to avoid the large signaling overhead associated with real-time IRS channel estimation.

[0039] Beneficial effects

[0040] In the above-mentioned technical solution provided by the present invention, a time-varying channel model between the IRS and the user is established to reflect the channel time-varying nature caused by user mobility; then a robust algorithm is used to optimize the beamforming and phase optimization problems with random terms. Compared with traditional non-robust algorithms, the algorithm design proposed by the present invention can maximize the weighted sum rate of each transmission time slot and reduce the capacity loss of system performance caused by channel time-varying nature; and it is proposed that the pilot period can be extended by using an optimization algorithm under a certain capacity loss, and the large-scale signaling overhead caused by real-time IRS channel estimation can be avoided by a reasonable pilot period setting method. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0042] Figure 1 This is a diagram of an IRS-assisted network scenario in the present invention considering a time-varying channel;

[0043] Figure 2 is a flow chart of the method implementation of the present invention;

[0044] Figure 3 It is a graph showing the relationship between system capacity performance and time under different algorithms; DETAILED DESCRIPTION

[0045] This paper proposes a method for setting the pilot period in an IRS-assisted network, taking into account the time-varying channel characteristics caused by user mobility. This method applies a robust beamforming and phase design algorithm to process random terms to achieve closed-form system capacity optimization, guiding the design of pilot periods. This method is not limited to a single base station scenario and is universally applicable to diverse scenarios.

[0046] like Figure 1As shown, in this scenario, considering the presence of buildings between the base station and the user, the introduction of the IRS introduces a cascaded channel. During the pilot signal transmission period, user movement actually causes the channel between the user and the IRS to constantly change. This leads to a mismatch between the beamforming and phase designed after receiving the pilot signal and the actual channel, resulting in performance loss. To further clarify the technical solution of the present invention, the following detailed description is provided with reference to specific embodiments and accompanying drawings.

[0047] like Figure 2 , is a flowchart of the pilot period setting method provided by an embodiment of the present invention.

[0048] Step 300: Acquire channel state information through pilot signals, and construct a time-varying channel model within a pilot period for the IRS-user channel.

[0049] Consider a flat fading MIMO system with a single configuration M t A base station with 1 antenna, a single smart metasurface with N reflective units and K with M r The K users are randomly distributed on a circle with a center at (x2, 0) and a radius of r in the initial time slot. The base station is located at the origin of the coordinate system, and the coordinates of the smart metasurface are (x1, y1). Users periodically transmit channel estimation pilot signals for channel estimation. To facilitate the design of the smart metasurface phase, sensors are configured on the IRS to receive and transmit pilot signals for channel estimation between the base station and IRS, and between the IRS and users, respectively.

[0050] Based on the above definition, the following system model is established: Define the t s The channel state information between the base station and the intelligent metasurface in the transmission time slot is: Since the base station and IRS are located at fixed positions, for simplicity, the channel between the base station and the smart metasurface during the SRS period is considered to be a quasi-static channel, i.e. Define t s The channel state information between the intelligent metasurface and user k in a transmission time slot is: The channel between the base station and user k is blocked, and the channel state information

[0051] To model the pilot period T ue The channel time-varying property is considered, and an autoregressive model including past known channel samples and random error matrix is introduced. The random error matrix attempts to apply statistical channel information modeling, thereby calculating the k-channel between the intelligent metasurface and the user. Modeling is performed to determine the matrix With random matrices The form of the sum, setting Updated once per transmission time slot, the following time-varying channel model can be obtained

[0052]

[0053] where ρ r,k As the time domain autocorrelation coefficient of the channel, it is modeled by the Jakes channel model, that is, ρ r,k =J0(2πf c Δt), Δt represents the transmission time interval, i.e., a single transmission time slot, J0 is the zero-order Bessel function, f c is the Doppler frequency shift, i.e. f c =vf d / c, where v is the user rate, f d is the carrier frequency, c is the speed of light. s ) represents a stable random process, which can be modeled by statistical channel information. The present invention establishes a characteristic beam model Perform statistical channel modeling, where U, Ω, and V can be expressed by −T ue ,-2T ue And the previous channel estimation sample value is determined, The elements in are Gaussian random variables with zero mean and unit variance.

[0054] In the above formula, t s =0 indicates the moment when channel estimation is completed, that is, the real-time channel state information is accurately estimated at this time, and the beamforming design based on this channel state information exactly matches the real-time channel;

[0055] And 0<t s <T ue Part 1 represents the channel model during the channel expiration phase within the SRS period, uses random errors to model the time-varying nature of the channel, and uses statistical channel information to model the channel. Its high-order moment information has been obtained at time t=0.

[0056] Step 310: For the multiple random terms introduced by the model in step 300, a method of eliminating random terms is used to solve the robust beamforming matrix and the IRS phase matrix to obtain the optimized capacity performance within the cycle.

[0057] Let the transmission time slot t s The phases of the N reflection units of the smart metasurface are in Transmission time slot t s The beamforming matrix of the base station to user k is expressed as and transmit data vector Then user k receives the signal It can be expressed as

[0058]

[0059] in is the noise vector at user k, satisfying Represents the noise variance. The total interference noise of each user is regarded as Gaussian noise Its covariance matrix can be expressed as

[0060]

[0061] Divide the channel matrix into deterministic matrices With random matrices but

[0062]

[0063] That is at this time To determine the matrix, the expected rate of user k can be expressed as

[0064]

[0065] The design goal of the algorithm of the present invention is to optimize the total network throughput of each transmission time slot, and the optimization variable is the beamforming matrix of each transmission time slot. With the phase matrix The optimization problem is established as follows

[0066]

[0067]

[0068]

[0069] The objective function of this optimization problem is α k Characterizing user fairness, the optimization objective function is the whole network throughput formula, and all users are fair, that is, α k are all equal to 1, and the constraint condition P is the total power of the base station, that is, the total power transmitted to the user needs to be lower than the total power of the base station.

[0070] Consider the transmission time slot t s The received signal at user k is passed through a linear receive filter The estimated signal vector is

[0071]

[0072] Due to the received signal and noise Independent of each other, the mean square error at user k can be expressed as

[0073]

[0074] Fix all transmit beamforming weight matrices By transforming the target problem into the form of minimum mean square error, the MMSE optimal linear receiving filter is solved

[0075]

[0076] Therefore use The MSE matrix is

[0077]

[0078] Further expressing the user rate expression after optimizing the receiving matrix, we observe that R k It is about the mean square error matrix Convex function, applying the first-order condition of convex function, we can establish the equivalent problem of the original optimization problem, as shown below

[0079] Observed R k It's about Convex function, applying the first-order condition of convex function, we can get

[0080]

[0081] in Denotes the variable beamforming matrix With the phase matrix If both the rightmost terms in the above formula are constants, the third term containing the variable is easier to handle than the original objective function, which is convenient for designing the optimization algorithm.

[0082] Therefore, an equivalent problem to the original optimization problem can be established as follows

[0083]

[0084]

[0085]

[0086] The present invention then solves this optimization problem by iteratively achieving convergence using an alternating optimization algorithm. To address limited pilot resources, reduce the signaling overhead of channel measurement, and ensure minimal capacity loss for users over longer pilot periods, the optimization problem is decomposed into two subproblems. Based on deterministic equivalence theory, a robust cooperative beamforming design is proposed, solving the robust beamforming weight design and robust beamforming phase design, respectively.

[0087] Fixed phase matrix, optimized beamforming matrix, using MM algorithm commonly used in convex optimization algorithm design, obtained the lower bound function of the objective function, and obtained about the beamforming matrix The convex function of further considering the power constraint condition can be solved by the Lagrange multiplier method.

[0088]

[0089] The optimization variables can be obtained by iterative solution Will Assign to For the random terms in the coefficients of the optimization expression, the deterministic equivalence theory is applied to obtain the asymptotically equivalent analytical expression of the user rate, which is further solved.

[0090] Then fix the optimized beamforming matrix and optimize the IRS phase. After a series of changes in the objective function and ignoring the irrelevant constant terms, the phase optimization objective expression can be decoupled from the above formula into the phase The function of

[0091]

[0092] The parameters can be expressed as

[0093]

[0094]

[0095]

[0096]

[0097]

[0098] Note that the problem expression contains multiple random coefficients and the calculation is complex. We also use the deterministic equivalent method to solve its closed-form expression, and then use the Riemann conjugate gradient descent method to solve the phase optimization matrix under the unit module constraint in the manifold space.

[0099] The process of the IRS-assisted network robust cooperative beamforming and phase design method based on deterministic equivalence theory proposed in this invention is summarized as follows:

[0100] The first step is to obtain IRS and user parameters and construct an IRS cascade time-varying channel model;

[0101] Step 2: Initialization on the base station side: alternating iteration number D = 0, beamforming matrix iteration number d1 = 0 and maximum convergence number Phase matrix iteration number d2 = 0 and maximum convergence number Beamforming matrix and satisfy them Phase matrix Φ (0,0) ;

[0102] The third step is to start the alternating iteration of the beamforming matrix and the phase matrix D = 0, 1, 2, ..., D max ;

[0103] Step 4: Fix the phase matrix Φ (D,0) , beamforming matrix iteration

[0104] Step 5: Substitute Calculate the covariance matrix of interference plus noise

[0105]

[0106] Step 6: Substitute Calculation coefficient

[0107]

[0108]

[0109] Step 7: Update

[0110] Step 8, d1←d1+1, if Then return to step 5, otherwise stop the iteration and go to the next step;

[0111] Step 9: Fix the beamforming matrix Phase Matrix Iteration

[0112] Step 10. Substitute According to the formula and Further, solve the Euclidean gradient And by obtaining the Riemann gradient, initialize η0=-gradf(φ (D,0) );

[0113] Step 11: Solve the Euclidean gradient And find the Riemann gradient

[0114] Step 12: Select the Armi jo backtracking line search step According to the formula Get the updated phase

[0115] Step 13: Obtain the conversion factor between tangent vectors in different tangent spaces according to the formula

[0116] Step 14: Select Polak-Ribiere parameters And find the direction vector

[0117] Step 15, d2←d21, if Then return to step 10, otherwise stop the iteration and go to the next step;

[0118] Step 16, If D≤D max -1, return to step 4, otherwise stop the iteration.

[0119] The symbols are as follows:

[0120]

[0121]

[0122]

[0123]

[0124]

[0125]

[0126]

[0127]

[0128]

[0129]

[0130]

[0131]

[0132]

[0133] Step 320: According to the change of performance over time, set a reasonable pilot period to update the IRS-user channel status information.

[0134] Using the above beamforming algorithm design, the performance graph of system capacity changing with time is obtained under different user mobility rates. Figure 3 As shown in the figure, it can be seen that after a period of time, the capacity drops significantly. This is because the channel characteristics at this time are different from the channel characteristics when the initial base station measures the pilot signal from the user. The time correlation of the channel is reduced. At this time, the transmission weights on the base station side will not match the channel at this moment, and the user and rate will drop.

[0135] The most appropriate pilot period is determined by analyzing capacity performance over time. When selecting the pilot period, a balance is struck between pilot overhead and user performance when capacity performance drops below a certain threshold, thereby improving overall system performance. If performance degrades during each update, the pilot period is reduced, trading pilot overhead for improved performance. If performance is good at that moment, with no significant degradation, the pilot period can be extended to avoid the large signaling overhead associated with real-time IRS channel estimation.

Claims

1. A method for setting the pilot period of an intelligent metasurface-assisted network under a time-varying channel, characterized in that: include: The channel state information is obtained through the pilot, and a time-varying channel model is constructed for the IRS-user channel within the pilot period. in, Represents the t s The channel between the transmission time slot IRS and user k, ρ r,k As the time domain autocorrelation coefficient of the channel, modeled by Jakes channel model, H r,k Represents t s = 0, that is, the channel is accurately estimated at the beginning of the SRS cycle, ζ(t s ) represents a stationary random process, which can be modeled by statistical channel information; the phase matrix is fixed, and the deterministic equivalent theory is used to obtain the asymptotic equivalent analytical expression of the user rate for the multiple random terms introduced by the model, and the optimized beamforming matrix is further obtained; then the optimized beamforming matrix is fixed and the IRS phase is optimized, wherein the phase optimization objective expression can be decoupled into the phase The function of The parameters can be expressed as The deterministic equivalence theory is also used to solve the closed-form expression of the intermediate variable, where Indicates the tth s The channel state information between the base station and the smart metasurface in each transmission time slot, Indicates the tth s The transmission time slot corresponds to the beamforming matrix of user k, with the superscript The number of alternating iterations is D and the number of beamforming iterations is the maximum Indicates the tth s The covariance matrix of interference plus noise at user k in the transmission time slot, E k ·and is the expectation operator; then the Riemann conjugate gradient descent method is used to solve the phase optimization matrix under the unit module constraint in the manifold space, and finally the average capacity performance optimized within the cycle is obtained through iterative optimization; according to the change of performance over time, a reasonable pilot period is set to update the IRS-user channel state information.

2. The method according to claim 1, characterized in that The IRS-user channel model includes an autoregressive model of past known channel samples and a random error matrix, wherein the random error matrix is constructed by establishing an eigenbeam model. Perform statistical channel modeling, where the statistics U, Ω, and V can be determined from the previous channel estimation samples. The elements in are Gaussian random variables with zero mean and unit variance, so the k-channel between the intelligent hypersurface and the user It can be expressed as a deterministic matrix With random matrices The form of the sum.

3. The method according to claim 1, characterized in that When the IRS-related cascade channel matrix is random, the following steps are used to optimize and solve the optimal beamforming: Fixed phase matrix, t s Time-slot optimized beamforming matrix The intermediate quantity is expressed as The superscript D,d1 corresponds to the number of alternating iterations D and the number of beamforming iterations d1. is the IRS phase matrix when the number of alternating iterations is D, represents the channel state information between the base station and the smart metasurface, is the covariance matrix of interference plus noise at user k after the last phase iteration with D alternating iterations and d1 beamforming iterations, E k ·and is the expectation operator, and the other coefficient symbols are calculated as follows For the above coefficient symbols, the deterministic equivalence theory is applied to obtain the asymptotically equivalent analytical expression of the user rate, which is further solved.

4. The method according to claim 1, wherein Analyze the capacity performance of the IRS-assisted system over time to determine the most appropriate pilot period. If performance degrades at a given moment, reduce the pilot period to improve performance at the expense of pilot overhead. If the performance at this moment is good and no serious performance degradation occurs, you can consider extending the pilot period to avoid the large-scale signaling overhead caused by real-time IRS channel estimation.

Images