Segmented channel estimation method, system and device of RIS auxiliary system and medium

Abstract

The invention discloses a segmented channel estimation method, system and device for an RIS auxiliary system and a medium, and the method comprises the steps: building an RIS auxiliary wireless communication system model, and enabling a base station and a user in the system not to have a direct link; estimating an RIS-base station channel in the RIS-assisted wireless communication system model by adopting a decomposition-optimization algorithm; and estimating a user-RIS channel in the RIS-assisted wireless communication system model according to the time correlation of the channel and a Kalman filtering algorithm. According to the method, the quasi-static characteristic from the RIS to the base station channel and the time-varying characteristic from the user to the RIS channel are fully considered, and different time scales are adopted for estimation, so that the pilot frequency overhead is reduced, the channel estimation accuracy is improved, the channel estimation complexity is reduced, and the channel estimation efficiency is improved. The technical problem of high pilot frequency overhead of RIS auxiliary communication system channel estimation is solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of wireless communications, and in particular relates to a segmented channel estimation method, system, device and medium for a RIS-assisted system. Background Art

[0002] As an emerging device, Reconfigurable Intelligent Surface (RIS) has great potential and has the advantages of low power consumption, passivity and low cost. It can be flexibly deployed between base stations and users to build effective connections to achieve stable and efficient information transmission. Its working principle is based on a large number of passive reflection units on the surface. By accurately and reasonably configuring the phase shift of these units, the incident wave can be finely controlled and accurately reflected to the designated receiving device. However, in actual communication scenarios, the mobility of users makes the channel between the intelligent metasurface and the user time-varying. However, although the channel changes over time, there is still a close time correlation between the channel at the current moment and the previous moment. In most existing channel estimation methods, this key time correlation is often ignored. These methods fail to fully explore the potential information of the channel in the time dimension, which restricts the accuracy and efficiency of channel estimation to a certain extent. Accurate channel estimation of the RIS communication system is the basis for the intelligent reflective surface to configure its reflection coefficient, but since the pilot overhead of channel estimation is proportional to the number of reflection units, it is of great significance to reduce the pilot overhead under the premise of accurate channel estimation.

[0003] Since intelligent reflecting surfaces cannot perform signal processing, the channel estimation of most RIS systems treats the cascaded channels from users to RIS and RIS to base stations as a whole and performs channel estimation at the base station, while ignoring the quasi-static characteristics of the RIS-base station channel. This makes the pilot overhead required for channel estimation in RIS communication systems very large, seriously reducing the communication performance of the system. In order to reduce the pilot overhead of channel estimation, compressed sensing is a common channel estimation method. For example, in the paper "Compressed channel estimation for intelligent reflecting surface-assisted millimeterwave systems", the sparsity of millimeter wave channels is used to reduce pilot overhead. However, this method is only applicable to sparse channel estimation and is not applicable to most Rayleigh channels. There are also a small number of documents that consider the high pilot overhead problem of channel estimation in RIS-assisted communication systems. For example, the document "Channelestimation for intelligent reflecting surface assisted multiuser communications: Framework, algorithms, and analysis" proposes a three-stage channel estimation method. The first stage estimates the direct link, the second stage estimates the cascade channel of a typical user, and the third stage estimates the channel using a lower pilot overhead based on the strong correlation between ordinary users and typical users. However, the accuracy of its channel estimation is not good. In patent application CN117614775A, a joint hardware damage and RIS channel estimation method based on Kalman filtering is proposed. Although this method has the characteristics of stability and low complexity in the presence of hardware damage, this method estimates the cascade channel as a whole during the channel estimation process, and does not consider the redundancy of the RIS to base station link channel estimation during the channel estimation process, and its pilot overhead is high.

[0004] Therefore, the prior art either ignores the pilot overhead problem and pursues the accuracy of channel estimation, or considers the high pilot overhead problem of channel estimation but the estimation accuracy is poor. Summary of the invention

[0005] In order to overcome the shortcomings of the above-mentioned prior art, the purpose of the present invention is to provide a segmented channel estimation method, system, device and medium for a RIS-assisted system, which fully considers the quasi-static characteristics of the RIS to base station channel and the time-varying characteristics of the user to RIS channel. When estimating the RIS-base station channel, a decomposition-optimization algorithm is used to decompose the channel estimation problem into N simple sub-problems, and the optimal solution for channel estimation is obtained by continuously optimizing the sub-problems; when estimating the user-RIS channel, Kalman filtering is used to perform channel estimation based on the correlation between the channel at the current moment and the channel at the previous moment; the present invention can ensure the accuracy of its channel estimation while reducing the pilot overhead, reduces the complexity of the channel estimation, and solves the technical problem of high pilot overhead in channel estimation of the RIS-assisted communication system.

[0006] In order to achieve the above object, the technical solution adopted by the present invention is as follows:

[0007] A segmented channel estimation method for a RIS-assisted system, comprising:

[0008] Step 1: Construct a RIS-assisted wireless communication system model in which there is no direct link between the base station and the user;

[0009] Step 2: Use a decomposition-optimization algorithm to estimate the RIS-base station channel in the RIS-assisted wireless communication system model;

[0010] Step 3: Estimate the user-RIS channel in the RIS-assisted wireless communication system model based on the time correlation of the channel and the Kalman filter algorithm.

[0011] The channels in the system obey Rayleigh fading and are independent of each other, that is,

[0012]

[0013] in, is the channel coefficient matrix from RIS to base station, referred to as RIS-base station channel;

[0014] is the channel coefficient vector from the kth user to RIS, referred to as user-RIS channel; ρ g , They represent the large-scale fading factors of the corresponding transmission paths; I MN , I N Respectively represent the square matrix of all 1s of the corresponding dimension; cascade channel

[0015] In step 2, at a large time scale T g Under the condition, the dual link pilot is used to estimate the RIS-base station channel; in the step 3, at a small time scale T fIn this case, the uplink pilot signal before data transmission is used to estimate the user-RIS channel.

[0016] The estimation process of the RIS-base station channel in step 2 is as follows:

[0017] Pilot transmission stage: The pilot transmission frame in the dual-link pilot transmission of the base station consists of (N+1) subframes, each subframe includes L time slots, and after each subframe is sent, the reflection coefficient of RIS is adjusted. Then, the pilot signal received by the m2th antenna at the base station can be expressed as:

[0018]

[0019] In the formula, represents the channel vector from RIS to the m2th antenna of the base station, T represents the transpose, represents the channel vector from the m1th antenna of the base station to the RIS, represents the reflection coefficient vector of RIS, represents the environmental reflection, Indicates the non-zero pilot transmitted by the m1th antenna of the base station in the lth (l=1,2,...,L)th time slot of the tth subframe, represents the residual self-interference, Indicates the noise received by the m2th antenna of the base station;

[0020] Defining index tuples Among them, m1 represents the index of the transmitting antenna, and m2 represents the index of the receiving antenna. After (N+1) subframes are sent, all pilot signals received in the transmission frame are obtained. For a given (m1,m2)∈S, when a non-zero pilot is sent When P BS represents the transmission power of the base station. The (N+1) pilot signals transmitted from the m1th transmitting antenna of the base station and received by the m2th receiving antenna after being reflected by the RIS can be expressed as:

[0021]

[0022] In the formula, The elements in contain the unknown quantities to be estimated, 1 1×(N+1) represents an N+1 column full-1 vector, represents the reflection coefficient matrix of RIS;

[0023] Channel estimation phase: The quasi-static RIS-base station channel is estimated in three phases;

[0024] 1) In the first stage, the RIS-base station channel estimation problem is divided into N independent sub-problems, where each sub-problem is to estimate the channel between the base station and a single reflection unit in the RIS; for a given Multiply formula (1.4) by The inverse matrix of Estimates:

[0025]

[0026] In the formula, express The elements in , and a single element represents an estimate of the channel product from the m1th antenna to the nth RIS reflection unit and from the m2th antenna to the nth RIS reflection unit;

[0027] Further rearranging formula (1.6) yields:

[0028]

[0029] In the formula, Indicates g m1,n g m2,n The estimate of in represents the channel from the m1th antenna of the base station to the nth reflection unit, represents the channel from the m2th antenna of the base station to the nth reflection unit, is the estimation error;

[0030] 2) In the second stage, the initial value of the nth subproblem is calculated;

[0031] For a given n-th reflection unit, the estimate of the n-th subproblem is expressed as:

[0032]

[0033] In the formula, the cost function

[0034] Take 1≤m1≤m2≤L,m3≠m1,m2, the initial value of the channel coefficient estimation from the m′th antenna of the base station to the nth reflection unit in the RIS, that is, the initial value expression of the nth sub-problem is:

[0035]

[0036] In the formula,

[0037] 3) In the third stage, the estimation of the RIS-base station channel coefficients is iteratively optimized using the coordinate descent method according to the initial value of the nth subproblem;

[0038] Then, each sub-problem repeats the second and third phases to obtain the estimated value of the RIS-base station channel

[0039] The estimation of the RIS-base station channel coefficient is iteratively optimized by the coordinate descent method, as follows:

[0040] In the i-th external iteration, the M RIS-base station channel coefficients are estimated in turn to obtain the estimated values ​​of the previous (m-1) RIS-base station channel coefficients The optimal solution in the i-th outer iteration, the estimated value of the next (Mm) RIS-base station channel coefficients Using the optimal solution obtained in the last external iteration, the mth internal iteration estimate for the ith external iteration is It can be expressed as:

[0041]

[0042] according to Find the optimal solution:

[0043]

[0044] Update g according to formula (1.12) m,n The estimated value of

[0045]

[0046] Then after M internal iterations, we know

[0047]

[0048] Then after I max After external iterations, we know

[0049]

[0050] J n The value of is bounded and monotonic, so the estimation of the RIS-base station channel It will eventually converge.

[0051] The estimation process of the user-RIS channel in step 3 is as follows:

[0052] Pilot transmission phase: The small time scale T of the user-RIS channel is f It is divided into data transmission and uplink pilot transmission. The uplink pilot transmission frame consists of B subframes, each of which includes K time slots. After the user sends the pilot signal and it is reflected by RIS, in the b-th subframe, the base station receives the pilot signal sent by the k-th user, which is expressed as:

[0053]

[0054] Where Y b is the pilot signal received by the base station, is the channel coefficient matrix from RIS to base station, is the channel coefficient vector from the kth user to RIS, represents the reflection coefficient vector of the b-th subframe in the uplink pilot transmission phase, x k is the pilot signal sent by the kth user, N b Additive noise introduced by the base station;

[0055] Channel estimation stage: The base station receives the pilot signal Y sent by the kth user b By right-multiplying the conjugate of the pilot sequence, the received signal of the kth user in the bth subframe is expressed as:

[0056]

[0057] in, * indicates conjugation in matrix operations, P UE Represents the signal power sent by each user;

[0058] Define the set of B subframes at the i-th hour time scale, the expression is as follows:

[0059]

[0060] Here, the subscript i is the index of the small time scale;

[0061] Then, in each small time scale, according to formula (1.18), the observation equation of the user-RIS channel is obtained:

[0062] Y i =A i f k +N i (1.21)

[0063] Among them, N i represents the measurement noise, i.e., the additive noise introduced by the base station, p(N)~CN(0,R), which conforms to the complex Gaussian noise distribution with mean 0 and covariance matrix R;

[0064] According to the time correlation of the time-varying channel, the user-RIS channel coefficient vector f is obtained k The state transition equation is:

[0065] f k,i =diag(α)f k,i-1 +u i-1 (1.22)

[0066] Where α=J0(2πf d T f ), J0 is the first kind zero-order Bessel function, f d represents the maximum Doppler shift, T f is the user-RIS channel coherence time, u i-1 represents process noise, p(u)~CN(0,Q), which conforms to the complex Gaussian noise distribution with mean 0 and covariance matrix Q;

[0067] Then the user-RIS channel is estimated based on the observation equation and state transfer equation of the Kalman filter algorithm.

[0068] The channel estimation process is estimated according to the observation equation and state transfer equation of the Kalman filter algorithm, as follows:

[0069] First, the user-RIS channel coefficient vector f k And the posterior covariance matrix is ​​initialized, that is,

[0070]

[0071]

[0072] In the formula, I N Represents a square matrix of all 1s with dimension N;

[0073] Then, using the correlation between the current and previous channel and the process noise u i-1 The covariance matrix Q of the user-RIS channel coefficient vector f k Make predictions and calculate the covariance matrix of the prior estimation error, i.e.

[0074]

[0075]

[0076] Then, according to the covariance matrix of the prior estimation error and the measurement noise N i The covariance matrix R of Kalman gain coefficient is calculated as follows:

[0077]

[0078] In the formula, I MB Represents a square matrix of all 1s with dimension MB;

[0079] Finally, according to the Kalman gain coefficient K i For the user-RIS channel coefficient vector f kThe posterior estimate of is corrected and the covariance matrix of its posterior estimate error is calculated, that is,

[0080]

[0081] Repeat the prediction and correction operations to obtain the estimated value of the user-RIS channel

[0082] The present invention also provides a segmented channel estimation system of a RIS-assisted system, comprising:

[0083] Model building module: used to build a RIS-assisted wireless communication system model, in which there is no direct link between the base station and the user;

[0084] RIS-base station channel estimation module: uses a decomposition-optimization algorithm to estimate the RIS-base station channel in the RIS-assisted wireless communication system model;

[0085] User-RIS channel estimation module: estimates the user-RIS channel in the RIS-assisted wireless communication system model based on the time correlation of the channel and the Kalman filter algorithm.

[0086] The present invention also provides a segmented channel estimation system device of a RIS-assisted system, comprising:

[0087] Memory: a computer program storing the above-mentioned segmented channel estimation system method of the RIS-assisted system, which is a computer-readable device;

[0088] Processor: used to implement the above-mentioned segmented channel estimation system method of a RIS-assisted system when executing the computer program.

[0089] The present invention also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the above-mentioned segmented channel estimation system method of the RIS-assisted system can be implemented.

[0090] Compared with the prior art, the present invention has the following beneficial effects:

[0091] 1) The present invention estimates the cascade channel in segments at different time scales, that is, the present invention proposes to estimate the two parts of the channel from user to RIS and from RIS to base station separately, by estimating the part of the channel from RIS to base station at a large time scale and estimating the time-varying channel from user to RIS more frequently at a small time scale, thereby reducing the redundancy of channel estimation and the pilot overhead of channel estimation.

[0092] 2) The present invention proposes a decomposition-optimization algorithm, which first decomposes the complex channel estimation problem into N simple sub-problems to reduce the complexity of channel estimation. When estimating the channel from RIS to the base station, since the channel from the base station to the nth reflection unit of the RIS is only related to the reflection unit and has nothing to do with other reflection units, the complex problem is converted into N simple sub-problems, and each sub-problem is represented as a channel estimation from the base station to a specific reflection unit. For each sub-problem, its cost function is a convex function of the estimated value. First, its initial value is roughly calculated, and then multiple external iterations are performed based on the initial value. In each external iteration, the optimal estimated value of the current coefficient is obtained by fixing (M-1) coefficients. Since the cost function is a convex function of the estimated value and it is bounded, the cost function will maintain a downward trend in each iteration until convergence. The present invention effectively reduces the complexity and difficulty of channel estimation by decomposing the complex problem into multiple simple sub-problems and then optimizing the sub-problems to obtain the optimal solution.

[0093] 3) The present invention makes full use of the time correlation of the channel. In actual communication scenarios, due to the mobility of users, the part of the channel from the user to the RIS exhibits time-varying characteristics. The present invention fully considers the time correlation of the channel and adopts the Kalman filter algorithm to estimate the part of the channel from the user to the RIS, thereby improving the accuracy of channel estimation and laying a solid foundation for achieving more efficient and stable communication.

[0094] 4) The present invention reduces pilot overhead. For the quasi-static RIS to base station channel that remains relatively stable on a large time scale, and the time-varying channel from user to RIS that changes frequently on a small time scale, the present invention uses different time scales for estimation. This method of estimating the cascaded channels separately can effectively reduce the redundant information in the channel estimation process, thereby significantly reducing the pilot overhead.

[0095] 5) The present invention improves the accuracy of channel estimation. For quasi-static RIS-base station channels, the channel estimation problem is broken down into N sub-problems. For each sub-problem, the optimal estimate is solved by iterative optimization, which can more accurately capture the characteristics of the channel and avoid the error accumulation that may occur during the overall estimation. For time-varying user-RIS channels, the present invention fully explores and utilizes the time correlation of the channel, and uses the Kalman filter algorithm to continuously optimize the estimation results. It can make full use of the changing law of the channel in the time dimension, greatly improve the accuracy of channel estimation, and thus significantly improve the estimation performance.

[0096] In summary, compared with the prior art, the present invention fully considers the quasi-static characteristics of the RIS to base station channel and the time-varying characteristics of the user to RIS channel. When estimating the RIS-base station channel, a decomposition-optimization algorithm is used to decompose the channel estimation problem into N simple sub-problems, and the optimal solution for channel estimation is obtained by continuously optimizing the sub-problems; when estimating the user-RIS channel, Kalman filtering is used to perform channel estimation based on the correlation between the channel at the current moment and the channel at the previous moment; the present invention can ensure the accuracy of its channel estimation while reducing the pilot overhead, reduces the complexity of the channel estimation, and solves the technical problem of high pilot overhead in the channel estimation of the RIS-assisted communication system. BRIEF DESCRIPTION OF THE DRAWINGS

[0097] Figure 1 This is a schematic diagram of the structure of the RIS-assisted wireless communication system model provided by the present invention.

[0098] Figure 2 This is a schematic diagram of the channel estimation frame structure provided by the present invention.

[0099] Figure 3 This is a structural diagram of the dual-link pilot transmission provided by the present invention.

[0100] Figure 4 This is a structural diagram of the uplink pilot transmission provided by the present invention.

[0101] Figure 5 The figure is a relationship diagram between the normalized mean square error of the cascaded channels and the transmission signal-to-noise ratio provided by the present invention.

[0102] Figure 6 A relationship diagram between the pilot overhead and the number of reflection units provided by the present invention.

[0103] Figure 7 A relationship diagram between the pilot overhead and the number of users provided by the present invention. DETAILED DESCRIPTION

[0104] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings.

[0105] A segmented channel estimation method for a RIS-assisted system specifically comprises the following steps:

[0106] Step 1: Construct a RIS-assisted wireless communication system model, in which the direct link is blocked and the user signal reaches the base station after being reflected by the RIS; Figure 1 As shown, the RIS-assisted wireless communication system model includes a full-duplex base station, an intelligent metasurface RIS and K mobile users; the base station includes M antennas, and the intelligent metasurface RIS includes N reflection units.

[0107] Normally, the base station and RIS are in fixed positions, so the channel from RIS to the base station (RIS-base station channel for short) is quasi-static and has a long coherence time. It is only necessary to estimate the RIS-base station channel on a large time scale, that is, to estimate the RIS-base station channel once over a long period of time. On the other hand, due to the mobility of the terminal, the channel from the user to the RIS (user-RIS channel for short) is time-varying and has a short coherence time, so it is necessary to estimate the user-RIS channel on a small time scale, that is, to estimate it once over a short period of time. Then, within a period of time, the estimation frequency of the RIS-base station channel is lower, thus reducing the high pilot overhead caused by the frequent estimation of the cascaded channel due to the mobility of the user.

[0108] Assume that the channels in the system obey Rayleigh fading and the channels are independent of each other, that is:

[0109]

[0110] in, is the channel coefficient matrix from RIS to base station, is the channel coefficient vector from the kth user to the RIS; ρ g , They represent the large-scale fading factors of the corresponding transmission paths; I MN , I N Respectively represent the square matrix of all 1s of the corresponding dimension; cascade channel

[0111] Figure 2 The base station sends a pilot signal to the RIS through the downlink channel, and the RIS returns to the base station again after being reflected through the uplink channel. This pilot transmission process is called dual-link pilot transmission. First, on a large time scale T g Under this condition, dual-link pilots are used to estimate the high-dimensional quasi-static RIS-base station channel; then, at a small time scale T f In this case, the uplink pilot before data transmission is used to estimate the low-dimensional mobile user-RIS channel; compared with the existing cascade channel estimation method, this scheme can significantly reduce the pilot overhead. On the one hand, since the estimation frequency of the RIS-base station channel is lower than the estimation frequency of the user-RIS channel, its average pilot overhead can be reduced in the long run. On the other hand, because the RIS-base station channel is a high-dimensional quasi-static channel with a total of M×N different channels, the minimum pilot overhead required to estimate these M×N channels is M×N. The user-RIS channel is low-dimensional, with N×1 channels. Therefore, NK pilots can be used to estimate the user-RIS channel {f k |1≤k≤K}, instead of using MNK pilots to estimate the concatenated channel {C k |1≤k≤K}, Ck is a concatenated channel. Therefore, the required pilot overhead can be greatly reduced.

[0112] Step 2: Use a decomposition-optimization algorithm to estimate the RIS-base station channel in the RIS-assisted wireless communication system model;

[0113] Pilot transmission stage: Since the RIS-base station channel needs to be estimated, the user end does not need to send or receive pilots, but the base station sends and receives pilot signals. The pilot signal is sent using the m1th antenna and reflected by the RIS. The pilot signal is received by the m2th antenna at the base station. Figure 3 As shown, the pilot transmission frame in the dual-link pilot transmission of the base station consists of (N+1) subframes, each of which includes L time slots. After each subframe is sent, the reflection coefficient of RIS is adjusted. Since the remaining (M-1) base station antennas send zero pilots, the pilot signal received by the m2th antenna at the base station can be expressed as:

[0114]

[0115] In the formula, represents the channel vector from RIS to the m2th antenna of the base station, T represents the transpose, represents the reflection coefficient vector of RIS, Indicates the non-zero pilot transmitted by the m1th antenna of the base station in the lth (l=1, 2, ..., L)th time slot of the tth subframe; represents the channel vector from the m1th antenna of the base station to the RIS; represents environmental reflections, since some objects other than the reflection units of the RIS can also reflect the pilot signal; Residual self-interference This is mainly caused by the fact that when the base station works in full-duplex mode, the signal is directly transmitted from the m1th antenna of the base station to the m2th antenna. represents the noise received by the m2th antenna of the base station. The current self-interference suppression method can reduce the interference to only about 3dB higher than the receiver noise. Therefore, assuming that the remaining self-interference after self-interference elimination is in and There is no significant order of magnitude difference between them.

[0116] Defining index tuples After (N+1) subframes are sent, all pilot signals in the received transmission frame are obtained. The index tuple of the definition Wherein, m1 represents the index of the transmitting antenna, and m2 represents the index of the receiving antenna.

[0117] For a given The (N+1) pilot signals transmitted from the m1th transmitting antenna of the base station and received by the m2th receiving antenna after being reflected by the RIS can be expressed as:

[0118]

[0119] When sending a non-zero pilot When, P BS represents the transmission power of the base station. Substituting formula (1.2) into formula (1.3), we can get:

[0120]

[0121] In the formula, The elements in contain the unknown quantities to be estimated, 1 1×(N+1) represents an N+1 column full-1 vector, represents the reflection coefficient matrix of RIS; Formula (1.4) can be further expressed as:

[0122]

[0123] Channel estimation phase: The quasi-static RIS-base station channel is estimated in three phases.

[0124] First, the problem is divided into N independent subproblems, where each subproblem is to estimate the channel between the base station and a single reflection unit in the RIS. Second, an initial estimate is calculated for each subproblem. Finally, the RIS-base station channel estimate is iteratively optimized using the coordinate descent method.

[0125] 1) In the first stage, the quasi-static RIS-base station channel estimation problem is divided into N independent sub-problems, where each sub-problem is to estimate the channel between the base station and a single reflection unit in the RIS; for a given By multiplying formula (1.5) by the inverse matrix of F, we can get Estimates:

[0126]

[0127] In the formula, express The elements in , and a single element represents the estimation of the channel product from the m1th antenna to the nth RIS reflection unit and from the m2th antenna to the nth RIS reflection unit; further rearranging formula (1.6), we can get

[0128]

[0129] In the formula, Indicates g m1,n g m2,n The estimate of in represents the channel from the m1th antenna of the base station to the nth reflection unit, represents the channel from the m2th antenna of the base station to the nth reflection unit, is the estimation error.

[0130] For a particular reflection unit, the variable It depends on the channel coefficient between the base station and the nth element of RIS, but has nothing to do with the channel coefficients related to other elements of RIS. In other words, the quasi-static RIS-base station channel estimation problem can be divided into N independent sub-problems, where the Nth sub-problem is The channel coefficients {g m,n |1≤m≤M}.

[0131] Next, solve these N sub-problems separately to estimate the quasi-static RIS-base station channel. For a given n-th reflection unit, the estimation of the n-th sub-problem is expressed as:

[0132]

[0133] In the formula, the cost function

[0134] 2) In the second stage, the initial value of the nth subproblem is calculated;

[0135] The initial value estimate is rough but easy to calculate, which helps the third stage to converge faster. Take 1≤m1≤m2≤L,m3≠m1,m2, then the initial value of the channel coefficient estimate from the m′th antenna of the base station to the nth reflection unit in the RIS, that is, the initial value expression of the nth subproblem is:

[0136]

[0137] In the formula,

[0138] 3) In the third stage, the estimation of the RIS-base station channel coefficients is iteratively optimized using the coordinate descent method according to the initial value of the nth subproblem;

[0139] The coordinate descent method is used to continuously iterate and optimize the estimated value. There are multiple outer iterations based on the coordinate descent method, and each outer iteration consists of M inner iterations. In each outer iteration, the estimates of all M coefficients from the first to the last are optimal. In each inner iteration, M-1 coefficients are fixed to estimate the optimal value of the remaining coefficient.

[0140] Specifically, in the i-th external iteration, the M RIS-base station channel coefficients are estimated in turn, and the estimated values ​​of the previous (m-1) RIS-base station channel coefficients are The optimal solution has been obtained in the i-th external iteration, and the estimated values ​​of the following (Mm) RIS-base station channel coefficients are Using the optimal solution obtained in the last external iteration, the mth internal iteration estimate for the ith external iteration is It can be expressed as:

[0141]

[0142] according to Find the optimal solution:

[0143]

[0144] By formula Knowable About g m,n is a convex function, so update g according to formula (1.11) m,n The estimated value of J n The value of will decrease or remain unchanged.

[0145]

[0146] Then after M internal iterations, we know

[0147]

[0148] Then after I max After external iterations, we know

[0149]

[0150] J n The value of is bounded and monotonic, so the estimation of the RIS-base station channel It will eventually converge.

[0151] Then repeat the second and third steps for each sub-problem to obtain the estimated value of the quasi-static RIS-base station channel

[0152] Step 3: Estimate the user-RIS channel in the RIS-assisted wireless communication system model based on the time correlation of the channel and the Kalman filter algorithm;

[0153] Pilot transmission stage: For estimating the user-RIS channel, an uplink pilot transmission scheme is adopted. Figure 3 As shown, the small time scale T fIt is further divided into two parts, one of which is used for pilot transmission and the other for data transmission. The pilot transmission frame consists of B subframes, each subframe includes K time slots, and the RIS reflection coefficient matrix between different subframes is different. In the kth time slot of each subframe, the pilot signal sent by the kth user is x k,k , whose structure is as follows Figure 4 shown.

[0154] In the pilot transmission phase, after the pilot sent by the user is reflected by the RIS, in the b-th subframe, the pilot signal received by the base station from the k-th user is expressed as:

[0155]

[0156] Where Y b is the pilot signal received by the base station, is the channel coefficient matrix from RIS to base station, is the channel coefficient vector from the kth user to RIS, represents the reflection coefficient vector of the b-th subframe in the uplink pilot transmission phase, x k is the pilot signal sent by the kth user, N b Additive noise introduced by the base station;

[0157] Channel estimation phase: After each subframe is sent, the base station receives the signals sent by K users and estimates the pilot signal Y received by the base station. b Right multiply the conjugate of the pilot sequence. Since the pilot sequences of different users are orthogonal to each other, that is, Among them, P UE represents the signal power sent by each user, so the received signal of the kth user in the bth subframe is expressed as:

[0158]

[0159] in, * indicates conjugation in matrix operations.

[0160] Define the set of B subframes at the i-th hour time scale, and the expression is as follows:

[0161]

[0162] Here, the subscript i is the index of the small time scale.

[0163] Then, in each small time scale, according to formula (1.16), the observation equation of the user-RIS channel is obtained:

[0164] Y i =A i f k +Ni (1.20)

[0165] Among them, N i represents the measurement noise, i.e., the additive noise introduced by the base station;

[0166] According to the time correlation of the time-varying channel, the user-RIS channel coefficient vector f is obtained k The state transition equation is:

[0167] f k,i =diag(α)f k,i-1 +u i-1 (1.21)

[0168] Where α=J0(2πf d T f ), J0 is the zero-order Bessel function of the first kind, f d represents the maximum Doppler shift, T f is the user-RIS channel coherence time, u i-1 represents process noise.

[0169] Combining formula (1.20) and formula (1.21) we can get the observation equation and state transfer equation of the Kalman filter algorithm

[0170]

[0171] Among them, N i is the measurement noise, p(N)~CN(0,R), which conforms to the complex Gaussian noise distribution with mean 0 and covariance matrix R; u i-1 is the process noise, p(u)~CN(0,Q), which conforms to the complex Gaussian noise distribution with mean 0 and covariance matrix Q.

[0172] Next, the user-RIS channel is estimated based on the observation equation and state transfer equation of the Kalman filter algorithm.

[0173] The specific steps of the Kalman filter channel estimation process are as follows:

[0174] First, the user-RIS channel coefficient vector f k And the posterior covariance matrix is ​​initialized, that is,

[0175]

[0176] In the formula, I N Represents a square matrix of all 1s with dimension N;

[0177] Then, the correlation between the current and previous channels is used to calculate the user-RIS channel coefficient vector f k Make a priori predictions, that is

[0178]

[0179] in, represents the prior prediction value at the i-th hourly scale, represents the posterior estimate at the i-1th hourly scale;

[0180] At the same time, the covariance matrix of the prior estimation error is calculated based on the prior estimation error, that is,

[0181]

[0182] in, Represents the covariance matrix of the prior estimation error at the i-1th hourly scale;

[0183] The prior estimation error can be expressed as

[0184]

[0185] Then, at the i-th hourly time scale, the user-RIS channel coefficient vector f k,i The posterior estimate, the posterior estimate error, and the covariance matrix of the posterior estimate error can be expressed as:

[0186]

[0187] Among them, f k,i is the true value of the user-RIS channel coefficient vector at the i-th hourly time scale, cov{,} represents the covariance operation, represents the expectation operator; K i represents the Kalman gain matrix, which determines the observed value Y i The weight used when updating the state estimate is to balance the prior estimate and the observed value Y by adjusting the weight i contribution to obtain the optimal state estimation.

[0188] Substituting the observation equation in equation (1.22) and equation (1.28) into equation (1.29), we can get:

[0189]

[0190] Then its posterior covariance matrix is

[0191]

[0192] According to the minimum mean square error criterion, About K i The partial derivative of , setting its partial derivative to zero, can get the optimal Kalman gain K i ,Right now

[0193]

[0194] The Kalman gain K can be obtained by rearranging i for

[0195]

[0196] In the formula, I MB Represents a square matrix of all 1s with dimension MB;

[0197] Finally, substitute equation (1.34) into equation (1.28) and equation (1.30) respectively to obtain the user-RIS channel coefficient vector f k The posterior estimate of is corrected, and the covariance matrix of the posterior estimate error is calculated as:

[0198]

[0199] in, is the covariance matrix of the prior estimation error;

[0200] Repeat the prediction and correction operations to obtain the estimated value of the user-RIS channel

[0201] In summary, the Kalman filtering process at the i-th hourly scale is as follows: first, in the prediction stage, the covariance matrix of the prior prediction value at the i-th hourly scale and the prior estimation error at the i-th hourly scale is calculated based on the posterior estimation value at the i-1th hourly scale; then, in the correction stage, the Kalman gain at the i-th hourly scale is calculated, and the covariance matrix of the posterior estimation value and the posterior estimation error is corrected based on the Kalman gain.

[0202] In the present invention, by performing 200 Monte Carlo experiments on Matlab, the relationship between the normalized mean square error of channel estimation and the transmission signal-to-noise ratio is as follows: Figure 5 , where M is the base station antenna, N is the RIS reflection unit, K is the number of users, L is the time slot in the large-scale channel estimation frame, and the large-scale channel fading is modeled as The large-scale fading factor ρ0 = -20dB, assuming α g =2.1,α fk =4.2, d g =10,d fk =20, the user moves at a speed of 10m / s, and the carrier frequency is 2.6GHz. And assuming that the quasi-static RIS-base station channel estimation frequency is 5 times the time-varying channel user-RIS channel estimation frequency, the channel estimation error comparison between the method of the present invention and the three-stage method and the compressed sensing method mentioned in the prior art is as follows: Figure 5It can be seen that the accuracy of the segmented channel estimation of the present invention is higher.

[0203] The present invention adopts different time scales for estimation, respectively, for quasi-static channels that remain relatively stable on a large time scale, and for time-varying channels from users to RIS that change frequently on a small time scale. In addition, the channel from RIS to the base station is an M×N high-dimensional channel, while the channel from the user to RIS is an N×1 low-dimensional channel. In traditional overall channel estimation methods, the estimation of high-dimensional channels often introduces a large amount of unnecessary calculations and resource consumption. Separating the high-dimensional channels from the low-dimensional channels can avoid such redundancy, thereby significantly reducing pilot overhead. When the number of users K=8 and the number of base station antennas M=32, the relationship between the three-stage method mentioned in the prior art and the method mentioned in patent CN117614775A and the channel estimation pilot overhead of the present invention regarding the RIS reflection unit is as follows: Figure 6 As shown, the pilot sequence length τ of patent CN117614775A is assumed to be 8. When the base station antennas M = 32 and the RIS reflection units N = 64, the relationship between the three-stage method of the prior art and the method mentioned in patent CN117614775A and the channel estimation pilot overhead of the present invention with respect to the number of users is as follows: Figure 7 As shown, it can be seen that the segmented channel estimation method of the present invention can effectively reduce redundant information in the channel estimation process, thereby significantly reducing pilot overhead.

[0204] The present invention also provides a segmented channel estimation system of a RIS-assisted system, comprising:

[0205] Model building module: used to build a RIS-assisted wireless communication system model, in which there is no direct link between the base station and the user;

[0206] RIS-base station channel estimation module: uses a decomposition-optimization algorithm to estimate the RIS-base station channel in the RIS-assisted wireless communication system model;

[0207] User-RIS channel estimation module: estimates the user-RIS channel in the RIS-assisted wireless communication system model based on the time correlation of the channel and the Kalman filter algorithm.

[0208] The present invention also provides a segmented channel estimation system device of a RIS-assisted system, comprising:

[0209] Memory: a computer program storing the above-mentioned segmented channel estimation system method of the RIS-assisted system, which is a computer-readable device;

[0210] Processor: used to implement the above-mentioned segmented channel estimation system method of a RIS-assisted system when executing the computer program.

[0211] The present invention also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the above-mentioned segmented channel estimation system method of the RIS-assisted system can be implemented.

Claims

1. A segmented channel estimation method for a RIS-assisted system, characterized in that: include: Step 1: Construct a RIS-assisted wireless communication system model in which there is no direct link between the base station and the user; Step 2: Use a decomposition-optimization algorithm to estimate the RIS-base station channel in the RIS-assisted wireless communication system model; Step 3: Estimate the user-RIS channel in the RIS-assisted wireless communication system model based on the time correlation of the channel and the Kalman filter algorithm.

2. The segmented channel estimation method of a RIS-assisted system according to claim 1, characterized in that: The channels in the system obey Rayleigh fading and are independent of each other, that is, in, is the channel coefficient matrix from RIS to base station, referred to as RIS-base station channel; is the channel coefficient vector from the kth user to RIS, referred to as user-RIS channel; ρ g , ρ fk They represent the large-scale fading factors of the corresponding transmission paths; I MN , I N Respectively represent the square matrix of all 1s of the corresponding dimension; cascade channel 3. The segmented channel estimation method of a RIS-assisted system according to claim 1, characterized in that: In step 2, at a large time scale T g Under the condition, the dual link pilot is used to estimate the RIS-base station channel; in the step 3, at a small time scale T f In this case, the uplink pilot signal before data transmission is used to estimate the user-RIS channel.

4. The segmented channel estimation method of a RIS-assisted system according to claim 3, characterized in that: The estimation process of the RIS-base station channel in step 2 is as follows: Pilot transmission stage: The pilot transmission frame in the dual-link pilot transmission of the base station consists of (N+1) subframes, each subframe includes L time slots, and after each subframe is sent, the reflection coefficient of RIS is adjusted. Then, the pilot signal received by the m2th antenna at the base station can be expressed as: In the formula, represents the channel vector from RIS to the m2th antenna of the base station, T represents the transpose, represents the channel vector from the m1th antenna of the base station to the RIS, represents the reflection coefficient vector of RIS, represents the environmental reflection, Indicates the non-zero pilot transmitted by the m1th antenna of the base station in the lth (l=1,2,...,L)th time slot of the tth subframe, represents the residual self-interference, Indicates the noise received by the m2th antenna of the base station; Defining index tuples Among them, m1 represents the index of the transmitting antenna, and m2 represents the index of the receiving antenna. After (N+1) subframes are sent, all pilot signals received in the transmission frame are obtained. For a given When sending a non-zero pilot When P BS represents the transmission power of the base station. The (N+1) pilot signals transmitted from the m1th transmitting antenna of the base station and received by the m2th receiving antenna after being reflected by the RIS can be expressed as: In the formula, The elements in contain the unknown quantities to be estimated, 1 1×(N+1) represents an N+1 column vector of all 1s, represents the reflection coefficient matrix of RIS; Channel estimation phase: The quasi-static RIS-base station channel is estimated in three phases; 1) In the first stage, the RIS-base station channel estimation problem is divided into N independent sub-problems, where each sub-problem is to estimate the channel between the base station and a single reflection unit in the RIS; for a given Multiply formula (1.4) by The inverse matrix of Estimates: In the formula, express The elements in , and a single element represents an estimate of the channel product from the m1th antenna to the nth RIS reflection unit and from the m2th antenna to the nth RIS reflection unit; Further rearranging formula (1.6) yields: In the formula, Indicates g m1,n g m2,n The estimate of in represents the channel from the m1th antenna of the base station to the nth reflection unit, represents the channel from the m2th antenna of the base station to the nth reflection unit, is the estimation error; 2) In the second stage, the initial value of the nth subproblem is calculated; For a given n-th reflection unit, the estimate of the n-th subproblem is expressed as: In the formula, the cost function Take 1≤m1≤m2≤L,m3≠m1,m2, the initial value of the channel coefficient estimation from the m′th antenna of the base station to the nth reflection unit in the RIS, that is, the initial value expression of the nth sub-problem is: In the formula, 3) In the third stage, the estimation of the RIS-base station channel coefficients is iteratively optimized using the coordinate descent method according to the initial value of the nth subproblem; Then, each sub-problem repeats the second and third phases to obtain the estimated value of the RIS-base station channel 5. The segmented channel estimation method of a RIS-assisted system according to claim 4, characterized in that: The estimation of the RIS-base station channel coefficient is iteratively optimized by the coordinate descent method, as follows: In the i-th external iteration, the M RIS-base station channel coefficients are estimated in turn to obtain the estimated values ​​of the previous (m-1) RIS-base station channel coefficients The optimal solution in the i-th outer iteration, the estimated value of the next (Mm) RIS-base station channel coefficients Using the optimal solution obtained in the last external iteration, the mth internal iteration estimate for the ith external iteration is It can be expressed as: according to Find the optimal solution: Update g according to formula (1.12) m,n The estimated value of Then after M internal iterations, we know Then after I max After external iterations, we know J n The value of is bounded and monotonic, so the estimation of the RIS-base station channel It will eventually converge.

6. The segmented channel estimation method of a RIS-assisted system according to claim 3, characterized in that: The estimation process of the user-RIS channel in step 3 is as follows: Pilot transmission phase: The small time scale T of the user-RIS channel is f It is divided into data transmission and uplink pilot transmission. The uplink pilot transmission frame consists of B subframes, each of which includes K time slots. After the user sends the pilot signal and it is reflected by RIS, in the b-th subframe, the base station receives the pilot signal sent by the k-th user, which is expressed as: Where Y b is the pilot signal received by the base station, is the channel coefficient matrix from RIS to base station, is the channel coefficient vector from the kth user to RIS, represents the reflection coefficient vector of the b-th subframe in the uplink pilot transmission phase, x k is the pilot signal sent by the kth user, N b Additive noise introduced by the base station; Channel estimation stage: The base station receives the pilot signal Y sent by the kth user b By right-multiplying the conjugate of the pilot sequence, the received signal of the kth user in the bth subframe is expressed as: in, * indicates conjugation in matrix operations, P UE Represents the signal power sent by each user; Define the set of B subframes at the i-th hour time scale, the expression is as follows: Here, the subscript i is the index of the small time scale; Then, in each small time scale, according to formula (1.18), the observation equation of the user-RIS channel is obtained: Y i =A i f k +N i (1.21) Among them, N i represents the measurement noise, i.e., the additive noise introduced by the base station, p(N)~CN(0,R), which conforms to the complex Gaussian noise distribution with mean 0 and covariance matrix R; According to the time correlation of the time-varying channel, the user-RIS channel coefficient vector f is obtained k The state transition equation is: f k,i =diag(α)f k,i-1 +u i-1 (1.22) Where α=J0(2πf d T f ), J0 is the first kind zero-order Bessel function, f d represents the maximum Doppler shift, T f is the user-RIS channel coherence time, u i-1 represents process noise, p(u)~CN(0,Q), which conforms to the complex Gaussian noise distribution with mean 0 and covariance matrix Q; Then the user-RIS channel is estimated based on the observation equation and state transfer equation of the Kalman filter algorithm.

7. The segmented channel estimation method of a RIS-assisted system according to claim 6, characterized in that: The channel estimation process is estimated according to the observation equation and state transfer equation of the Kalman filter algorithm, as follows: First, the user-RIS channel coefficient vector f k And the posterior covariance matrix is ​​initialized, that is, In the formula, I N Represents a square matrix of all 1s with dimension N; Then, using the correlation between the current and previous channel and the process noise u i-1 The covariance matrix Q of the user-RIS channel coefficient vector f k Make predictions and calculate the covariance matrix of the prior estimation error, i.e. Then, according to the covariance matrix of the prior estimation error and the measurement noise N i The covariance matrix R of Kalman gain coefficient is calculated as follows: In the formula, I MB Represents a square matrix of all 1s with dimension MB; Finally, according to the Kalman gain coefficient K i For the user-RIS channel coefficient vector f k The posterior estimate of is corrected and the covariance matrix of its posterior estimate error is calculated, that is, Repeat the prediction and correction operations to obtain the estimated value of the user-RIS channel 8. A segmented channel estimation system for a RIS-assisted system based on the method of claims 1-7, characterized in that: include: Model building module: used to build a RIS-assisted wireless communication system model, in which there is no direct link between the base station and the user; RIS-base station channel estimation module: uses a decomposition-optimization algorithm to estimate the RIS-base station channel in the RIS-assisted wireless communication system model; User-RIS channel estimation module: estimates the user-RIS channel in the RIS-assisted wireless communication system model based on the time correlation of the channel and the Kalman filter algorithm.

9. A segmented channel estimation system device for a RIS-assisted system, characterized in that: include: Memory: a computer program storing a segmented channel estimation system method of a RIS-assisted system according to any one of claims 1 to 7, which is a computer-readable device; Processor: used to implement the segmented channel estimation system method of a RIS-assisted system as described in any one of claims 1-7 when executing the computer program.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, it can implement the segmented channel estimation system method of a RIS-assisted system as described in any one of claims 1 to 7.

Images