IRS-assisted quantization bit optimization method for intra-cluster channel estimation in C-RAN

Abstract

The present invention discloses an IRS-assisted C-RAN intra-cluster channel estimation quantization bit optimization method. A single-antenna user UE communicates with a baseband processing unit BBU pool through a radio remote head RRH, and an IRS is deployed between the RRH and the UE to assist in communication. The RRH quantizes the signal from the UE and transmits it to the BBU pool through a fronthaul link. The present invention is directed to a three-stage channel estimation method in an IRS-assisted C-RAN uplink transmission system. In each stage, the RRH quantizes the received UE signal before transmitting it to the BBU pool, and proposes an optimization quantization bit method to minimize the estimation error of each stage under the fronthaul link capacity limitation.

Description

Technical Field

[0001] The present invention relates to the field of wireless communication technology, and in particular to a method for optimizing RRH quantization bits for channel estimation within a C-RAN cluster assisted by an intelligent reflecting surface IRS. Background Art

[0002] 5G technology has been widely used in social production, but with the rapid growth of the number of mobile devices, users have higher and higher requirements for communication systems. Therefore, in order to meet the rapid growth of mobile Internet, researchers have proposed a new antenna deployment architecture for cloud radio access network (C-RAN). C-RAN is composed of baseband processing unit (BBU) and radio remote head (RRH). BBU acts as a digital unit to realize base station functions from baseband processing to data packet processing, while RRH performs radio functions, including frequency conversion, amplification, and A / D and D / A conversion.

[0003] However, too many buildings in the city will affect the communication quality. Introducing intelligent reflective surface (IRS) technology into the C-RAN system can effectively improve the communication quality of the system. IRS is a new passive reflective array technology proposed in recent years. Each IRS surface is arranged with multiple passive reflective units made of metamaterials, and the reflection characteristics of each reflective unit are digitally controllable. Therefore, IRS can intelligently reflect the incident RF signal to form the desired reflected signal.

[0004] Quantization, in the field of digital signal processing, refers to the process of approximating the continuous value of a signal to a finite number of discrete values. During C-RAN fronthaul, quantizing the signal to be transmitted can effectively improve the fronthaul rate. Channel estimation is a technology that obtains channel state information between the user and the receiving end. It can greatly improve communication performance and increase the probability of the receiving end correctly demodulating the transmitted signal. Summary of the invention

[0005] In view of the problems existing in the prior art, the present invention provides an IRS-assisted C-RAN intra-cluster channel estimation and quantization bit optimization method. When quantization noise is introduced, the direct channel and reflected channel from the user UE to the remote radio head (RRH) and IRS providing service are estimated in three stages and the MSE expression is obtained. Then, under the restriction of the sum of RRH quantization bits, an optimization problem is proposed with the MSE minimization in each stage, and the optimization algorithm is used to solve it, and the channel estimation error in each stage is minimized using the optimization algorithm.

[0006] The technical solution of the present invention is as follows:

[0007] An IRS-assisted C-RAN intra-cluster channel estimation quantization bit optimization method is provided. For an uplink communication system of a cloud radio access network assisted by an intelligent reflector, the quantization bits of each RRH and each antenna are allocated according to the limitation of the intra-cluster fronthaul capacity to minimize the channel estimation error. The specific steps are as follows:

[0008] 1.1) K single-antenna user UEs communicate with the BBU pool through I RRHs, where the RRHs are divided into two sets and gather The communication between RRH and UE in the system requires IRS assistance. The RRH in the network does not need IRS to assist in communication. Each RRH has M antennas, and an IRS with N reflective units is deployed between the user and the RRH. The RRH compresses the received signal and then transmits it to the BBU pool through the fronthaul link.

[0009] 1.2) In the C-RAN system with IRS assistance only, at time t in the channel estimation phase, user k sends a signal to the RRH Each RRH receives the signal sent by the user through the direct link and the reflection link of the IRS. We assume that the reflection signal received by the i-th (i=1,…,I1) RRH is assisted by the i-th IRS, where. Therefore, the signal received by RRHi at time t is expressed as:

[0010]

[0011] in represents the direct channel from the kth user to RRHi, h i,k The covariance matrix of Indicates h i,k The path loss of p is the same as the transmission power of the user. i,n,t represents the reflection coefficient of the nth reflection unit of the ith IRS at time t for the considered coherent block, satisfying

[0012]

[0013] g k,n,i =s k,i,n r i,n represents the UE-IRS-RRH reflection channel from the kth user to the i-th RRH through the n-th reflection unit of the i-th IRS. is the channel from user k to the i-th IRS, where s k,i The covariance matrix of Indicates k,i of path loss. is the channel from the ith IRS to the ith RRH, where R i The covariance matrix of Represents R i of path loss. represents the additive white Gaussian noise (AWGN) at RRHi at time t, and its covariance matrix is

[0014] 1.3) Channel estimation is divided into three stages. In the first stage, all IRS reflection units are turned off, that is, φ i,n,t = 0, estimate the direct channel from all users to the RRH providing service. Definition is the pilot sequence sent by user k in the first phase, where q 2 = -1, τ1 ≥ K, Then the received signal of RRHi at the first stage at time t (t=1, ···, τ1) is

[0015]

[0016] Therefore, at time t (t=1,···,τ1), the signal received by the mth antenna of RRHi is

[0017]

[0018] Among them, H i,m =[h i,1,m ,···,h i,K,m ],

[0019] 1.4) After receiving the signal from the user, each RRH quantizes it and then forwards the quantized signal to the BBU pool. This quantization method can be achieved by adjusting the quantizer resolution. First, the signal received by the mth antenna of RRHi is normalized to in is a scaling factor:

[0020]

[0021] Normalized signal After passing through the quantizer, it can be quantized into The quantized signal can be calculated from the normalized signal as

[0022]

[0023] The symbol represents the integer closest to x, Q i,m represents the resolution of the quantizer when quantizing the received signal of the mth antenna of RRHi. The quantization error at the mth antenna of RRHi is and It can be approximated as [-δ i,m ,δ i,m ] uniform random variable, where has a mean of zero and a variance of The quantized signal is then sent to the BBU pool via an error-free fronthaul link.

[0024] 1.5) In the first stage, the signal received at the BBU pool at time t (t=1, ···, τ1) is

[0025]

[0026] in is compression noise.

[0027] The signal received at the BBU pool in the entire first stage is multiplied by both sides at the same time Available

[0028]

[0029] in

[0030] Then, using the linear minimum mean square error method, we can get h i,k The estimate is

[0031]

[0032] So h i,k The MSE is

[0033]

[0034] in

[0035]

[0036] The total error of the channel estimation in one stage is obtained as

[0037]

[0038] 1.6) In the second stage, we arbitrarily select one of the users (assuming it is user 1) to send a pilot symbol to RRHi, that is, Estimate the reflected channel to the serving RRHi and IRSi. Definition is the pilot sequence of user 1 in the second stage, then the received signal of RRHi at time t (t = τ1+1,…,τ1+τ2) in the second stage is

[0039]

[0040] where g 1,n,i =[g 1,n,i,1 ,···,g 1,n,i,M ] T .

[0041] Then at time t (t = τ1+1, ..., τ1+τ2), the signal received by the mth antenna of RRHi is

[0042]

[0043] therefore

[0044]

[0045] So we can get has a mean of zero and a variance of

[0046]

[0047] 1.7) Throughout the second stage, the effective received signal at the BBU pool is expressed as

[0048]

[0049] in

[0050] make As the total noise, we can get

[0051]

[0052] in

[0053]

[0054] Redefine Then, using the LMMSE method, we can get [g 1,1,i ,···,g 1,N,i ] is estimated to be

[0055]

[0056] So we can get the two-stage estimate [g 1,1,i ,···,g 1,N,i The MSE of ] is

[0057]

[0058] The total MSE of the two stages is

[0059]

[0060] 1.8) In the third stage of channel estimation, the remaining users send pilot symbols to RRHi, defining k=2,···,K is the pilot sequence of user k, then the received signal of RRHi at the third stage time t (t=τ1+τ2+1,…,τ1+τ2+τ3) is

[0061]

[0062] Because the UE-IRS-RRH reflection channels of user 1 and other users have the following relationship:

[0063]

[0064] Therefore, the received signal of RRHi at the third stage time t (t = τ1 + τ2 + 1, ..., τ1 + τ2 + τ3) can be re-expressed as follows:

[0065]

[0066] When t≥τ1+τ2+1, it is defined that at each time t, only the user with the serial number k t ,2≤k t ≤K users can transmit their pilot symbols to the RRH, and only M of the N IRS reflection units t ≤M, using the set Δ t Indicates that M t =|Δ t |≤M, turned on to reflect pilot symbols so that RRH can estimate in

[0067]

[0068]

[0069] in Therefore, for user number k t Users, assigned time slot to estimate its reflection channel. In the time slot, M IRS reflectors are turned on, and in the last time slot, the remaining IRS reflector units are turned on. Time slot estimation

[0070] 1.9) At time We set

[0071]

[0072]

[0073] At this time, the signal received by the mth antenna of RRHi is

[0074]

[0075] therefore

[0076]

[0077] because

[0078]

[0079]

[0080] therefore

[0081]

[0082] So we can get has a mean of zero and a variance of

[0083]

[0084] 1.10) In the third stage, assume that the channel estimation error Due to h i,k and g 1,n,i The estimation is imperfect. The effective received signal of the BBU pool at time t (t = τ1 + τ2 + 1, ..., τ1 + τ2 + τ3) is

[0085]

[0086] in

[0087] From the above formula, we can get

[0088]

[0089] in

[0090]

[0091]

[0092] definition Then, using the LMMSE method, we can get The estimate is

[0093]

[0094] Then, given G 1,t,i In the case of The MSE is

[0095]

[0096] By average G 1,t,i All channels of the estimated The MSE is

[0097]

[0098] So the total MSE of the third stage is

[0099]

[0100] Furthermore, for the uplink communication system of the cloud wireless access network assisted by the intelligent reflector, the quantization bits of each RRH and each antenna are allocated according to the limitation of the fronthaul capacity within the cluster to minimize the channel estimation error. The specific steps are as follows:

[0101] 2.1) In each stage of channel estimation, the number of quantization bits for the mth antenna of RRHi is Q i,m The RRH antenna quantization bit limit can be expressed as Where C is the sum of the number of quantized bits of the RRH antennas. Taking the sum of the number of quantized bits of the RRH antennas as a constraint, the optimization problem of minimizing the MSE at each stage can be expressed as follows (P1), (P2) and (P3):

[0102] (P1)

[0103] st

[0104] (C2):Q i,m ≥0 and Q i,m is an integer

[0105] (P2)

[0106] st

[0107] (C2):Q i,m ≥0 and Q i,m is an integer

[0108] (P3)

[0109] st

[0110] (C2):Q i,m ≥0 and Q i,m is an integer

[0111] The total channel estimation error ε of stages 1, 2, and 3 Ι , ε ΙΙ , ε ΙΙΙ Given by steps 1.6), 1.7), and 1.10) respectively, the optimization steps are as follows.

[0112] 2.2) For the optimization problem P1, initialize (each element is greater than or equal to 0), d 0 , δ 0 , D. Let p∈[0,d 0 ], f best =f(x 0 ), where the function f(x) is ε Ι About x 0 The function of . Define the upper limit of iteration as steps, and the number of iterations t≤steps. Define the vector set of random search directions as: Represents the direction of the i-th exploration.

[0113] 2.3) First determine the search direction By x t Get the position x of the left and right points lt and x rt ,in Substituting into the function f(x) we can get the function value f at the left and right points. l =f(x lt ) and f r =f(x rt ).

[0114] 2.4) If f best ≤f l &&f best ≤f r , then the updated search direction is recorded as According to x t Get the new left and right point positions x lt and x rt , compare the function values ​​of the new left and right point positions with f best Compare. If both f best ≤f l &&f best ≤f r &&i≤D, then continue to update And continue to look for new left and right points; if one of the conditions is not satisfied, stop looking for new left and right points.

[0115] 2.5) According to step 2.4), the function value f(x) of the left and right points is obtained. lt )、f(x rt ) and x t You can get xt+1 ,in The distance between the left and right points is d t+1 =ad t +b, a,b∈(0,1). And let f best =f(x t+1 ), the number of iterations becomes t=t+1, and the update step size is δ t+1 =uδ t ,u∈(0,1). If t≤steps at this time, return to step 2.3); otherwise, stop the iteration and the last obtained x t It is the optimal solution, the smallest estimated error value ε in problem P1 Ι is equal to the function value f(x t ).

[0116] 2.6) The steps for solving optimization problems P2 and P3 are the same as those for P1, so the optimal solutions of the three optimization problems can be solved using the above steps.

[0117] The beneficial effects of the present invention are as follows: under the limitation of the fronthaul capacity within the cluster, the present invention allocates the quantization bits of each RRH and each antenna to minimize the channel estimation error. Therefore, the method can better estimate the channel and obtain more accurate channel parameters, thereby effectively improving the communication quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0118] Figure 1 is the channel estimation error achieved by the method of the present invention and the comparative scheme under different fronthaul capacity constraints;

[0119] Figure 2 This is a model diagram of the IRS-assisted C-RAN single cluster uplink communication system in the present invention. DETAILED DESCRIPTION

[0120] The present invention is further described below in conjunction with the accompanying drawings.

[0121] An IRS-assisted C-RAN intra-cluster channel estimation quantization bit optimization method is provided. For an uplink communication system of a cloud radio access network assisted by an intelligent reflector, the quantization bits of each RRH and each antenna are allocated according to the limitation of the intra-cluster fronthaul capacity to minimize the channel estimation error. The specific steps are as follows:

[0122] 1.1) K single-antenna user UEs communicate with the BBU pool through I RRHs, where the RRHs are divided into two sets and gather The communication between RRH and UE in the system requires IRS assistance. The RRH in the network does not need IRS to assist in communication. Each RRH has M antennas, and an IRS with N reflective units is deployed between the user and the RRH. The RRH compresses the received signal and then transmits it to the BBU pool through the fronthaul link.

[0123] 1.2) In the C-RAN system with IRS assistance only, at time t in the channel estimation phase, user k sends a signal to the RRH Each RRH receives the signal sent by the user through the direct link and the reflection link of the IRS. We assume that the reflection signal received by the i-th (i=1,…,I1) RRH is assisted by the i-th IRS, where. Therefore, the signal received by RRHi at time t is expressed as:

[0124]

[0125] in represents the direct channel from the kth user to RRHi, h i,k The covariance matrix of Indicates h i,k The path loss of p is the same as the transmission power of the user. i,n,t represents the reflection coefficient of the nth reflection unit of the ith IRS at time t for the considered coherent block, satisfying

[0126]

[0127] g k,n,i =s k,i,n r i,n represents the UE-IRS-RRH reflection channel from the kth user to the i-th RRH through the n-th reflection unit of the i-th IRS. is the channel from user k to the i-th IRS, where s k,i The covariance matrix of Indicates k,i of path loss. is the channel from the ith IRS to the ith RRH, where R i The covariance matrix of Represents R i of path loss. represents the additive white Gaussian noise (AWGN) at RRHi at time t, and its covariance matrix is

[0128] 1.3) Channel estimation is divided into three stages. In the first stage, all IRS reflection units are turned off, that is, φ i,n,t = 0, estimate the direct channel from all users to the RRH providing service. Definition is the pilot sequence sent by user k in the first phase, where q 2 =-1, Then the received signal of RRHi at the first stage at time t (t=1, ···, τ1) is

[0129]

[0130] Therefore, at time t (t=1,···,τ1), the signal received by the mth antenna of RRHi is

[0131]

[0132] Among them, H i,m =[h i,1,m ,···,h i,K,m ],

[0133] 1.4) After receiving the signal from the user, each RRH quantizes it and then forwards the quantized signal to the BBU pool. This quantization method can be achieved by adjusting the quantizer resolution. First, the signal received by the mth antenna of RRHi is normalized to in is a scaling factor:

[0134]

[0135] Normalized signal After passing through the quantizer, it can be quantized into The quantized signal can be calculated from the normalized signal as

[0136]

[0137] The symbol represents the integer closest to x, Q i,m represents the resolution of the quantizer when quantizing the received signal of the mth antenna of RRHi. The quantization error at the mth antenna of RRHi is and It can be approximated as [-δ i,m ,δ i,m ] uniform random variable, where has a mean of zero and a variance of The quantized signal is then sent to the BBU pool via an error-free fronthaul link.

[0138] 1.5) In the first stage, the signal received at the BBU pool at time t (t=1, ···, τ1) is

[0139]

[0140] in is compression noise.

[0141] The signal received at the BBU pool in the entire first stage is multiplied by both sides at the same time Available

[0142]

[0143] in

[0144] Then, using the linear minimum mean square error method, we can get h i,k The estimate is

[0145]

[0146] So h i,k The MSE is

[0147]

[0148] in

[0149]

[0150] The total error of the channel estimation in one stage is obtained as

[0151]

[0152] 1.6) In the second stage, we arbitrarily select one of the users (assuming it is user 1) to send a pilot symbol to RRHi, that is, Estimate the reflected channel to the serving RRHi and IRSi. Definition is the pilot sequence of user 1 in the second stage, then the received signal of RRHi at time t (t = τ1+1,…,τ1+τ2) in the second stage is

[0153]

[0154] where g 1,n,i =[g 1,n,i,1 ,···,g 1,n,i,M ] T .

[0155] Then at time t (t = τ1+1, ..., τ1+τ2), the signal received by the mth antenna of RRHi is

[0156]

[0157] therefore

[0158]

[0159] So we can get has a mean of zero and a variance of

[0160]

[0161] 1.7) Throughout the second stage, the effective received signal at the BBU pool is expressed as

[0162]

[0163] in

[0164] make As the total noise, we can get

[0165]

[0166] in

[0167]

[0168] Redefine Then, using the LMMSE method, we can get [g 1,1,i ,···,g 1,N,i ] is estimated to be

[0169]

[0170] So we can get the two-stage estimate [g 1,1,i ,···,g 1,N,i The MSE of ] is

[0171]

[0172] The total MSE of the two stages is

[0173]

[0174] 1.8) In the third stage of channel estimation, the remaining users send pilot symbols to RRHi, defining k=2,···,K is the pilot sequence of user k, then the received signal of RRHi at the third stage time t (t=τ1+τ2+1,…,τ1+τ2+τ3) is

[0175]

[0176] Because the UE-IRS-RRH reflection channels of user 1 and other users have the following relationship:

[0177]

[0178] Therefore, the received signal of RRHi at the third stage time t (t = τ1 + τ2 + 1, ..., τ1 + τ2 + τ3) can be re-expressed as follows:

[0179]

[0180] When t≥τ1+τ2+1, it is defined that at each time t, only the user with the serial number k t ,2≤k t ≤K users can transmit their pilot symbols to the RRH, and only M of the N IRS reflection units t ≤M, using the set Δ t Indicates that M t =|Δ t |≤M, turned on to reflect pilot symbols so that RRH can estimate in

[0181]

[0182]

[0183] in Therefore, for user number k t Users, assigned time slot to estimate its reflection channel. In the time slot, M IRS reflectors are turned on, and in the last time slot, the remaining IRS reflector units are turned on. Time slot estimation

[0184] 1.9) At time We set

[0185]

[0186]

[0187] At this time, the signal received by the mth antenna of RRHi is

[0188]

[0189] therefore

[0190]

[0191] because

[0192]

[0193]

[0194] therefore

[0195]

[0196] So we can get has a mean of zero and a variance of

[0197]

[0198] 1.10) In the third stage, assume that the channel estimation error Due to h i,k and g 1,n,i The estimation is imperfect. The effective received signal of the BBU pool at time t (t = τ1 + τ2 + 1, ..., τ1 + τ2 + τ3) is

[0199]

[0200] in

[0201] From the above formula, we can get

[0202]

[0203] in

[0204]

[0205]

[0206] definition Then, using the LMMSE method, we can get The estimate is

[0207]

[0208] Then, given G 1,t,i In the case of The MSE is

[0209]

[0210] By average G 1,t,i All channels of the estimated The MSE is

[0211]

[0212] So the total MSE of the third stage is

[0213]

[0214] For the uplink communication system of the cloud wireless access network assisted by the intelligent reflector, the quantization bits of each RRH and each antenna are allocated according to the limitation of the fronthaul capacity within the cluster to minimize the channel estimation error. The specific steps are as follows:

[0215] 2.1) In each stage of channel estimation, the number of quantization bits for the mth antenna of RRHi is Q i,m The RRH antenna quantization bit limit can be expressed as Where C is the sum of the number of quantized bits of the RRH antennas. Taking the sum of the number of quantized bits of the RRH antennas as a constraint, the optimization problem of minimizing the MSE at each stage can be expressed as follows (P1), (P2) and (P3):

[0216] (P1)

[0217] st

[0218] (C2):Q i,m ≥0 and Q i,m is an integer

[0219] (P2)

[0220] st

[0221] (C2):Q i,m ≥0 and Q i,m is an integer

[0222] (P3)

[0223] st

[0224] (C2):Q i,m ≥0 and Q i,m is an integer

[0225] The total channel estimation error ε of stages 1, 2, and 3 Ι , ε ΙΙ , ε ΙΙΙ Given by steps 1.6), 1.7), and 1.10) respectively, the optimization steps are as follows.

[0226] 2.2) For the optimization problem P1, initialize (each element is greater than or equal to 0), d 0 , δ 0 , D. Let p∈[0,d 0 ], f best =f(x 0 ), where the function f(x) is ε ΙAbout x 0 The function of . Define the upper limit of iteration as steps, and the number of iterations t≤steps. Define the vector set of random search directions as: Represents the direction of the i-th exploration.

[0227] 2.3) First determine the search direction By x t Get the position x of the left and right points lt and x rt ,in Substituting into the function f(x) we can get the function value f at the left and right points. l =f(x lt ) and f r =f(x rt ).

[0228] 2.4) If f best ≤f l &&f best ≤f r , then the updated search direction is recorded as According to x t Get the new left and right point positions x lt and x rt , compare the function values ​​of the new left and right point positions with f best Compare. If both f best ≤f l &&f best ≤f r &&i≤D, then continue to update And continue to look for new left and right points; if one item is not satisfied, stop looking for new left and right points.

[0229] 2.5) According to step 2.4), the function value f(x) of the left and right points is obtained. lt )、f(x rt ) and x t You can get x t+1 ,in The distance between the left and right points is d t+1 =ad t +b, a,b∈(0,1). And let f best =f(x t+1 ), the number of iterations becomes t=t+1, and the update step size is δ t+1 =uδ t ,u∈(0,1). If t≤steps at this time, return to step 2.3); otherwise, stop the iteration and the last obtained x t It is the optimal solution, the smallest estimated error value ε in problem P1 Ι is equal to the function value f(xt ).

[0230] 2.6) The steps for solving optimization problems P2 and P3 are the same as those for P1, so the optimal solutions of the three optimization problems can be solved using the above steps.

Claims

1. An IRS-assisted C-RAN intra-cluster channel estimation quantization bit optimization method, characterized in that: for an uplink communication system of a cloud radio access network assisted by an intelligent reflector, according to the limitation of the intra-cluster fronthaul capacity, the quantization bits of each RRH and each antenna are allocated to minimize the channel estimation error, and the specific steps are as follows: 1.1) K single-antenna user UEs communicate with the BBU pool through I RRHs, where the RRHs are divided into two sets and gather The communication between RRH and UE in the system requires IRS assistance. The RRH in the system does not need IRS to assist in communication. Each RRH has M antennas, and an IRS equipped with N reflective units is deployed between the user and the RRH. The RRH compresses the received signal and then transmits it to the BBU pool through the fronthaul link. 1.2) In the C-RAN system with IRS assistance only, at time t in the channel estimation phase, user k sends a signal to the RRH Each RRH receives the signal sent by the user through the direct link and the reflection link of the IRS. Assume that the reflection signal received by the i-th (i=1,…,I1) RRH is assisted by the i-th IRS, where the signal received by RRHi at time t is expressed as: in represents the direct channel from the kth user to RRHi, h i,k The covariance matrix of Indicates h i,k The path loss of p represents the same transmission power of the user, φ i,n,t represents the reflection coefficient of the nth reflection unit of the i-th IRS at time t for the considered coherent block, satisfying: g k,n,i =s k,i,n r i,n represents the UE-IRS-RRH reflection channel from the kth user through the nth reflection unit of the ith IRS to the ith RRH, where is the channel from user k to the i-th IRS, where s k,i The covariance matrix of Indicates k,i The path loss is is the channel from the ith IRS to the ith RRH, where R i The covariance matrix of Represents Ri The path loss is represents the additive white Gaussian noise AWGN at RRHi at time t, and its covariance matrix is 1.3) Channel estimation is divided into three stages. In the first stage, all IRS reflection units are turned off, that is, φ i,n,t = 0, estimate the direct channel from all users to the RRH providing service, and define is the pilot sequence sent by user k in the first phase, where q 2 = -1, τ1 ≥ K, Then the received signal of RRHi at time t (t=1, ···, τ1) in the first stage is: Therefore, at time t (t=1,···,τ1), RRH i The signal received by the mth antenna is: Among them, H i,m =[h i,1,m ,···,h i,K,m ], 1.4) After receiving the signal from the user, each RRH quantizes it and then forwards the quantized signal to the BBU pool; First, the signal received by the mth antenna of RRH i is normalized to in is the noise power received by RRHi, is a scaling factor: Normalized signal After passing through the quantizer, it is quantized to The quantized signal can be calculated from the normalized signal as: The symbol represents the integer closest to x, Q i,m represents the resolution of the quantizer when quantizing the received signal of the mth antenna of RRHi; the quantization error at the mth antenna of RRHi is and Approximately [-δ i,m ,δ i,m ] uniform random variable, where has a mean of zero and a variance of The quantized signal is then sent to the BBU pool via an error-free fronthaul link; 1.5) The signal received at the BBU pool at time t (t=1, ···, τ1) in the first stage is: in is compression noise; The signal received at the BBU pool in the entire first stage is multiplied by both sides at the same time have to: in Then the linear minimum mean square error method is used to obtain h i,k The estimate is: So h i,k The MSE is: in The total error of the channel estimation in one stage is obtained as: 1.6) In the second stage, one of the users is randomly selected, assuming it is user 1, and a pilot symbol is sent to RRHi, that is, Estimate the reflected channel to the serving RRHi and IRSi and define is the pilot sequence of user 1 in the second stage, then the received signal of RRH i at time t (t = τ1+1,…,τ1+τ2) in the second stage is: where g 1,n,i =[g 1,n,i,1 ,···,g 1,n,i,M ] T ; Then at time t (t = τ1+1, ..., τ1+τ2), the signal received by the mth antenna of RRHi is: therefore So we can get has a mean of zero and a variance of: 1.7) Throughout the second phase, the effective received signal at the BBU pool is expressed as: in make As the total noise, we get: in is the expectation of the matrix in the brackets, and Redefine Then, the LMMSE method is used to obtain [g 1,1,i ,···,g 1,N,i ] is estimated as: So we get the second stage estimate [g 1,1,i ,···,g 1,N,i The MSE of ] is: The total MSE of the second stage is: 1.8) In the third stage of channel estimation, the remaining users send pilot symbols to RRH i, defining is the pilot sequence of user k, then the received signal of RRHi at the third stage time t (t = τ1 + τ2 + 1, ..., τ1 + τ2 + τ3) is: Because the UE-IRS-RRH reflection channels of user 1 and other users have the following relationship: where s k,i,n is the channel from user k to the nth reflection unit of the ith IRS; Therefore, the received signal of RRHi at the third stage time t (t = τ1 + τ2 + 1, ..., τ1 + τ2 + τ3) is re-expressed as follows: When t≥τ1+τ2+1, it is defined that at each time t, only the user with the serial number k t ,2≤k t ≤K users can transmit their pilot symbols to the RRH, and only M of the N IRS reflection units t ≤M, using the set Δ t Indicates that M t =|Δ t |≤M, turned on to reflect pilot symbols for RRH estimation in: in Therefore, for user number k t Users, assigned time slot to estimate its reflection channel; in the first In the time slot, M IRS reflectors are turned on, and in the last time slot, the remaining IRS reflector units are turned on; Time slot estimation 1.9) At time set up At this time, the signal received by the mth antenna of RRHi is: therefore because therefore So has a mean of zero and a variance of: 1.10) In the third stage, assume that the channel estimation error Due to h i,k and g 1,n,i The estimation is imperfect, and the effective received signal of the BBU pool at time t (t = τ1 + τ2 + 1, ..., τ1 + τ2 + τ3) is: in From the above formula, we can get: in definition Then, the LMMSE method is used to obtain The estimate is: Then, given G 1,t,i In the case of The MSE is: By average G 1,t,i All channels of the realization, estimated The MSE is: So the total MSE of the third stage is: For the uplink communication system of the cloud wireless access network assisted by the intelligent reflector, according to the limitation of the fronthaul capacity within the cluster, the quantization bits of each RRH and each antenna are allocated to minimize the channel estimation error. The specific steps are as follows: 2.1) In each stage of channel estimation, the number of quantization bits for the mth antenna of RRH i is Q i,m ; The RRH antenna quantization bit limit is expressed as Where C is the sum of the number of quantized bits of the RRH antennas. Taking the sum of the number of quantized bits of the RRH antennas as a constraint, the optimization problem of minimizing the MSE at each stage is expressed as follows (P1), (P2) and (P3): (C2):Q i,m ≥0 and Q i,m is an integer (C2):Q i,m ≥0 and Q i,m is an integer (C2):Q i,m ≥0 and Q i,m is an integer The total channel estimation error ε of stage 1, 2, and 3 Ι , ε ΙΙ , ε ΙIΙ Given by steps 1.6), 1.7), and 1.10), the optimization steps are as follows: 2.2) For the optimization problem P1, initialize d 0 , δ 0 , D; d 0 is the initial setting value of the distance between the left and right points, δ 0 is the initial step size, D represents the maximum number of vectors in the random search direction vector set at each iteration; let p∈[0,d 0 ], f best =f(x 0 ), where the function f(x) is ε Ι About x 0 function; define the upper limit of iteration as steps, the number of iterations t≤steps; define the vector set of random search directions as: represents the direction of the i-th exploration; 2.3) First determine the search direction By x t Get the position x of the left and right points lt and x rt ,in Substituting into the function f(x) we can get the function value f at the left and right points. l =f(x lt ) and f r =f(x rt ); 2.4) If f best ≤f l &&f best ≤f r , then the updated search direction is recorded as According to x t Get the new left and right point positions x lt and x rt , compare the function values ​​of the new left and right point positions with f best Compare; if both f are satisfied best ≤f l &&f best ≤f r &&i≤D, then continue to update And continue to look for new left and right points; if one of the conditions is not satisfied, stop looking for new left and right points; 2.5) According to step 2.4), the function value f(x) of the left and right points is obtained. lt )、f(x rt ) and x t You can get x t+1 ,in The distance between the left and right points is d t+1 =ad t +b, a,b∈(0,1); and let f best =f(x t+1 ), the number of iterations becomes t=t+1, and the update step size is δ t+1 =uδ t ,u∈(0,1); if t≤steps at this time, return to step 2.3); otherwise, stop the iteration and the last obtained x t It is the optimal solution, the smallest estimated error value ε in problem P1 Ι is equal to the function value f(x t ); 2.6) The steps for solving optimization problems P2 and P3 are the same as those for P1, so the above steps can be used to solve the optimal solutions of the three optimization problems.

Images