A massive MIMO cross-frequency cooperative robust transmission method

By reconstructing the CSI of the high-frequency channel with statistical CSI of low-frequency channels and considering the error source in precoding, the problem of channel error and performance degradation in traditional methods under non-ideal conditions is solved, and higher sum of performance and robustness.

CN114143896BActive Publication Date: 2024-11-29SOUTHEAST UNIV
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202111505864.5
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-12-10
Publication Date
2024-11-29
Estimated Expiration
2041-12-10

AI Technical Summary

Technical Problem

In the cross-band MIMO communication system, the traditional linear precoding method is based on the instantaneous CSI design, which is greatly affected by time and noise, resulting in a degradation of channel error and precoding performance under non-ideal conditions.

Method used

By utilizing the statistical CSI of the low-frequency channel, the channel state information CSI of the high-frequency channel is reconstructed and substituted into and rate maximizing during the precoding process, the error source is considered to achieve robust transmission.

Benefits of technology

The system and rate performance are improved, and the robustness and transmission efficiency are greatly improved compared to traditional methods under non-ideal conditions.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114143896B_ABST
    Figure CN114143896B_ABST
Patent Text Reader

Abstract

The present invention is a large-scale MIMO cross-frequency collaborative robust transmission method. The method first uses the Lagrange multiplier method and the MM iteration method under ideal conditions for accurate CSI on high-frequency channels to derive a sum rate-maximizing precoding matrix expression. Then, under non-ideal conditions, two types of error sources, uplink detection and channel obsolescence, are analyzed. Based on the basic principles of LS and MMSE estimation and the first-order Markov process, the second-order statistics of the error are modeled with the help of the statistical CSI of the low-frequency channel. Finally, the channel information is reconstructed using the error statistics and channel estimation values, and substituted into the sum rate-maximizing precoding matrix expression to achieve downlink robust transmission of high-frequency channels. Compared with traditional downlink precoding transmission methods, the present invention has significant sum rate performance gain.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention specifically relates to modeling channel errors under non-ideal conditions by means of statistical channel state information (CSI) of low-frequency channels in a cross-band communication system, reconstructing channel CSI using error modeling values ​​and channel estimation values, substituting them into a rate maximization precoding process, and realizing downlink robust transmission, which belongs to the field of communications. Background Art

[0002] Massive Multiple-Input Multiple-Output (MIMO) communication technology is constantly developing. As the demand for channel capacity continues to increase, spectrum scarcity is also a problem that needs to be solved in the design of future communication systems. The application of multi-frequency communication systems has enabled the exploration of higher frequency resources. Typical multi-frequency communication systems include collaborative communications of millimeter wave and sub-6GHz channels, dual-frequency wireless local area networks (WLAN) (2.4GHz and 5GHz), etc. In addition to solving the problem of spectrum scarcity, multi-frequency communication systems can send information through co-located channel arrays at different frequencies, and improve system capacity through mutual collaboration between frequency points.

[0003] In the cross-band MIMO communication system, the base station side precodes the original data stream of the downlink transmission based on the detected CSI. The traditional linear precoding method designs the precoding matrix based on the instantaneous CSI of the channel. Due to its low complexity, it is widely used in the downlink transmission process. However, under most conditions, on the one hand, the mobility of the channel causes the instantaneous information to change rapidly; on the other hand, under low signal-to-noise ratio conditions, the channel detection value also has errors due to the influence of noise, which will lead to the decline of precoding performance. Compared with the instantaneous CSI, which is greatly affected by time and noise, the statistical CSI reflects the long-term characteristics of the channel and is less affected by the mobility of the terminal. How to reasonably use the statistical CSI to make up for the performance error caused by the inaccurate instantaneous CSI to a certain extent is also one of the important contents that need to be studied in the downlink channel transmission. In the same cross-band communication system, the statistical CSI corresponding to the high-frequency channel is difficult to obtain directly through measurement, while the statistical CSI of different frequency channels presents similar properties in terms of angle. Therefore, the statistical CSI of the low-frequency channel can be used to replace the statistical CSI of the high-frequency channel, thereby completing the downlink robust precoding transmission of the high-frequency channel. To this end, the present invention proposes a large-scale MIMO cross-frequency collaborative robust transmission method, which uses low-frequency statistical CSI to assist robust transmission on high-frequency channels, thereby avoiding the huge overhead of measuring and calculating statistical CSI on high-frequency channels, and taking channel errors into account to correct the accuracy of the precoding matrix, thereby improving the system and rate performance. Summary of the invention

[0004] Purpose of the invention: For cross-band MIMO systems, the present invention provides a large-scale multi-input multi-output cross-frequency collaborative robust transmission method, which uses the statistical CSI of the low-frequency channel to complete the downlink robust precoding with maximum sum rate on the high-frequency channel, and greatly improves the sum rate performance compared with the traditional precoding method under non-ideal conditions, proving the feasibility of cross-frequency collaboration.

[0005] Technical solution: A large-scale multi-input multi-output cross-frequency collaborative robust transmission method of the present invention comprises the following steps:

[0006] Step 1. In the cross-band MIMO communication system, the transmitting antenna arrays corresponding to different frequencies are co-located and arranged in parallel on the base station side, and the base station communicates with all users in the cell at the same time; each frequency corresponding channel uses the time division duplex (TDD) transmission mode. Assuming that the user side sends a pilot to the base station, the base station can estimate the instantaneous value of the uplink channel based on the received uplink detection information. Based on the reciprocal characteristics of uplink and downlink, the uplink channel state information CSI estimation value obtained by uplink detection is directly used in the design process of downlink precoding;

[0007] Step 2. Under ideal conditions, the base station side uses the accurate channel state information CSI at each moment for precoding, takes system sum rate maximization as the optimization goal, uses the Lagrange multiplier method to establish the problem model, and uses the minimum-maximum (MM, Minorize-Maximize) iteration method to obtain the expression of the sum rate maximization precoding matrix;

[0008] Step 3. Under non-ideal conditions, first, in the process of uplink detection, due to the influence of channel additive noise, there are errors in channel estimation. In the same cross-band MIMO system, for different channel estimation methods on the high-frequency channel, the second-order statistics of the estimation error can be modeled with the help of noise power and the statistical channel state information CSI of the low-frequency channel;

[0009] Step 4. Under non-ideal conditions, it takes time to detect the uplink channel in each time block, so the channel estimation value cannot be used by the base station for real-time precoding design; considering the scenario where there is a relative movement speed between the base station side and the user side, the channel changes over time and also causes an error between the uplink detection value and the ideal channel state information CSI. By establishing a first-order Markov model, the outdated error on the high-frequency channel is modeled and represented by the statistical channel state information CSI of the low-frequency channel;

[0010] Step 5. Reconstruct the channel state information CSI of the high-frequency channel based on the channel estimation value of the uplink detection and the estimation error and obsolete error modeled by the low-frequency channel statistical channel state information CSI in steps 3 and 4, and substitute the reconstructed channel state information CSI into the high-frequency channel and rate maximization precoding process. Use the low-frequency channel state information CSI to take all error sources into account to achieve cross-frequency collaborative robust transmission under non-ideal conditions.

[0011] in:

[0012] In step 2, the rate maximizing precoding matrix is ​​obtained. If the transmission power constraint of the entire system is P max , let the original number of data streams be S, the number of antennas at the receiving and transmitting ends be N and M respectively, and the channel matrix corresponding to user k is The precoding matrix is in represents a complex matrix of a certain dimension, and the information transmitted by user k is x k And satisfy the orthogonal condition, the noise vector is n k , then the inter-user interference and additive noise suffered by user k can be combined into an interference term The covariance matrix of the interference term can be expressed as

[0013]

[0014] In the formula Indicates the average, I N represents the N×N dimensional identity matrix, σ 2 represents the noise vector n k The power, (·) H Represents the conjugate transpose of the matrix; define R sum is the sum rate of the system, then according to Shannon’s theorem, the expected rate of user k is

[0015]

[0016] Where det(·) represents the determinant. Assuming the total number of users in the system is K, the system sum rate is

[0017]

[0018] In summary, the problem of solving the precoding matrix that maximizes the system traversal and rate is expressed as

[0019]

[0020]

[0021] where tr(·) represents the trace of the matrix, It represents the optimal solution that satisfies and maximizes the rate. Since the original optimization problem is relatively complex to solve, the original objective function is transformed into an optimization problem that is easier to solve, and the MM algorithm is used to iteratively solve it.

[0022] The MM algorithm is used for iteration. Let the precoding matrices corresponding to all users in the iteration be Then the covariance of the interference term corresponding to user k is

[0023]

[0024] According to Shannon's formula and the properties of the determinant, the rate value of user k at this time is

[0025]

[0026] Among them I S is the unit matrix of dimension S×S, and this method is used to represent the unit matrix below.

[0027]

[0028] At the same time, define the matrix in the nth iteration express

[0029]

[0030] Then in this iteration, the user k defines a constant as follows

[0031]

[0032] And the operator

[0033]

[0034]

[0035]

[0036] Then we get the optimal precoding matrix that satisfies the sum rate maximization condition in the n+1th iteration. for

[0037]

[0038] Where μ (n) is the precoding matrix The power constraint factor is adjusted by (n) Make The power constraints are met.

[0039] The second-order statistics of the estimation error in step 3 are modeled by using the noise power and the CSI of the low-frequency channel. For the least squares LS channel estimation method, the covariance matrix of the estimation error is expressed by R k,LS It means that the calculation method is

[0040] R k,LS =σ 2 I MN ,

[0041] Among them I MN is the identity matrix of dimension MN×MN. For the minimum mean square error MMSE channel estimation method, the covariance matrix of the estimation error is expressed by R k,MMSE It means that the calculation method is

[0042] R k,MMSE =σ 2 R k (R k +σ 2 I MN ) -1 ,

[0043] Where R k is the covariance matrix of the channel between the base station and user k. Based on the correlation between different frequency points in the cross-frequency communication system, R on the high-frequency channel k The covariance matrix of the low-frequency channel can be used for reconstruction.

[0044] In step 4, the outdated error on the high-frequency channel is modeled using the statistical channel state information CSI of the low-frequency channel. The error caused by outdated channel between the base station and user k at time t is It can be expressed as follows with the help of statistical CSI:

[0045]

[0046] Where V M and U N are the array response matrices of the receiving and transmitting ends respectively, the subscripts M and N represent the number of antennas of the receiving and transmitting ends respectively, Ω k Represents the discrete power distribution in the angle domain, W k,t represents a phase matrix composed of elements satisfying the standard Gaussian distribution at time t, β k is the correlation coefficient of the channel matrix between the uplink detection value at time t and the previous time.

[0047] The error reconstruction in step 5, the channel impulse response matrix H at time t k,t Using the detection value at the previous moment Estimation error statistics And the angular power distribution Ω k Reconstructed as

[0048]

[0049] Where W k,t-1 and W k,t are the angle domain phase matrices at time t-1 and time t respectively; the reconstructed H k,t Substitute into the robust precoding expression and calculate the operator required for robust transmission design at time t as well as Then substitute

[0050]

[0051] That is, through the first n The precoding vector Get the sum rate maximizing precoding vector obtained at the n+1th iteration at time t A similar method is used to iterate until convergence to obtain the final value of the precoding matrix, thus completing the design of the downlink robust transmission precoding matrix.

[0052] In the cross-frequency cooperative robust transmission in step 5, for the cross-frequency communication system, the angle domain statistical parameters Ω on the low-frequency and high-frequency channels k and The estimation error and outdated error on the high-frequency channel can be modeled using the statistical parameters on the low-frequency channel. In the design of the robust transmission method for the high-frequency channel, since the higher the frequency, the greater the noise, the angle domain statistics of the low-frequency channel CSI are used to complete the H k,t The reconstruction further improves H k,t accuracy, thereby achieving cross-frequency collaborative transmission.

[0053] Beneficial effects: Compared with the prior art, the present invention has the following advantages:

[0054] 1. In the multi-user scenario, the proposed precoding method takes the maximum system and rate as the optimization condition, considers the interference between users and the interference between data streams, and has a certain performance gain under the quasi-static channel condition compared with the traditional method.

[0055] 2. The proposed precoding method takes detection error and staleness error into consideration. In non-ideal scenarios, the error is represented by statistical CSI modeling, which has obvious gains compared to the traditional method based on instantaneous CSI.

[0056] 3. In cross-frequency communication systems, in order to solve the problem that statistical CSI of high-frequency channels is difficult to obtain, the proposed method makes full use of the statistical CSI of low-frequency channels to achieve cross-frequency collaborative robust transmission, which has obvious gain compared with the traditional precoding method based on instantaneous CSI on high-frequency channels. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 The figure is a flow chart of a method according to an embodiment of the present invention.

[0058] Figure 2 These are the implementation steps of the neutralization rate maximization robust precoding method of the present invention. DETAILED DESCRIPTION

[0059] In order to achieve the above-mentioned object of the invention, the massive MIMO cross-frequency cooperative robust transmission method of the present invention comprises the following steps:

[0060] (1) In the cross-band MIMO communication system, the transmit antenna arrays corresponding to different frequencies are co-located and arranged in parallel on the base station side, and the base station communicates with all users in the cell at the same time. Each frequency corresponding channel uses the time division duplex (TDD) transmission mode. Assuming that the user side sends a pilot signal to the base station, the base station can estimate the instantaneous value of the uplink channel based on the received uplink detection information. Based on the reciprocity of uplink and downlink, the uplink CSI estimation value obtained by uplink detection can be directly used in the design process of downlink precoding.

[0061] (2) Under ideal conditions, the base station side uses the accurate CSI at each moment for precoding, takes system sum rate maximization as the optimization goal, uses the Lagrange multiplier method to establish the problem model, and uses the MM iteration method to obtain the expression of the sum rate maximization precoding matrix.

[0062] (3) Under non-ideal conditions, firstly, in the uplink detection process, due to the influence of channel additive noise, there are errors in channel estimation. For different channel estimation methods, the second-order statistics of the estimation error can be expressed with the help of noise power and the statistical CSI of the low-frequency channel.

[0063] (4) Under non-ideal conditions, it takes time to detect the uplink channel in each time block, so the channel estimation value cannot be used by the base station for real-time precoding design. Considering the scenario where there is a relative movement speed between the base station side and the user side, the channel changes over time and causes an error between the uplink detection value and the ideal CSI. By establishing a first-order Markov model, the outdated error can be represented by the statistical CSI model of the low-frequency channel.

[0064] (5) The channel CSI is reconstructed based on the channel estimation value, estimation error, and outdated error of the uplink detection, and the reconstructed CSI is substituted into the process of rate maximization precoding. Compared with the traditional method, all error sources are taken into account to achieve robust transmission under non-ideal conditions.

[0065] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0066] The specific steps of the embodiment of the present invention are described below in conjunction with specific scenarios:

[0067] 1) Massive MIMO system model:

[0068] In a massive MIMO communication system, assume that the number of transmitting antennas is M, the number of receiving antennas is N, and the total number of users in the system is K. For user k, assume that the channel matrix is The precoding matrix is The number of original signal data streams sent by the base station is S, and the original signal is represented by vector In theory, the receiving end signal vector can be expressed as

[0069] y k =H k P k x k +n k ,

[0070] Where n k represents the additive noise vector.

[0071] Considering the interference between users under multi-user conditions, the receiving signal can be further expressed as

[0072]

[0073] The above formula shows that in a multi-user MIMO system, the error of the receiving end signal comes from two parts: inter-user interference and additive noise. In the actual channel transmission process, before the downlink channel precoding is performed, the channel CSI referenced by the precoding needs to be obtained through detection. Considering the time division duplex (TDD) transmission mode, assuming that the user side sends a pilot to the base station, the base station can estimate the instantaneous value of the uplink channel based on the received uplink detection information. Based on the reciprocal characteristics of uplink and downlink, the uplink channel estimation value obtained by uplink detection can be directly used in the design process of downlink precoding.

[0074] Since it takes time to detect the uplink channel in each time block, the channel estimation value cannot be used by the base station for real-time precoding design. Therefore, the total time sampling length is divided into several time slots, each of which includes several sampling points. Assume that the number of sampling points in each time slot is N B, each sampling point corresponds to a time block, the first time block is used for uplink detection of the channel, and the next N B -1 sampling point will refer to the channel estimation value detected in the first time block to design precoding and realize downlink transmission.

[0075] 2) Downlink precoding with maximum rate

[0076] Assume that the transmission power constraint of the entire system is P max , let the original data stream be S, the number of antennas at the receiving and transmitting ends be N and M respectively, and the channel matrix corresponding to user k is The precoding matrix is At the same time, assume that the transmitted information is x k And the orthogonality condition is met, the noise vector is n k , then the interference between users and additive noise can be combined into an interference term The covariance matrix of the interference term can be expressed as

[0077]

[0078] Define R sum is the sum rate of the system, then according to Shannon’s theorem, the sum rate of user k is

[0079]

[0080] where det(·) represents the determinant, and the system and rate are

[0081]

[0082] In summary, the problem of solving the precoding matrix that maximizes the system traversal and rate can be expressed as

[0083]

[0084]

[0085] where tr(·) represents the trace of the matrix, (·) H Represents the conjugate transpose of the matrix. Since the original optimization problem is relatively complex to solve, the original objective function is transformed into an optimization problem that is easier to solve and is solved iteratively using the MM algorithm.

[0086] In the nth MM algorithm iteration, the precoding matrix corresponding to this iteration is When , the corresponding interference term covariance is

[0087]

[0088] According to Shannon's formula and the properties of the determinant, the rate value of user k at this time is

[0089]

[0090] Among them I S is the unit matrix of dimension S×S, and this method is used to represent the unit matrix below.

[0091]

[0092] At the same time, define the matrix in the nth iteration express

[0093]

[0094] Then in this iteration, the user k defines a constant as follows

[0095]

[0096] At the same time, define the following operators

[0097]

[0098]

[0099]

[0100]

[0101] Then define the precoding groups P1,…,P K Function as follows

[0102]

[0103] This function is the system and rate function in the precoding group The above gives an MM function for which the Shannon and rate formulas are applicable. Then the function g constructed in the nth iteration is * (P1,…,P K ) can be used to construct an optimization problem

[0104]

[0105]

[0106] Then in each iteration, a function g can be constructed * (P1,…,P K ), the optimization problem in the above formula is used to find the precoding group that can achieve the maximum traversal and rate in each iteration. Using the Lagrange multiplier method, the optimization problem can be converted to

[0107]

[0108] where μ (n) Denotes the Lagrange multiplier, and applies the above formula to any precoding matrix P in k∈[1,K] k Derivation, due to

[0109]

[0110] Then the derivative of the Lagrangian function is

[0111]

[0112] If the above formula is 0, the optimal precoding matrix that meets the conditions can be obtained. for

[0113]

[0114] 3) Massive MIMO robust transmission method

[0115] In the traversal and rate maximization method, the operator in each iteration is compared with the actual channel H k In actual large-scale MIMO systems, in order to reflect different sampling time blocks, let the instantaneous CSI of the kth user in the t time block be H k,t , then in the t-th time block, the covariance matrix of the channel interference in the n-th iteration is The obtained robust precoding matrix can be expressed as

[0116]

[0117] In the formula The calculation method is

[0118]

[0119]

[0120]

[0121]

[0122] The calculation of the above operators all depends on the instantaneous channel information H of the t time block k,tIn actual communication systems, since the channel CSI required for base station precoding is obtained through uplink detection, there is a channel estimation error in channel detection. At the same time, there is a time difference between channel detection and downlink transmission. When there is a relative movement speed between the base station and the user, the channel has differences at different sampling points. Therefore, there is an error between the uplink detected channel CSI and the channel on the downlink transmission time block due to channel aging. Based on the above error sources, it is assumed that the detected channel The error between the actual channel and the detected channel is ΔH k,t ,Right now

[0123]

[0124] Where ΔH k,t Including channel estimation error and outdated error caused by channel aging, in the process of operator calculation, For example

[0125]

[0126] The first term of this formula can be directly substituted into the channel detection value for calculation, while the second term involves the estimation error on each downlink detection time block. k,t It is impossible to obtain it directly in each time block, so we can consider calculating the error term from the perspective of statistical CSI. For the convenience of representation, assume that for any N×N dimensional conjugate matrix Φ1 and M×M dimensional conjugate matrix Φ2, we define

[0127]

[0128]

[0129] According to the sparsity of the channel in the angle domain, the above operator can be transformed into

[0130]

[0131] Where V M is the angle domain sampling matrix of the origin, and its dimension is M×B T , and define the dimension as N×B at the receiving end R The oversampling matrix U N ,B T and B R are the number of angular domain samples at the receiving and transmitting ends respectively. The two oversampling matrices are usually constructed based on the form of DFT matrix. Λ′ is B T ×B T The diagonal matrix of , whose diagonal elements can be obtained by B in the angle domain R ×B TEnergy coupling matrix Ω k The sum of the elements in each column is calculated. Similarly, the operator η 2,k,t (Φ2) can also use the energy coupling matrix Ω k In summary, the operator used in the precoding matrix calculation can be further expressed as

[0132]

[0133] Also use the following approximation

[0134]

[0135] Use D k,t Represents, then the operator The calculation can be transformed into

[0136]

[0137] From a statistical point of view, we can is an approximation of It can be expressed as

[0138]

[0139] In summary, given the known channel estimate And the error ΔH k,t Under the condition of angle domain statistical CSI, downlink robust precoding can be completed. The flowchart of the downlink robust transmission method is as follows Figure 2 shown.

[0140] 4) Cross-frequency collaborative robust transmission method

[0141] In the time division duplex (TDD) transmission mode, the channel impulse response matrix H at time t is k,t The detection value at the previous moment can be used Estimation error statistics And the angular power distribution Ω k Reconstructed as

[0142]

[0143] make In the known Estimation error energy distribution And the outdated error energy distribution Ω k In the unified cross-frequency communication system, the energy distribution of channels at different frequencies in the angle domain is approximately the same. Assuming that the low frequency and high frequency are represented by f1 and f2 respectively, the angle domain power distribution of the high frequency channel is Angular domain power distribution of low frequency channels can be used Approximately, without estimating the high-frequency statistical CSI, the channel matrix of f2 at time t is approximately calculated for

[0144]

[0145] Then the detection channel term can be expressed as

[0146]

[0147] The error term can be expressed as

[0148]

[0149] Since the robust transmission algorithm design needs to detect the channel item at the f2 frequency point, and the error term The angle domain statistical CSI is taken as a known condition. According to the above two equations, It can be obtained through channel detection. The angle domain statistical CSI can use the statistical parameters on the low frequency channel Calculate and then use the low-frequency statistical information to complete the robust precoding transmission method on the high-frequency channel.

Claims

1. A large-scale multi-input multi-output cross-frequency cooperative robust transmission method, characterized by: The transmission method comprises the following steps: Step 1. In the cross-band MIMO communication system, the transmitting antenna arrays corresponding to different frequencies are co-located and arranged in parallel on the base station side, and the base station communicates with all users in the cell at the same time; each frequency corresponding channel uses the time division duplex TDD transmission mode. Assuming that the user side sends a pilot to the base station, the base station can estimate the instantaneous value of the uplink channel based on the received uplink detection information. Based on the reciprocal characteristics of uplink and downlink, the uplink channel state information CSI estimation value obtained by uplink detection is directly used in the design process of downlink precoding; Step 2. Under ideal conditions, the base station side uses the accurate channel state information CSI at each moment for precoding, takes system and rate maximization as the optimization goal, uses the Lagrange multiplier method to establish the problem model, and uses the minimum-maximum iteration method to obtain the expression of the rate maximization precoding matrix; Step 3. Under non-ideal conditions, first, in the process of uplink detection, due to the influence of channel additive noise, there are errors in channel estimation. In the same cross-band MIMO system, for different channel estimation methods on the high-frequency channel, the second-order statistics of the estimation error can be modeled with the help of noise power and the statistical channel state information CSI of the low-frequency channel; Step 4. Under non-ideal conditions, it takes time to detect the uplink channel in each time block, so the channel estimation value cannot be used by the base station for real-time precoding design; considering the scenario where there is a relative movement speed between the base station side and the user side, the channel changes over time and also causes an error between the uplink detection value and the ideal channel state information CSI. By establishing a first-order Markov model, the outdated error on the high-frequency channel is modeled and represented by the statistical channel state information CSI of the low-frequency channel; Step 5. Reconstruct the channel state information CSI of the high-frequency channel based on the channel estimation value of the uplink detection and the estimation error and obsolete error modeled by the low-frequency statistical channel state information CSI in steps 3 and 4, and substitute the reconstructed channel state information CSI into the high-frequency channel and the rate maximization precoding process, and use the low-frequency channel state information CSI to take all error sources into account, so as to achieve cross-frequency collaborative robust transmission under non-ideal conditions; In step 2, the rate maximizing precoding matrix is ​​obtained. If the transmission power constraint of the entire system is P max , let the original number of data streams be S, the number of antennas at the receiving and transmitting ends be N and M respectively, and the channel matrix corresponding to user k is The precoding matrix is Where C represents a complex matrix of a certain dimension, and the information transmitted by user k is x k And satisfy the orthogonal condition, the noise vector is n k , then the inter-user interference and additive noise suffered by user k can be combined into an interference term The covariance matrix of the interference term can be expressed as Where E{·} represents the average, I N represents the N×N dimensional identity matrix, σ 2 represents the noise vector n k The power, (·) H Represents the conjugate transpose of the matrix; define R sum is the sum rate of the system, then according to Shannon’s theorem, the expected rate of user k is Where det(·) represents the determinant. Assuming the total number of users in the system is K, the system sum rate is In summary, the problem of solving the precoding matrix that maximizes the system traversal and rate is expressed as where tr(·) represents the trace of the matrix, It represents the optimal solution that satisfies the maximum sum rate, and the MM algorithm is used to iteratively solve it; The MM algorithm is used for iteration. Let the precoding matrices corresponding to all users in the iteration be Then the covariance of the interference term corresponding to user k is According to Shannon's formula and the properties of the determinant, the rate value of user k in the nth iteration is Among them I S is the unit matrix with dimension S×S. This method is used to represent the unit matrix in the following. At the same time, define the matrix in the nth iteration express Then in this iteration, the user k defines a constant as follows And the operator Then we get the optimal precoding matrix that satisfies the sum rate maximization condition in the n+1th iteration. for Where μ (n) is the precoding matrix The power constraint factor is adjusted by (n) Make Meet power constraints; The second-order statistics of the estimation error in step 3 are modeled by using the noise power and the CSI of the low-frequency channel. For the least squares LS channel estimation method, the covariance matrix of the estimation error is expressed by R k,LS It means that the calculation method is R k,LS =s 2 I MN , Among them I MN is the identity matrix of dimension MN×MN. For the minimum mean square error MMSE channel estimation method, the covariance matrix of the estimation error is expressed by R k,MMSE It means that the calculation method is R k,MMSE =s 2 R k (R k +s 2 I MN ) -1 , Where R k is the covariance matrix of the channel between the base station and user k. Based on the correlation between different frequency points in the cross-frequency communication system, R on the high-frequency channel k The covariance matrix of the low-frequency channel can be used for reconstruction; In step 4, the outdated error on the high-frequency channel is modeled using the statistical channel state information CSI of the low-frequency channel. The error caused by outdated channel between the base station and user k at time t is It can be expressed as follows using statistical CSI: Where V M and U N are the array response matrices of the receiving and transmitting ends respectively, the subscripts M and N represent the number of antennas of the receiving and transmitting ends respectively, Ω k Represents the discrete power distribution in the angular domain, W k,t represents a phase matrix composed of elements satisfying the standard Gaussian distribution at time t, β k is the correlation coefficient of the channel matrix between the uplink detection value at time t and the previous time; The error reconstruction in step 5, the channel impulse response matrix H at time t k,t Using the detection value at the previous moment Estimation error statistics And the angular power distribution Ω k Reconstructed as Where W k,t-1 and W k,t are the angle domain phase matrices at time t-1 and time t respectively; the reconstructed H k,t Substitute into the robust precoding expression and calculate the operator required for robust transmission design at time t as well as Then substitute That is, through the nth precoding vector at time t Get the sum rate maximizing precoding vector obtained at the n+1th iteration at time t A similar method is used to iterate until convergence to obtain the final value of the precoding matrix, thus completing the design of the downlink robust transmission precoding matrix.

2. The method for large-scale multi-input multi-output cross-frequency coordinated robust transmission according to claim 1, characterized in that: In the cross-frequency cooperative robust transmission in step 5, for the cross-frequency communication system, the angle domain statistical parameters Ω on the low-frequency and high-frequency channels k and The estimation error and outdated error on the high-frequency channel can be modeled using the statistical parameters on the low-frequency channel. In the design of the robust transmission method for the high-frequency channel, since the higher the frequency, the greater the noise, the angle domain statistics of the low-frequency channel CSI are used to complete the H k,t reconstruction, thereby achieving cross-frequency collaborative transmission.

Citation Information

Patent Citations

  • Large-scale MIMO robust precoding transmission method

    CN107483088A

  • Beamforming in cellular systems using same feedback information for different physical channels

    CN111989869A