Massive MIMO high-speed rail communication channel prediction and transmission method

Abstract

The present invention discloses a large-scale MIMO high-speed rail communication channel prediction and transmission method. The base station obtains the quadruple beam domain channel information of the pilot segment through uplink channel detection, and predicts the spatial-frequency-time domain channel information of the data segment between the pilot symbols; the obtained quadruple beam domain channel information of the pilot segment is used to construct a quadruple beam basis posterior channel model, and the posterior spatial-frequency-time domain channel vector of the data segment after the pilot symbol is predicted; in uplink communication, the base station uses the predicted spatial-frequency-time domain channel information of the data segment between the pilot symbols to design the receiver; in downlink communication, the base station uses the predicted posterior spatial-frequency-time domain channel vector of the data segment after the pilot symbol to design the precoding matrix. The present invention can improve the accuracy of the channel information obtainable on the base station side without increasing the pilot overhead, can efficiently design the precoding matrix, effectively solve the serious problem of channel aging in high-speed rail communication, and improve performance.

Description

Technical Field

[0001] The present invention belongs to the field of wireless communication technology and relates to a large-scale MIMO high-speed rail communication channel prediction and transmission method and a related system. Background Art

[0002] With the rapid deployment of high-speed rail (HSR) worldwide, communication systems tailored for HSR systems have attracted widespread attention from academia and industry. Simultaneously, driven by the continuous evolution of wireless technology, the demand for higher transmission rates and improved quality of service in HSR communications is growing exponentially. Massive Multiple-Input Multiple-Output (MIMO), a key technology in 5G mobile communications, is one of the key technologies that can meet the communication needs of HSR. By equipping the base station (BS) with a large number of antennas, Massive MIMO can simultaneously serve multiple users using the same time-frequency resources, achieving multiplexing gain and significantly improving spectrum and energy efficiency.

[0003] However, providing services to multiple users simultaneously will cause severe inter-user interference. To address this problem, precoding technology has been widely studied and applied in high-speed rail communications due to its ability to effectively suppress interference and improve received signal strength. In fact, the design of the downlink precoder is highly dependent on the channel state information (CSI) available on the base station side. Typically, the channel state information on the base station side needs to be periodically estimated in the pilot segment through the uplink sounding process. The estimated CSI can be used for a period of time within the channel coherence time to ensure accuracy while avoiding excessive pilot overhead. However, due to the severe Doppler effect in high-speed rail communications, the spatial domain channel estimated from the pilot segment will quickly become outdated, resulting in severe channel aging effects and serious loss of precoder performance.

[0004] Although the spatial domain channel changes rapidly over time, its beam domain channel changes relatively slowly. Therefore, the relationship between the beam domain channel and the spatial channel can be used to predict the channel, thereby obtaining more accurate CSI to combat the channel aging effect. However, due to the limited resolution of the beam domain, the prediction error becomes more significant as the time interval between the predicted data symbol and the pilot symbol increases. In addition, most existing precoding design methods involve the inversion of large-dimensional matrices, which increases the computational complexity and is not easy to implement. Therefore, how to construct a channel model to better predict the channel and characterize the prediction error, and use the predicted channel for low-complexity precoding design, is an urgent problem to be solved in large-scale MIMO high-speed rail communications. Summary of the Invention

[0005] Purpose of the Invention: The present invention aims to provide a method for predicting channels for massive MIMO high-speed rail communications. This method utilizes a quadruple-beam-based channel model and a posterior quadruple-beam-based channel model for channel prediction, enabling the acquisition of more accurate CSI in the data segment. Furthermore, the present invention provides a method for massive MIMO high-speed rail communication transmission. This method utilizes the predicted channel for uplink reception and downlink precoding design, further incorporating manifold optimization to achieve low-complexity precoding design. The design process does not involve expectation operations or large-dimensional matrix inversion.

[0006] Technical solution: In order to achieve the above-mentioned purpose, the present invention provides the following technical solution:

[0007] Massive MIMO high-speed rail communication channel prediction method, including:

[0008] Based on the quadruple-beam basis channel model of large-scale MIMO high-speed rail communication, the base station obtains estimated quadruple-beam domain channel information through uplink channel detection, and predicts the space-frequency-time domain channel information of the data segment between pilot symbols; in the quadruple-beam basis channel model, the space-frequency-time domain channel vector is expressed as the product of the quadruple-beam matrix and the quadruple-beam domain channel vector; based on the quadruple-beam domain channel information of the pilot segment and the quadruple-beam basis a posteriori channel model, the a posteriori space-frequency-time domain channel information of the data segment after the current pilot symbol, or the a posteriori space-frequency-time domain channel information and the a posteriori angle domain channel information are predicted.

[0009] Furthermore, the quadruple beam matrix is ​​obtained by the Kronecker product of the receiving end spatial domain beam matrix, the transmitting end spatial domain beam matrix, the frequency domain beam matrix and the time domain beam matrix; the receiving end spatial domain beam matrix, the transmitting end spatial domain beam matrix, the frequency domain beam matrix and the time domain beam matrix are matrices spliced ​​from the sampling steering vectors corresponding to a group of receiving angle direction cosines, transmitting angle direction cosines, time delay and Doppler frequency sampling points selected by the base station; each column of the quadruple beam matrix corresponds to a quadruple beam, which is obtained by the Kronecker product of the sampled receiving end spatial domain steering vector, the sampled transmitting end spatial domain steering vector, the sampled frequency domain steering vector and the sampled time domain steering vector.

[0010] Furthermore, the sampling range of the receiving angle direction cosine is -1 to 1, the sampling range of the transmitting angle direction cosine is -1 to 1, the sampling range of the delay is 0 to the maximum delay extension, and the sampling range of the Doppler frequency is from the negative maximum Doppler frequency to the positive maximum Doppler frequency; the sampling method is uniform sampling; the number of sampling points for the receiving angle direction cosine, the transmitting angle direction cosine, the delay and the Doppler frequency is greater than, equal to or less than the number of user-side receiving antennas, the number of base station-side transmitting antennas, the equivalent delay extension point number and the equivalent Doppler extension point number; the equivalent delay extension point number is obtained by multiplying the ratio of the number of effective subcarriers to the total number of subcarriers by the cyclic prefix length and rounding up, and the equivalent Doppler extension point number is obtained by multiplying twice the maximum Doppler frequency by the total duration of a subframe.

[0011] Furthermore, the spatial-frequency-time domain channel information of the data segment between the pilot symbols is predicted by right-multiplying the quadruple beam matrix by the estimated quadruple beam domain channel vector.

[0012] Furthermore, the quadruple-beam basis posterior channel model includes two parts: deterministic and random. The deterministic part is the Hadamard product of a column vector and the quadruple-beam basis channel estimated in the previous time slot, and the column vector can be obtained through the Pearson correlation coefficient. The random part is a random vector in which each element obeys an independent Gaussian variable with zero mean and different variance.

[0013] Furthermore, the a posteriori space-frequency-time domain channel vector of the data segment after the pilot symbol is predicted by right-multiplying the quadruple beam matrix by the a posteriori quadruple beam domain channel vector; the a posteriori angle domain channel information is obtained by left-multiplying the covariance matrix of the posterior quadruple beam domain channel vector by the double beam matrix, and then right-multiplying it by the conjugate transpose of the double beam matrix, wherein the double beam matrix is ​​obtained by the Kronecker product of the receiving antenna dimension unit matrix, the transmitting antenna dimension unit matrix, the sampling frequency domain beam matrix and the sampling time domain beam matrix.

[0014] Furthermore, in the uplink communication of the high-speed railway, the base station side predicts the spatial domain channel information of the data segment between the pilot symbols, designs a receiving matrix, and right-multiplies the receiving matrix by the receiving vector to restore the data symbols sent by the user; the spatial domain channel information of the data segment is obtained by intercepting the elements of the predicted spatial-frequency-time domain channel vector at a given subcarrier and symbol; a precoding transmission method is adopted in the downlink communication of the high-speed railway, and the base station side uses the data segment a posteriori spatial domain channel information and a posteriori angle domain channel information predicted by the method according to claim 6 to design a precoding matrix for all service users, and right-multiplies it by the data symbol vector of the corresponding user, accumulates the precoded symbols of all users, and generates a sending signal of the base station on a subcarrier in the selected subcarrier group.

[0015] Furthermore, the massive MIMO high-speed rail communication downlink precoding transmission method includes:

[0016] Transform the constrained precoding matrix design problem in Euclidean space into manifold space Unconstrained precoding matrix design problem;

[0017] Solving manifold space using Riemann conjugate gradient method Unconstrained precoding matrix design problem.

[0018] The search direction of the RCG method is the weighted sum of the reverse direction of the Riemann gradient of the current iteration point and the previous search direction, translated to the tangent space of the current iteration point via a vector. The new iteration point is ensured to remain on the manifold through contraction mapping. The search step size is determined using the backtracking method. Each iteration first obtains the channel matrix of each user terminal and multiplies it by the precoding matrix of the current iteration point. Then, the user channel matrix is ​​calculated and multiplied by the search direction. Thus, the multiplication of the user channel matrix and the precoding matrix obtained with different search step sizes is converted into the sum of known quantities. The channel information used by the Riemann conjugate gradient algorithm is the predicted a posteriori spatial-frequency-time domain channel information and the a posteriori angular domain channel information, including the following steps:

[0019] Step a): Obtain the precoding initial value matrix using the regularized zero-forcing method;

[0020] Step b): Set the initial search step size to be greater than 0 and calculate the Euclidean gradient;

[0021] Step c): Obtaining the Riemann gradient;

[0022] Step d): Obtain conjugate gradient search direction;

[0023] Step e): Calculate the weighted sum rate of the test points obtained by the current search step;

[0024] Step f): If the precoding matrix does not sufficiently improve the weighted sum rate, adjust the search step size and return to step e);

[0025] Step g): Obtain the precoding matrix for this iteration;

[0026] Step h): If converged, output the current precoding matrix; if not, return to step b).

[0027] A large-scale MIMO high-speed rail communication system includes a base station and multiple user terminals carried on the high-speed rail, the base station and the user terminals are both equipped with antenna arrays, the base station is used for large-scale MIMO high-speed rail communication channel prediction, the large-scale MIMO high-speed rail communication channel prediction includes, based on a quadruple beam basis channel model, obtaining estimated quadruple beam domain channel information through uplink channel detection, and predicting the data segment space-frequency domain-time domain channel information between pilot symbols; in the quadruple beam basis channel model, the space-frequency domain-time domain channel vector is expressed as the product of a quadruple beam matrix and a quadruple beam domain channel vector; based on the quadruple beam domain channel information of the pilot segment and the quadruple beam basis posterior channel model, predicting the data segment after the current pilot symbol a posteriori space-frequency domain-time domain channel information. domain channel information, or, a posteriori space-frequency-time domain channel information and a posteriori angle domain channel information; the user terminal is used to send a pilot sequence in the pilot segment of the wireless frame in the uplink; or, the base station is also used to design a receiving matrix using the predicted data segment a posteriori space domain channel information between the pilot symbols in the high-speed rail uplink communication, and right-multiply the receiving matrix by the receiving vector to recover the data symbols sent by the user; and / or, adopt a precoding transmission method in the high-speed rail downlink communication, use the predicted data segment a posteriori space domain channel information to design a precoding matrix for all service users, and right-multiply it by the data symbol vector of the corresponding user, accumulate the precoded symbols of all users, and generate a sending signal of the base station on a subcarrier in the selected subcarrier group.

[0028] Beneficial effects: The large-scale MIMO high-speed rail communication channel prediction and transmission method proposed in the present invention can effectively deal with the serious problem of channel aging in high-speed rail communications, predict the spatial domain channel of the wireless frame data segment, and provide more accurate CSI for precoding design. The design process does not require expectation operations and does not involve inversion of large-dimensional matrices. It can more efficiently design precoding matrices and improve high-speed rail communication and rate performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 Schematic diagram of the massive MIMO high-speed rail communication precoding transmission method and propagation environment proposed in the present invention.

[0030] Figure 2 Schematic diagram of the wireless frame structure of massive MIMO high-speed rail communication in an embodiment of the present invention.

[0031] Figure 3 This is a diagram of the deployment of large-scale MIMO high-speed rail in an embodiment of the present invention.

[0032] Figure 4 Graph showing the NMSE performance of the predicted channel in an embodiment of the present invention.

[0033] Figure 5 This is a QB domain channel energy distribution diagram in an embodiment of the present invention.

[0034] Figure 6 This is a graph showing the convergence curve of RCG precoding in an embodiment of the present invention.

[0035] Figure 7 This is a performance comparison chart of the RCG precoding method in an embodiment of the present invention and the existing method.

[0036] Figure 8 This is a performance comparison diagram of the RCG precoding method under different channel prediction times in an embodiment of the present invention. DETAILED DESCRIPTION

[0037] The technical solutions provided by the present invention will be described in detail below with reference to specific embodiments. It should be understood that the following specific embodiments are only used to illustrate the present invention and are not used to limit the scope of the present invention.

[0038] An embodiment of the present invention discloses a large-scale MIMO high-speed railway communication channel prediction method, comprising: based on a large-scale MIMO high-speed railway communication quadruple-beam basis channel model, a base station obtains estimated quadruple-beam domain channel information through uplink channel detection, and predicts the data segment space-frequency domain-time domain channel information between pilot symbols; based on the quadruple-beam domain channel information of the pilot segment and the quadruple-beam basis posterior channel model, predicts the data segment after the current pilot symbol, or, the posterior space-frequency domain-time domain channel information and the posterior angle domain channel information. In the quadruple-beam basis channel model, the space-frequency-time domain channel vector is expressed as the product of the quadruple-beam matrix and the quadruple-beam basis channel vector; the quadruple-beam matrix is ​​obtained by the Kronecker product of the receiving end space domain beam matrix, the transmitting end space domain beam matrix, the frequency domain beam matrix and the time domain beam matrix; the receiving end space domain beam matrix, the transmitting end space domain beam matrix, the frequency domain beam matrix and the time domain beam matrix are matrices spliced ​​together by the sampling rudder vectors corresponding to a group of receiving angle direction cosines, transmitting angle direction cosines, time delay and Doppler frequency sampling points selected by the base station; each column of the quadruple-beam matrix corresponds to a quadruple beam, which is obtained by the Kronecker product of the sampled receiving end space domain rudder vector, the sampled transmitting end space domain rudder vector, the sampled frequency domain rudder vector and the sampled time domain rudder vector; the space-frequency-time domain channel information of the data segment between the pilot symbols is predicted by the quadruple-beam domain channel vector estimated by right multiplication of the quadruple-beam matrix. The quadruple beam basis posterior channel model is composed of a quadruple beam basis channel vector estimated in the previous time slot and a random vector whose elements obey independent and identically distributed Gaussian variables.

[0039] The embodiment of the present invention also discloses a large-scale MIMO high-speed rail communication transmission method. In the uplink communication of the high-speed rail, the base station side uses the predicted data segment a posteriori spatial domain channel information between the pilot symbols to design a receiving matrix, and right-multiplies the receiving matrix with the receiving vector to restore the data symbols sent by the user; in the downlink communication of the high-speed rail, a precoding transmission method is adopted. The base station side uses the predicted data segment a posteriori spatial domain channel information and the posteriori angle domain channel information to design a precoding matrix for all service users, and right-multiplies the corresponding user's data symbol vector, accumulates the precoded symbols of all users, and generates a transmission signal of the base station on a subcarrier in the selected subcarrier group. The precoding design is to design a precoding matrix for the user using the a posteriori spatial domain channel obtained by the a posteriori quadruple beam basis channel prediction and the Riemann conjugate gradient algorithm. The constrained precoding matrix design problem on the Euclidean space is converted into a manifold space. The unconstrained precoding matrix design problem; the Riemann conjugate gradient method is used to solve the manifold space Unconstrained precoding matrix design problem.

[0040] The method of the present invention is primarily applicable to Massive MIMO high-speed rail communication systems, where the base station is equipped with a large-scale antenna array to simultaneously serve multiple user terminals on the high-speed rail. The specific implementation of the Massive MIMO high-speed rail channel prediction and transmission method of the present invention is described in detail below, using a specific communication system example. It should be noted that the method of the present invention is applicable not only to the specific system model exemplified below, but also to system models with other configurations.

[0041] 1. System Configuration

[0042] We consider a single-cell massive MIMO-OFDM system in a high-speed railway (HSR) scenario. Specifically, we focus on the most common elevated bridge scenario in HSR communications, which accounts for more than 40% of railway lines and has a transmission environment dominated by direct path (LoS). The base station serves U users in a high-speed railway scenario with a moving speed of v km / h. Represents the set of all users. The base station side and each user terminal are equipped with M t and M r There are a total of P uniform linear array antennas (ULA) from the base station to user u. u paths, including a LoS path and (P u -1) non-direct (NLoS) paths. All non-direct paths of user u form a set The pth path is composed of Q p,u These subpaths are called subpath clusters of path p. In particular, for direct paths, Q LoS,u =1.

[0043] In high-speed railway communications, the massive MIMO-OFDM system operates in time division duplex (TDD) mode. Each subframe contains N b time slots, each time slot consists of N s OFDM symbols. Let N = N b N s Indicates the total number of OFDM symbols in each subframe. Each OFDM symbol has N c subcarriers, of which N v The effective subcarriers are used for data transmission, and their index set is f c represents the center frequency, λ c =c / f c is the wavelength, Δf is the subcarrier spacing. The sampling interval of the system is T s =1 / (N c Δf), the length of the cyclic prefix (CP) is N g . T g =N g T s and T sym =(N c +N g )T s In general, it is assumed that the duration of the cyclic prefix is ​​T g Greater than the maximum channel delay of each user terminal.

[0044] We assume that the channel parameters, including Doppler shift, delay, angle of departure (AoD) and angle of arrival (AoA), are almost constant within each subframe. In addition, we assume that the CSI is constant over the time duration T of an OFDM symbol. sym It remains constant within the symbol, but due to the severe Doppler effect, it will change between symbols. For simplicity, it is assumed that there are only uplink training phase and downlink transmission phase. In each time slot, the uplink sounding reference signal is only used during the nth time slot. p OFDM symbols are sent, and the remaining symbols are used for downlink transmission. Assume that the CSI of the user end is perfectly known.

[0045] 2. Quadruple Beam-Based Channel Modeling

[0046] let Indicates the mth user terminal UTu r The arrival angle of the qth subpath corresponding to the pth path is The antenna response is let Indicates the mth t The starting angle of the qth subpath of the pth path corresponding to the antenna is The antenna response is and It can be written as

[0047]

[0048] Among them, Δ r and Δ t denote the normalized receiving antenna spacing and transmitting antenna spacing respectively. Let α u,p,q ,τ u,p,q and v u,p,q denote the propagation loss, propagation delay, and Doppler shift of the qth sub-path of the uth UT along the pth path, respectively. The Doppler shift is mainly caused by the movement of the high-speed train, while the movement of users in the carriage can be ignored. We assume that all sub-paths of the main path have the same delay and Doppler shift. Therefore, we have τ u,p =τ u,p,q and ν u,p =ν u,p,q .

[0049] Note that the viaduct scenario is dominated by the LoS path. Since the high-speed rail has high-precision location information and regular movement, the Doppler shift of the LoS path can be easily compensated to mitigate inter-carrier interference (ICI). The complex-valued gain is defined as After Doppler compensation, the mth t The mth transmitting antenna r The channel impulse response between the receiving antennas can be expressed as

[0050]

[0051] in represents the Doppler frequency of the UTu LoS path after Doppler shift compensation, where is the Doppler frequency shift of the LoS path of UT u. After OFDM operation, the mth t The mth transmitting antenna r The channel frequency response of the receiving antenna at the mth subcarrier of the nth OFDM symbol can be written as

[0052]

[0053] We further consider the entire space-frequency-time (SFT) domain channel between BS and UTu over N OFDM symbols, called the SFT domain channel vector Its index is (nN v M t M r +(m-m0)M t Mr +m t M r +m r ) elements. Let

[0054]

[0055] Represent the sampling receiving end spatial domain, sampling sending end spatial domain, sampling frequency domain and sampling time domain steering vector respectively. Note that each path of the channel is represented by the parameter (Ω A ,Ω D ,τ,ν). Let

[0056]

[0057] represents the steering vector in the SFT domain. Therefore, the SFT domain channel can be expressed as

[0058]

[0059] Channel parameters τ u,p and Restricted to the set and Among them, τ max =N g / N c Δf is the maximum delay, v max =v / λ c =N d / (2NT sym ) is the maximum Doppler shift. We divide these sets evenly into multiple non-overlapping subsets:

[0060]

[0061] in,

[0062]

[0063] Here, N r =F r M r ,N t =F t M t ,N de =F de N τ , N do =F do N d , where F r ,F t ,F do ,F dois the adjustment factor. Therefore, formula (6) can be approximated as

[0064]

[0065] in, is the sampled steering vector, which is approximated as the sampled steering vector in As the number of partition subsets increases, the accuracy of this approximation will improve. We then define a transformation matrix S u , which consists of all the sampled pointing vectors of UT u, where the first (n do N de N t N r +n de N t N r +n t N r +n r ) S u The form is:

[0066]

[0067] in have have have have Therefore, the SFT domain channel (9) can be expressed as

[0068]

[0069] in, represents the quad-beam (QB) domain channel, S u It is the transformation matrix from the QB domain channel of UTu to the SFT domain channel, that is, the quadruple beam matrix, which consists of the receiving end spatial domain beam matrix U, the transmitting end spatial domain beam matrix V, the frequency domain beam matrix F and the time domain beam matrix D u From the above definition, we can know that the receiving end spatial domain beam matrix U, the transmitting end spatial domain beam matrix V, the frequency domain beam matrix F and the time domain beam matrix D u are matrices formed by splicing the sampling rudder vectors corresponding to a set of receiving angle direction cosines, transmitting angle direction cosines, time delay and Doppler frequency sampling points selected by the base station; each column of the quadruple beam matrix corresponds to a quadruple beam, which is composed of the sampling receiving end space domain rudder vector u(Ω A ), sampling the sending end space domain steering vector v(Ω D ), the Kronecker product of the sampled frequency domain rudder vector f(τ) and the sampled time domain rudder vector d(v) is obtained. Specifically, The definition of

[0070]

[0071] The steering vector of each sample, S u Each column of ,corresponds to a physical quadruple beam in the SFT domain.,Therefore, we refer to (11) as the QB-based channel model.

[0072] let represents the stacked channel matrix of all users in the system. SFT The form is

[0073]

[0074] Where S=blkdiag(S1,…,S U ), Note that in the viaduct scenario, the LoS path dominates, indicating that h QB The power of h is dominated by the LoS element. In addition, due to the sparse scattering environment of the overpass scene, the number of propagation paths and the angular spread are limited. These characteristics lead to h QB It is a sparse vector in the Doppler dimension, time delay dimension and departure angle dimension. Assume that h QB The elements in h follow independent complex Gaussian distributions with zero mean and different variances, then h QB The covariance matrix of

[0075]

[0076] This is called statistical CSI in the QB domain, where is the statistical QB domain CSI of user u. is a sparse diagonal matrix and changes slowly over time. We assume that the BS side knows

[0077] 3. Channel Prediction

[0078] We consider pilot-based channel estimation. The sliding window method is widely used to integrate the current time slot with the previous time slot into a complete subframe. In this process, pilot segments are allocated for estimation. Therefore, every time a new time slot arrives, the QB domain channel estimation is performed to obtain higher accuracy. Defined by the nth b The subframe consisting of the time slot and the previous time slot is the nth b Virtual subframes. b The QB domain channel estimated by the virtual subframe is expressed as Then, in the nth b In a virtual subframe, the corresponding SFT domain channel is

[0079]

[0080] According to formula (15), we can obtain the nth b The channel of all OFDM symbols in a virtual subframe. In the quasi-static scenario, since the channel remains almost constant within a time slot, the channel estimated from the pilot segment is usually directly applied to the precoding of the remaining symbols. In this case, the estimated SFT domain at the pilot segment can be obtained by selecting the matrix S u Set to S u,p To obtain, specifically

[0081]

[0082] Among them S u,p represents the transformation matrix from the QB domain channel to the pilot segment SFT domain channel, where D u,p =Θ p D u , where Θ p is a matrix whose nth b Row (n b =1,2,…,Nb) is I N The first (n b N s +n p )OK.

[0083] However, in HSR communication, the channel changes rapidly. When the mobile speed v and the center frequency f c When increases, the channel may change even every other OFDM symbol. In this case, we can use and The relationship between is used to predict the channel between the pilot segments of interest.

[0084]

[0085] is the transformation matrix from the QB domain channel to the spatial domain channel on the mth subcarrier of the nth OFDM symbol, where n elements of are 1, and the rest are 0; The mth element of S is 1, and the rest are 0. u,n,m By selecting the matrix S u Set to S u,n,m , the predicted channel at the m-th subcarrier of the n-th OFDM symbol can be written as

[0086]

[0087] That is, the space-frequency-time domain channel vector obtained by intercepting the prediction The spatial domain channel information of the data segment is obtained by reconstructing the elements at a given subcarrier m and symbol n. We can get the channel matrix of m subcarriers of the nth OFDM symbol for precoder design As shown below:

[0088]

[0089] 4. Quadruple Beam-Based A Posteriori Channel Model

[0090] In practical applications, direct use approximate It may be affected by the prediction error caused by the limited resolution of the Doppler domain. In the HSR scenario, due to the high speed movement, this error is inevitable. Even worse, this inaccuracy is caused when the symbol to be predicted n is far away from the pilot symbol n. p To further improve the performance, we consider the QB domain channels estimated from different virtual subframe periodicities as a time series, which helps to derive the QB domain channels that better match the subsequent time slots, thereby better predicting the channels between pilot segments.

[0091] It should be noted that within a subframe, the channel parameters are assumed to be almost time-invariant, which means that the QB domain channels estimated from different virtual subframes can be regarded as a time series with significant time correlation. In addition, due to The elements in are assumed to be independent of each other. We consider the prediction error and propose a posterior QB domain channel model based on an element-wise first-order Gaussian Markov process, which is of the form

[0092]

[0093] in, is the (n b +1) virtual subframes, It follows an independent and identically distributed complex Gaussian distribution with mean 0 and variance 1. It reflects the element-by-element temporal correlation between channels in the QB domain and can be obtained from the Pearson linear correlation coefficient. is the statistical prediction error, satisfying

[0094]

[0095] let

[0096]

[0097] The transformation matrix representing the QB domain channel to the angle domain channel of the mth subcarrier of the nth OFDM symbol, i.e., the double beam matrix, is composed of the unit matrix of the number of receiving angle direction cosine sampling points Unit array of emission angle direction cosine sampling points Sampling frequency domain beam matrix at a given subcarrier and the sampling time domain beam matrix at a given symbol The Kronecker product is obtained. b No. and n b The a posteriori angle domain channel vector of user u on the mth subcarrier of the nth OFDM symbol can be expressed as

[0098]

[0099] in and are the deterministic and statistical components, respectively. is a zero-mean, covariance matrix

[0100] A complex Gaussian random vector of .

[0101]

[0102] Easy to verify is still a diagonal matrix. Accordingly, The mean is The covariance matrix is The autocorrelation matrix is

[0103] A complex Gaussian random vector, That is, the a posteriori angle domain channel information.

[0104] Based on the a posteriori angle-domain channel (23), the a posteriori spatial-domain channel vector has the form:

[0105]

[0106] Its autocorrelation matrix is:

[0107]

[0108] Furthermore, on the m-th subcarrier of the n-th OFDM symbol, the a posteriori spatial domain channel matrix is:

[0109]

[0110] Where (a) is Come to. yes The matrix form of is defined as follows:

[0111]

[0112] It is called the angle domain energy matrix.

[0113] Set the time interval between the current pilot segment and the virtual subframe to be predicted We can describe the channel uncertainty under different channel conditions. The a posteriori spatial domain channel (29) can be simplified to the prediction channel (19). The precoder is designed using only statistical CSI. The a posteriori channel model can predict the channel for a longer period of time.

[0114] 5. Problem Construction and Riemann Elements

[0115] We study the precoder design for channel aging and prediction error in high HSR scenarios. For the sake of simplicity, the subcarrier subscripts are omitted. Represents the signal transmitted to the kth UT on a given subcarrier of the nth symbol, satisfying where d k is the dimension of the data stream sent to user k. The received signal of the kth UT in the nth OFDM symbol is

[0116]

[0117] Among them H kn and denote the channel matrix and precoding matrix of the kth user in the nth OFDM symbol, respectively, z kn To obey The interference plus noise is assumed to be Gaussian distributed, and the covariance matrix is

[0118]

[0119] Assume that the user knows R kn , the ergodic rate of user k can be written as

[0120]

[0121] The ergodic weighted sum rate maximization problem can be expressed as

[0122]

[0123] Among them, P t is the total power constraint. However, due to the existence of the expectation operator, Without a closed-form expression, it is very difficult to directly process (34). Note that when A is a positive definite matrix, logdet(A) is a concave function. Therefore, we can use Jensen's inequality on (33) to obtain an upper bound

[0124]

[0125] The optimization problem can be reformulated as

[0126]

[0127] Let P n =[P 1n ,…,P Un ] represents the user stacked precoding matrix, noting

[0128]

[0129] Linear manifold The Riemannian submanifold of

[0130]

[0131] here and yes In P n Therefore, the constraint problem (36) can also be restated as Unconstrained Problems on

[0132]

[0133] in is the objective function.

[0134] Each nonlinear manifold P on n are connected to a The linear tangent space of P n By introducing contraction, it can be ensured that P n Moving along the tangent vector while still on the nonlinear manifold Above. Riemann submanifold is a sphere whose contraction is defined as

[0135]

[0136] An effective search direction is usually associated with the negative Riemann gradient. Let

[0137]

[0138] f(Pn ) has a Euclidean gradient of

[0139]

[0140] in

[0141]

[0142] f(P n ) in the tangent space The Riemann gradient on is

[0143]

[0144] in

[0145]

[0146] For vector transmission Indicates that it can transform the tangent vector from Transport to another space To facilitate the operation of tangent vectors in different tangent spaces. You can do this by Orthogonal projection to Come up and realize it. The form of vector transmission is

[0147]

[0148] in

[0149]

[0150] Represents the tangent vector ξ P from Orthogonal projection to

[0151] 6. RCG Precoding Design Method

[0152] For the sake of clarity, any symbol with a superscript p represents the symbol in the p-th outer iteration, and any symbol with a superscript pair (p,q) represents the symbol in the q-th inner iteration in the p-th outer iteration. After that, you need to search along the search direction with a step length of α p , to find an effective point, which can be effectively achieved by the Backtracking method. The update formula for the upper search step is as follows:

[0153]

[0154] where α p,q is the trial step size in the (p,q)th iteration. In the inner iteration, (48) needs to be repeated until a satisfactory α is obtained. p =α p,q ,and Directly used as input for the next outer iteration.

[0155] In this process, it is necessary to repeatedly compare the size of the objective function. The objective function in the (p,q)th iteration can be rewritten as

[0156]

[0157] in

[0158]

[0159] In order to speed up the convergence, the search direction is chosen to be the Riemann conjugate gradient (RCG), which can be expressed as

[0160]

[0161] in, is the RCG update parameter, different choices will lead to different RCG methods. In order to avoid lag, β p The modified Polak-Ribi`ere parameter (PRP) is chosen and is defined as follows:

[0162]

[0163] in

[0164]

[0165] definition for:

[0166]

[0167] In order to reduce the complexity of finding the expectation, two operators are defined as

[0168]

[0169]

[0170] in Note that U and V are part of the discrete Fourier transform matrix, indicating that (55) and (56) can be computed efficiently.

[0171] In addition, by pre-calculating some elements, the computational complexity can be significantly reduced, making the algorithm complexity independent of the number of internal iterations. With the help of (48), we have

[0172]

[0173] therefore It can be expressed as

[0174]

[0175] in

[0176]

[0177] Notice that and There is no index in the inner iteration, which means that (58) does not need to be repeatedly calculated in the inner iteration. In addition, let According to (48), It can be calculated by the following formula

[0178]

[0179] Calculate in advance and back, The number of internal iterations of the calculation is irrelevant. and can be rewritten as

[0180]

[0181] and

[0182]

[0183] By pre-calculating and (60), (61), and (62) can be derived efficiently, and these matrices are weighted sums of known matrices, showing that the objective function can be efficiently calculated during the internal iteration process.

[0184] After getting a good enough point in the (p,q)th iteration, we can directly make As the input of the next iteration. Therefore, the Euclidean gradient in the (p+1)th iteration can be further expressed as

[0185]

[0186] The RCG precoding design method is implemented as follows:

[0187] Step a): Use the regularized zero forcing (RZF) method to obtain the precoding initial value matrix P 0 , set the initial step size α 0 >0, select constants r∈(0,1), c∈(0,1);

[0188] Step b): Calculation

[0189] Step c): Using (63), calculate the Euclidean gradient

[0190] Step d): Use orthogonal projection to obtain the Riemann gradient

[0191] Step e): Use (51) to obtain the search direction and calculate

[0192] Step f): Calculation and

[0193] Step g): α n,0 =α 0 ;

[0194] Step h): Using (48) and (60), we get and

[0195] Step i): Using (49), (61) and (62), calculate q←q+1;

[0196] Step j): If

[0197] Return to step h);

[0198] Step k):

[0199] Step 1): If converged, output If there is no convergence, go back to step b).

[0200] VII. Implementation Effect

[0201] To help those skilled in the art better understand the present invention, a performance comparison of the massive MIMO high-speed rail communication precoder in this embodiment with the existing precoder using a pilot channel design is given below under a specific system configuration, as well as a comparison of the rate performance and algorithm complexity of the manifold optimization method with the existing method.

[0202] In order to establish a simulation environment that can reproduce the real characteristics of HSR, we use the widely used QuaDRiGa channel model to generate simulation scenarios, in which "3GPP 37.885 Highway LOS" is considered to simulate the viaduct scenario in the high-speed railway. The high-speed railway is set to travel eastward along the x-axis at a speed of 300 km / h. There are 20 user devices evenly distributed on the high-speed railway with an interval of 10 meters. The initial position of the first UT is set to (-200,0), the vertical base station track distance is set to 200 meters, and the base station position is (0,-200). The layout of the simulation scene is as follows Figure 3 As shown in the figure, the duration of each subframe is set to 1 millisecond, with a total of 2048 subcarriers, of which 120 are effective subcarriers, the subcarrier spacing is set to 60KHz, and the CP length is 144. In this case, each subframe contains N b = 4 time slots, each time slot consists of N s =14 OFDM symbols. The carrier frequency is 4.8GHz, and each base station is equipped with M t = 128 antennas, each UT has M r = 2 antennas, number of data streams per user terminal d k =2. Assuming that the CSI remains constant within the time of one OFDM symbol and the high-speed train moves uniformly on the track, a channel is generated for each OFDM symbol within 2ms. The channel power generated by QuaDRiGa is normalized and the transmit signal-to-noise ratio (SNR) is defined as in

[0203] The comparison methods include the RCG method that uses the pilot channel for precoding design, and the robust precoding method based on maximum-minimum (MM) optimization proposed in the literature "A.-A.Lu, X.Gao and C.Xiao,"Robust Linear Precoder Design for 3D Massive MIMO Downlink With APosteriori Channel Model," in IEEE Transactions on Vehicular Technology, vol.71, no.7, pp.7274-7286, July 2022". The energy distribution of the QB domain channel in the high-speed rail communication scenario is as follows: Figure 4 As shown in the figure, it can be seen that the QB domain channel is relatively sparse in the high-speed rail scenario. The NMSE performance of channel prediction based on the QB domain channel is shown in Figure 5 As shown in , it can be seen that the predicted channel has higher accuracy than the unpredicted channel. The convergence curve of the RCG robust precoding method is shown in Figure 6 As shown in Figure 2, it can be seen that the RCG method converges faster. Figure 7 As shown in , it can be seen that the RCG robust design method based on the QB domain posterior channel model has better sum rate performance. The sum rate performance of different prediction time lengths is shown in Figure 8 As shown in the figure, "RCG-robust0.125ms" represents the robust precoder designed for the data segment 0.125ms away from the pilot symbol using the RCG method and the channel predicted by the a posteriori QB domain channel, and "RCG-pilot 0.125ms" represents the precoder designed for the data segment 0.125ms away from the pilot symbol using the RCG method and the channel at the pilot. It can be seen that the RCG robust design method based on the QB domain a posteriori channel model can predict the channel for a longer time while ensuring the sum rate performance.

[0204] Based on the same inventive concept, an embodiment of the present invention discloses a computer program product comprising a computer program / instructions that, when executed by a processor, implement the steps of each of the above methods. The computer program product may be stored in the memory of a computer device. In a specific implementation, the device includes a processor, a communication bus, a memory, and a communication interface. The processor may be a general-purpose central processing unit (CPU), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits for controlling the execution of the program of the present invention. The communication bus may include a path for transmitting information between the above components. The communication interface, using any device such as a transceiver, is used to communicate with other devices or a communication network. The memory may be a read-only memory (ROM) or other type of static storage device that can store static information and instructions, a random access memory (RAM) or other type of dynamic storage device that can store information and instructions, an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical storage, a disk storage medium or other magnetic storage device, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but is not limited thereto. The memory can be independent and connected to the processor via a bus, or it can be integrated with the processor.

[0205] The memory is used to store application code for executing the solution of the present invention, and the execution is controlled by the processor. The processor is used to execute the application code stored in the memory, thereby realizing the user-centric network large-scale MIMO communication and downlink precoding manifold optimization design method provided by the above embodiment. The processor may include one or more CPUs, or may include multiple processors, each of which may be a single-core processor or a multi-core processor. The processor here may refer to one or more devices, circuits, and / or processing cores for processing data (such as computer program instructions).

[0206] Based on the same inventive concept, an embodiment of the present invention discloses a large-scale MIMO high-speed rail communication system, including a base station and multiple user terminals carried on the high-speed rail, both the base station and the user terminals are equipped with antenna arrays, the base station is used for large-scale MIMO high-speed rail communication channel prediction, the large-scale MIMO high-speed rail communication channel prediction includes, based on the quadruple beam basis channel model, obtaining estimated quadruple beam domain channel information through uplink channel detection, and predicting the data segment space-frequency domain-time domain channel information between pilot symbols; in the quadruple beam basis channel model, the space-frequency domain-time domain channel vector is expressed as the product of the quadruple beam matrix and the quadruple beam domain channel vector; based on the obtained quadruple beam domain channel information of the pilot segment and the quadruple beam basis posterior channel model, predicting the data after the current pilot symbol The user terminal is configured to transmit a pilot sequence in the pilot segment of a radio frame in the uplink using the predicted data segment a posteriori spatial domain channel information between pilot symbols to design a receiving matrix, and right-multiply the receiving matrix by the receiving vector to recover the data symbols sent by the user in the uplink communication of the high-speed rail, and / or, in the downlink communication of the high-speed rail, a precoding transmission method is used to design a precoding matrix for all service users using the predicted data segment a posteriori spatial domain channel information, and right-multiply the data symbol vector of the corresponding user, accumulate the precoded symbols of all users, and generate a signal transmitted by the base station on a subcarrier in the selected subcarrier group. In the embodiments provided in this application, it should be understood that the disclosed method can be implemented in other ways without exceeding the spirit and scope of this application. The current embodiment is only an illustrative example and should not be used as a limitation. The specific content provided should not limit the purpose of this application. For example, some features can be ignored or not implemented.

[0207] The technical means disclosed in the solutions of the present invention are not limited to those disclosed in the above-mentioned embodiments, but also include technical solutions composed of any combination of the above-mentioned technical features. It should be noted that those skilled in the art may make various improvements and modifications without departing from the principles of the present invention, and such improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for predicting massive MIMO high-speed rail communication channels, characterized in that: include: Based on the quadruple-beam basis channel model for massive MIMO high-speed rail communications, the base station obtains estimated quadruple-beam domain channel information through uplink channel sounding and predicts the spatial-frequency-time domain channel information of the data segment between pilot symbols. In the quadruple-beam basis channel model, the spatial-frequency-time domain channel vector is expressed as the product of a quadruple-beam matrix and a quadruple-beam domain channel vector. The quadruple-beam matrix is ​​obtained by the Kronecker product of the receiving end spatial domain beam matrix, the transmitting end spatial domain beam matrix, the frequency domain beam matrix, and the time domain beam matrix. Based on the quadruple beam domain channel information of the pilot segment and the quadruple beam basis posterior channel model, the posterior space-frequency-time domain channel information of the data segment after the current pilot symbol, or the posterior space-frequency-time domain channel information and the posterior angle domain channel information are predicted; the quadruple beam basis posterior channel model includes two parts, a deterministic part and a random part; the deterministic part is a Hadamard product of a column vector reflecting time correlation and the quadruple beam domain channel estimated in the previous time slot; the random part is a random vector whose elements obey independent Gaussian variables with zero mean and different variances.

2. The method for predicting a massive MIMO high-speed railway communication channel according to claim 1, wherein: The receiving end spatial domain beam matrix, the transmitting end spatial domain beam matrix, the frequency domain beam matrix and the time domain beam matrix are matrices formed by splicing the sampling steering vectors corresponding to a group of receiving angle direction cosines, transmitting angle direction cosines, time delay and Doppler frequency sampling points selected by the base station; each column of the quadruple beam matrix corresponds to a quadruple beam, which is obtained by the Kronecker product of the sampled receiving end spatial domain steering vector, the sampled transmitting end spatial domain steering vector, the sampled frequency domain steering vector and the sampled time domain steering vector.

3. The method for predicting a massive MIMO high-speed railway communication channel according to claim 2, wherein: The sampling range of the receiving angular direction cosine is -1 to 1, the sampling range of the transmitting angular direction cosine is -1 to 1, the sampling range of the delay is 0 to the maximum delay spread, and the sampling range of the Doppler frequency is from the negative maximum Doppler frequency to the positive maximum Doppler frequency; the sampling method is uniform sampling; The number of sampling points for the receiving angular direction cosines, transmitting angular direction cosines, delay, and Doppler frequency, respectively, is greater than, equal to, or less than the number of user-side receiving antennas, the number of base station-side transmitting antennas, the number of equivalent delay extension points, and the number of equivalent Doppler extension points; the equivalent delay extension point number is obtained by multiplying the ratio of the number of effective subcarriers to the total number of subcarriers by the cyclic prefix length and rounding up; the equivalent Doppler extension point number is obtained by multiplying twice the maximum Doppler frequency by the total duration of one subframe.

4. The method for predicting a massive MIMO high-speed railway communication channel according to claim 1, wherein: The data segment space-frequency-time domain channel information between the pilot symbols is predicted by right-multiplying the quadruple beam matrix by the estimated quadruple beam domain channel vector.

5. The method for predicting a massive MIMO high-speed railway communication channel according to claim 1, wherein: The a posteriori spatial-frequency-time domain channel information of the data segment after the pilot symbol is predicted by right-multiplying the quadruple beam matrix by the a posteriori quadruple beam domain channel vector; the a posteriori angle domain channel information is obtained by left-multiplying the covariance matrix of the a posteriori quadruple beam domain channel vector by the double beam matrix, and then right-multiplying it by the conjugate transpose of the double beam matrix, wherein the double beam matrix is ​​obtained by the Kronecker product of the unit matrix of the receiving angular direction cosine sampling points, the unit matrix of the transmitting angular direction cosine sampling points, the sampling frequency domain beam matrix at a given subcarrier, and the sampling time domain beam matrix at a given symbol.

6. A massive MIMO high-speed rail communication transmission method, characterized in that: In the uplink communication of high-speed rail, the base station side predicts the spatial domain channel information of the data segment between the pilot symbols, designs the receiving matrix, and right-multiplies the receiving matrix by the receiving vector to restore the data symbols sent by the user; the spatial domain channel information of the data segment is obtained by intercepting the elements of the spatial-frequency-time domain channel vector at the given subcarrier and symbol; wherein, based on the quadruple-beam basis channel model of large-scale MIMO high-speed rail communication, the base station obtains the estimated quadruple-beam domain channel information through uplink channel detection, and predicts the spatial-frequency-time domain channel information of the data segment between the pilot symbols; in the quadruple-beam basis channel model, the spatial-frequency-time domain channel vector is expressed as the product of the quadruple-beam matrix and the quadruple-beam domain channel vector; the quadruple-beam matrix is ​​obtained by the Kronecker product of the spatial domain beam matrix of the receiving end, the spatial domain beam matrix of the transmitting end, the frequency domain beam matrix and the time domain beam matrix; A precoding transmission method is adopted in high-speed rail downlink communication. The base station side uses the predicted data segment posterior spatial domain channel information and posterior angular domain channel information to design a precoding matrix for all service users, and multiplies the data symbol vector of the corresponding user on the right, accumulates the precoded symbols of all users, and generates the base station's transmission signal on a subcarrier in the selected subcarrier group; the data segment posterior spatial domain channel information is obtained by intercepting the elements of the posterior spatial-frequency-time domain channel vector at a given subcarrier and symbol; wherein, based on the quadruple beam domain channel information of the pilot segment and the quadruple beam basis posterior channel model, the base station predicts the data segment posterior spatial-frequency-time domain channel information after the current pilot symbol, or the posterior spatial-frequency-time domain channel information and the posterior angular domain channel information; the quadruple beam basis posterior channel model includes two parts: deterministic and random; the deterministic part is a column vector reflecting time correlation and the Hadamard product of the quadruple beam domain channel estimated in the previous time slot; the random part is a random vector whose elements obey independent Gaussian variables with zero mean and different variances.

7. The massive MIMO high-speed rail communication transmission method according to claim 6, characterized in that: The precoding transmission method used in high-speed rail downlink communication includes: Transform the constrained precoding matrix design problem in Euclidean space into manifold space Unconstrained precoding matrix design problem; The Riemann conjugate gradient (RCG) method is used to solve the manifold space. Unconstrained precoding matrix design problem.

8. The massive MIMO high-speed rail communication transmission method according to claim 7, characterized in that: The objective function of the constrained precoding matrix design problem on the Euclidean space is the upper bound of the maximum weighted sum rate problem, and the constraint condition is the total power constraint of the base station; the manifold space It is the set of all precoding matrices that meet the total power constraint of the base station.

9. The massive MIMO high-speed rail communication transmission method according to claim 7, characterized in that: The RCG method includes: The search direction of the RCG method is the weighted sum of the reverse direction of the Riemann gradient of the current iteration point and the previous search direction, translated to the tangent space of the current iteration point via a vector. The new iteration point is ensured to remain on the manifold through contraction mapping. The search step size is determined using the backtracking method. The channel information used by the Riemann conjugate gradient algorithm is the posterior spatial-frequency-time domain channel information and the posterior angular domain channel information, including the following steps: Step a): Obtain the precoding initial value matrix using the regularized zero-forcing method; Step b): Set the initial search step size to be greater than 0 and calculate the Euclidean gradient; Step c): Obtaining the Riemann gradient; Step d): Obtain conjugate gradient search direction; Step e): Calculate the weighted sum rate of the test points obtained by the current search step; Step f): If the precoding matrix does not sufficiently improve the weighted sum rate, adjust the search step size and return to step e); Step g): Obtain the precoding matrix for this iteration; Step h): If converged, output the current precoding matrix; if not, return to step b).

10. A massive MIMO high-speed rail communication system, comprising a base station and multiple user terminals mounted on the high-speed rail, wherein both the base station and the user terminals are equipped with antenna arrays, characterized in that: The base station is used for large-scale MIMO high-speed railway communication channel prediction, and the large-scale MIMO high-speed railway communication channel prediction includes, based on the quadruple beam basis channel model, obtaining estimated quadruple beam domain channel information through uplink channel detection, and predicting the data segment space-frequency domain-time domain channel information between pilot symbols; in the quadruple beam basis channel model, the space-frequency domain-time domain channel vector is expressed as the product of the quadruple beam matrix and the quadruple beam domain channel vector; the quadruple beam matrix is ​​obtained by the Kronecker product of the receiving end space domain beam matrix, the transmitting end space domain beam matrix, the frequency domain beam matrix and the time domain beam matrix; based on the quadruple beam domain channel information of the pilot segment and the quadruple beam basis posterior channel model, predicting the data segment after the current pilot symbol, or, the posterior space-frequency domain-time domain channel information and the posterior angle domain channel information; the quadruple beam basis posterior channel model includes two parts, a deterministic and a random part; the deterministic part is a reflection of time The column vector of the correlation and the Hadamard product of the quadruple beam domain channel estimated in the previous time slot; the random part is a random vector of independent Gaussian variables with zero mean and different variance for each element; the user terminal is used to send a pilot sequence in the pilot segment of the wireless frame in the uplink; or, the base station is also used to design a receiving matrix using the predicted data segment posterior spatial domain channel information between the pilot symbols in the high-speed rail uplink communication, and right-multiply the receiving matrix by the receiving vector to recover the data symbols sent by the user; and / or, adopt a precoding transmission method in the high-speed rail downlink communication, use the predicted data segment posterior spatial domain channel information to design a precoding matrix for all service users, and right-multiply the data symbol vector of the corresponding user, accumulate the precoded symbols of all users, and generate a sending signal of the base station on a subcarrier in the selected subcarrier group; the spatial domain channel information is obtained by the elements of the space-frequency-time domain channel vector at a given subcarrier and symbol.

11. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instructions are executed by a processor, the steps of the method according to any one of claims 1 to 9 are implemented.

Images