Mobile communication base station and communication method

Abstract

The application discloses a kind of mobile communication base station and communication method, communication method includes: communication base station sends pilot configuration instruction to user, carries out estimation to Doppler frequency offset, obtains phase-compensated received pilot signal, and is converted into vectorization received signal;Linear regression model is constructed, adaptive sparse channel estimation is carried out based on variational bayes inference, and output downlink frequency domain channel matrix, angle domain channel matrix;Angle domain channel matrix is quantized using Lloyd-Max non-uniform quantizer, and the global channel matrix of base station end is obtained;Output optimal analog precoding matrix and digital precoding matrix;Communication base station generates radio frequency signal based on precoding after transmission signal vector, and sends to user.Communication base station includes distributed antenna unit, baseband processing unit, radio frequency link and central control unit.The technical scheme of the application covers the whole process signal processing link of mobile communication base station, and the calculation complexity is low, and can be directly adapted to existing 3GPP standard and hardware platform.

Description

Technical Field

[0001] This invention relates to the field of communication technology, specifically to a mobile communication base station and a communication method. Background Technology

[0002] With the advancement of 5G-Advanced commercialization and the research and development of 6G technology, ultra-dense heterogeneous networking, massive MIMO, and millimeter-wave / terahertz communication have become core technologies for next-generation mobile communication systems, enabling peak data rates in the Tbps range, end-to-end latency in the millisecond range, and connection densities in the millions. However, existing mobile communication base station communication methods still have many technical shortcomings: Channel estimation performance and complexity cannot be balanced: traditional LS channel estimation has poor anti-interference capability, MMSE estimation requires known prior channel statistics, and in high-speed mobile scenarios, Doppler frequency shift exacerbates the time-varying characteristics of the channel, resulting in a significant decrease in the estimation accuracy of existing algorithms, making them unsuitable for high-speed scenarios such as high-speed rail and vehicle-to-everything (V2X) networks. There are bottlenecks in the design of hybrid precoding for large-scale MIMO: all-digital precoding has the best performance but requires an RF chain of equal number of antennas, resulting in extremely high hardware cost and power consumption; most existing hybrid precoding algorithms are non-convex optimization problems with high iterative complexity, and cannot achieve the optimal balance between performance and complexity under constant modulus hardware constraints. The ultra-dense network has prominent interference and energy consumption problems: the co-frequency interference between adjacent base stations is serious, the signal-to-interference-plus-noise ratio (SINR) of edge users is extremely low, the centralized architecture has high signaling overhead and high response latency, and cannot adapt to dynamic service scenarios. To address the aforementioned shortcomings, this invention proposes a novel mobile communication base station and communication method. Summary of the Invention

[0003] To address the aforementioned shortcomings in the prior art, this invention provides a mobile communication base station and communication method, and constructs a closed-loop adaptive mobile communication system and method.

[0004] To achieve the above-mentioned objectives, the technical solution adopted by this invention is as follows: A mobile communication method is provided, comprising the following steps: S1: The communication base station sends a pilot configuration command to the user, the user sends an uplink pilot signal, the communication base station receives the frequency domain baseband pilot signal, estimates the Doppler frequency offset of the user in the high-speed mobile scenario based on the maximum likelihood criterion, obtains the phase-compensated received pilot signal, and converts it into a vectorized received signal. S2: Construct a linear regression model using vectorized received signals, perform adaptive sparse channel estimation based on variational Bayesian inference, and output the downlink frequency domain channel matrix in TDD mode and the angle domain channel matrix in FDD mode. S3: The angle domain channel matrix is ​​quantized using a Lloyd-Max non-uniform quantizer and then input into the encoder network of a lightweight autoencoder to compress it into a low-dimensional bitstream; the user feeds back the low-dimensional bitstream to the communication base station, and the communication base station reconstructs the low-dimensional bitstream to obtain the global channel matrix at the base station. S4: Calculate the signal-to-interference-plus-noise ratio (SINR) of the user on the subcarrier based on the downlink frequency domain channel matrix of the full subcarrier, optimize the analog precoding matrix and digital precoding matrix, and output the optimal analog precoding matrix and digital precoding matrix. S5: Based on the angle domain channel matrix, extract the user's horizontal and vertical departure angle information, design the transmit spatial beamforming matrix, and combine the optimal analog precoding matrix and digital precoding matrix to calculate the final concatenated full-link precoding matrix, thereby obtaining the precoded transmit signal vector of the subcarrier. The communication base station generates radio frequency signals based on the precoded transmit signal vector and sends them to the user.

[0005] Further, step S1 includes: S11: The communication base station sends pilot configuration instructions to all users through the broadcast channel. Users transmit uplink pilot signals on the specified uplink time-frequency resources. The uplink orthogonal pilot matrix of the uplink pilot signal is... Uplink orthogonal pilot matrix Satisfying orthogonality constraints: ; in, This represents the conjugate transpose operation. The number of pilot symbols, for K An identity matrix of order 1. K identity matrix of order The diagonal elements are 1, and the rest are 0. K The number of users per antenna; S12: Communication base station The root receiving antenna receives the uplink pilot signal sent by the user. After completing RF down-conversion and analog-to-digital conversion, the frequency domain baseband pilot signal is obtained. The frequency domain baseband pilot matrix of the frequency domain baseband pilot signal is: ; ; Among them, the frequency domain baseband pilot matrix The latitude is , latitude is The uplink frequency domain channel matrix, This refers to the number of distributed remote antenna units (RAUs) in a communication base station. M The number of transmit antennas for a single distributed remote antenna unit (RAU). latitude is The additive white Gaussian noise matrix at the receiver; S13: Based on the frequency domain baseband pilot matrix The element obtains the received pilot signals for each user on different antennas and pilot symbols. Based on the maximum likelihood criterion, the Doppler frequency offset of each user in a high-speed moving scenario is estimated, and the estimated value of the user's Doppler frequency offset is obtained. ; ; in, k Number the user. For the traversal search variable of Doppler frequency offset, m Number the antenna. n Numbering the pilot symbols, For users k In the n Conjugate operation of transmit pilot symbols on each pilot symbol j The imaginary unit, OFDM symbol period; S14: Based on Doppler frequency offset estimation Phase compensation is performed on the received pilot signal for each user to eliminate the linear phase rotation caused by Doppler, resulting in a phase-compensated received pilot signal. ; ; S15: All users' received pilot signals By splicing the components, we obtain the globally compensated frequency domain baseband pilot matrix. ; S16: Convert the baseband pilot matrix in the frequency domain Vectorization processing is performed to obtain the vectorized received signal. Simultaneously, a sensing matrix for channel estimation is constructed. ; ; in, For matrix vectorization operations, for An identity matrix of order 1. This represents the Kronecker product operation.

[0006] Further, step S2 includes: S21: Based on the perception matrix and vectorized received signal Construct a linear regression model for sparse channel estimation, and apply it to the uplink channel vector. Each channel element in Assign an independent Gaussian prior distribution, and assign a Gamma conjugate prior distribution to the precision parameter of each channel element; S22: Define noise accuracy For noise accuracy By assigning a Gamma conjugate prior distribution and minimizing the KL divergence between the variational distribution and the true posterior distribution, the uplink channel vector is solved. The optimal variational posterior distribution; S23: Fixed uplink channel vector posterior mean and posterior covariance matrix Update accuracy parameters and noise accuracy The variational posterior distribution is given, and the convergent posterior mean is output. and the posterior mean Reconstructed into uplink frequency domain channel matrix And converting it into an angle-domain channel matrix. .

[0007] Further, step S21 includes: S211: Based on the perception matrix and vectorized received signal Construct a linear regression model for sparse channel estimation; ; in, The uplink channel vector is a vectorized uplink channel vector. The latitude is Uplink channel vector For the target to be estimated, The vectorized noise vector The latitude is ; S212: For uplink channel vector Each channel element in Assign independent Gaussian prior distributions: ; in, i Channel number, uplink channel vector Element number in , For the first i Channel elements The accuracy parameters; S213: Assign a Gamma conjugate prior distribution to the precision parameter of each channel element: ; in, These are the conjugate prior distributions of Gamma. Shape parameters and rate parameters.

[0008] Further, step S22 includes: S221: Define noise accuracy For noise accuracy Assign a Gamma conjugate prior distribution: To achieve noise variance Adaptive estimation; in, These are the conjugate prior distributions of Gamma. Shape parameters and rate parameters, The standard deviation of noise; S222: Solve for the uplink channel vector by minimizing the KL divergence between the variational distribution and the true posterior distribution. The optimal variational posterior distribution is a complex Gaussian distribution: ; ; in, uplink channel vector The posterior mean, with dimension . , uplink channel vector The posterior covariance matrix, with dimension . , Indicates noise accuracy Mathematical expectation operation, For noise accuracy The precision matrix, It is a diagonal matrix. For the precision matrix The Middle Noise accuracy.

[0009] Further, step S23 includes: S231: Fixed uplink channel vector posterior mean and posterior covariance matrix Update precision parameters and noise accuracy The variational posterior distribution is calculated, and the accuracy parameter is determined. and noise accuracy The expected result is that after the update is complete, the precision matrix will be refreshed. ; Update precision parameters variational posterior distribution The method is as follows: ; Precision parameters The expectation is: ; Update noise accuracy variational posterior distribution The method is as follows: ; Noise accuracy The expectation is: ; S232: Obtain the uplink channel vector before and after the update. posterior mean Calculate the magnitude change of the channel mean vector. ; ; in, The 2-norm of a vector; S233: Set the threshold for the magnitude change ,like If the number of updates reaches the maximum value, then stop updating and output the converged posterior mean. Otherwise, continue updating the precision parameters. and noise accuracy The variational posterior distribution; S234: The convergent posterior mean Reconstructed into uplink frequency domain channel matrix ; In TDD mode, channel reciprocity is utilized to output the downlink frequency domain channel matrix. This yields the downlink channel matrix for all subcarriers; In FDD mode, the downlink frequency domain channel matrix is ​​transformed using DFT. Transform to the angle domain to obtain the angle domain channel matrix. ; ; in, The DFT matrix of the base station transmitter has dimensions of Used to convert spatial channels to the angular domain. The DFT matrix at the user receiver and the channel matrix in the angle domain. Indicates the first k The user in the first n Angle-domain channel matrix on each subcarrier.

[0010] Further, step S3 includes: S31: Using a Lloyd-Max non-uniform quantizer to quantize the angle-domain channel matrix The amplitude and phase of the mid-angle domain channel elements are quantized separately to minimize quantization distortion, and the quantized angle domain channel matrix is ​​output. ; ; in, x Angle domain channel matrix The amplitude or phase of the channel element in the mid-angle domain. Let be the probability density function. The first u The upper and lower boundaries of each quantization interval u The number of the quantization interval. The quantization level of the quantization interval. Q This represents the total number of quantization levels. S32: The user terminal will quantize the angle domain channel matrix. In the encoder network of the lightweight autoencoder, the high-dimensional angle domain channel matrix is ​​input. Compressed into a low-dimensional bitstream ; ; in, For encoder network functions, For encoder network parameters; S33: The user transmits the low-dimensional code stream via the uplink control channel. The data is fed back to the communication base station, which then receives the low-dimensional bitstream. The angle-domain channel matrix is ​​reconstructed by inputting a lightweight autoencoder into a decoder network. ; S34: Transform the angle domain channel matrix using the inverse DFT. Transforming back to the frequency domain, we obtain the full subcarrier downlink frequency domain channel matrix reconstructed at the base station. ; ; S35: Transform the full subcarrier downlink frequency domain channel matrix By concatenating the matrices, the global channel matrix at the base station is obtained. ; .

[0011] Further, step S4 includes: S41: Based on the full subcarrier downlink frequency domain channel matrix Calculate the first k The user in the first n Signal-to-interference-plus-noise ratio on each subcarrier ; ; in, v For simulating the precoding matrix Column number, To simulate the precoding matrix, For the first k The user in the first n Digital precoding matrix on each subcarrier; S42: Optimize the analog precoding matrix with the goal of maximizing system performance and rate. and digital precoding matrix Output the optimal analog precoding matrix and digital precoding matrix ; ; in, For the system's total throughput and rate, The number of subcarriers, Total transmission power, For simulating the precoding matrix The amount of elements.

[0012] Further, step S5 includes: S51: Based on the angle domain channel matrix Extract the user's horizontal and vertical departure angle information to design the transmit spatial beamforming matrix. ; ; in, w These are the numbers of adjacent base stations. For the first w The number of adjacent communication base stations to the first k Angle-domain channel matrix for each user For the first w Transmit beamforming matrix of adjacent communication base stations; S52: Utilizing a transmit spatial beamforming matrix Optimal analog precoding matrix and digital precoding matrix Calculate the final concatenated full-link precoding matrix ; ; S53: Perform link-layer scheduling, LDPC / Turbo coding, QAM modulation, and resource mapping on the downlink data for each user to obtain the modulation symbol matrix. ; S54: Based on modulation symbol matrix and full-link precoding matrix After obtaining the precoded subcarrier, the transmitted signal vector is obtained. ; ; S55: Using a time-domain shaping filter to shape the precoded transmit signal vector Time-domain shaping filtering is performed to obtain the time-domain transmit baseband signal. The time-domain transmit baseband signal is then input into the radio frequency link of the communication base station to form a radio frequency signal. The radio frequency signal is then transmitted to the user through the transmit antenna array of the distributed remote antenna unit (RAU) of the communication base station.

[0013] A communication base station is provided for performing the above-described mobile communication method, comprising a distributed antenna unit, a baseband processing unit, a radio frequency link, and a central control unit; Distributed antenna elements include N Each distributed remote antenna unit (RAU) is configured with one distributed remote antenna unit (RAU). M Root transmitting antenna; The baseband processing unit is used to generate time-domain transmit baseband signals; The radio frequency link is used to convert time-domain baseband signals into radio frequency signals.

[0014] The beneficial effects of this invention are as follows: The communication base station of this invention adopts a distributed antenna architecture, which breaks through the coverage limitations of traditional centralized antennas and reduces path loss; it adopts a hybrid precoding architecture to balance hardware cost and array gain; and it accurately characterizes the channel characteristics in high-speed mobile scenarios based on a millimeter-wave broadband time-varying channel model.

[0015] By leveraging the inherent sparsity of millimeter-wave channels, the sparse structure of the channel is adaptively learned through a Bayesian hierarchical prior model without requiring prior knowledge of the channel's statistical characteristics. Combined with Doppler frequency offset compensation, the phase rotation caused by time-varying channels is eliminated, achieving low-complexity and high-precision channel estimation.

[0016] The non-convex joint optimization problem is decoupled into two convex subproblems: digital precoding and analog precoding. The digital domain achieves multi-user interference suppression, while the analog domain achieves array gain enhancement.

[0017] The technical solution of this invention covers the entire signal processing process of mobile communication base stations, has low computational complexity, and can be directly adapted to existing 3GPP standards and hardware platforms. Attached Figure Description

[0018] Figure 1 This is a flowchart of a mobile communication method. Detailed Implementation

[0019] The specific embodiments of the present invention are described below to enable those skilled in the art to understand the present invention. However, it should be understood that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, various changes are obvious as long as they are within the spirit and scope of the present invention as defined and determined by the appended claims. All inventions utilizing the concept of the present invention are protected.

[0020] like Figure 1 As shown, a mobile communication method includes the following steps: S1: The communication base station sends a pilot configuration command to the user, the user sends an uplink pilot signal, the communication base station receives the frequency domain baseband pilot signal, estimates the user's Doppler frequency offset in high-speed mobile scenarios based on the maximum likelihood criterion, obtains the phase-compensated received pilot signal, and converts it into a vectorized received signal.

[0021] Step S1 specifically includes the following steps: S11: The communication base station sends pilot configuration instructions to all users through the broadcast channel. Users transmit uplink pilot signals on the specified uplink time-frequency resources. The uplink orthogonal pilot matrix of the uplink pilot signal is... Uplink orthogonal pilot matrix Satisfying orthogonality constraints: ; in, This represents the conjugate transpose operation. The number of pilot symbols, for K An identity matrix of order 1. K identity matrix of order The diagonal elements are 1, and the rest are 0. K The number of users per antenna; Orthogonality constraints ensure that pilot sequences from different users are uncorrelated, thus avoiding multi-user pilot interference.

[0022] S12: Communication base station The root receiving antenna receives the uplink pilot signal sent by the user. After completing RF down-conversion and analog-to-digital conversion, the frequency domain baseband pilot signal is obtained. The frequency domain baseband pilot matrix of the frequency domain baseband pilot signal is: ; ; Among them, the frequency domain baseband pilot matrix The latitude is , latitude is The uplink frequency domain channel matrix, where each column corresponds to one uplink channel vector from a user to the communication base station. This refers to the number of distributed remote antenna units (RAUs) in a communication base station. M The number of transmit antennas for a single distributed remote antenna unit (RAU). latitude is The additive white Gaussian noise matrix at the receiver; S13: Based on the frequency domain baseband pilot matrix The element obtains the received pilot signals for each user on different antennas and pilot symbols. Based on the maximum likelihood criterion, the Doppler frequency offset of each user in a high-speed moving scenario is estimated, and the estimated value of the user's Doppler frequency offset is obtained. ; ; in, k Number the user. For the traversal search variable of Doppler frequency offset, m Number the antenna. n Numbering the pilot symbols, For users k In the n Conjugate operation of transmit pilot symbols on each pilot symbol j The imaginary unit, OFDM symbol period; S14: Based on Doppler frequency offset estimation Phase compensation is performed on the received pilot signal for each user to eliminate the linear phase rotation caused by Doppler, resulting in a phase-compensated received pilot signal. ; ; S15: All users' received pilot signals By splicing the components, we obtain the globally compensated frequency domain baseband pilot matrix. ; S16: Convert the baseband pilot matrix in the frequency domain Vectorization processing is performed to obtain the vectorized received signal. Simultaneously, a sensing matrix for channel estimation is constructed. ; ; in, For matrix vectorization operations, for An identity matrix of order 1. This represents the Kronecker product operation; By using maximum likelihood estimation with matched filtering, the Doppler frequency offset in high-speed mobile scenarios is accurately obtained. Phase compensation is used to eliminate the linear phase rotation caused by Doppler, thus avoiding serious phase errors in channel estimation. This embodiment can control the frequency offset estimation error within 50Hz in high-speed scenarios of 350km / h, laying the foundation for subsequent high-precision channel estimation.

[0023] S2: Construct a linear regression model using vectorized received signals, perform adaptive sparse channel estimation based on variational Bayesian inference, and output the downlink frequency domain channel matrix in TDD mode and the angle domain channel matrix in FDD mode.

[0024] The core of this step is to leverage the natural sparsity of millimeter-wave channels to achieve high-precision channel estimation without prior information through variational Bayesian inference, outputting a full-band frequency domain channel matrix, and providing core channel state information for subsequent precoding, beamforming, and resource allocation.

[0025] Step S2 specifically includes the following steps: S21: Based on the perception matrix and vectorized received signal Construct a linear regression model for sparse channel estimation, and apply it to the uplink channel vector. Each channel element in Assign an independent Gaussian prior distribution, and assign a Gamma conjugate prior distribution to the precision parameter of each channel element; Step S21 specifically includes the following steps: S211: Based on the perception matrix and vectorized received signal Construct a linear regression model for sparse channel estimation; ; in, The uplink channel vector is a vectorized uplink channel vector. The latitude is Uplink channel vector For the target to be estimated, The vectorized noise vector The latitude is ; In millimeter-wave scenarios, the number of channel multipaths Therefore, the uplink channel vector It has natural sparsity, with only a few elements having non-zero values ​​and the rest having zero values, providing a theoretical basis for Bayesian sparse inference.

[0026] S212: For uplink channel vector Each channel element in Assign independent Gaussian prior distributions: ; in, i Channel number, uplink channel vector Element number in , For the first i Channel elements The accuracy parameters; Non-zero channel elements correspond to high-precision parameters Zero element corresponds to small precision parameter Channel sparsity constraints are achieved through adaptive updates of precision parameters.

[0027] S213: Assign a Gamma conjugate prior distribution to the precision parameter of each channel element: ; in, These are the conjugate prior distributions of Gamma. Shape parameters and rate parameters; By allocating a Gamma conjugate prior distribution, no informational prior is achieved, eliminating the need to manually set sparsity. In this embodiment, the shape and rate parameters are initialized to values ​​of... This is without prior information and does not introduce subjective bias.

[0028] S22: Define noise accuracy For noise accuracy By assigning a Gamma conjugate prior distribution and minimizing the KL divergence between the variational distribution and the true posterior distribution, the uplink channel vector is solved. The optimal variational posterior distribution.

[0029] Step S22 specifically includes the following steps: S221: Define noise accuracy For noise accuracy Assign a Gamma conjugate prior distribution: To achieve noise variance Adaptive estimation; in, These are the conjugate prior distributions of Gamma. Shape parameters and rate parameters, The standard deviation of noise; S222: Solve for the uplink channel vector by minimizing the KL divergence between the variational distribution and the true posterior distribution. The optimal variational posterior distribution is a complex Gaussian distribution: ; ; in, uplink channel vector The posterior mean, with dimension . , uplink channel vector The posterior covariance matrix, with dimension . , Indicates noise accuracy Mathematical expectation operation, For noise accuracy The precision matrix, It is a diagonal matrix. For the precision matrix The Middle Noise accuracy.

[0030] S23: Fixed uplink channel vector posterior mean and posterior covariance matrix Update accuracy parameters and noise accuracy The variational posterior distribution is given, and the convergent posterior mean is output. and the posterior mean Reconstructed into uplink frequency domain channel matrix And converting it into an angle-domain channel matrix. .

[0031] Step S23 specifically includes the following steps: S231: Fixed uplink channel vector posterior mean and posterior covariance matrix Update precision parameters and noise accuracy The variational posterior distribution is calculated, and the accuracy parameter is determined. and noise accuracy The expected result is that after the update is complete, the precision matrix will be refreshed. ; Update precision parameters variational posterior distribution The method is as follows: ; Precision parameters The expectation is: ; Update noise accuracy variational posterior distribution The method is as follows: ; Noise accuracy The expectation is: ; S232: Obtain the uplink channel vector before and after the update. posterior mean Calculate the magnitude change of the channel mean vector. ; ; in, The 2-norm of a vector; S233: Set the threshold for the magnitude change ,like If the number of updates reaches the maximum value, then stop updating and output the converged posterior mean. Otherwise, continue updating the precision parameters. and noise accuracy The variational posterior distribution; S234: The convergent posterior mean Reconstructed into uplink frequency domain channel matrix ; In TDD mode, channel reciprocity is utilized to output the downlink frequency domain channel matrix. This yields the downlink channel matrix for all subcarriers; In FDD mode, the downlink frequency domain channel matrix is ​​transformed using DFT. Transform to the angle domain to obtain the angle domain channel matrix. ; ; in, The DFT matrix of the base station transmitter has dimensions of Used to convert spatial channels to the angular domain. This is the DFT matrix of the user receiver, with a single antenna user dimension of 1×1 and a value of 1, and the angle domain channel matrix. Indicates the first k The user in the first n The angle-domain channel matrix on each subcarrier has dimensions of It exhibits significant sparsity, with only a few angles corresponding to non-zero elements.

[0032] This scheme leverages the inherent sparsity of millimeter-wave channels and adaptively learns the sparse structure of the channel through a Gaussian-Laplace hierarchical prior model, without requiring prior statistical information such as the channel's covariance. By using variational Bayesian inference, the complex posterior probability calculation is decoupled into alternating iterations, transforming the non-convex problem into a convex optimization problem, thus significantly reducing computational complexity.

[0033] Compared to traditional LS estimation, the mean square error of channel estimation is significantly reduced; compared to traditional MMSE estimation, no prior channel information is required, reducing computational complexity; combined with Doppler compensation, the mean square error of channel estimation is small in high-speed mobile scenarios at 350km / h, perfectly adapting to high-speed mobile communication scenarios.

[0034] In FDD mode, the uplink and downlink channels are not reciprocal, requiring users to feed back downlink channel state information (CSI) to the communication base station. While ensuring reconstruction accuracy, the overhead of CSI feedback is greatly reduced, enabling collaborative processing between the user end and the communication base station.

[0035] S3: The angle domain channel matrix is ​​quantized using a Lloyd-Max non-uniform quantizer and then input into the encoder network of a lightweight autoencoder to compress it into a low-dimensional bitstream; the user feeds back the low-dimensional bitstream to the communication base station, and the communication base station reconstructs the low-dimensional bitstream to obtain the global channel matrix at the base station.

[0036] Step S3 specifically includes the following steps: S31: Using a Lloyd-Max non-uniform quantizer to quantize the angle-domain channel matrix The amplitude and phase of the mid-angle domain channel elements are quantized separately to minimize quantization distortion, and the quantized angle domain channel matrix is ​​output. ; ; in, x Angle domain channel matrix The amplitude or phase of the channel element in the mid-angle domain. Let be the probability density function. The first u The upper and lower boundaries of each quantization interval u The number of the quantization interval. The quantization level of the quantization interval. Q This represents the total number of quantization levels. S32: The user terminal will quantize the angle domain channel matrix. In the encoder network of the lightweight autoencoder, the high-dimensional angle domain channel matrix is ​​input. Compressed into a low-dimensional bitstream ; ; in, For encoder network functions, For encoder network parameters; S33: The user transmits the low-dimensional code stream via the uplink control channel. The data is fed back to the communication base station, which then receives the low-dimensional bitstream. The angle-domain channel matrix is ​​reconstructed by inputting a lightweight autoencoder into a decoder network. ; S34: Transform the angle domain channel matrix using the inverse DFT. Transforming back to the frequency domain, we obtain the full subcarrier downlink frequency domain channel matrix reconstructed at the base station. ; ; S35: Transform the full subcarrier downlink frequency domain channel matrix By concatenating the matrices, the global channel matrix at the base station is obtained. ; .

[0037] This scheme leverages the sparsity of millimeter-wave channels in the angular domain to remove channel redundancy through DFT transformation; it adapts to the amplitude distribution of channel elements using Lloyd-Max non-uniform quantization, achieving lower quantization distortion with the same number of bits; and it implements end-to-end compression and reconstruction through a lightweight autoencoder, utilizing the nonlinear fitting capabilities of the encoder and decoder to maintain high reconstruction accuracy at high compression rates. This perfectly solves the engineering pain point of huge CSI feedback overhead in large-scale MIMO.

[0038] S4: Calculate the signal-to-interference-plus-noise ratio (SIR) of the user on the subcarrier based on the downlink frequency domain channel matrix of the full subcarrier, optimize the analog precoding matrix and digital precoding matrix, and output the optimal analog precoding matrix and digital precoding matrix.

[0039] Step S4 specifically includes the following steps: S41: Based on the full subcarrier downlink frequency domain channel matrix Calculate the first k The user in the first n Signal-to-interference-plus-noise ratio on each subcarrier ; ; in, v For simulating the precoding matrix Column number, To simulate the precoding matrix, For the first k The user in the first n Digital precoding matrix on each subcarrier; Signal-to-interference-to-noise ratio In the calculation process, the numerator represents the desired user signal's useful received power after precoding, and the denominator represents multi-user interference and noise from other users, thus fully depicting the competitive relationship between signal and interference in a large-scale MIMO system.

[0040] S42: Optimize the analog precoding matrix with the goal of maximizing system performance and rate. and digital precoding matrix Output the optimal analog precoding matrix and digital precoding matrix ; ; in, For the system's total throughput and rate, The number of subcarriers, Total transmission power, For simulating the precoding matrix The amount of elements.

[0041] Optimize the analog precoding matrix and digital precoding matrix The three constraints are total transmit power constraint, analog phase shifter constant mode constraint, and RF chain number constraint. By establishing an optimization model with the goal of maximizing system and rate, the transmission relationship between signal and interference in multi-user MIMO channels is accurately characterized, providing a unified mathematical framework for hybrid precoding design. This model can maximize system throughput and suppress inter-user interference while ensuring hardware feasibility, thereby improving spectral efficiency and system stability.

[0042] Total transmit power constraint: Ensure that the transmit power of the communication base station does not exceed the hardware limit to avoid power amplifier saturation and excessive power consumption; Analog phase shifter constant mode constraint: adapting to actual RF hardware structure, the analog precoding matrix can only adjust the phase, not the amplitude.

[0043] Radio frequency link quantity constraint: Ensure that the number of radio frequency links supports the number of users that can be scheduled simultaneously, in accordance with the hardware architecture.

[0044] S5: Based on the angle domain channel matrix, extract the user's horizontal and vertical departure angle information, design the transmit spatial beamforming matrix, and combine the optimal analog precoding matrix and digital precoding matrix to calculate the final concatenated full-link precoding matrix, thereby obtaining the precoded transmit signal vector of the subcarrier. The communication base station generates radio frequency signals based on the precoded transmit signal vector and sends them to the user.

[0045] Step S5 specifically includes the following steps: S51: Based on the angle domain channel matrix Extract the user's horizontal and vertical departure angle information to design the transmit spatial beamforming matrix. ; ; in, w These are the numbers of adjacent base stations. For the first w The number of adjacent communication base stations to the first k Angle-domain channel matrix for each user For the first w Transmit beamforming matrix of adjacent communication base stations; S52: Utilizing a transmit spatial beamforming matrix Optimal analog precoding matrix and digital precoding matrix Calculate the final concatenated full-link precoding matrix ; ; S53: Perform link-layer scheduling, LDPC / Turbo coding, QAM modulation, and resource mapping on the downlink data for each user to obtain the modulation symbol matrix. ; S54: Based on modulation symbol matrix and full-link precoding matrix After obtaining the precoded subcarrier, the transmitted signal vector is obtained. ; ; S55: Using a time-domain shaping filter to shape the precoded transmit signal vector Time-domain shaping filtering is performed to obtain the time-domain transmit baseband signal. The time-domain transmit baseband signal is then input into the radio frequency link of the communication base station to form a radio frequency signal. The radio frequency signal is then transmitted to the user through the transmit antenna array of the distributed remote antenna unit (RAU) of the communication base station.

[0046] A communication base station for performing the above-described mobile communication method includes a distributed antenna unit, a baseband processing unit, a radio frequency link, and a central control unit. Distributed antenna elements include N Each distributed remote antenna unit (RAU) is configured with one distributed remote antenna unit (RAU). M Root transmitting antenna; The baseband processing unit is used to generate time-domain transmit baseband signals; The radio frequency link is used to convert time-domain baseband signals into radio frequency signals.

[0047] The technical solution of this invention covers the entire signal processing process of mobile communication base stations, has low computational complexity, and can be directly adapted to existing 3GPP standards and hardware platforms.

Claims

1. A mobile communication method, characterized in that, Includes the following steps: S1: The communication base station sends a pilot configuration command to the user, the user sends an uplink pilot signal, the communication base station receives the frequency domain baseband pilot signal, estimates the Doppler frequency offset of the user in the high-speed mobile scenario based on the maximum likelihood criterion, obtains the phase-compensated received pilot signal, and converts it into a vectorized received signal. S2: Construct a linear regression model using vectorized received signals, perform adaptive sparse channel estimation based on variational Bayesian inference, and output the downlink frequency domain channel matrix in TDD mode and the angle domain channel matrix in FDD mode. S3: The angle domain channel matrix is ​​quantized using a Lloyd-Max non-uniform quantizer and then input into the encoder network of a lightweight autoencoder to compress it into a low-dimensional bitstream; the user feeds back the low-dimensional bitstream to the communication base station, and the communication base station reconstructs the low-dimensional bitstream to obtain the global channel matrix at the base station. S4: Calculate the signal-to-interference-plus-noise ratio (SINR) of the user on the subcarrier based on the downlink frequency domain channel matrix of the full subcarrier, optimize the analog precoding matrix and digital precoding matrix, and output the optimal analog precoding matrix and digital precoding matrix. S5: Based on the angle domain channel matrix, extract the user's horizontal and vertical departure angle information, design the transmit spatial beamforming matrix, and combine the optimal analog precoding matrix and digital precoding matrix to calculate the final concatenated full-link precoding matrix, thereby obtaining the precoded transmit signal vector of the subcarrier. The communication base station generates radio frequency signals based on the precoded transmit signal vector and sends them to the user.

2. The mobile communication method according to claim 1, characterized in that, Step S1 includes: S11: The communication base station sends pilot configuration instructions to all users through the broadcast channel. Users transmit uplink pilot signals on the specified uplink time-frequency resources. The uplink orthogonal pilot matrix of the uplink pilot signal is... Uplink orthogonal pilot matrix Satisfying orthogonality constraints: ； in, This represents the conjugate transpose operation. The number of pilot symbols, for K An identity matrix of order 1. K identity matrix of order The diagonal elements are 1, and the rest are 0. K The number of users per antenna; S12: Communication base station The root receiving antenna receives the uplink pilot signal sent by the user. After completing RF down-conversion and analog-to-digital conversion, the frequency domain baseband pilot signal is obtained. The frequency domain baseband pilot matrix of the frequency domain baseband pilot signal is: ; ； Among them, the frequency domain baseband pilot matrix The latitude is , latitude is The uplink frequency domain channel matrix, This refers to the number of distributed remote antenna units (RAUs) in a communication base station. M The number of transmit antennas for a single distributed remote antenna unit (RAU). latitude is The additive white Gaussian noise matrix at the receiver; S13: Based on the frequency domain baseband pilot matrix The element obtains the received pilot signals for each user on different antennas and pilot symbols. Based on the maximum likelihood criterion, the Doppler frequency offset of each user in a high-speed moving scenario is estimated, and the estimated value of the user's Doppler frequency offset is obtained. ; ； in, k Number the user. For the traversal search variable of Doppler frequency offset, m Number the antenna. n Numbering the pilot symbols, For users k In the n Conjugate operation of transmit pilot symbols on each pilot symbol j The imaginary unit, OFDM symbol period; S14: Based on Doppler frequency offset estimation Phase compensation is performed on the received pilot signal for each user to eliminate the linear phase rotation caused by Doppler, resulting in a phase-compensated received pilot signal. ; ； S15: All users' received pilot signals By splicing the components, we obtain the globally compensated frequency domain baseband pilot matrix. ; S16: Convert the baseband pilot matrix in the frequency domain Vectorization processing is performed to obtain the vectorized received signal. Simultaneously, a sensing matrix for channel estimation is constructed. ; ； in, For matrix vectorization operations, for An identity matrix of order 1. This represents the Kronecker product operation.

3. The mobile communication method according to claim 2, characterized in that, Step S2 includes: S21: Based on the perception matrix and vectorized received signal Construct a linear regression model for sparse channel estimation, and apply it to the uplink channel vector. Each channel element in Assign an independent Gaussian prior distribution, and assign a Gamma conjugate prior distribution to the precision parameter of each channel element; S22: Define noise accuracy For noise accuracy By assigning a Gamma conjugate prior distribution and minimizing the KL divergence between the variational distribution and the true posterior distribution, the uplink channel vector is solved. The optimal variational posterior distribution; S23: Fixed uplink channel vector posterior mean and posterior covariance matrix Update accuracy parameters and noise accuracy The variational posterior distribution is given, and the convergent posterior mean is output. and the posterior mean Reconstructed into uplink frequency domain channel matrix And converting it into an angle-domain channel matrix. .

4. The mobile communication method according to claim 3, characterized in that, Step S21 includes: S211: Based on the perception matrix and vectorized received signal Construct a linear regression model for sparse channel estimation; ； in, The uplink channel vector is a vectorized uplink channel vector. The latitude is Uplink channel vector For the target to be estimated, The vectorized noise vector The latitude is ; S212: For uplink channel vector Each channel element in Assign independent Gaussian prior distributions: ; in, i Channel number, uplink channel vector Element number in , For the first i Channel elements The accuracy parameters; S213: Assign a Gamma conjugate prior distribution to the precision parameter of each channel element: ; in, These are the conjugate prior distributions of Gamma. Shape parameters and rate parameters.

5. The mobile communication method according to claim 4, characterized in that, Step S22 includes: S221: Define noise accuracy For noise accuracy Assign a Gamma conjugate prior distribution: To achieve noise variance Adaptive estimation; in, These are the conjugate prior distributions of Gamma. Shape parameters and rate parameters, The standard deviation of noise; S222: Solve for the uplink channel vector by minimizing the KL divergence between the variational distribution and the true posterior distribution. The optimal variational posterior distribution is a complex Gaussian distribution: ; ； in, uplink channel vector The posterior mean, with dimension . , uplink channel vector The posterior covariance matrix, with dimension . , Indicates noise accuracy Mathematical expectation operation, For noise accuracy The precision matrix, It is a diagonal matrix. For the precision matrix The Middle Noise accuracy.

6. The mobile communication method according to claim 5, characterized in that, Step S23 includes: S231: Fixed uplink channel vector posterior mean and posterior covariance matrix Update precision parameters and noise accuracy The variational posterior distribution is calculated, and the accuracy parameter is determined. and noise accuracy The expected result is that after the update is complete, the precision matrix will be refreshed. ; Update precision parameters variational posterior distribution The method is as follows: ； Precision parameters The expectation is: ; Update noise accuracy variational posterior distribution The method is as follows: ； Noise accuracy The expectation is: ; S232: Obtain the uplink channel vector before and after the update. posterior mean Calculate the magnitude change of the channel mean vector. ; ； in, The 2-norm of a vector; S233: Set the threshold for the magnitude change ,like If the number of updates reaches the maximum value, then stop updating and output the converged posterior mean. Otherwise, continue updating the precision parameters. and noise accuracy The variational posterior distribution; S234: The convergent posterior mean Reconstructed into uplink frequency domain channel matrix ; In TDD mode, channel reciprocity is utilized to output the downlink frequency domain channel matrix. This yields the downlink channel matrix for all subcarriers; In FDD mode, the downlink frequency domain channel matrix is ​​transformed using DFT. Transform to the angle domain to obtain the angle domain channel matrix. ; ； in, The DFT matrix of the base station transmitter has dimensions of Used to convert spatial channels to the angular domain. The DFT matrix at the user receiver and the channel matrix in the angle domain. Indicates the first k The user in the first n Angle-domain channel matrix on each subcarrier.

7. The mobile communication method according to claim 6, characterized in that, Step S3 includes: S31: Using a Lloyd-Max non-uniform quantizer to quantize the angle-domain channel matrix The amplitude and phase of the mid-angle domain channel elements are quantized separately to minimize quantization distortion, and the quantized angle domain channel matrix is ​​output. ; ； in, x Angle domain channel matrix The amplitude or phase of the channel element in the mid-angle domain. Let be the probability density function. The first u The upper and lower boundaries of each quantization interval u The number of the quantization interval. The quantization level of the quantization interval. Q This represents the total number of quantization levels. S32: The user terminal will quantize the angle domain channel matrix. In the encoder network of the lightweight autoencoder, the high-dimensional angle domain channel matrix is ​​input. Compressed into a low-dimensional bitstream ; ； in, For encoder network functions, For encoder network parameters; S33: The user transmits the low-dimensional code stream via the uplink control channel. The data is fed back to the communication base station, which then receives the low-dimensional bitstream. The angle-domain channel matrix is ​​reconstructed by inputting a lightweight autoencoder into a decoder network. ; S34: Transform the angle domain channel matrix using the inverse DFT. Transforming back to the frequency domain, we obtain the full subcarrier downlink frequency domain channel matrix reconstructed at the base station. ; ； S35: Transform the full subcarrier downlink frequency domain channel matrix By concatenating the matrices, the global channel matrix at the base station is obtained. ; 。 8. The mobile communication method according to claim 7, characterized in that, Step S4 includes: S41: Based on the full subcarrier downlink frequency domain channel matrix Calculate the first k The user in the first n Signal-to-interference-plus-noise ratio on each subcarrier ; ； in, v For simulating the precoding matrix Column number, To simulate the precoding matrix, For the first k The user in the first n Digital precoding matrix on each subcarrier; S42: Optimize the analog precoding matrix with the goal of maximizing system performance and rate. and digital precoding matrix Output the optimal analog precoding matrix and digital precoding matrix ; ； in, For the system's total throughput and rate, The number of subcarriers, Total transmission power, For simulating the precoding matrix The amount of elements.

9. The mobile communication method according to claim 8, characterized in that, Step S5 includes: S51: Based on the angle domain channel matrix Extract the user's horizontal and vertical departure angle information to design the transmit spatial beamforming matrix. ; ； in, w These are the numbers of adjacent base stations. For the first w The number of adjacent communication base stations to the first k Angle-domain channel matrix for each user For the first w Transmit beamforming matrix of adjacent communication base stations; S52: Utilizing a transmit spatial beamforming matrix Optimal analog precoding matrix and digital precoding matrix Calculate the final concatenated full-link precoding matrix ; ； S53: Perform link-layer scheduling, LDPC / Turbo coding, QAM modulation, and resource mapping on the downlink data for each user to obtain the modulation symbol matrix. ; S54: Based on modulation symbol matrix and full-link precoding matrix After obtaining the precoded subcarrier, the transmitted signal vector is obtained. ; ； S55: Using a time-domain shaping filter to shape the precoded transmit signal vector Time-domain shaping filtering is performed to obtain the time-domain transmit baseband signal. The time-domain transmit baseband signal is then input into the radio frequency link of the communication base station to form a radio frequency signal. The radio frequency signal is then transmitted to the user through the transmit antenna array of the distributed remote antenna unit (RAU) of the communication base station.

10. A communication base station for executing the mobile communication method according to any one of claims 1-9, characterized in that, It includes distributed antenna units, baseband processing units, radio frequency links, and a central control unit; The distributed antenna unit includes N Each of the distributed remote antenna units (RAUs) is configured with a specified number of distributed remote antenna units (RAUs). M Root transmitting antenna; The baseband processing unit is used to generate time-domain transmitted baseband signals; The radio frequency link is used to convert the time-domain transmitted baseband signal into a radio frequency signal.

Images