Physical channel fingerprint acquisition and map construction method for massive MIMO-OFDM system based on virtual antenna array

Abstract

This invention proposes a method for acquiring physical channel fingerprints and constructing physical channel maps for large-scale MIMO-OFDM systems based on virtual antenna arrays. This method targets each grid within a target communication area, controlling a measurement terminal to move along a preset trajectory within the grid to synthesize a virtual antenna array to collect channel samples in the spatial-frequency-time domain. Based on the mapping relationship between physical channel fingerprints and beam-domain statistical channel state information, the channel samples measured by the virtual antenna array are processed to construct and solve an optimization problem using beam-domain statistical channel state information as an auxiliary variable and physical channel fingerprints as the target variable, thus acquiring the physical channel fingerprints. The position coordinate indexes of each grid are associated and stored with their corresponding physical channel fingerprints to construct a physical channel map with position indexing functionality. This invention, through flexible measurement terminal deployment, achieves reliable physical channel fingerprint acquisition and can be used for applications such as auxiliary statistical channel information acquisition, instantaneous channel estimation, and precoding.

Description

Technical Field

[0001] This invention belongs to the field of wireless communication technology and proposes a method for constructing physical channel fingerprints and spectra based on virtual antenna arrays for large-scale MIMO-OFDM systems. Background Technology

[0002] As 5G mobile communication systems evolve towards 6G, while inheriting key technologies such as massive MIMO-OFDM from 5G systems, the systems need to simultaneously meet new demands such as higher user terminal connection density, higher data transmission rates, lower latency, and higher-precision positioning. The synergistic improvement of these multi-dimensional performance indicators significantly increases the design complexity of wireless communication systems, and the coupling between different technical modules becomes increasingly tight. Against this backdrop, location-aware communication has gradually become an important means of improving system performance. To enhance the system's environmental awareness capabilities, channel map technology has gained increasing attention. This type of method constructs a channel fingerprint database with location indexes within a target area, enabling direct lookup of the corresponding channel fingerprint when the terminal location is known. It leverages prior information generated from fingerprints to assist in instantaneous channel estimation and precoding design, thereby reducing the online training and detection overhead of acquiring prior channel information. The effectiveness of a channel map largely depends on the form of the channel fingerprints it stores. Physical channel fingerprints, through statistical modeling of multipath propagation characteristics in the propagation environment, characterize the physical propagation structure of the channel with a set of low-dimensional parameters and can be further mapped to generate different forms of statistical channel state information, exhibiting good versatility and stability.

[0003] However, acquiring physical channel fingerprints typically relies on high-resolution estimation of parameters such as angle and latency. In practical large-scale measurement scenarios, lightweight, single-antenna or low-antenna measurement terminals are more feasible, but their limited spatial aperture makes it difficult to directly support high-precision fingerprint acquisition. Summary of the Invention

[0004] Purpose of the invention: This invention aims to provide a method for physical channel fingerprint acquisition and spectrum construction of large-scale MIMO-OFDM systems based on virtual antenna arrays, achieving high-precision physical channel fingerprint acquisition and location index channel spectrum construction, and providing reliable support for applications such as statistical channel acquisition, channel estimation and precoding based on channel spectrum.

[0005] Technical Solution: To achieve the above-mentioned objectives, the present invention provides a method for acquiring physical channel fingerprints and constructing a physical channel map for a large-scale MIMO-OFDM system based on a virtual antenna array, comprising the following steps: dividing the target communication area into multiple spatial grids of predetermined size, establishing a unique position coordinate index for each grid, and determining the composition of the physical channel fingerprint within the area; for each grid, controlling the measurement terminal to move within the grid according to a preset trajectory, synthesizing a virtual antenna array to collect channel samples in the spatial-frequency-time domain; processing the channel samples measured by the virtual antenna array according to the mapping relationship between the physical channel fingerprint and the beam domain statistical channel state information, constructing and solving an optimization problem with the beam domain statistical channel state information as an auxiliary variable and the physical channel fingerprint as the target variable, thereby acquiring the physical channel fingerprint; associating and storing the position coordinate index of each grid with its corresponding physical channel fingerprint, and constructing a physical channel map with position indexing function.

[0006] Furthermore, the physical channel fingerprint refers to the set of statistical characteristic parameters corresponding to all physical propagation path clusters between the user terminal and the base station; based on the physical channel fingerprint, the joint power distribution function in the angle-delay domain can be reconstructed to characterize the statistical characteristics of the channel; the statistical characteristic parameters include the mean, standard deviation, or variance of cluster power, delay, and angle parameters.

[0007] Furthermore, the virtual antenna array is constructed by a mobile measurement terminal moving at a constant speed vector along a preset trajectory. The trajectory formed by the spatial position of the measurement terminal at different times is equivalent to a virtual array with a larger equivalent aperture.

[0008] Furthermore, when the measurement terminal equipped with a single antenna moves at a constant speed and direction, the virtual antenna array is a uniform linear array, and the displacement of the measurement terminal within a single transmission time slot constitutes the equivalent element spacing of the virtual antenna array; the equivalent element spacing satisfies the spatial Nyquist sampling theorem to provide unambiguous estimation resolution of the terminal-side angular parameters.

[0009] Furthermore, the beam domain is a quantized form of a continuous angle-delay-Doppler domain. The angle domain, delay domain, and Doppler domain after sampling and quantization are respectively called the spatial beam domain, frequency beam domain, and time beam domain. Each quantized grid point in the beam domain corresponds to a spatial-frequency-time domain beam with a specific angle, delay, and Doppler frequency shift. The beam domain statistical channel state information refers to the power distribution tensor of the channel in the beam domain. The power distribution tensor is a quantized form of the three-dimensional power distribution function in the angle-delay-Doppler domain.

[0010] Furthermore, the mapping relationship between the physical channel fingerprint and the beam domain statistical channel state information refers to the following: each beam domain grid point corresponds to a quantization interval in the angle-delay domain; the integral value of the joint power distribution function in the angle-delay domain over the quantization interval is the power value at the beam domain grid point; and the power values ​​at all beam domain grid points together constitute the beam domain statistical channel state information.

[0011] Furthermore, based on the above mapping relationship, a two-step method can be used to construct and solve the physical channel fingerprint optimization problem. The steps include: First, based on the measured spatial-frequency-time domain channel samples, construct a maximum likelihood optimization problem, using beam domain statistical channel state information as the optimization objective, and obtain the beam domain statistical channel state information through iterative solution; Second, based on the beam domain statistical channel state information obtained in the first step, construct a Kullback-Leibler divergence minimization problem and perform iterative clustering and statistical parameter fitting to obtain the estimation result of the physical channel fingerprint.

[0012] On the other hand, based on the mapping relationship between the physical channel fingerprint and the beam domain statistical channel state information, the physical channel fingerprint optimization problem can be constructed and solved using the joint expectation-maximization method. The steps include: constructing a maximum joint likelihood optimization problem based on the space-frequency-time domain channel samples obtained from the measurement; and alternately performing the following steps under the problem: updating the beam domain statistical channel state information by the expectation-maximization method while keeping the current physical channel fingerprint parameters fixed; and updating the estimation result of the physical channel fingerprint while keeping the updated beam domain statistical channel state information fixed.

[0013] Furthermore, the physical channel map is a database composed of physical channel fingerprints indexed by location; the map is queried according to the location information of the user terminal to obtain the corresponding physical channel fingerprint; based on the physical channel fingerprint, combined with the terminal's mobility characteristics, beam domain statistical channel information adapted to the current system bandwidth and antenna configuration is generated, which can be used for applications such as instantaneous channel estimation and precoding.

[0014] The present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is loaded onto the processor, it implements the method for physical channel fingerprint acquisition and map construction of a large-scale MIMO-OFDM system based on a virtual antenna array.

[0015] Beneficial effects: This invention introduces a virtual antenna array mechanism into a large-scale MIMO-OFDM system. Without deploying multi-antenna measurement equipment, it utilizes the controlled movement of the measurement terminal to reconstruct an equivalent multi-antenna array, achieving high-resolution sensing of physical parameters such as channel angles. This reduces the hardware cost and deployment complexity of physical channel fingerprint acquisition. By establishing a mapping relationship between physical channel fingerprints and beam domain statistical channel state information, the complex continuous-domain statistical parameter estimation problem is transformed into an iteratively solvable optimization problem, ensuring the reliability and accuracy of physical channel fingerprint acquisition. Furthermore, the constructed channel map can complete the main calculations offline and generate statistical channel information adapted to different large-scale MIMO-OFDM system configurations and terminal movement states on demand during the online phase, effectively reducing the overhead of online channel detection and calculation. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the method flow according to an embodiment of the present invention.

[0017] Figure 2 This is a schematic diagram of a cluster-based geometric random channel model in an embodiment of the present invention.

[0018] Figure 3 This is a schematic diagram of the virtual antenna array architecture in an embodiment of the present invention.

[0019] Figure 4 This is a performance diagram of the physical channel fingerprint acquisition method proposed in an embodiment of the present invention. Detailed Implementation

[0020] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0021] like Figure 1 As shown in the embodiments of the present invention, the method for acquiring physical channel fingerprints and constructing maps of large-scale MIMO-OFDM systems based on virtual antenna arrays includes the following steps:

[0022] The target communication area is divided into multiple spatial grids with predetermined sizes, and a unique location coordinate index is established for each grid. At the same time, the composition of the physical channel fingerprint within the area is determined.

[0023] For each grid, the control measurement terminal moves within the grid according to a preset trajectory to synthesize a virtual antenna array to collect channel samples in the spatial-frequency-time domain;

[0024] Based on the mapping relationship between physical channel fingerprint and beam domain statistical channel state information, the channel samples measured by the virtual antenna array are processed, and an optimization problem is constructed and solved with beam domain statistical channel state information as auxiliary variable and physical channel fingerprint as target variable, so as to obtain physical channel fingerprint.

[0025] The location coordinate index of each grid is associated with its corresponding physical channel fingerprint and stored to construct a physical channel map with location indexing function.

[0026] Figure 2 A schematic diagram of a propagation scenario according to an embodiment of the present invention is provided, in which a base station (BS) and multiple user terminals (UTs) on the ground communicate with each other. Different propagation paths between the BS and the UTs are represented by different physical propagation path clusters, and each cluster contains multipath components with similar propagation characteristics.

[0027] Figure 3 A schematic diagram of the synthesized virtual antenna array in an embodiment of the present invention is given. The mobile measurement terminal (MT) sends pilot signals to the BS according to the given frame structure. The positions of the MT at all pilot transmission times within the same data frame can be regarded as virtual antenna elements. The positions of the MT at different pilot transmission times together constitute a virtual uniform linear array (ULA). The spacing between the array elements is the moving distance of the MT within a single pilot transmission time slot.

[0028] The following describes in detail the specific implementation process of the physical channel fingerprint acquisition and spectrum construction method for a large-scale MIMO-OFDM system based on a virtual antenna array, using a specific communication system example. It should be noted that the method of the present invention is not only applicable to the specific system model mentioned in the example below, but also to system models with other configurations.

[0029] I. System Model

[0030] Consider a device operating at a carrier frequency This is a single-cell TDD massive MIMO-OFDM system. To capture location-dependent channel characteristics, the cell is divided into multiple small grids. Within each small grid, large-scale parameters remain constant, satisfying the wide stationarity assumption. The BS side is equipped with a... The ULA of the root antenna has an element spacing of... ,in It is the carrier wavelength. OFDM modulation uses... Subcarriers, with a cyclic prefix (CP) length of [length missing]. The sampling interval is The subcarrier spacing is The duration of a single OFDM symbol is Select one of them There are 1 valid training / data subcarriers, with the index 1. In uplink channel detection, a single antenna MT moves along a predefined trajectory, and its velocity vector is... The magnitude and direction angle of the velocity are expressed as follows: and (Measured counterclockwise from the reference axis, such as...) Figure 2 (As shown). The MT periodically sends probe pilot signals to the BS, thereby obtaining channel measurements at specific locations.

[0031] like Figure 2 As shown, the grid where the base station is located With the current MT's grid The propagation environment between them consists of multiple multipath components (MPCs). These MPCs can be well approximated as several clusters, with MPCs within the same cluster originating from similar scattering regions, exhibiting similar power levels, and showing strong correlation in time delay and angular parameters. Let... express and The number of clusters between, Indicates the first The number of MPCs within a cluster. (Subscript) This clearly indicates that these quantities are location-dependent, and that these location-dependent quantities may differ between different grids.

[0032] When the mobile terminal is in its current grid When moving along a certain trajectory, the BS side... The time-varying channel impulse response (CIR) between the root antenna and the MT can be represented by the following geometric-based stochastic model (GBSM):

[0033] (1)

[0034] in This represents the q-th MPC in the c-th cluster. Represents the current path complex gain, where Cluster power, It is a random phase. It is the path delay between the m-th antenna element on the MT and BS sides at the current moment. The above formula constitutes the CIR formula based on GBSM.

[0035] Suppose at time t, Let represent the propagation delay of the q-th MPC in the c-th cluster between the first BS-side antenna element and the MT, i.e., the reference delay. Under the far-field assumption, the path delay in equation (1) is... It can be expressed as about The function, and also related to the first The arrival angle (AoA) and departure angle (AoD) of each multipath component are related and are denoted as follows: and Previous studies have shown that within sufficiently short observation intervals (i.e., corresponding to grid areas) (local trajectory segment in the middle), parameters , and The time-varying delay is negligible, so for the sake of symbolic simplicity, we omit the explicit time dependency, and the time-varying delay in equation (1) is simplified to:

[0036] (2)

[0037] in The speed of light is represented. In equation (2), the three terms correspond to the reference delay, the BS-side delay caused by AoA, and the time-varying delay term caused by MT mobility, respectively. This time-varying term is related to the magnitude of AoD. Figure 2 The corresponding geometric interpretation is shown: in the time interval Within the device, the movement of the mobile terminal introduces an incremental propagation distance. This leads to a change in time delay, which is consistent with the third term in equation (2). Substituting into the time delay expression in (2), CIR can be further expressed as:

[0038] (3)

[0039] According to equation (3), CIR is essentially the superposition of all MPC contributions. Although the parameters of a single MPC may change over a long period, the underlying cluster geometry makes its angles and delays statistically regular. Typically, the AoA, AoD, and delay within a cluster are approximated by a Gaussian distribution: , , Among them, for , and Let represent the mean and standard deviation of the time delay / AoA / AoD distributions, respectively. Then, we define the following parameter set.

[0040] (4)

[0041] It captured the area Statistical characteristics of the internal channel. Unlike instantaneous MPC parameters, This is known as the Physical Channel Fingerprint (PCF). As defined above, the PCF is a set of statistical characteristic parameters corresponding to all physical propagation path clusters. Based on the PCF, the joint power distribution function in the angle-delay domain can be reconstructed to characterize the statistical properties of the channel. For example, the channel statistical characteristics in the spatial-frequency-time (SFT) domain, including time variations dominated by the Doppler effect, can be derived from the PCF. This allows PCFs at different locations to form a robust, storage- and computationally efficient channel chart (CC).

[0042] The PCF is used to characterize the statistical distribution of MPC parameters, and its acquisition usually requires extracting parameters from channel measurements. In large-scale MIMO-OFDM systems, this is typically done on channel transfer function (CTF) samples in the spatial-frequency (SF) domain. The CTF formula for large-scale MIMO-OFDM systems is derived from the CIR in equation (2), assuming that the channel state information remains constant within an OFDM symbol, while changing between different OFDM symbols due to the Doppler effect. Consider an observation interval containing multiple OFDM symbols. When the conditions are met When the CTF corresponding to the k-th subcarrier of the n-th OFDM symbol is approximately equal to...

[0043] (5)

[0044] For the nth OFDM symbol, the CTF of all effective subcarriers and the base station antenna together form a matrix. ,in Vector This indicates the MT speed. However, for a single-antenna MT, The inability to provide spatial diversity at the terminal side makes it impossible to directly resolve the transmission angle from the observed channel transfer function matrix alone. .

[0045] To overcome this fundamental limitation, we adopted a Virtual Antenna Array (VAA) architecture, which synthesizes a virtual antenna aperture by controlling the movement trajectory of the MT (Mobile Measurement Terminal). The virtual antenna array is constructed by the mobile measurement terminal moving at a constant velocity vector along a preset trajectory. The trajectory formed by the spatial position of the measurement terminal at different times is equivalent to a virtual array with a larger equivalent aperture. Its implementation principle is based on the Doppler frequency shift of the MPC (Multi-Purpose Controller). This introduces a phase difference between OFDM symbols, and this phase difference is determined by its AoD, i.e. The uniqueness is determined. This effect is reflected in the second multiplication term in the summation term on the right side of equation (5), namely the term related to the OFDM symbol index n. It should be noted that the derivation from equation (3) to equation (5) is essentially assuming that the Doppler frequency shift in equation (3) does not change with time within an OFDM symbol, and replacing the time variable t on the right side of equation (3) with .

[0046] Utilizing this characteristic, this embodiment uses a constant velocity vector This is achieved by guiding the MT (motor unit) along a predetermined trajectory. For example... Figure 3 As shown, the observation interval is set to a range containing A frame in 1 time slot, each time slot containing One OFDM symbol. Pilots are transmitted from the MT to the BS in the first OFDM symbol of each time slot. Since the time interval between adjacent pilot transmissions is... The movement of the MT causes consecutive pilot transmission times to correspond to different mobile terminal positions on the trajectory. Within one frame, the set of all OFDM symbol indices corresponding to pilot transmissions is... The time-varying phase terms caused by the Doppler frequency shift corresponding to these pilot transmission times correspond to a... Time-domain rudder vector Its nth element is

[0047] (6)

[0048] in It is the distance the MT travels within a continuous pilot transmission interval, i.e., within one time slot. The key point is that, although... Originating from the phase evolving over time, it exhibits characteristics similar to that with element spacing. of The array response vectors of the ULAs are identical to those of the Vandermonde structure. Therefore, the position of the MT at each pilot transmission moment can be considered as a virtual antenna element, and these virtual elements together form a VAA at the terminal side.

[0049] As shown in (6), when the MT with a single antenna is controlled to move at a constant speed and direction, the VAA is a ULA, and the displacement of the MT in a single transmission time slot constitutes the equivalent element spacing of the virtual ULA. The equivalent element spacing satisfies the spatial Nyquist sampling theorem, i.e. This provides unambiguous estimation resolution of the terminal-side angle parameters.

[0050] Using the aforementioned VAA, this embodiment introduces a corresponding channel representation. All The corresponding SF domain channel matrices are stacked to obtain a tensor. ,in The third dimension of this tensor corresponds not only to the OFDM symbol dimension but also to the element dimension of the virtual synthesis array. Vectorizing this tensor yields the corresponding vector representation:

[0051] (7)

[0052] in ,symbol Denotes the Kronecker product of vectors, and

[0053] (8)

[0054] When the spatial broadband effect is negligible, the above vector model can be further simplified to:

[0055] (9)

[0056] In the SFT domain channel representation described above, the vector embedding incorporates the virtual spatial response. and There is a one-to-one correspondence, therefore for Processing enables joint estimation of AoD, AoA, and time delay. Although VAA is introduced in this embodiment to overcome the limitations of single-antenna MT in angle estimation, this method can also be extended to multi-antenna mobile terminals to further increase the virtual antenna aperture size, thereby improving the angle estimation accuracy.

[0057] Because the VAA architecture supports complete MPC parameter estimation, and PCF is a statistical property of the physical parameters corresponding to MPC, there is a certain mapping relationship between PCF and Channel State Information (CSI) in the SFT domain under the VAA architecture. This represents the power spectrum in the angle-delay (AD) domain. Under the uncorrelated scattering approximation,

[0058] (10)

[0059] in , and Let represent the probability density functions (PDFs) of AoD, AoA, and delay for all MPCs in the p-th cluster, respectively. These PDFs can be modeled as Gaussian distributions. For example, the PDF of AoA is given by the following equation:

[0060] (11)

[0061] Then the covariance matrix of the channel in the SFT domain The following relationships exist with PCF:

[0062] (12)

[0063] Equation (12) establishes a forward mapping from PCF to SFT domain statistics. This makes the corresponding inverse problem, namely estimating PCF from VAA-based channel observations, theoretically well-posed and forms the subsequent objective of the embodiment. However, directly estimating PCF under the original GBSM is computationally difficult because PCF is defined in a continuous parameter space. Therefore, this embodiment instead uses a Beam Based Channel Model (BBCM) as an alternative for channel representation, which is defined in a discrete parameter space. Subsequent embodiments will derive the relationship between PCF and channel statistics under the VAA architecture and the BBCM.

[0064] II. Relationship between Beam Domain Statistical CSI and PCF and PCF Acquisition Based on Two-Step Method

[0065] This embodiment first discretizes the continuous physical parameter space to construct a three-dimensional beam domain. The beam domain can be considered as a quantized form of the continuous angle-delay-Doppler domain. The sampled and quantized angle domain, delay domain, and Doppler domain are respectively called the spatial beam domain, frequency beam domain, and time beam domain. Each quantized grid point in the beam domain corresponds to a spatial-frequency-time domain beam with a specific angle, delay, and Doppler shift. In large-scale MIMO-OFDM systems, the channel gains between different beams can be approximated as independent. The beam domain statistical channel state information specifically refers to the channel power distribution tensor in the beam domain, which can be considered as a quantized form of the angle-delay-Doppler domain three-dimensional power distribution function.

[0066] Specifically, the cosine value of AoA is defined as The beam domain is determined by the cosine of the angle of arrival (θ). ), delay ( ) and Doppler ( The dimensions are uniformly quantized as follows: , and The set of grid points formed by these grid points is denoted as: , and ,in , , Each grid point Represents a discrete beam and is associated with a quantization interval: , and ,in . It is known as the triple beam domain.

[0067] We now define the beam domain channel tensor Each element in the tensor aggregates the MPC complex gain of the corresponding parameters falling into the corresponding three-dimensional beam element:

[0068] (13)

[0069] Typically, beam domain resolution , and These are set to integer multiples of the number of antennas, the number of effective subcarriers, and the number of pilot symbols, respectively. , , , where F is also called the refinement factor. When , and When large enough, any All can use the corresponding grid points A good approximation. In this case, the entire SFT domain channel tensor can be approximated as...

[0070] (14)

[0071] in The third-order Einstein product of a tensor is represented by the tensor. The elements are given by the following formula

[0072] (15)

[0073] make Represents the beam domain power tensor (second-order scatterer channel state information), operators This represents the dot product. Under the common uncorrelated scattering assumption, each element represents the total power of all multipath components within the corresponding beam domain. The beam domain is constructed by uniformly quantizing the angle-of-arrival cosine, time delay, and Doppler shift. AoA cosine quantization value. and Doppler quantization value It is possible and Mapping to candidate physical angles. However, due to the periodicity of the cosine function, this mapping is inherently many-to-one: a single Doppler value corresponds to multiple possible departure angles, making it impossible to identify physical parameters solely from beam-domain observations. To establish a unique and physically interpretable mapping, we impose an identifiability condition by restricting the analysis to the principal monotonic intervals of the inverse cosine function, i.e., requiring... This ensures that each beam interval corresponds to a unique set of physical angle intervals, which is crucial for unambiguously recovering the power coupling function parameters. Under this condition, the beam interval and Simplified to a well-defined physical angle range and ,in and For channel estimation, a moderate refinement factor (e.g., F=2) typically yields good results. However, for beam domain power distribution tensor acquisition, a sufficiently refined beam domain is required to accurately approximate the actual continuous power spectrum. Therefore, subsequent PCF acquisition processes are based on the assumption of a sufficiently large refinement factor. Under this assumption, along with the uncorrelated scattering assumption, the following mapping relationship can be obtained between PCF and beam domain statistical CSI: when the beam domain dimension is sufficiently large, each beam domain grid point corresponds to a fine quantization interval in the angle-delay domain; the integral value of the joint power distribution function in the angle-delay domain over this quantization interval is the power value at that beam domain grid point; the power values ​​at all beam domain grid points together constitute the beam domain statistical channel state information. Mathematically, this can be expressed as: beam domain statistical CSI satisfies...

[0074] (16)

[0075] in , ,and The probability density functions of all MPCs in the c-th cluster falling within the current beam interval are as follows:

[0076] (17)

[0077] (18)

[0078] (19)

[0079] This represents the probability accumulation function of the standard normal distribution.

[0080] The above conclusions establish a deterministic forward mapping from PCF to beam domain statistical CSI. Therefore, in the PCF estimation process, beam domain statistical CSI can be used as an auxiliary variable, leading to the following two-step PCF acquisition method. That is, based on the mapping relationship between the physical channel fingerprint and beam domain statistical channel state information, the beam domain statistical channel state information can be used as an auxiliary variable to solve the physical channel fingerprint acquisition problem. Specifically, a two-step method can be adopted. The steps of constructing and solving the physical channel fingerprint optimization problem include: First, based on the measured SFT domain channel samples, construct a maximum likelihood optimization problem, using beam domain statistical CSI as the optimization objective, and obtain the beam domain CSI through iterative solution; Second, based on the beam domain CSI obtained in the first step, construct a KL divergence minimization problem and perform iterative clustering and statistical parameter fitting to obtain the PCF estimation result.

[0081] The following uses mathematical formulas to explain the detailed process of the two-step method. Beam domain statistical CSI is first estimated from the SFT domain channel observation results; then, a PCF is determined based on this estimation result, denoted as . This makes it possible to The beam domain statistical CSI obtained by substituting into equations (16)-(19) best matches the beam domain statistical CSI estimated in the first step. The estimation of the beam domain statistical CSI in the first step adopts the expectation maximum (EM) algorithm. Specifically, the S S SFT domain channel samples obtained by measurement are denoted as The estimation problem in the first step can be expressed as:

[0082] (20)

[0083] In solving the above problem, we introduce the beam domain channel tensor. As a latent variable. In the EM algorithm... In this iteration, the Q function constructed in the E-step is as follows:

[0084] (twenty one)

[0085] This is the conditional expectation of the log-likelihood. The subsequent M steps, by maximizing this Q-function, yield the following update rule:

[0086] (twenty two)

[0087] Iterating through these two steps yields an estimate of the auxiliary variable, denoted as . This estimate will serve as the input for the PCF fitting stage described below. Since the channel data dimension is large in large-scale MIMO-OFDM systems, the a posteriori channel information calculation in (22) can be performed using the Information Geometry Approach (IGA) to reduce computational complexity.

[0088] The second step, fitting and solving, still follows an iterative method. In the... In the next iteration, let Let the power allocation tensor be defined to represent the PCF updated in the last iteration. Each of its elements is

[0089] (twenty three)

[0090] in , and By respectively Substituting into (17)-(19), we obtain the result. Then, the power is updated according to the following rules:

[0091] (twenty four)

[0092] Angle and time delay parameters are updated according to the same rules. Taking the time delay parameter as an example:

[0093] (25)

[0094] (26)

[0095] when , , At that time, the iterative updates in equations (24)-(26) constitute a process, which makes and The Kullback-Leibler (KL) divergence between them is monotonically decreasing and converges to a stationary point in the following optimization problem:

[0096] (27)

[0097] Since the number of clusters is usually unknown in advance, this embodiment of the invention determines the number of clusters by minimizing the Davies-Bouldin (DB) exponent, and the initial value for PCF fitting in the second step is determined by the K-Power-Means algorithm. This two-stage design provides a conceptually simple and computationally efficient solution by decoupling beam domain statistical CSI estimation from PCF parameter fitting. However, due to this decoupling, the Gaussian prior distribution of the angle-delay parameter is only explicitly included in the second stage. Therefore, although the beam domain statistical CSI obtained in the first stage maximizes the likelihood function in equation (20), it cannot guarantee that it is optimally consistent with the assumed Gaussian structure of the physical parameters. In order to make fuller use of the Gaussian structure of the physical parameters and further improve the estimation efficiency, a PCF acquisition framework based on joint EM will be introduced later.

[0098] III. PCF Acquisition Based on Joint EM Algorithm

[0099] This embodiment also proposes a method based on joint EM to obtain PCF. The steps include: constructing a maximum joint likelihood optimization problem based on the measured spatial-frequency-time domain channel samples; under the problem, alternately performing the following steps: updating the beam domain statistical CSI by EM method while keeping the current PCF parameters fixed; updating the PCF while keeping the updated beam domain statistical CSI fixed.

[0100] Specifically, this method differs from the two-step method, in which beam domain statistics (CSI) and PCF are estimated sequentially. The subsequent EM method embeds the two stages of the two-step method into a unified EM framework, ensuring that the Gaussian structure of the physical parameters always dominates the PCF acquisition process.

[0101] This method directly addresses PCF acquisition as the following optimization problem.

[0102] (28)

[0103] In solving the above optimization problem using the EM method, assume that the PCF estimate obtained after the first (t-1) iterations is... First, update the beam domain statistics CSI according to the EM algorithm:

[0104] (29)

[0105] Then use (23)-(26) to The projection is a set of PCFs, denoted as It is important to note that the standard EM algorithm requires ensuring the monotonicity of the likelihood function during optimization. However, in each iteration, the parameter projection process disrupts the monotonically increasing nature of the likelihood function in (28). Therefore, a damped update process needs to be introduced during the iteration:

[0106] (30)

[0107] Where the damping factor In the implementation of the above algorithm, the a posteriori channel information calculation in (29) can still be performed using the information geometry algorithm.

[0108] After obtaining the PCF using the proposed method, a corresponding physical channel map can be constructed. The physical channel map is a database composed of PCFs indexed by location. The map is queried based on the location information of the user terminal to obtain the corresponding PCF. Based on the PCF and combined with the terminal's mobility characteristics, a beam domain statistical CSI adapted to the current large-scale MIMO-OFDM system configuration is generated, which can be used for applications such as instantaneous channel estimation and precoding.

[0109] IV. Implementation Results

[0110] Consider a large-scale MIMO-OFDM system with the following system parameters: carrier frequency Number of subcarriers , length of the cyclic prefix Subcarrier spacing Number of effective subcarriers Number of antenna array elements on the base station side The array element spacing is half a wavelength, and the number of users in the system is... Number of time slots within a frame along the time domain Number of OFDM symbols in a single time slot The user terminal is moving at a speed of 15 km / h. Oversampling factor. This invention employs a 6G Universal Channel Model (6GPCM) generator to generate channel data. The specific implementation steps are as follows: First, a PCF map corresponding to the current cell is constructed based on the PCF acquisition method under the VAA architecture. Then, according to the proposed technical solution, the PCF is converted into sCSI under BBCM under specific system configuration and UT mobility state, and compared with the sCSI obtained by the traditional online detection method. Both sCSIs are applied to the channel estimation process. If their channel estimation performance is comparable, it confirms that the PCF obtained by the proposed method can serve as reliable information to assist in statistical channel acquisition and instantaneous channel estimation.

[0111] The specific simulation process includes two key steps: extracting SFT domain channel samples and obtaining the PCF from the channel measurement results under the VAA architecture, and generating the corresponding sCSI from the PCF based on the UT's current location and movement state to assist in instantaneous channel estimation. In the PCF acquisition stage, the number of OFDM symbols per time slot is set to 44 in the frame structure configuration, and each frame contains 8 time slots to ensure that it can effectively generate an array with an element spacing of half a wavelength. After the PCF acquisition is completed, this embodiment verifies the acquisition of statistical channel information and instantaneous channel estimation based on PCF in a scenario where multiple single-antenna UTs communicate with BSs. It should be noted that in the PCF acquisition stage, the refinement factor is set to 8 to improve the parameter estimation performance, while in the subsequent channel estimation stage, the refinement factor is set to 2.

[0112] In the channel estimation stage, this embodiment considers a large-scale MIMO-OFDM channel acquisition scenario with U=150. The initial position of UT is randomly distributed within the scenario area, the moving speed is 15km / h, and the moving direction is randomly selected within the range of 0 to π. Simulation verification is performed under the typical scenario of "3GPP_38.901_UMa_NLOS". Figure 4 Simulation results show that the PCF obtained by the two methods proposed under the VAA architecture can be used as reliable prior information to assist subsequent instantaneous channel estimation, and the performance of obtaining statistical channel information to assist instantaneous channel estimation is close to that of the ideal online detection method.

[0113] Based on the same inventive concept, the present invention discloses a computer device including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is loaded onto the processor, it implements the method for acquiring physical channel fingerprints and constructing maps of a large-scale MIMO-OFDM system based on a virtual antenna array.

[0114] In a specific implementation, the device includes a processor, a communication bus, a memory, and a communication interface. The processor can be a general-purpose central processing unit (CPU), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits used to control the execution of the program of the present invention. The communication bus may include a path for transmitting information between the aforementioned components. The communication interface, using any transceiver-like device, is used for communicating with other devices or communication networks. The memory can be a read-only memory (ROM) or other types of static storage devices capable of storing static information and instructions, random access memory (RAM) or other types of dynamic storage devices capable of storing information and instructions, or an electrically erasable programmable read-only memory (EEPROM), a read-only optical disc (CD-ROM) or other optical disc storage, disk storage media, or other magnetic storage devices, 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 exist independently and be connected to the processor via a bus. The memory can also be integrated with the processor. The memory is used to store the application code that executes the present invention and is controlled by the processor for execution. The processor executes application code stored in memory to implement the channel acquisition method provided in the above embodiments. The processor may include one or more CPUs, or multiple processors, each of which may be a single-core processor or a multi-core processor. Here, "processor" may refer to one or more devices, circuits, and / or processing cores for processing data (e.g., computer program instructions).

[0115] In the embodiments provided in this application, it should be understood that the disclosed methods can be implemented in other ways without departing from the spirit and scope of this application. The current embodiments are merely exemplary examples and should not be considered limiting, nor should the specific content given limit the purpose of this application. For example, some features may be omitted or not implemented.

[0116] The technical means disclosed in this invention are not limited to those disclosed in the above embodiments, but also include technical solutions composed of any combination of the above technical features. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of this invention, and these improvements and modifications are also considered within the scope of protection of this invention.

Claims

1. A method for acquiring physical channel fingerprints and constructing maps for large-scale MIMO-OFDM systems based on virtual antenna arrays, characterized in that: Includes the following steps: The target communication area is divided into multiple spatial grids with predetermined sizes, and a unique location coordinate index is established for each grid. At the same time, the composition of the physical channel fingerprint within the area is determined. For each grid, the control measurement terminal moves within the grid according to a preset trajectory to synthesize a virtual antenna array to collect channel samples in the spatial-frequency-time domain; Based on the mapping relationship between physical channel fingerprint and beam domain statistical channel state information, the channel samples measured by the virtual antenna array are processed, and an optimization problem is constructed and solved with beam domain statistical channel state information as auxiliary variable and physical channel fingerprint as target variable, so as to obtain physical channel fingerprint. The location coordinate index of each grid is associated with its corresponding physical channel fingerprint and stored to construct a physical channel map with location indexing function.

2. The method for acquiring physical channel fingerprints and constructing maps for large-scale MIMO-OFDM systems based on virtual antenna arrays according to claim 1, characterized in that: The physical channel fingerprint refers to the set of statistical characteristic parameters corresponding to all physical propagation path clusters between the user terminal and the base station. Based on the physical channel fingerprint, the joint power distribution function in the angle-delay domain can be reconstructed to characterize the statistical characteristics of the channel. The statistical characteristic parameters include the mean, standard deviation, or variance of the cluster power, delay, and angle parameters.

3. The method for acquiring physical channel fingerprints and constructing maps for large-scale MIMO-OFDM systems based on virtual antenna arrays according to claim 1, characterized in that: The virtual antenna array is constructed by a mobile measurement terminal moving at a constant speed vector along a preset trajectory. The trajectory formed by the spatial position of the measurement terminal at different times is equivalent to a virtual array with a larger equivalent aperture.

4. The method for acquiring physical channel fingerprints and constructing maps for large-scale MIMO-OFDM systems based on virtual antenna arrays according to claim 3, characterized in that: When a measurement terminal equipped with a single antenna moves at a constant speed and direction, the virtual antenna array is a uniform linear array, and the displacement of the measurement terminal within a single transmission time slot constitutes the equivalent element spacing of the virtual antenna array; the equivalent element spacing satisfies the spatial Nyquist sampling theorem to provide unambiguous estimation resolution of the terminal-side angular parameters.

5. The method for acquiring physical channel fingerprints and constructing maps for large-scale MIMO-OFDM systems based on virtual antenna arrays according to claim 1, characterized in that: The beam domain is a quantized form of a continuous angle-delay-Doppler domain. The angle domain, delay domain, and Doppler domain after sampling and quantization are called the spatial beam domain, frequency beam domain, and time beam domain, respectively. Each quantized grid point in the beam domain corresponds to a spatial-frequency-time domain beam with a specific angle, delay, and Doppler frequency shift. The beam domain statistical channel state information refers to the power distribution tensor of the channel in the beam domain. The power distribution tensor is a quantized form of the three-dimensional power distribution function in the angle-delay-Doppler domain.

6. The method for acquiring physical channel fingerprints and constructing maps for large-scale MIMO-OFDM systems based on virtual antenna arrays according to claim 1, characterized in that: The mapping relationship between the physical channel fingerprint and the beam domain statistical channel state information refers to the following: each beam domain grid point corresponds to a quantization interval in the angle-delay domain; the integral value of the joint power distribution function in the angle-delay domain over the quantization interval is the power value at the beam domain grid point; the power values ​​at all beam domain grid points together constitute the beam domain statistical channel state information.

7. The method for acquiring physical channel fingerprints and constructing maps for large-scale MIMO-OFDM systems based on virtual antenna arrays according to claim 1, characterized in that: A two-step method is used to construct and solve the physical channel fingerprint optimization problem, including: The first step is to construct a maximum likelihood optimization problem based on the measured spatial-frequency-time domain channel samples, using beam domain statistical channel state information as the optimization objective, and obtain the beam domain statistical channel state information by iterative solution. The second step involves constructing a Kullback-Leibler divergence minimization problem based on the beam domain statistical channel state information obtained in the first step, followed by iterative clustering and statistical parameter fitting to obtain the estimation results of the physical channel fingerprint.

8. The method for acquiring physical channel fingerprints and constructing maps for a large-scale MIMO-OFDM system based on a virtual antenna array according to claim 1, characterized in that: The physical channel fingerprint optimization problem is constructed and solved using the joint expectation-maximization method, including: Based on the measured spatial-frequency-time domain channel samples, construct the maximum joint likelihood optimization problem; Under the aforementioned problem, the following steps are performed alternately: with the current physical channel fingerprint parameters fixed, the beam domain statistical channel state information is updated using the expectation-maximization method; with the updated beam domain statistical channel state information fixed, the estimation result of the physical channel fingerprint is updated.

9. The method for acquiring physical channel fingerprints and constructing maps for a large-scale MIMO-OFDM system based on a virtual antenna array according to claim 1, characterized in that: The physical channel map is a database composed of physical channel fingerprints indexed by location; the map is queried based on the location information of the user terminal to obtain the corresponding physical channel fingerprint; Based on this physical channel fingerprint, combined with the terminal's mobility characteristics, beam domain statistical channel information adapted to the current large-scale MIMO-OFDM system configuration is generated for instantaneous channel estimation and precoding.

10. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the computer program is loaded into the processor, it implements the method for acquiring physical channel fingerprints and constructing maps for large-scale MIMO-OFDM systems based on virtual antenna arrays as described in claims 1-9.

Images