Communication-aware integrated massive MIMO channel state and target parameter acquisition method

Abstract

The present invention discloses a method for acquiring channel states and target parameters of large-scale MIMO with integrated communication and perception. By parameterizing high-dimensional communication channels into a small number of physical parameters, an intrinsic correspondence with the perceived target parameters is established from the angle, delay and Doppler dimension. A shared pilot mode is proposed under the same time-frequency resources, so that channel estimation and target parameter estimation can be expressed in the same mathematical form as a structured tensor multiple decomposition model. Based on this model, the uniqueness conditions of the tensor decomposition problem and the maximum number of solvable targets are analyzed using the Vandermonde structure of the factor matrix. On this basis, a unified channel and target estimation algorithm is proposed to jointly estimate parameters including arrival angle / departure angle, delay, Doppler frequency shift, channel coefficient / reflection coefficient. The unified channel state and target parameter acquisition method proposed by the present invention has great advantages in reducing training overhead and improving perception resolution.

Description

Technical Field

[0001] The present invention belongs to MIMO technology, and specifically is a method for acquiring large-scale MIMO channel status and target parameters with integrated communication perception. Background Art

[0002] With the increasing shortage of spectrum resources and the decline in energy efficiency, Integrated Communications and Awareness (ISAC) has emerged. It aims to realize both radar detection and communications awareness on the same hardware platform. It is widely regarded as one of the key technologies for the next generation of 6G wireless communications. Massive MIMO technology uses a large number of antennas in base stations, has significant spectrum and energy efficiency gains, and has become an important enabling technology for current and future wireless cellular networks. The huge degree of freedom and inherent perception capabilities make it possible to combine massive MIMO with ISAC (MIMO-ISAC) to achieve high-quality wireless communications and high-resolution perception performance at the same time, and can be applied to scenarios such as autonomous driving in intelligent transportation, Wi-Fi perception in smart homes, and drone networks in smart cities.

[0003] In order to realize the functions of radar detection and communication perception at the same time, many studies have been conducted. In the early days, a common radar waveform for communication or a common communication waveform for radar was used, such as using a multi-beam framework to simultaneously generate fixed sub-beams for communication and variable scanning sub-beams for perception, or using OFDM signals widely used for communication for target perception; recent studies have jointly designed an integrated waveform by using the additional degrees of freedom and flexibility of MIMO technology to balance the two functions. Although the dual functions of communication and sensing are integrated into one platform to a certain extent, the above-mentioned ISAC technology usually assumes accurate communication channel state information or a priori information of some known target parameters. Therefore, the target parameter estimation problem has been widely studied. Regarding the direction of arrival (DOA) estimation, the current multiple signal classification (MUSIC), estimation of signal parameters by rotation invariance technology (ESPRIT) and compressed sensing have super-resolution and high-precision estimation performance, but the communication transmission of the above methods requires channel state information, which consumes a large number of pilot resources in the massive MIMO-ISAC system with a large number of antennas deployed at the base station. For channel estimation, broadband systems in quasi-static and highly mobile scenarios use the channel structure of a pure communication system of massive MIMO to reduce channel training overhead. At present, a tensor-based signal processing solution has been developed using the multi-dimensional characteristics of massive MIMO channels to achieve accurate channel estimation. However, although the pilot overhead of channel estimation can be greatly reduced using perceptual echoes, existing research is only applicable to line-of-sight channels, and there is currently no effective method to solve the target parameter estimation and multipath time-varying channel estimation problems under the unified architecture of massive MIMO-ISAC. Summary of the invention

[0004] In order to solve the above technical defects in the prior art, the present invention proposes a method for acquiring the channel state and target parameters of a large-scale MIMO with integrated communication perception.

[0005] The technical solution to achieve the purpose of the present invention is: a method for acquiring the channel state and target parameters of a large-scale MIMO integrated with communication perception, the specific steps of which are:

[0006] Step 1: Construct a communication transmission process model;

[0007] Step 2: Construct radar perception process model;

[0008] Step 3: Construct a unified channel model of massive MIMO-ISAC based on the communication transmission process model and radar perception process model;

[0009] Step 4: Model the channel and target parameter estimation in CPD form: construct a precoding training structure to express the received signal as a third-order tensor and remove the pilot signal, then accumulate to obtain the k-th subcarrier signal containing N OFDM training symbols, concatenate the received K subcarriers and express them in the form of a third-order tensor, then decompose them into CPD form, and then obtain the target parameter estimation problem model and the channel parameter estimation problem model;

[0010] Step 5: With the help of the uniqueness condition of CPD, the factor matrix B of the tensor decomposition obtained by analyzing step 4 is (1) , B (2) and B (3) Estimating target parameters;

[0011] Step 6: Expand the matrix along the first dimension of the received observation tensor and perform a transpose operation to obtain the processed signal. Using the Vandermonde structure of the factor matrix, a spatial smoothing technique is used to further expand the dimension of the processed signal to obtain a dimensionally expanded matrix, perform singular value decomposition, and reconstruct the factor matrix.

[0012] Step 7: Construct the relationship between the estimated factor matrix and the true factor matrix, from the factor matrix B (1) Each column of estimates AoA separately, and the problem of estimating AoA is formulated as estimating the factor matrix The qth column and the receiving array steering vector a re (θ q ) is maximized by estimating the factor matrix The generalized eigenvalues ​​on the unit circle yield the time delay;

[0013] Step 8: Construct an alternating optimization method to iteratively refine the azimuth angle of the qth target relative to the transmitter and Doppler shift The estimation problem of the i-th iteration is obtained, and then the estimated values ​​of AoD and Doppler frequency shift are obtained until the iterative update process converges to a stable point and the Doppler frequency shift is obtained. The azimuth of the qth target relative to the transmitter The azimuth of the qth target relative to the receiver and delay estimates;

[0014] Step 9: Expand the initial tensor modulo 3 and transpose it, and then estimate the reflection coefficient according to the LS criterion;

[0015] Step 10: Estimate the multipath channel parameters on the UE side using steps 8-10 Using the estimated five-tuple parameter set, reconstruct the entire channel matrix

[0016] Preferably, the specific method of constructing the communication transmission process model is:

[0017] Step 1.1: For the mth ue The carrier frequency is removed from the received signal of each antenna to obtain the baseband received signal. The baseband received signal is specifically:

[0018]

[0019] In the formula, is the passband path gain, α p is the equivalent baseband path gain, is a continuous-time bandpass signal, v p ∈[-v max / 2,v max / 2) is the Doppler shift, v max is the Doppler spread of the channel, f c is the carrier frequency, From the mth bs Antenna to the mth ue The delay of the pth path of the antenna is, and are the delays introduced by the corresponding transmitting and receiving antennas, respectively;

[0020] Step 1.2: For the mth bs The BS antenna is at the mth ue The channel response of each UE antenna is Fourier transformed to obtain the frequency domain time-varying channel response, which is specifically:

[0021]

[0022] In the formula, α p is the equivalent baseband path gain, vp ∈[-v max / 2,v max / 2) is the Doppler shift, v max is the Doppler spread of the channel, τ p is the free space propagation delay between the first transmit and receive antennas, f c is the carrier frequency, and are the delays introduced by the corresponding transmitting and receiving antennas, respectively;

[0023] Step 1.3: Superimpose the elements of all BS antennas and UE antennas to obtain the communication channel. The communication channel is specifically:

[0024]

[0025] In the formula, α p is the equivalent baseband path gain, v p ∈[-v max / 2,v max / 2) is the Doppler shift, v max is the Doppler spread of the channel, τ p is the free space propagation delay between the first transmit and receive antennas, a bs (θ p ,f) and a ue (θ p ,f) is the broadband array steering vector, expressed as

[0026]

[0027]

[0028] Among them, f c is the carrier frequency, d is the array element spacing, represents the carrier wavelength, θ p is the AoA of the p-th path.

[0029] Preferably, the specific method of constructing the radar perception process model is:

[0030] Step 2.1: Remove the carrier frequency from the received target echo signal and demodulate it to obtain a demodulated target echo signal. The demodulated target echo signal is specifically:

[0031]

[0032] in, is the reflected signal of the qth target, f c is the carrier frequency, R q is the distance of the qth point target, V qis the radial velocity of the qth point target, φ q and θ q They represent the azimuth angle of the qth target relative to the transmitting and receiving arrays, d is the array element spacing, Indicates the mth bs The baseband signal of the transmitting antenna is is the complex reflection coefficient, τ q and v q denote the round-trip delay and Doppler shift respectively;

[0033] Step 2.2: The equivalent radar perception channel model is obtained from the demodulated target echo signal:

[0034]

[0035] In the formula, is the complex reflection coefficient, is the reflected signal of the qth target, f c is the carrier frequency, τ q is the round trip delay, v q is the Doppler shift, φ q and θ q They represent the azimuth angle of the qth target relative to the transmitting and receiving arrays, d is the array element spacing, and c is the speed of light;

[0036] Step 2.3: Perform Fourier transform on the equivalent radar perception channel model formula and superimpose the channels of all transmitting and receiving antennas to obtain the radar perception channel. The radar perception channel is specifically:

[0037]

[0038] In the formula, is the complex reflection coefficient, is the reflected signal of the qth target, f c is the carrier frequency, τ q is the round trip delay, v q is the Doppler shift, φ q and θ q denote the azimuth angle of the qth target relative to the transmitting and receiving arrays, respectively, and a bs (φ q ,f) is the transmitting array steering vector, a re (θ q ,f) is M re ×1-dimensional broadband array steering vector, expressed as

[0039]

[0040] Preferably, in the unified channel model of massive MIMO-ISAC, the perceived channel matrix at the k-th subcarrier of the n-th OFDM symbol is specifically:

[0041]

[0042] Among them, G n,k Equivalent to The phase change caused by Doppler frequency shift in an OFDM symbol is approximately constant, that is, Where T sym =T eff +T cp , T eff and T cp are the duration of effective OFDM symbol and cyclic prefix respectively, K0 is the number of Fourier transform points, is the complex reflection coefficient, is the reflected signal of the qth target, f c is the carrier frequency, τ q is the round trip delay, f s is the sampling frequency, v q is the Doppler shift, a bs (φ q ) and a re (θ q ) represent the transmitting and receiving array steering vectors, φ q and θ q represent the azimuth angle of the qth target relative to the transmitting and receiving arrays respectively.

[0043] Preferably, the specific method of performing CPD modeling on the channel and target parameter estimation in step 4 is:

[0044] Step 4.1: Construct a precoding training structure. The precoding structure is as follows:

[0045] s n,k =p n x k

[0046] in is the trained precoder for the nth OFDM symbol, x k is satisfied|x k | 2 =1 for the pilot symbol at the kth subcarrier.

[0047] Step 4.2: With the help of the precoding training structure proposed in step 4.1, the received signal is expressed as a third-order tensor and the pilot signal is removed to obtain the received signal without the pilot signal. Then, the signals of N OFDM training symbols are accumulated and obtained. The signals of N OFDM training symbols are specifically:

[0048]

[0049] in contains all measurements received at the kth subcarrier, The nth diagonal element is The diagonal matrix of ; represents the training pre-encoder, each element of which is randomly generated by the unit circle; represents the noise matrix; is the complex reflection coefficient, is the reflected signal of the qth target, f c is the carrier frequency, τ q is the round-trip delay, a ue (θ q ) and a bs (φ q ) is the array steering vector, φ q and θ q represent the azimuth angle of the qth target relative to the transmitting and receiving arrays, respectively, and f s is the sampling frequency, K0 is the number of Fourier transform points;

[0050] Step 4.3: Cascade the received K subcarriers and represent the received echo signal as a third-order tensor Its (m,n,k)th item is [Y k ] m,n , tensor It can be decomposed into CPD forms. The specific CPD forms are:

[0051]

[0052] in is the noise tensor, and the (m,n,k)th item is [N k ] m,n ; a re (θ q ) is the array steering vector, b bs (φ q ,ν q )=Γ(ν q ) T a bs (φ q ) is the Doppler frequency shift v q The equivalent steering vector of c(β q ,τ q )=Γ(ν q ) T a bs (φ q ), The nth diagonal element is The diagonal matrix ofbs (φ q ) is the array steering vector, τ q is the round-trip delay; represents the training precoder, each element of which is randomly generated from the unit circle; φ q and θ q denote the azimuth angle of the qth target relative to the transmitting and receiving arrays, respectively, is the complex reflection coefficient, is the reflected signal of the qth target, f c is the carrier frequency, T sym =T eff +T cp , T eff and T cp are the durations of valid OFDM symbols and cyclic prefix respectively;

[0053] Step 4.4: The target parameter estimation problem model is obtained from steps 4.1-4.3. The target parameter estimation problem model is specifically:

[0054]

[0055] Among them, a re (θ q ) is the array steering vector, b bs (φ q ,v q )=Γ(ν q ) T a bs (φ q ) is the Doppler frequency shift v q The equivalent steering vector of c(β q ,τ q )=Γ(v q ) T a bs (φ q ) is related to the time delay τ q The relevant steering vector, φ q and θ q denote the azimuth angle of the qth target relative to the transmitting and receiving arrays, respectively, The nth diagonal element is The diagonal matrix of represents the training precoder, each element of which is randomly generated by the unit circle, a bs (φ q ) is the array steering vector, is the complex reflection coefficient, is the reflected signal of the qth target, f c is the carrier frequency;

[0056] Step 4.5: The channel parameter estimation problem model is as follows:

[0057]

[0058] in, is the UE-side observation tensor used for channel estimation, a ue (θ p ) is the array steering vector, b bs (φ p ,ν p ) is the Doppler frequency shift v p The equivalent steering vector of c(α p ,τ p ) is related to the time delay τ p The relevant steering vector, φ p and θ p They represent the physical AoA and AoD of the p-th path respectively.

[0059] Preferably, the factor matrix is ​​specifically:

[0060]

[0061]

[0062]

[0063] In the formula, a re (θ q ) represents the receiving array steering vector, b bs (φ q ,v q ) is the Doppler frequency shift v q The equivalent launch steering vector, c(β q ,τ q ) is related to the time delay τ q The relevant steering vector, β q is the scattering intensity of the perceived target, φ q and θ q represent the azimuth angle of the qth target relative to the transmitting and receiving arrays respectively.

[0064] Preferably, the matrix is ​​expanded along the first dimension of the received observation tensor and a transposition operation is performed to obtain a processed signal. According to Theorem 1, the Vandermonde structure of the factor matrix is ​​used, and a spatial smoothing technique is used to further expand the dimension of the processed signal to obtain a dimensionally expanded matrix, and singular value decomposition is performed. The specific steps of reconstructing the factor matrix are as follows:

[0065] Step 7.1: Along the received observation tensor The first-dimensional expanded matrix and transpose operation are performed to obtain the processed signal:

[0066]

[0067] Among them, B (1) , B (2) , B (3) is a factor matrix, is the corresponding noise matrix;

[0068] Step 7.2: Using the factor matrix B (3) The Vandermonde structure is enlarged by spatial smoothing technology The dimension of the matrix with enlarged dimension is:

[0069]

[0070] in is a cyclic selection matrix, the integer pair (K3, L3) satisfies K3+L3=K+1, is the observation tensor received along The first-dimensional expansion matrix of is transposed to obtain the processed signal, B is the factor matrix, is the corresponding noise matrix;

[0071] Step 7.3: Compute the singular value decomposition where Y s is the matrix with expanded dimension, and the Q main singular vectors in U and V respectively form the signal subspace U s and V s ,Σ s It is a diagonal matrix consisting of Q largest singular values;

[0072] Step 7.4: Define the submatrices and Calculate EVD as Reconstruction factor matrix and The qth column of is:

[0073]

[0074]

[0075]

[0076] in is the generalized eigenvalue estimate on the unit circle, [Z] q,q is the qth diagonal element of Z, m q is the qth column of M, t q It is M -T The qth column, U s and Vs The Q principal singular vectors in the matrices U and V generated after singular value decomposition respectively form the signal subspace.

[0077] Preferably, the relationship between the estimated factor matrix and the true factor matrix is ​​constructed, from the factor matrix B (1) Each column of estimates AoA separately, and the problem of estimating AoA is formulated as estimating the factor matrix The qth column and the receiving array steering vector a re (θ q ) is maximized through the factor matrix The specific method of estimating the time delay using the generalized eigenvalue on the unit circle is:

[0078] Step 8.1: Analyze the relationship between the estimated factor matrix and the true factor matrix, specifically:

[0079]

[0080] in is the estimated factor matrix, B (i) is the real factor matrix, Δ i is to satisfy Δ1Δ2Δ3=I Q The unknown diagonal matrix of is an unknown permutation matrix common to all factor matrices, E i represents the corresponding estimation error;

[0081] Step 8.2: From the factor matrix B (1) Each column of estimates AoA separately, and the azimuth angle θ of the qth target relative to the receiving array is estimated q The problem is stated as estimating the factor matrix The qth column and the receiving array steering vector a re (θ q ), specifically:

[0082]

[0083] in is the estimated factor matrix qth column of ;

[0084] Step 8.3: Through the factor matrix Generalized Eigenvalue Estimates on the Unit Circle Get delay Specifically:

[0085]

[0086] Where ∠ represents the phase angle extraction operator, f s is the sampling frequency, K0 is the number of Fourier transform points.

[0087] Preferably, an alternating optimization method is constructed to iteratively refine the azimuth angle of the qth target relative to the transmitter and Doppler shift The estimation problem of the i-th iteration is obtained, and then the estimated values ​​of AoD and Doppler frequency shift are obtained until the iterative update process converges to a stable point and the Doppler frequency shift is obtained. The azimuth of the qth target relative to the transmitter The azimuth of the qth target relative to the receiver and delay The specific method of estimating is:

[0088] Step 9.1: Construct an alternating optimization method to iteratively refine the azimuth of the qth target relative to the transmitter and Doppler shift The Doppler shift in the (i-1)th iteration is known. The estimation of AoD in the i-th iteration is obtained. The Doppler shift estimation problem of the i-th iteration is obtained. The estimation problem of AoD in the i-th iteration is specifically:

[0089]

[0090] In the formula, is the estimated factor matrix The qth column of q is the qth diagonal element of Δ2, where Δ2 satisfies Δ1Δ2Δ3=I Q The unknown diagonal matrix of is a diagonal matrix, is the estimate of Doppler frequency shift, P is the training precoder, a bs (φ q ) is the array steering vector, φ q is the azimuth of the qth target relative to the transmitter,

[0091] The Doppler shift estimation problem for the i-th iteration is specifically:

[0092]

[0093] In the formula, is the estimated factor matrix The qth column of q is the qth diagonal element of Δ2, and Δ2 satisfies

[0094] Δ1Δ2Δ3=I Q The unknown diagonal matrix, Γ(ν q ) is a diagonal matrix, ν q is the Doppler shift, P is the training precoder, is the array steering vector, is the estimated azimuth of the qth target relative to the transmitter, Q is the diagonal element The diagonal matrix of do (ν q ) is Γ(ν q ) is a vector consisting of the diagonal elements of ;

[0095] Step 9.2: Using the array to guide vector a bs (φ) and Γ(ν q ) is a vector consisting of the diagonal elements of do (v) The Vandermonde structure, for the AoD estimation problem of the i-th iteration, is The guide vector a bs (φ) is expressed as This results in the following expression:

[0096]

[0097] in a(z φ ) is the array steering vector, is the diagonal matrix Γ, Doppler frequency shift estimate Estimated factor matrix The qth column The quantity concerned, P is the training precoder;

[0098] Step 9.3: Consider the case of no noise, then the numerator of the above equation is zero, and 2(M bs -1) roots, let is the root of unit amplitude in the unit circle, then we can find Get the solution Get the estimated value of AoD, specifically:

[0099]

[0100] Where ∠ represents the phase angle extraction operator, λ c is the wavelength, d is the array element spacing;

[0101] Step 9.4: Similar to steps 9.2-9.3, obtain the Doppler shift estimate for the i-th iteration, specifically:

[0102]

[0103] Where T symis the duration of an OFDM symbol;

[0104] Step 9.5: Based on the above steps, in alternating cases, the azimuth of the qth target relative to the transmitter and Doppler shift The estimation is performed until the iterative update process converges to a stable point and the Doppler frequency shift is obtained. The azimuth of the qth target relative to the transmitter The azimuth of the qth target relative to the receiver and delay Estimates.

[0105] Preferably, the tensor After expansion and transposition, it is specifically:

[0106]

[0107] where β=[β1,...,β Q ] T , A td is related to the delay τ q The array steering vector matrix is ​​concerned, B is the factor matrix, and N is the noise matrix;

[0108] The reflection coefficient is estimated as follows:

[0109]

[0110] in and Based on delay Doppler shift and the azimuth of the qth target relative to the transmitter and the azimuth of the qth target relative to the receiver structure.

[0111] Compared with the prior art, the present invention has the following significant advantages:

[0112] The present invention solves the target parameter estimation and multipath time-varying channel estimation problems under the unified architecture of large-scale MIMO-ISAC, and has great advantages in reducing training overhead and improving perception resolution.

[0113] Other features and advantages of the present invention will be described in the following description, and part of them will become obvious from the description, or will be understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description, claims, and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0114] The drawings are only for the purpose of illustrating particular embodiments and are not to be considered limiting of the present invention. Like reference symbols denote like components throughout the drawings.

[0115] Figure 1 Select channels for hybrid MIMO architectures and time-varying frequency.

[0116] Figure 2 The target parameters for the proposed method are estimated performance.

[0117] Figure 3 To correctly distinguish the success rate of the target.

[0118] Figure 4 The RMSE performance of distance and speed estimation varies with SNR.

[0119] Figure 5 This is the curve of the NMSE performance of channel estimation changing with SNR. DETAILED DESCRIPTION

[0120] It is easy to understand that according to the technical solution of the present invention, without changing the essential spirit of the present invention, a person skilled in the art can imagine various embodiments of the present invention. Therefore, the following specific embodiments and drawings are only exemplary descriptions of the technical solution of the present invention, and should not be regarded as the whole of the present invention or as a limitation or limitation of the technical solution of the present invention. On the contrary, the purpose of providing these embodiments is to enable those skilled in the art to understand the present invention more thoroughly. The preferred embodiments of the present invention are described in detail below in conjunction with the accompanying drawings, wherein the accompanying drawings constitute a part of this application and are used together with the embodiments of the present invention to illustrate the innovative concept of the present invention.

[0121] The present invention is conceived as a method for acquiring a communication-aware integrated large-scale MIMO channel state and target parameters, comprising:

[0122] 1. Communication transmission process modeling

[0123] Assume a continuous-time bandpass signal from the mth bs The transmitting antenna of the base station reaches the user UE located in the far field through P multipaths, then the mth ue The received signal of the antenna is:

[0124]

[0125] in is the passband path gain, v p ∈[-v max / 2,v max / 2) is the Doppler shift, v max is the (bilateral) Doppler spread of the channel, From the mthbs Antenna to the mth ue The delay of the pth path of the antenna is expressed as

[0126]

[0127] where τ p is the free space propagation delay between the first transmit antenna and the receive antenna, θ p and φ p denote the physical AoA and AoD of the pth path, d and c denote the distance between adjacent antennas and the speed of light, respectively. The baseband received signal is:

[0128]

[0129] in is the equivalent baseband path gain, and are the delays caused by the corresponding transmitting and receiving antennas. bs The BS antenna is at the mth ue The channel response of each UE antenna is

[0130]

[0131] Perform Fourier transform on the above equation to obtain the frequency domain time-varying channel response:

[0132]

[0133] The last equation holds true because v p <<f c By superimposing the elements of all BS antennas and UE antennas, the communication channel can be expressed as

[0134]

[0135] where a bs (θ p ,f) and a ue (θ p ,f) is the broadband array steering vector, expressed as

[0136]

[0137]

[0138] in Indicates the carrier wavelength.

[0139] 2. Radar perception process modeling

[0140] Assume that the qth point target is located at a distance Rq and the radial velocity is V q , at the base station end m re The echo of Q receiving antennas and Q targets is expressed as

[0141]

[0142] in It is the mth bs The transmitted signal of each BS antenna is is the qth reflected signal of the target (proportional to the radar cross section), φ q and θ q Represent the azimuth of the transmitting and receiving arrays relative to the qth target. At the receiver, the demodulated signal (with the carrier frequency removed) ) is expressed as

[0143]

[0144] in is the complex reflection coefficient, and denote the round-trip delay and Doppler shift respectively. The second equation is established by using s(t-τ q )≈s(t-τ q +2V q t / c), because the target speed is assumed to be V q <<c, a constant delay is allowed within the coherent processing interval (CPI) time. Then the equivalent radar sensing channel model is

[0145]

[0146] in It can be regarded as a linear time-varying filter. Perform Fourier transform on the above equation and superimpose the channels of all transmitting and receiving antennas. The radar sensing channel can be expressed as

[0147]

[0148] where a re (θ q ,f,v q ) is M re ×1-dimensional broadband array steering vector, expressed as

[0149]

[0150] 3. Unified Channel Model for Massive MIMO-ISAC

[0151] From the above analysis, we can see that the communication channel and the sensing channel have similar geometric structures, so the channel and target parameters can be estimated in the same framework. The similarity can be understood as:

[0152] On the one hand, in addition to estimating the entire channel matrix, the communication perspective can also choose to estimate the limited physical parameters inside the channel, that is, These limited parameters include angle, delay and Doppler shift, and target parameters This provides the premise for using the same signal for channel and target parameter estimation, which is very attractive for reducing training overhead and improving spectrum efficiency.

[0153] On the other hand, the mathematical expressions of the scattering characteristics of the communication and sensing channels are the same, but the parameters inside H and G have different physical meanings. For example, α p Characterizes the scattering characteristics in the wireless propagation environment, and β p Represents the scattering intensity of the perceived target. τ p and v p represents the one-way Doppler frequency shift and delay associated with the communication propagation path, τ q and v q represents the round-trip Doppler shift and delay caused by the target.

[0154] For the special case where the bandwidth effect can be ignored, that is, the system bandwidth is not large enough, the steering vector degenerates to a case that is independent of frequency, that is, a re (θ q )=a re (θ q ,f)∣ f＝0 ,a bs (φ q )=a bs (φ q ,f)∣ f＝0 The perceived channel matrix at the kth subcarrier of the nth OFDM symbol is

[0155]

[0156] Among them G n,k Equivalent to The phase change caused by Doppler frequency shift in an OFDM symbol is approximately constant, that is, Where T sym =T eff +T cp , T eff and T cp are the durations of a valid OFDM symbol and a cyclic prefix, respectively.

[0157] 4. CPD modeling of received signal

[0158] Assumptions is the transmission training sequence at the kth subcarrier of the nth OFDM symbol, where k = 1, 2, ..., K, n = 1, 2, ..., N. Using M bs The parallel K0 point inverse Fourier transform transforms the symbol block to the time domain. The cyclic prefix (CP) is used to suppress the inter-symbol interference (ISI) caused by the maximum multipath delay and the delay corresponding to the maximum detection range. After removing the CP and performing a discrete Fourier transform, the received signal at the kth subcarrier of the nth OFDM symbol is obtained as

[0159]

[0160] Among them G n,k represents the sensing channel matrix, is a circularly symmetric complex Gaussian distributed additive noise vector.

[0161] However, the observation data with multiple OFDM symbols and subcarriers cannot be decoupled into independent dimensions, so it does not conform to the CPD model. In order to facilitate the estimation of channel and target parameters, a precoding training structure is proposed, namely s n,k =p n x k ,in is the trained precoder for the nth OFDM symbol, x k is satisfied|x k | 2 =1. n As different OFDM symbols change, x k The proposed training structure allows the received signal to be expressed as a third-order tensor. At the receiver, the signal after removing the pilot symbol is expressed as

[0162]

[0163] in is the corresponding noise vector. Collect the received signals of N training OFDM symbols to obtain

[0164]

[0165] in contains all measurements received at the kth subcarrier, The nth diagonal element is The diagonal matrix of ; represents the training pre-encoder, each element of which is randomly generated by the unit circle; represents the noise matrix.

[0166] By concatenating the received K subcarriers, the echo signal is represented as a third-order tensor Its (m,n,k)th item is [Y k ] m,n . Tensor The three patterns of represent receive antennas, training OFDM symbols, and training subcarriers, which are associated with the received signal along the space / time / frequency domain. Accepts the following CPD format, which decomposes a tensor into a sum of rank-one component tensors, i.e.

[0167]

[0168] in is the noise tensor, where the (m,n,k)th item is [N k ] m,n ; b bs (φ q ,v q )=Γ(v q ) T a bs (φ q ) is the Doppler frequency shift v q The equivalent steering vector of c(β q ,τ q )=Γ(v q ) T a bs (φ q ),in

[0169]

[0170] where a td (τ q ) is considered as the time delay τ q The guiding vector.

[0171] The goal of the proposed method is to transform the observation tensor Estimating unknown parameters in Stated as the following question:

[0172]

[0173] Since the unknown parameters {θ q ,φ q ,τ q ,β q ,ν q} are nonlinearly coupled together, so it is very challenging to directly jointly optimize them. However, the uniqueness condition of CPD is an effective parameter estimation method, namely

[0174]

[0175]

[0176]

[0177] The received UE-side pilot signal can be expressed as a third-order tensor according to the above modeling method, so the channel estimation problem is expressed as

[0178]

[0179] in is the UE-side observation tensor used for channel estimation.

[0180] 5. Estimation of factor matrix

[0181] The uniqueness condition of CP decomposition is a necessary condition for blind signal separation, which can ensure that the obtained factor matrix contains accurate information of unknown parameters. In the target parameter estimation or channel estimation problem considered, the factor matrix B (3) because The resulting translation invariant vector a td (τ q ) has the Vandermonde property. Using this structure, we can obtain the uniqueness condition of CPD relaxation.

[0182] Theorem 1. Let is a matrix containing factors A tensor, where B (3) By different generators shows the Vandermonde structure. If

[0183]

[0184] Among them, K3+L3=I3+1, then The rank of is Q and the CPD is unique. In general, the above formula is

[0185]

[0186] Theorem 1 shows that the CPD problem with Vandermonde structure constraints has strong identifiability and can bring the following advantages to the parameter estimation of large-scale MIMO-ISAC systems. On the one hand, Theorem 1 does not require k(B (2) )=1 and / or k(B (3))=1, and the uniqueness condition applies when the factor matrix has independent columns. This makes the parameter estimation reliable even when the directions of different targets are close, thus providing a higher spatial resolution of target perception. On the other hand, given the dimension of the factor matrix, Theorem 1 is able to handle CP decomposition problems with higher tensor ranks than the traditional unstructured constraint CPD model. This means that the maximum number of identifiable targets can be improved when considering the Vandermonde constraint.

[0187] Using B (3) Based on the Vandermonde property of , a spatial smoothing technique is used to propose an improved channel parameter estimation scheme according to Theorem 1. Take the received observation tensor By expanding the matrix along the first dimension and performing a transpose operation, we can obtain

[0188]

[0189] in is the corresponding noise matrix. Since B (3) The Vandermonde structure is further enlarged using spatial smoothing techniques. Select the integer pair (K3, L3) that meets the condition K3+L3=K+1, and construct the cyclic selection matrix Sequentially Perform spatial smoothing to obtain

[0190]

[0191] in is the corresponding noise matrix. Next, we use a method similar to ESPRIT to estimate the factor matrix. Calculate the singular value decomposition The Q principal singular vectors in U and V respectively form the signal subspace U s and V s ,Σ s is a diagonal matrix consisting of the Q largest singular values. Define the submatrix and Calculate EVD as Therefore, the factor matrix The qth column of can be reconstructed as

[0192]

[0193] in is the generalized eigenvalue estimate on the unit circle, [Z] q,q is the qth diagonal element of Z.

[0194] and The qth column of can be constructed as

[0195]

[0196]

[0197] Where m q is the qth column of M, t q It is M -T The qth column of . Get the estimated factor matrix Then, the unknown parameter θ q ,φ q , τ q , β q and v q Estimates.

[0198] 6. Estimation of unknown parameters

[0199] The relationship between the estimated factor matrix and the true factor matrix is

[0200]

[0201] where Δ i is to satisfy Δ1Δ2Δ3=I Q The unknown diagonal matrix of . The matrix is an unknown permutation matrix common to all factor matrices, E i represents the corresponding estimation error.

[0202] Because B (1) Each column of q As a feature, we can (1) Each column of θ estimates AoA separately. That is, find θ q The problem can be stated as and a re (θ q ) is maximized, that is,

[0203]

[0204] in is the estimated factor matrix The solution of the above equation can be obtained by q Perform a one-dimensional search to obtain the delay It can be directly obtained through the factor matrix Generalized Eigenvalue Estimates on the Unit Circle Get, that is

[0205]

[0206] Where ∠ represents the phase angle extraction operator.

[0207] Next, we estimate the AoD and Doppler shift. Specifically, we estimate the factor matrix The qth column Expressed as

[0208]

[0209] definition δ q is the qth diagonal element of Δ2, e q is the qth column of E2. The goal of the proposed method is to Joint estimation of AoDφ q and Doppler shift v q , then the problem is stated as

[0210]

[0211] From the above formula, we can see that the three unknowns δ q , v q and φ q Coupled together, it is extremely difficult to estimate the three parameters simultaneously. In addition, due to the normalized Doppler shift v q The phase offset introduced by the correlation method will lead to the degradation of the angle and Doppler estimation performance if the phase offset correlation method is ignored. To solve the above problems, the present invention proposes an alternating optimization method, which can iteratively refine the and Doppler shift and can significantly improve the estimation performance.

[0212] The Doppler shift in the (i-1)th iteration is known. The estimation problem of AoD in the i-th iteration can be expressed as

[0213]

[0214] The second equation comes from At the same time, it is known that the estimated The Doppler shift is estimated as

[0215]

[0216] The second equality holds because we define a diagonal matrix Q whose diagonal elements are And extract Γ(v q ) has diagonal elements of

[0217]

[0218] From the above analysis, we can see that the joint optimization problem is transformed into two independent estimation problems for each variable. bs (φ) and ado The Vandermonde structure of (v) can be derived as a closed-form expression using the polynomial method. For the AoD estimation problem of the i-th iteration, The guide vector a bs (φ) is expressed as but

[0219]

[0220] in The molecule of the above formula can be further expressed as φ The polynomial of is zero in the noise-free case, that is,

[0221]

[0222] There are 2 (M bs -1) roots, half of which are inside the unit circle. is the root of unit amplitude in the unit circle, then we can find Get the solution Therefore, the AoD estimate is

[0223]

[0224] where ∠ denotes the phase angle extraction operator. Due to the similar mathematical form, the solution to the AoD estimation problem in the ith iteration is the same as the solution to the Doppler estimation problem in the ith iteration. Define the matrix The generator is The guiding vector You can do this by P, W P,q and a(z φ ) are replaced by Q, W Q,q and a(z v ) to get an estimate of the Doppler shift generator. Let To estimate the generalized eigenvalue on the unit circle, the Doppler shift estimate for the i-th iteration can be obtained as

[0225]

[0226] Based on the above steps, in the alternating case, and Estimation is performed until the iterative update process converges to a stable point. and After that, the reflection coefficient is estimated. From the initial observation tensor model, take By expanding modulo 3 and performing a transposition operation, we can obtain

[0227]

[0228] where β=[β1,...,β Q ] T , a td (τ q ) is A td According to the LS criterion, the estimation of the reflection coefficient can be expressed as

[0229]

[0230] in and Based on and as well as constructed.

[0231] Thus, the unknown parameters of the target can be obtained at the BS end, and the UE can estimate the multipath channel parameters in the same steps. Using the estimated five-tuple parameter set, the entire channel matrix can be reconstructed

[0232] Example

[0233] The present invention is mainly verified by computer simulation method, and all steps and conclusions are verified to be correct on MATLAB-R2019a.

[0234] 1) Initialization of system related parameters

[0235] In the simulation, the carrier frequency f c =28GHz, the transmitter has N bs = 64 antennas, the number of antennas at the UE is N re =8, N ue =8, the number of subcarriers K0 = 128, the sampling rate is 100MHz, the elements of P are randomly generated from the unit circle, and the CP duration of an OFDM symbol is Assume that the BS detects Q = 4 targets and the channel experiences L = 4 multipaths. The reflection coefficient and path gain are given by The AoA and AoD of the target and channel are randomly generated from [-π / 3,π / 3], and the speed v of each target is q Uniformly distributed in [-v max ,v max ], maximum speed v max =30m / s, the maximum Doppler shift of the channel The root mean square error (RMSE) is used to evaluate the performance of the target parameters, that is,

[0236]

[0237] in and χ represent the estimated parameters and true parameters respectively, q ∈{θ q ,φ q ,τ q ,ν q}. The channel estimation performance at the nth training OFDM symbol is measured by the normalized MSE (NMSE), expressed as

[0238]

[0239] The channel estimation performance is evaluated by NMSE = NMSE(N), and the signal-to-noise ratio is defined as 2) Method performance simulation analysis

[0240] The target parameter estimation performance of the method proposed in the present invention is as follows: Figure 2 As shown in Figure 1, the proposed method can accurately estimate all angles and related distance-velocity parameters using only a small amount of time-frequency training resources. The existing method can only accurately recover some target parameters. Figure 2 As can be seen in (a), two adjacent targets located at azimuth angles of 31.87° and 35.01° cannot be identified by the MUSIC method, which is due to the limited aperture of the receiving antenna array. Figure 2 From (e), we can see that although the distance parameter can be effectively estimated, there is a large error in the estimation of the target speed. This is because in massive MIMO-ISAC, the number of OFDM pilot symbols used for target perception is less than the number of pilot symbols of the transmitting antenna, resulting in a short observation time and low Doppler resolution.

[0241] Figure 3 The success rate of correctly identifying the target based on the ALS method and the MUSIC algorithm is described. If the AoA estimation error between the estimated angle and the true parameter satisfies The target is considered to be successfully identified. The simulation parameters are set as SNR = 15dB, N = 16, K = 16. Figure 3 It can be seen that the success rate of the method proposed in the present invention is significantly higher than that of the comparison method. When the number of targets to be detected exceeds 11, the success rate of the MUSIC method is zero. The conventional ALS method cannot detect correctly when the number of targets is 17, while the method proposed in the present invention can provide an accuracy rate of more than 79.1%.

[0242] Figure 4(a) shows how the RMSE performance of distance estimation varies with SNR under the same simulation settings. Compared with the traditional MF-based method, the RMSE performance of the proposed method and the ALS-based method improves with the increase of SNR. This is due to the fact that the tensor-based method has a more accurate estimate of the delay, and the delays caused by multiple targets can be estimated independently from the factor matrix. Figure 4 (b) shows the RMSE performance of speed estimation as a function of SNR. The conventional ALS method has the worst RMSE performance for speed estimation because it cannot obtain the value from the factor matrix Recovery q and φ q Doppler shift in case of mutual coupling. The MF-based method can achieve improved velocity estimation performance relative to the conventional ALS method as the SNR gradually increases. By considering Doppler compensation, the algorithm can achieve accurate velocity estimation performance as the SNR gradually increases.

[0243] Figure 5 The NMSE performance of the proposed algorithm's channel estimation varies with SNR, where the number of OFDM training symbols is N=10 and N=16, respectively. It can be seen that the channel estimation performance of the proposed method improves with the increase of SNR. In contrast, the NMSE of the ALS method gradually saturates with the increase of SNR because it does not consider the Doppler shift. As N increases, the channel estimation performance based on the ALS method is poor, which is due to the more serious cumulative phase distortion caused by the Doppler shift. Compared with the ALS-based method, the NMSE performance of the proposed algorithm gradually improves with the increase of the number of training frames, which shows that the proposed algorithm is more robust to the phase distortion caused by the Doppler effect.

[0244] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by any technician familiar with the technical field within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention.

[0245] It should be understood that in order to simplify the present invention and help those skilled in the art understand the various aspects of the present invention, in the above description of the exemplary embodiments of the present invention, the various features of the present invention are sometimes described in a single embodiment or described with reference to a single figure. However, the present invention should not be interpreted as the features included in the exemplary embodiments are all necessary technical features of the patent claims.

[0246] It should be understood that the modules, units, components, etc. included in the device of an embodiment of the present invention can be adaptively changed to be set in a device different from the embodiment. The different modules, units, or components included in the device of the embodiment can be combined into one module, unit, or component, or they can be divided into multiple sub-modules, sub-units, or sub-components.

Claims

1. A method for acquiring channel status and target parameters of a large-scale MIMO integrated with communication perception, characterized in that: The specific steps are: Step 1: Construct a communication transmission process model; Step 2: Construct radar perception process model; Step 3: Construct a unified channel model of massive MIMO-ISAC based on the communication transmission process model and radar perception process model; Step 4: Model the channel and target parameter estimation in CPD form: construct a precoding training structure to express the received signal as a third-order tensor and remove the pilot signal, then accumulate to obtain the k-th subcarrier signal containing N OFDM training symbols, concatenate the received K subcarriers and express them in the form of a third-order tensor, then decompose them into CPD form, and then obtain the target parameter estimation problem model and the channel parameter estimation problem model; Step 5: With the help of the uniqueness condition of CPD, the factor matrix B of the tensor decomposition obtained by analyzing step 4 is (1) , B (2) and B (3) Estimating target parameters; Step 6: Expand the matrix along the first dimension of the received observation tensor and perform a transpose operation to obtain the processed signal. Using the Vandermonde structure of the factor matrix, a spatial smoothing technique is used to further expand the dimension of the processed signal to obtain a dimensionally expanded matrix, perform singular value decomposition, and reconstruct the factor matrix. Step 7: Construct the relationship between the estimated factor matrix and the true factor matrix, from the factor matrix B (1) Each column of estimates AoA separately, and the problem of estimating AoA is formulated as estimating the factor matrix The qth column and the receiving array steering vector a re (θ q ) is maximized by estimating the factor matrix The generalized eigenvalues ​​on the unit circle yield the time delay; Step 8: Construct an alternating optimization method to iteratively refine the azimuth angle of the qth target relative to the transmitter and Doppler shift The estimation problem of the i-th iteration is obtained, and then the estimated values ​​of AoD and Doppler frequency shift are obtained until the iterative update process converges to a stable point and the Doppler frequency shift is obtained. The azimuth of the qth target relative to the transmitter The azimuth of the qth target relative to the receiver and delay estimates; Step 9: Expand the initial tensor modulo 3 and transpose it, and then estimate the reflection coefficient according to the LS criterion; Step 10: Estimate the multipath channel parameters on the UE side using steps 8-9 Using the estimated five-tuple parameter set, reconstruct the entire channel matrix 2. The method for acquiring the channel state and target parameters of the communication-aware integrated large-scale MIMO according to claim 1, characterized in that: The specific method of constructing the communication transmission process model is: Step 1.1: For the mth ue The carrier frequency is removed from the received signal of each antenna to obtain the baseband received signal. The baseband received signal is specifically: In the formula, is the passband path gain, α p is the equivalent baseband path gain, is a continuous-time bandpass signal, v p ∈[-v max / 2,v max / 2) is the Doppler shift, v max is the Doppler spread of the channel, f c is the carrier frequency, From the mth bs Antenna to the mth ue The delay of the pth path of the antenna is, and are the delays introduced by the corresponding transmitting and receiving antennas, respectively; Step 1.2: For the mth bs The BS antenna is at the mth ue The channel response of each UE antenna is Fourier transformed to obtain the frequency domain time-varying channel response, which is specifically: In the formula, α p is the equivalent baseband path gain, v p ∈[-v max / 2,v max / 2) is the Doppler shift, v max is the Doppler spread of the channel, τ p is the free space propagation delay between the first transmit and receive antennas, f c is the carrier frequency, and are the delays introduced by the corresponding transmitting and receiving antennas, respectively; Step 1.3: Superimpose the elements of all BS antennas and UE antennas to obtain the communication channel. The communication channel is specifically: In the formula, α p is the equivalent baseband path gain, v p ∈[-v max / 2,v max / 2) is the Doppler shift, v max is the Doppler spread of the channel, τ p is the free space propagation delay between the first transmit and receive antennas, a bs (θ p ,f) and a ue (θ p ,f) is the broadband array steering vector, expressed as Among them, f c is the carrier frequency, d is the array element spacing, represents the carrier wavelength, θ p is the AoA of the p-th path.

3. The method for acquiring the channel state and target parameters of a large-scale MIMO with integrated communication and perception according to claim 1, characterized in that: The specific method of constructing the radar perception process model is: Step 2.1: Remove the carrier frequency from the received target echo signal and demodulate it to obtain a demodulated target echo signal. The demodulated target echo signal is specifically: in, is the reflected signal of the qth target, f c is the carrier frequency, R q is the distance of the qth point target, V q is the radial velocity of the qth point target, φ q and θ q They represent the azimuth angle of the qth target relative to the transmitting and receiving arrays, d is the array element spacing, Indicates the mth bs The baseband signal of the transmitting antenna is is the complex reflection coefficient, τ q and v q denote the round-trip delay and Doppler shift respectively, and c is the speed of light; Step 2.2: The equivalent radar perception channel model is obtained from the demodulated target echo signal: In the formula, is the complex reflection coefficient, is the reflected signal of the qth target, f c is the carrier frequency, τ q is the round trip delay, v q is the Doppler shift, φ q and θ q denote the azimuth angle of the qth target relative to the transmitting and receiving arrays, respectively, and d is the array element spacing; Step 2.3: Perform Fourier transform on the equivalent radar perception channel model formula and superimpose the channels of all transmitting and receiving antennas to obtain the radar perception channel. The radar perception channel is specifically: In the formula, is the complex reflection coefficient, is the reflected signal of the qth target, f c is the carrier frequency, τ q is the round trip delay, v q is the Doppler shift, φ q and θ q denote the azimuth angle of the qth target relative to the transmitting and receiving arrays, respectively, and a bs (φ q ,f) is the transmitting array steering vector, a re (θ q ,f) is M re ×1-dimensional broadband array steering vector, expressed as 4. The method for acquiring the channel state and target parameters of the communication-aware integrated large-scale MIMO according to claim 1, characterized in that: The unified channel model of massive MIMO-ISAC, the perceived channel matrix at the k-th subcarrier of the n-th OFDM symbol is specifically: Among them, G n,k Equivalent to The phase change caused by Doppler frequency shift in an OFDM symbol is approximately constant, that is, t∈[nT sym ,(n+1)T sym ], where T sym =T eff +T cp , T eff and T cp are the duration of effective OFDM symbol and cyclic prefix respectively, K0 is the number of Fourier transform points, is the complex reflection coefficient, is the reflected signal of the qth target, f c is the carrier frequency, τ q is the round trip delay, f s is the sampling frequency, v q is the Doppler shift, a bs (φ q ) and a re (θ q ) represent the transmitting and receiving array steering vectors, φ q and θ q represent the azimuth angle of the qth target relative to the transmitting and receiving arrays respectively.

5. The method for acquiring channel status and target parameters of large-scale MIMO with integrated communication and perception according to claim 1, characterized in that: The specific method of performing CPD modeling on the channel and target parameter estimation in step 4 is: Step 4.1: Construct a precoding training structure. The precoding structure is as follows: s n,k =p n x k in is the trained precoder for the nth OFDM symbol, x k is satisfied|x k | 2 =1 for the pilot symbol at the kth subcarrier; Step 4.2: With the help of the precoding training structure proposed in step 4.1, the received signal is expressed as a third-order tensor and the pilot signal is removed to obtain the received signal without the pilot signal. Then, the signals of N OFDM training symbols are accumulated and obtained. The signals of N OFDM training symbols are specifically: in contains all measurements received at the kth subcarrier, The nth diagonal element is The diagonal matrix of ; represents the training pre-encoder, each element of which is randomly generated by the unit circle; represents the noise matrix; is the complex reflection coefficient, is the reflected signal of the qth target, f c is the carrier frequency, τ q is the round-trip delay, a ue (θ q ) and a bs (φ q ) is the array steering vector, φ q and θ q represent the azimuth angle of the qth target relative to the transmitting and receiving arrays, respectively, and f s is the sampling frequency, K0 is the number of Fourier transform points; Step 4.3: Cascade the received K subcarriers and represent the received echo signal as a third-order tensor Its (m,n,k)th item is [Y k ] m,n , tensor It can be decomposed into CPD forms. The specific CPD forms are: in is the noise tensor, and the (m,n,k)th item is [N k ] m,n ; a re (θ q ) is the array steering vector, b bs (φ q ,v q )=Γ(v q ) T a bs (φ q ) is the Doppler frequency shift v q The equivalent steering vector of c(β q ,τ q )=Γ(ν q ) T a bs (φ q ), The nth diagonal element is The diagonal matrix of bs (φ q ) is the array steering vector, τ q is the round-trip delay; represents the training precoder, each element of which is randomly generated from the unit circle; φ q and θ q denote the azimuth angle of the qth target relative to the transmitting and receiving arrays, respectively, is the complex reflection coefficient, is the reflected signal of the qth target, f c is the carrier frequency, T sym =T eff +T cp , T eff and T cp are the durations of valid OFDM symbols and cyclic prefix respectively; Step 4.4: The target parameter estimation problem model is obtained from steps 4.1-4.

3. The target parameter estimation problem model is specifically: Among them, a re (θ q ) is the array steering vector, b bs (φ q ,v q )=Γ(v q ) T a bs (φ q ) is the Doppler frequency shift v q The equivalent steering vector of c(β q ,τ q )=Γ(v q ) T a bs (φ q ) is related to the time delay τ q The relevant steering vector, φ q and θ q denote the azimuth angle of the qth target relative to the transmitting and receiving arrays, respectively, The nth diagonal element is The diagonal matrix of represents the training precoder, each element of which is randomly generated by the unit circle, a bs (φ q ) is the array steering vector, is the complex reflection coefficient, is the reflected signal of the qth target, f c is the carrier frequency; Step 4.5: The channel parameter estimation problem model is as follows: in, is the UE-side observation tensor used for channel estimation, a ue (θ p ) is the array steering vector, b bs (φ p ,ν p ) is the Doppler frequency shift v p The equivalent steering vector of c(α p ,τ p ) is related to the time delay τ p The relevant steering vector, φ p and θ p They represent the physical AoA and AoD of the p-th path respectively.

6. The method for acquiring channel status and target parameters of large-scale MIMO with integrated communication and perception according to claim 1, characterized in that: The factor matrix is ​​specifically: In the formula, a re (θ q ) represents the receiving array steering vector, b bs (φ q ,v q ) is the Doppler frequency shift v q The equivalent launch steering vector, c(β q ,τ q ) is related to the time delay τ q The relevant steering vector, β q is the scattering intensity of the perceived target, φ q and θ q represent the azimuth angle of the qth target relative to the transmitting and receiving arrays respectively.

7. The method for acquiring channel status and target parameters of large-scale MIMO with integrated communication and perception according to claim 6, characterized in that: The matrix is ​​expanded along the first dimension of the received observation tensor and the transposition operation is performed to obtain the processed signal. The Vandermonde structure of the factor matrix is ​​used to further expand the dimension of the processed signal by using spatial smoothing technology to obtain a dimensionally expanded matrix, perform singular value decomposition, and reconstruct the factor matrix. The specific steps are: Step 7.1: Along the received observation tensor The first-dimensional expanded matrix and transpose operation are performed to obtain the processed signal: Among them, B (1) , B (2) , B (3) is a factor matrix, is the corresponding noise matrix; Step 7.2: Using the factor matrix B (3) The Vandermonde structure is enlarged by spatial smoothing technology The dimension of the matrix with enlarged dimension is: in l3∈{1,...,L3} is a cyclic selection matrix, and the integer pair (K3,L3) satisfies K3+L3=K+1. is the observation tensor received along The first-dimensional expansion matrix of is transposed to obtain the processed signal, B is the factor matrix, is the corresponding noise matrix; Step 7.3: Compute the singular value decomposition where Y s is the matrix with expanded dimension, and the Q main singular vectors in U and V respectively form the signal subspace U s and V s ,Σ s It is a diagonal matrix consisting of Q largest singular values; Step 7.4: Define the submatrices and Calculate EVD as Reconstruction factor matrix and The qth column of is: in is the generalized eigenvalue estimate on the unit circle, [Z] q,q is the qth diagonal element of Z, m q is the qth column of M, t q It is M -T The qth column, U s and V s The Q principal singular vectors in the matrices U and V generated after singular value decomposition respectively form the signal subspace.

8. The method for acquiring channel status and target parameters of large-scale MIMO with integrated communication and perception according to claim 1, characterized in that: Construct the relationship between the estimated factor matrix and the true factor matrix, from the factor matrix B (1) Each column of estimates AoA separately, and the problem of estimating AoA is formulated as estimating the factor matrix The qth column of and the receiving array steering vector a re (θ q ) is maximized through the factor matrix The specific method of estimating the time delay using the generalized eigenvalue on the unit circle is: Step 8.1: Analyze the relationship between the estimated factor matrix and the true factor matrix, specifically: in is the estimated factor matrix, B (i) is the real factor matrix, Δ i is to satisfy Δ1Δ2Δ3=I Q The unknown diagonal matrix of is an unknown permutation matrix common to all factor matrices, E i represents the corresponding estimation error; Step 8.2: From the factor matrix B (1) Each column of estimates AoA separately, and the azimuth angle θ of the qth target relative to the receiving array is estimated q The problem is stated as estimating the factor matrix The qth column of and the receiving array steering vector a re (θ q ), specifically: in is the estimated factor matrix qth column of ; Step 8.3: Through the factor matrix Generalized Eigenvalue Estimates on the Unit Circle Get delay Specifically: Where ∠ represents the phase angle extraction operator, f s is the sampling frequency, K0 is the number of Fourier transform points.

9. The method for acquiring channel status and target parameters of large-scale MIMO with integrated communication and perception according to claim 1, characterized in that: Construct an alternating optimization method to iteratively refine the azimuth of the qth target relative to the transmitter and Doppler shift The estimation problem of the i-th iteration is obtained, and then the estimated values ​​of AoD and Doppler frequency shift are obtained until the iterative update process converges to a stable point and the Doppler frequency shift is obtained. The azimuth of the qth target relative to the transmitter The azimuth of the qth target relative to the receiver and delay The specific method of estimating is: Step 9.1: Construct an alternating optimization method to iteratively refine the azimuth of the qth target relative to the transmitter and Doppler shift The Doppler shift in the (i-1)th iteration is known. The estimation of AoD in the i-th iteration is obtained. The Doppler shift estimation problem of the i-th iteration is obtained. The estimation problem of AoD in the i-th iteration is specifically: In the formula, is the estimated factor matrix The qth column of q is the qth diagonal element of Δ2, where Δ2 satisfies Δ1Δ2Δ3=I Q The unknown diagonal matrix of is a diagonal matrix, is the estimate of Doppler frequency shift, P is the training precoder, a bs (φ q ) is the array steering vector, φ q is the azimuth of the qth target relative to the transmitter, The Doppler shift estimation problem for the i-th iteration is specifically: In the formula, is the estimated factor matrix The qth column of q is the qth diagonal element of Δ2, where Δ2 satisfies Δ1Δ2Δ3=I Q The unknown diagonal matrix, Γ(v q ) is a diagonal matrix, v q is the Doppler shift, P is the training precoder, is the array steering vector, is the estimated azimuth of the qth target relative to the transmitter, Q is the diagonal element The diagonal matrix of do (ν q ) is Γ(ν q ) is a vector consisting of the diagonal elements of ; Step 9.2: Using the array to guide vector a bs (φ) and Γ(v q ) is a vector consisting of the diagonal elements of do (v) The Vandermonde structure, for the AoD estimation problem of the i-th iteration, is The guide vector a bs (φ) is expressed as This results in the following expression: in a(z φ ) is the array steering vector, is the diagonal matrix Γ, Doppler frequency shift estimate Estimated factor matrix The qth column of The quantity concerned, P is the training precoder; Step 9.3: Consider the case of no noise, then the numerator of the above equation is zero, and 2(M bs -1) roots, let is the root of unit amplitude in the unit circle, then we can find Get the solution Get the estimated value of AoD, specifically: Where ∠ represents the phase angle extraction operator, λ c is the wavelength, d is the array element spacing; Step 9.4: According to steps 9.2-9.3, the Doppler shift estimate of the i-th iteration is obtained, which is: Where T sym is the duration of an OFDM symbol; Step 9.5: Based on the above steps, in alternating cases, the azimuth of the qth target relative to the transmitter and Doppler shift The estimation is performed until the iterative update process converges to a stable point and the Doppler frequency shift is obtained. The azimuth of the qth target relative to the transmitter The azimuth of the qth target relative to the receiver and delay Estimates.

10. The method for acquiring channel status and target parameters of large-scale MIMO with integrated communication and perception according to claim 1, characterized in that: Tensor After expansion and transposition, it is specifically: where β=[β1,...,β Q ] T , A td is related to the delay τ q The array steering vector matrix is ​​concerned, B is the factor matrix, and N is the noise matrix; The reflection coefficient is estimated as follows: in and Based on delay Doppler shift and the azimuth of the qth target relative to the transmitter and the azimuth of the qth target relative to the receiver structure.

Images