Distributed xl-mimo near-field channel estimation method and system based on cooperation mechanism

Abstract

This invention discloses a distributed XL-MIMO near-field channel estimation method and system based on a cooperative mechanism. In the distributed XL-MIMO near-field system, reference access points and non-reference access points are divided. A unified near-field parameterized expression model for Loss-of-Stake (LoS) and Non-LoS (NLoS) components is constructed by characterizing the coupling relationship between angle and distance parameters. A channel parameter correlation mapping mechanism is established based on the access point deployment topology. LoS component cooperative parameter estimation is completed using a uniform quantization codebook and continuous domain gradient correction. After cancelling the LoS components, a sparse representation model of the NLoS components is constructed, and NLoS component cooperative parameter estimation is achieved through sparse reconstruction and continuous domain gradient correction. Finally, the near-field channel of each access point is reconstructed. This invention reduces parameter dimensionality and pilot overhead, improving the channel estimation accuracy and robustness of the distributed XL-MIMO near-field system.

Description

Technical Field

[0001] This invention belongs to the field of wireless communication technology, and further relates to a cooperative mechanism-based distributed XL-MIMO (Extra-Large Scale Multiple-Input Multiple-Output) near-field channel estimation method and system in the field of near-field channel state information estimation. This invention can be used for near-field spherical wave propagation modeling, parameter correlation derivation between access points (APs), and distributed cooperative channel estimation in near-field multiple-input multiple-output wireless communication systems with distributed array architectures. Background Technology

[0002] With the increasing application of distributed array architectures in wireless communication systems, XL-MIMO systems have improved system coverage and spatial resolution by expanding array aperture and spatial degrees of freedom. As array size increases and communication distances shorten, channel propagation gradually transitions from traditional far-field planar wave characteristics to near-field spherical wave propagation characteristics. The channel response is simultaneously influenced by array geometry, propagation distance, and angular parameters. In such distributed XL-MIMO near-field systems, multiple access points typically share the same user location and some scattering environment information, resulting in a natural correlation mapping of their channel parameters at the topological level. However, existing channel modeling and estimation methods often follow a single-access-point or centralized architecture design approach, treating the channel parameters of different access points as independently modeled and estimated, without establishing topological correlation mappings between access points during the modeling phase. Furthermore, angular and distance parameters are typically searched and estimated separately as independent high-dimensional parameter sets, without explicitly characterizing their coupling relationship. Under distributed deployment conditions, as the number of access points, array size, and multipath propagation increases, each access point repeatedly models and estimates highly correlated channel parameters, leading to a rapid expansion of the parameters to be estimated. Especially under conditions of limited pilot resources, high-dimensional parameter search not only significantly increases pilot overhead and computational complexity, but also places high demands on storage resources, limiting the scalability and engineering implementation capabilities of the system under large-scale deployment conditions. In addition, parameter search methods based on fixed discrete codebooks are prone to in-lattice parameter mismatch errors, which further affect estimation accuracy and algorithm stability under near-field high-resolution modeling conditions.

[0003] Shandong University disclosed a channel estimation method suitable for near-field XL-MIMO scenarios in its patent application "A Method and System for Near-Field Channel Estimation of Ultra-Large-Scale MIMO" (Application No.: 202311463501.9, Publication No.: CN 117478251 A). This method proposes to parameterize the near-field channel by introducing a spherical wave propagation model and to achieve channel estimation by combining angle and distance parameters. This method improves the near-field modeling accuracy in single-access-point scenarios. However, it still has shortcomings because it is designed based on a single access point or centralized architecture and does not consider the parameter correlation between access points in a distributed architecture. In multi-access-point scenarios, each access point still needs to perform a complete angle and distance parameter search and estimation process separately, and parameter redundancy remains. Furthermore, this method uses a discrete spectrum search method based on a preset search space to obtain angle and distance parameters in the parameter estimation stage. When the actual path parameters do not fall on discrete sampling grid points, the estimation results will be limited by the sampling interval, resulting in parameter quantization errors. When the array aperture is large or the near-field resolution requirement is high, the aforementioned discretization error may affect the estimation accuracy and stability. Therefore, although this type of method reduces the complexity of the two-dimensional search in single access point scenarios, it still does not eliminate the accuracy bottleneck caused by the discrete search at the level of continuous parameter modeling.

[0004] Southeast University disclosed a distributed near-field XL-MIMO channel estimation method in its patent application "A Distributed Estimation Method for Near-Field CSI for Broadband XL-MIMO" (Application No.: 202510507522.9, Publication No.: CN 120434081 A). This method proposes joint processing of channel information from multiple access points in a distributed architecture to improve system coverage. However, a drawback is that the method still relies on centralized fusion or collaborative processing of results after estimation, employing a separate modeling and estimation approach for each access point, without establishing a parameter mapping or correlation mechanism between access points from the modeling stage. Therefore, as the number of access points increases, the dimension of channel parameters and the size of the estimation codebook continue to grow rapidly, making it difficult to effectively reduce pilot overhead and computational complexity, thus limiting the scalability of the system under distributed deployment conditions.

[0005] In summary, the following technical issues still exist in the distributed near-field XL-MIMO system architecture:

[0006] 1) The coupling relationship between angle and distance parameters was not depicted during the modeling stage, which limited the accuracy of near-field modeling;

[0007] 2) The lack of a channel parameter correlation mapping mechanism between access points leads to severe redundant estimation and parameter redundancy;

[0008] 3) Under finite pilot conditions, high-dimensional parameter search leads to higher pilot overhead and computational complexity;

[0009] 4) Parameter search based on a fixed discrete codebook suffers from parameter mismatch error, which affects the estimation accuracy and stability. Summary of the Invention

[0010] The purpose of this invention is to address the problems existing in the prior art by providing a distributed XL-MIMO near-field channel estimation method and system based on a cooperative mechanism. This invention addresses the channel topology correlation characteristics caused by multiple access points sharing user locations and scattering environments in distributed XL-MIMO near-field systems, aiming to solve the following technical problems existing in the prior art in this scenario: (1) the near-field channel modeling does not explicitly characterize the coupling relationship between angle and distance parameters, resulting in limited modeling accuracy; (2) repeated modeling and estimation of highly correlated channel parameters by each access point, leading to parameter redundancy; (3) high pilot overhead and computational complexity caused by high-dimensional parameter search under finite pilot conditions; and (4) the in-grid parameter mismatch caused by discrete codebook search affecting estimation accuracy and stability.

[0011] To achieve the above objectives, the technical approach of this invention is as follows: This invention employs a unified near-field parameterized expression modeling: based on the near-field spherical wave propagation mechanism, it explicitly characterizes the coupling relationship between angle parameters and distance parameters during array propagation distance expansion, describing the line-of-sight (LoS) component and the non-line-of-sight (NLoS) component under the same parameterized structure, enabling the channel to be described by a unified parameterized expression. Its key feature is that it directly characterizes the coupling relationship between angle parameters and distance parameters during the modeling stage, constructing a unified near-field channel parameterized expression model. Since angle parameters and distance parameters are essentially determined by the same topological relationship in near-field propagation, unified parameterized expression modeling avoids model mismatch caused by independent modeling of angle and distance parameters, thereby improving near-field modeling accuracy and solving the problem of limited near-field modeling accuracy. This invention also employs a channel parameter correlation mapping mechanism between reference access points and non-reference access points: arbitrarily selecting an access point in the distributed network as the reference access point. Other access points are designated as non-reference access points. ( =1,2,…, -1), based on the known deployment topology between access points, an angle and distance parameter mapping expression is established between reference access points and non-reference access points. This allows some LosS / NLoS component parameters of non-reference access points to be directly derived from the estimation results of reference access points, thereby avoiding independent complete parameter searches for the same physical path by each access point. Its characteristic lies in achieving cross-access point parameter derivation through a parameter association mapping mechanism, reducing the independent execution of complete parameter search processes by each access point. This solves the problems of redundant estimation and parameter redundancy. This invention adopts a segmented cooperative parameter estimation technology scheme, dividing the channel parameter estimation process into a staged cooperative estimation process. First, the LosS component parameters are cooperatively estimated to form network-level parameter constraints. Then, under these constraints, the NLoS component parameters are estimated. During the estimation process, the joint parameter search dimension and channel estimation codebook dimension are reduced based on the cooperative mechanism. Its characteristic is that channel estimation is transformed from a high-dimensional search to a staged low-dimensional parameter search, using the estimated path parameters to constrain the subsequent estimation space, and sharing estimation information under the cooperative mechanism to reduce redundant searches. Since the original high-dimensional parameter space is decomposed into multiple low-dimensional subspaces and constraints are formed by the estimated parameters, the search dimensionality can be significantly reduced, thereby reducing pilot overhead and computational complexity, and solving the resource consumption problem caused by high-dimensional parameter search. This invention employs a parameter continuous domain gradient correction optimization technique. Based on discrete codebook coarse estimation, a continuous domain parameter iterative update mechanism is introduced to continuously correct the angle and distance parameters, transforming the estimation results from discrete sampled values ​​to continuously optimized values. Its key feature is the combination of discrete coarse estimation and continuous domain optimization, with iterative updates in the continuous parameter space, overcoming the limitations of discrete codebook sampling resolution. Because the actual path parameters are distributed in a continuous space, continuous domain parameter correction avoids quantization errors caused by discrete sampling, thereby reducing parameter mismatch errors, improving estimation accuracy and stability, and solving the estimation accuracy and stability problems caused by fixed discrete codebook parameter search.

[0012] Based on the above technical approach, the technical solution provided by this invention is to divide the access points in a distributed XL-MIMO near-field system, which consists of multiple spatially distributed access points and user terminals, into one reference access point and multiple non-reference access points; the access points are connected through a fronthaul link, and the coarse estimation and continuous domain correction of the LoS and NLoS components of the wireless channel of each access point are estimated in segments; the non-reference access points, under the cooperation mechanism, use the parameter estimation results of the reference access point to complete their own channel parameter estimation based on the constructed channel parameter association mapping mechanism.

[0013] The steps of the distributed XL-MIMO near-field channel estimation method based on a cooperative mechanism of the present invention include the following:

[0014] Step 1: In the distributed XL-MIMO near-field system, arbitrarily select one access point as the reference access point and the rest as non-reference access points; by characterizing the coupling relationship between angle parameters and distance parameters in the model building stage, construct a unified near-field parameterized expression model for LoS components and NLoS components under the same parameterized structure.

[0015] Step 2: Based on the deployment topology relationship between access points in the distributed XL-MIMO system, establish a channel parameter association mapping mechanism between reference access points and non-reference access points;

[0016] Step 3: Each access point constructs an observation signal based on the pilot signal sent by the received user terminal; the LoS component parameters are coarsely estimated using a uniformly quantized codebook, and the parameter continuous domain gradient is corrected using the coarse estimate from the discrete codebook as the initial value; the estimated LoS component parameters are shared to each non-reference access point through the fronthaul link to limit the parameter search space of the non-reference access points.

[0017] Step 4: Based on the channel parameter association mapping mechanism between access points, each non-reference access point performs discrete codebook coarse estimation and continuous domain gradient correction of the necessary low-dimensional AoA parameters of the LosS component within the limited parameter search space. The parameters of the remaining LosS components are derived from the estimation results of the reference access point and the topology information.

[0018] Step 5: After offsetting the influence of the LoS component, a sparse representation model is constructed for the NLoS component. A coarse estimate is obtained through a sparse reconstruction algorithm. The angle parameter and distance parameter are optimized by continuous domain gradient correction. The reference access point shares the critical path parameters of the NLoS component with each non-reference access point to support them in completing the NLoS component estimation and channel reconstruction in the dimensionality reduction search space.

[0019] Step 6: Based on the parameter estimation results of the access point and the non-reference access point, reconstruct the LoS component and NLoS component of each access point, and output the channel estimation results of each access point in the distributed XL-MIMO near-field system.

[0020] Furthermore, the coupling relationship between the angle parameter and the distance parameter is characterized in the model building stage as follows:

[0021] ;

[0022] in, This represents the coupling relationship between the angle and distance parameters of the access point and the user terminal when establishing the near-field channel model between the access point and the user terminal. Indicates the signal carrier wavelength. Indicating the number of Loss transmitters in the access point array The spatial projection ratio of the Loss component on the origin side. In the LoS receiver array of the access point, the first... Spatial projection ratio of the Loss component on the receiving end side This indicates the LoS propagation distance between the access point and the user terminal.

[0023] Furthermore, the steps for constructing a unified near-field parameterized representation model are as follows:

[0024] The first step is to model the LoS components of each access point and the user terminal into the following near-field parameterized expression based on the coupling relationship between the angle and distance parameters of each access point, according to the following formula:

[0025] ;

[0026] in, This indicates the Loss component channel of the access point. Indicates the Loss component path gain at the access point. The phase rotation factor representing the Loss component. This represents the near-field steering vector on the receiving end side. This represents the near-field steering vector at the origin. This indicates the transpose operation. This indicates an element-wise multiplication operation. Indicates the coupling relationship The angle and distance coupled modulation matrix is ​​composed of;

[0027] The second step is to model the NLoS components of each access point and user terminal as follows:

[0028] ;

[0029] in, Indicates the first NLoS component of the access point Channels corresponding to each scatterer Indicates the first NLoS component of the access point The path gain corresponding to each scatterer Indicates the first NLoS component Phase rotation factor corresponding to each scatterer Indicates the distance from the center of the user terminal array to the... The propagation distance of each scatterer Indicates the distance from the center of the access point array to the... The propagation distance of each scatterer This indicates the center of the user terminal array and the center of the access point array.

[0030] Furthermore, the steps for establishing the channel parameter association mapping mechanism between the reference access point and the non-reference access point are as follows:

[0031] The first step is to establish a channel angle parameter association mapping mechanism based on the LoS component angle parameters of the reference access point and non-reference access points, as follows:

[0032] ;

[0033] in, , Representing the reference access point and the first The LoS components of a non-reference access point: departure angle and arrival angle. The radian system used to express 90°. , These represent the departure angle and arrival angle of the LoS components of the reference access point, respectively. , Both indicate that the system has known topological angle information;

[0034] The second step, based on the known topological distance information of the system, establishes a mapping mechanism for the LosS component distance parameters between the reference access point and the non-reference access point as follows:

[0035] ;

[0036] in, , These represent the channel propagation distances of the LoS components of the non-reference access point and the reference access point, respectively. This indicates that the system has known topological distance information. Represents the cosine function;

[0037] The third step is to share the near-field steering vector of the NLoS component transmitter side of the reference access point with the non-reference access point; and to establish a channel distance parameter association mapping mechanism between the NLoS components of the reference access point and the non-reference access point in the same way as the first and second steps.

[0038] Furthermore, the step of coarsely estimating the LosS component parameters using a uniformly quantized codebook is as follows:

[0039] The first step is to establish the observation signals of each access point after the user terminal sends the pilot signal within a coherent time period:

[0040] ;

[0041] in, This represents the observed signal at the access point. This represents the radio frequency combining matrix at the receiving end. Indicates the total number of channels at the access point. This represents the pilot signal on the transmitting side of the user terminal. This represents zero-mean complex Gaussian white noise;

[0042] The second step is to generate the LoS channel parameter quantization codebook for the reference access point:

[0043]

[0044] in, This represents the quantization codebook of the LosS component channel parameters of the reference access point. , , These represent the sampled values ​​of the Loss component distance, AoD, and AoA parameters of the reference access point, respectively. , These represent the maximum and minimum distance quantization values ​​of the LoS component of the reference access point, respectively. , These represent the maximum and minimum values ​​of the AoD quantization for the Loss component, respectively. , These represent the maximum and minimum values ​​of the AoA quantization for the Loss component, respectively. , , These represent the quantization step size of the LosS component distance, AoD, and AoA parameters, respectively.

[0045] The third step is based on the least squares method. Coarse estimation of the LoS component parameters of the reference access point in the search.

[0046] Furthermore, the step of performing continuous-domain gradient correction of parameters using the discrete codebook coarse estimate as the initial value is as follows:

[0047] The first step is to use the Frobenius norm squared error between the observed signal and the model reconstructed signal as the objective function, that is, to construct a continuous domain parameter optimization problem by minimizing the error between the observed signal and the reconstructed channel response; the optimization variables include the LosS component distance, AoA, and AoD of the reference access point.

[0048] The second step is to use the coarse estimate of the LoS component parameters as the initial value for continuous domain gradient correction, and calculate the gradient of the objective function with respect to each optimization parameter to obtain the parameter update direction.

[0049] The third step involves adaptively adjusting the gradient descent step size using a simplified Armijo criterion. After setting the initial step size, it is determined whether the updated parameters satisfy the descent condition that the current iteration's objective function value is no greater than the previous iteration's objective function value. If not, the step size is reduced by a preset ratio; this process is repeated until the descent condition is met.

[0050] The fourth step is to update each parameter along the gradient descent direction using a determined step size. When the normalization error of each LosS component parameter is less than the preset threshold or the maximum number of iterations is reached, the iteration is terminated and the fifth step is executed; otherwise, the process returns to the second step.

[0051] The fifth step is to use the continuously corrected LoS component parameter estimation results as the final LoS component parameter estimation values ​​for the reference access point.

[0052] Furthermore, the steps for performing discrete codebook coarse estimation and continuous domain gradient correction on the necessary low-dimensional AoA parameters for the LoS components are as follows:

[0053] The first step is to share the final estimation results of the LoS component parameters of the reference access point with each non-reference access point. After constructing the observation signals of the non-reference access points, the non-reference access points only need to estimate the AoA parameter in the LoS component, and the other LoS component parameters are determined by the estimation results of the reference access point and the known topology information of the system.

[0054] The second step involves constructing a discrete angle search codebook for the AoA parameters only for the non-reference access point, and calculating the error between the observed signal and the reconstructed channel within this discrete search space, thereby obtaining a coarse estimate of the AoA parameters through a traversal search.

[0055] The third step involves using the coarse discrete estimation result as the initial value and employing the same continuous domain gradient correction method as the reference access point to refine the AoA parameter. The optimization ends when the normalization error of the AoA parameter is less than the preset threshold or the maximum number of iterations is reached, thus obtaining the final AoA parameter estimation result for the non-reference access point.

[0056] Furthermore, the derivation of the remaining LoS component parameters from the reference access point estimation results and topology information means that after performing continuous domain correction optimization on the AoA parameters of the non-reference access point to obtain the final estimation result, combining the LoS component parameter estimation results of the reference access point with the established channel parameter correlation mapping mechanism between the reference access point and the non-reference access point, the final parameter estimation results of the AoD parameters and distance parameters of the non-reference access point can be derived, thus completing the LoS component estimation of the non-reference access point.

[0057] Furthermore, the steps of obtaining a coarse estimate through a sparse reconstruction algorithm and optimizing the angle and distance parameters using continuous domain gradient correction are as follows:

[0058] The first step is that after each access point completes the LoS component estimation, it removes the estimated LoS component from the observed signal to obtain a residual signal containing only NLoS component and noise, which is used for subsequent NLoS component parameter estimation.

[0059] The second step involves constructing near-field polar domain discrete codebooks for each access point based on the system's preset angle and distance parameter sampling rules, and characterizing the near-field steering vector under different combinations of angle and distance parameters; and generating the corresponding sensing matrix based on the constructed near-field polar domain discrete codebooks.

[0060] The third step involves using an orthogonal matching pursuit algorithm combined with compressed sensing to iteratively select several support components in the sensing matrix, determine the set of angle and distance parameter indices corresponding to the NLoS components, and thus obtain a coarse estimation result for the NLoS components.

[0061] The fourth step involves using the coarse estimation result as the initial value to construct a continuous domain gradient correction optimization problem with the residual signal reconstruction error as the objective function, and further optimizing the angle and distance parameters of the NLoS component. During the parameter update process, the Armijo criterion is used to adaptively adjust the step size until the change in the objective function value between two consecutive iterations is less than a preset threshold or the maximum number of iterations is reached, at which point the optimization ends and the continuously corrected NLoS component parameter estimation result is obtained.

[0062] The fifth step is to share the critical path parameters of the NLoS components estimated by the reference access point to the non-reference access point. The non-reference access point performs dimensionality reduction on the near-field polar domain discrete codebook of the receiving end based on the shared parameters of the transmitting end and the known topology information of the system, and completes the NLoS component estimation by adopting the sparse reconstruction and continuous domain gradient correction process consistent with the reference access point.

[0063] The sixth step is to reconstruct the corresponding NLoS component channel matrix based on the optimized NLoS component parameters of each access point, and then superimpose it with the estimated LoS components to obtain the final near-field channel estimation results for each access point.

[0064] This invention discloses a distributed XL-MIMO near-field channel estimation system based on a cooperative mechanism, which is implemented using a distributed XL-MIMO near-field channel estimation method; it includes the following modules:

[0065] The distributed near-field parameterized modeling module is used to characterize the coupling relationship between angle and distance parameters during the model building phase, constructing a unified near-field parameterized expression model for LoS and NLoS components. The access point parameter association mapping module, based on the constructed unified near-field parameterized expression model and combined with the deployment topology relationships between access points, establishes a channel parameter association mapping mechanism between reference access points and non-reference access points. The LoS component cooperative parameter estimation module is used to perform coarse estimation and continuous domain gradient correction of the reference access point's LoS component parameters, and shares the LoS component parameter estimation results of the reference access point with the non-reference access points, completing the non-... The system performs a coarse estimation of the low-dimensional AoA parameters of the reference access point and continuous domain gradient correction to obtain the LoS component parameter estimation results for each access point. The NLoS component cooperative parameter estimation module performs coarse estimation of the NLoS component parameters of the reference access point and continuous domain gradient correction, and shares critical path parameters with non-reference access points. This completes the coarse estimation of the NLoS component parameters of the non-reference access points in the reduced-dimensional search space and continuous domain gradient correction, obtaining the NLoS component parameter estimation results for each access point. The channel reconstruction module reconstructs the near-field channel matrix based on the LoS and NLoS component parameter estimation results of each access point and outputs the final channel estimation results.

[0066] Compared with the prior art, the present invention has the following advantages:

[0067] First, the method of the present invention overcomes the problem of inconsistent model expression caused by independent modeling of angle and distance parameters in existing methods by unifying the near-field parameterized expression modeling and explicitly characterizing the coupling relationship between angle and distance parameters during the model building stage. This enables the channel response of the present invention to be consistently represented by a unified set of physical parameters, thereby improving the consistency and accuracy of near-field channel expression from the modeling level and significantly improving the modeling accuracy of channel parameter estimation under near-field conditions.

[0068] Second, the method of the present invention establishes a channel parameter correlation mapping mechanism between reference access points and non-reference access points, enabling some LoS / NLoS component parameters of non-reference access points to be directly derived from the estimation results of reference access points. This overcomes the parameter redundancy and dimensionality expansion problems caused by non-reference access points performing independent and complete parameter searches and repeated estimations for the same physical path in distributed XL-MIMO near-field systems. As a result, the present invention can reduce the scale of parameters and codebook dimensions that non-reference access points need to search independently during the modeling stage, thereby significantly reducing the degree of parameter redundancy under multi-access point deployment conditions.

[0069] Third, the method of the present invention uses a continuous domain gradient correction mechanism on the basis of discrete codebook coarse estimation to transition the parameter estimation result from discrete approximation value to continuous correction value, which overcomes the parameter quantization error caused by parameter search based on fixed discrete codebook. This allows the present invention to break through the limitation of discrete sampling resolution and improve the accuracy and stability of parameter estimation from the estimation level, thereby achieving more accurate near-field channel reconstruction under different signal-to-noise ratio and pilot length configuration conditions.

[0070] Fourth, the system of the present invention combines the parameter derivation mechanism of the access point with the segmented collaborative parameter estimation mechanism through the collaborative mode of "parameter derivation + dimensionality reduction estimation". This overcomes the problem of high pilot overhead and computational complexity caused by joint search of high-dimensional parameters under limited pilot conditions in existing systems. As a result, the system of the present invention significantly reduces pilot resource consumption, computational complexity and fronthaul link information transmission burden while ensuring estimation accuracy, thereby improving the scalability and engineering implementation capability of the distributed XL-MIMO near-field system under large-scale deployment conditions. Attached Figure Description

[0071] Figure 1 This is a flowchart of the method of the present invention;

[0072] Figure 2 This is a block diagram of the distributed XL-MIMO near-field system of the present invention;

[0073] Figure 3 This is a schematic diagram of the LoS component channel correlation mapping mechanism proposed in this invention;

[0074] Figure 4 This is a schematic diagram of the NLoS component channel association mapping mechanism proposed in this invention;

[0075] Figure 5 This is a schematic diagram of the polar-domain discrete codebook dimensionality reduction based on the proposed channel correlation mapping mechanism of this invention;

[0076] Figure 6 This is a comparison of the channel estimation performance curves of the proposed algorithm and the traditional algorithm under a fixed signal-to-noise ratio as the pilot length changes;

[0077] Figure 7 This is a comparison of the channel estimation performance curves of the proposed algorithm and the traditional algorithm under a fixed pilot length, as the signal-to-noise ratio changes. Detailed Implementation

[0078] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments.

[0079] Reference Figure 1 The core steps and Figure 2The system block diagram further describes the implementation steps of the distributed XL-MIMO near-field channel estimation method and system embodiment based on the cooperative mechanism of the present invention.

[0080] This invention applies to distributed XL-MIMO near-field systems, where each access point communicates with a user terminal via a wireless channel. In this embodiment, each access point and user terminal is equipped with a very large-scale linear antenna array with an element spacing of half a wavelength; the channel in the system is a hybrid channel of LoS and NLoS components. Due to the distributed deployment structure, different access points and the same user terminal share the user terminal's location and some scattering environment information.

[0081] Step 1: The distributed near-field parametric modeling module arbitrarily selects one access point in the system as the reference access point, and marks the remaining access points as non-reference access points. The propagation distance between the receiving antenna elements and the transmitting antenna elements of each access point and the user terminal is performed using Taylor expansion, and decomposed into receiver-side correlation terms, transmitter-side correlation terms, and the coupling relationship between angle parameters and distance parameters.

[0082]

[0083] in, This represents the coupling relationship between the angle and distance parameters of the access point and the user terminal when establishing the near-field channel model between the access point and the user terminal. Indicates the signal carrier wavelength. Indicating the number of Loss transmitters in the access point array The spatial projection ratio of the Loss component on the origin side. In the LoS receiver array of the access point, the first... Spatial projection ratio of the Loss component on the receiving end side This indicates the LoS propagation distance between the access point and the user terminal.

[0084] Based on the coupling relationship between the angle and distance parameters of each access point, the LoS component of each access point is modeled as the following near-field parameterized expression:

[0085]

[0086] in, This indicates the Loss component channel of the access point. Indicates the Loss component path gain at the access point. The phase rotation factor representing the Loss component. This represents the near-field steering vector on the receiving end side. This represents the near-field steering vector at the origin. This indicates the transpose operation. This indicates an element-wise multiplication operation. Indicates the coupling relationship The angle and distance coupled modulation matrix is ​​formed.

[0087] The NLoS components of each access point and user terminal are modeled as follows in a near-field parameterized form:

[0088]

[0089] in, Indicates the first NLoS component of the access point Channels corresponding to each scatterer Indicates the first NLoS component of the access point The path gain corresponding to each scatterer Indicates the first NLoS component Phase rotation factor corresponding to each scatterer Indicates the distance from the center of the user terminal array to the... The propagation distance of each scatterer Indicates the distance from the center of the access point array to the... The propagation distance of each scatterer This represents the center of the user terminal array and the center of the access point array. It allows different components to be described under a unified parameterized structure, significantly improving the consistency of channel representation and modeling accuracy in distributed XL-MIMO near-field systems.

[0090] Step 2: The access point parameter association mapping module is based on the LoS component angle parameters of the reference access point and the non-reference access point.

[0091] Reference Figure 3 Based on the system topology, the channel angle parameter correlation mapping mechanism is established as follows:

[0092]

[0093] in, , Representing the reference access point and the first The LoS components of a non-reference access point: departure angle and arrival angle. The radian system used to express 90°. , These represent the departure angle and arrival angle of the LoS components of the reference access point, respectively. , Both represent known topological angle information of the system. Similarly, by... Figure 3 Based on the system topology and known topological distance information, the following mechanism is established to associate and map the LoS component distance parameters between reference access points and non-reference access points:

[0094]

[0095] in, , These represent the channel propagation distances of the LoS components of the non-reference access point and the reference access point, respectively. This indicates that the system has known topological distance information. This represents the cosine function. Based on the aforementioned angle parameter association mapping mechanism and distance parameter association mapping mechanism for the LoS components, non-reference access points only need to re-estimate the AoA parameter in the LoS components.

[0096] For NLoS components, refer to Figure 4 Different access points observe the same set of dominant scatterers. Based on this physical characteristic, the access point parameter correlation mapping module shares the near-field steering vector pair at the transmitting end of the NLoS component of the reference access point with the non-reference access points. Simultaneously, using the same mechanism as the parameter correlation mapping mechanism in the LoS component, a distance parameter correlation mapping mechanism is established between the near-field steering vectors at the receiving end of the NLoS component of the reference and non-reference access points. Non-reference access points only need to estimate a small number of AoA parameters in the NLoS component.

[0097] Step 3: After the user terminal transmits the pilot signal within one coherent time period, the LoS component cooperative parameter estimation module establishes the observation signals of each access point:

[0098]

[0099] in, This represents the observed signal at the access point. This represents the radio frequency combining matrix at the receiving end. Indicates the total number of channels at the access point. This represents the pilot signal on the transmitting side of the user terminal. Represents zero-mean complex Gaussian white noise. Generate the LoS channel parameter quantization codebook for the reference access point:

[0100]

[0101] in, This represents the quantization codebook of the LosS component channel parameters of the reference access point. , , These represent the sampled values ​​of the Loss component distance, AoD, and AoA parameters of the reference access point, respectively. , These represent the maximum and minimum distance quantization values ​​of the LoS component of the reference access point, respectively. , These represent the maximum and minimum values ​​of the AoD quantization for the Loss component, respectively. , These represent the maximum and minimum values ​​of the AoA quantization for the LoS component, respectively. , , These represent the quantization step size for the LosS component distance, AoD, and AoA parameters, respectively. The quantization step size is calculated as follows:

[0102] ;

[0103] in, , , These represent the distance to the Loss component, the number of quantized parameters for AoD, and AoA, respectively. Based on the least squares method... Coarse estimation of the LoS component parameters of the reference access point in the search.

[0104] After obtaining the coarse estimation results of the LosS component parameters of the reference access point, the Frobenius norm squared error between the observed signal and the model reconstructed signal of the reference access point is used as the objective function. That is, by minimizing the error between the observed signal and the reconstructed channel response, a continuous domain parameter optimization problem is constructed. The optimization variables include the LosS component distance, AoA, and AoD of the reference access point. The gradient of the objective function with respect to each optimization parameter is calculated to obtain the parameter update direction. The simplified Armijo criterion is used to adaptively adjust the gradient descent step size. After setting the initial step size, it is determined whether the updated parameters make the objective function satisfy the descent condition that the current iteration objective function value is not greater than the previous iteration objective function value. If not, the step size is reduced by a preset ratio; the above process is repeated until the descent condition is met. Using the determined step size, each parameter is updated along the gradient descent direction. When the normalization error of each LosS component parameter is less than a preset threshold or the maximum number of iterations is reached, the iteration is terminated, and the final estimation result of the reference access point LosS component parameters after continuous domain correction is obtained.

[0105] The final estimation results of the LosS component parameters of the reference access point are shared with each non-reference access point. After constructing the observation signals of the non-reference access points, the non-reference access points only need to estimate the AoA parameter in the LosS component; the parameters of the remaining LosS components are determined by the estimation results of the reference access point and the known topology information of the system. The non-reference access points construct a discrete angle search codebook only for the AoA parameter.

[0106] ;

[0107] in, This represents the quantization codebook of the LosS component channel parameters for non-reference access points. This represents the sampled value of the AoA parameter of the LosS component of the non-reference access point. The error between the observed signal and the reconstructed channel is calculated within this discrete search codebook, providing a coarse estimate of the AoA parameter obtained through traversal search. After obtaining the discrete coarse estimate, this value is used as the initial value, and the AoA parameter is finely optimized using the same continuous domain gradient correction method as the reference access point. Optimization ends when the normalization error of the AoA parameter is less than a preset threshold or the maximum number of iterations is reached. The final AoA parameter estimation result for the non-reference access point is obtained. After obtaining the final estimation result by continuous domain correction optimization of the AoA parameter of the non-reference access point, combined with the LosS component parameter estimation result of the reference access point and the established channel parameter correlation mapping mechanism between the reference and non-reference access points, the final parameter estimation results of the AoD parameter and distance parameter of the non-reference access point can be derived, completing the LosS component estimation of the non-reference access point.

[0108] Step 4: After each access point completes the LoS component estimation, the NLoS component cooperative parameter estimation module eliminates the estimated LoS components from the observed signals of each access point, obtaining a residual signal containing only the NLoS component and noise.

[0109] ;

[0110] in This represents the residual signal containing NLoS components and noise. This represents the estimated Loss component of the access point. This residual signal is used for subsequent NLoS component parameter estimation.

[0111] For the reference access point, sampling rules based on preset angle and distance parameters are applied, where uniform quantization is used for the angle parameters and non-uniform quantization for the distance parameters. Near-field polar discrete codebooks are constructed for both the receiving and transmitting ends to characterize the near-field steering vector under different combinations of angle and distance parameters. Based on the constructed near-field polar discrete codebooks, corresponding sensing matrices are generated.

[0112] ;

[0113] in, This represents the sensing matrix at the receiving end of the reference access point. This represents the near-field polar discrete codebook at the receiving end of the reference access point. This represents the sensing matrix at the reference access point transmitting side. This represents the near-field polar discrete codebook at the reference access point transmitter. This represents the conjugate transpose operation on the matrix. Using an orthogonal matching pursuit algorithm combined with compressed sensing, several support components are iteratively selected from the sensing matrices at both the transmitting and receiving ends to determine the set of angle and distance parameter indices corresponding to the NLoS components of the reference access point, thereby obtaining a coarse estimate of the NLoS components of the reference access point.

[0114] Using the coarse estimation result of the NLoS component of the reference access point as the initial value, a continuous domain gradient correction optimization problem is constructed with the residual signal reconstruction error as the objective function to further optimize the angle and distance parameters of the NLoS component of the reference access point. During the parameter update process, the Armijo criterion is used to adaptively adjust the step size until the change of the objective function value in two consecutive iterations is less than a preset threshold or the maximum number of iterations is reached, and the optimization ends, thus obtaining the continuous domain corrected NLoS component parameter estimation result of the reference access point.

[0115] The critical path parameters of the NLoS components estimated by the reference access point are shared with non-reference access points, as shown in the following formula: Non-reference access points share the originating-side support components of the reference access point:

[0116]

[0117] in, This represents the sensing matrix at the non-reference access point transmitting side. This indicates the supporting component index of the reference access point's transmitting side.

[0118] Reference Figure 5 Based on the shared parameter information and the known topology information of the system, and combined with the proposed channel parameter association mapping mechanism, the non-reference access point only needs to uniformly quantize the angle parameter at the deduced and located distance parameter sampling value to obtain the dimension-reduced receiver-side near-field polar domain discrete codebook and the corresponding receiver-side sensing matrix. Then, the non-reference access point completes the NLoS component parameter estimation using the same sparse reconstruction and continuous domain gradient correction process as the reference access point.

[0119] Step 5: The channel reconstruction module reconstructs the corresponding LoS component and NLoS component channel matrix based on the parameter estimation results of each access point's LoS component and NLoS component, and superimposes the two to obtain the final near-field channel estimation result for each access point.

[0120] The effects of this invention can be further illustrated by the following simulation.

[0121] 1. Simulation experimental conditions.

[0122] The simulation software platform for this invention is: Windows 10 operating system and Matlab R2022a. 2. Simulation content and result analysis.

[0123] There are two simulation experiments for this invention.

[0124] The distributed XL-MIMO near-field system used in simulation experiment 1 of this invention operates in the millimeter-wave band with a carrier wavelength of 0.0075m. Each access point and user terminal uses a 128-element uniform linear array. The receiver has 16 RF chains. The user terminal operates at a range of 30m. 2 The pilots are randomly generated within the coverage area, and the propagation distance satisfies the near-field propagation condition. Both the pilot matrix and the receiver combining matrix are generated using a normalized Rademacher random sequence. The signal-to-noise ratio is fixed at 3dB, and the pilot length is incremented from 4 to 20 in steps of 2.

[0125] Simulation Experiment 1 of this invention uses the method of this invention and three existing technologies to obtain the normalized mean square error (MSE) values ​​for channel estimation at nine pilot lengths. The relationship between the obtained MSE and the pilot length is then plotted. Figure 6 The four curves shown.

[0126] In simulation experiment 1, the three existing technologies used are:

[0127] The prior art 1 refers to the low-complexity least squares channel estimation method proposed by Huang Bo et al. in their paper "OFDM Channel Estimation Based on Least Squares Algorithm" (Ship Electronic Engineering, 2015, 35(06): 51-53+72).

[0128] Existing technology 2 refers to the OMP channel estimation method based on the far-field angular domain transformation matrix proposed by Lee J et al. in their paper "Channel estimation via orthogonal matching pursuit for hybrid MIMO systems in millimeter wavecommunications" (IEEE Transactions on Communications, 2016, 64(6): 2370-2386).

[0129] Existing technology 3 refers to the OMP channel estimation method based on the near-field polar domain transformation matrix proposed by Dai Linglong et al. in their published paper "Channel estimation for extremely large-scale MIMO: Far-field or near-field?" (IEEE transactions on communications, 2022, 70(4): 2663-2677).

[0130] The distributed XL-MIMO system used in simulation experiment 2 of this invention has a fixed pilot length of 32, and the signal-to-noise ratio is increased from -6dB to 4dB in steps of 2. The other parameter configurations are the same as those in simulation experiment 1.

[0131] Simulation Experiment 2 of this invention uses the method of this invention and three existing technologies to obtain the normalized mean square error (MSE) values ​​of channel estimation at six different signal-to-noise ratios (SNRs). The relationship between the obtained MSE and the SNR is then plotted. Figure 7 The four curves shown.

[0132] In simulation experiment 2, the three existing technologies used are the same as those in simulation experiment 1.

[0133] The following is combined Figure 6 and Figure 7 The simulation diagrams further illustrate the effects of the present invention.

[0134] Figure 6 The horizontal axis in the figure represents the pilot length for channel estimation in a distributed XL-MIMO system, in units of pilots, and the vertical axis represents the normalized mean square error of channel estimation. Figure 6 The black curve represents the relationship between the normalized mean square error and the pilot length obtained by simulation using prior art 1; the light blue curve represents the relationship between the normalized mean square error and the pilot length obtained by simulation using prior art 2; the dark blue curve represents the relationship between the normalized mean square error and the pilot length obtained by simulation using prior art 3; and the green curve represents the relationship between the normalized mean square error and the pilot length obtained by the method proposed in this invention.

[0135] from Figure 6 As can be seen, the nine normalized mean square errors obtained by the method of this invention continuously decrease with the increase of pilot length. Under the same pilot length condition, the normalized mean square error obtained by the method of this invention is always smaller than the normalized mean square errors obtained by three existing simulations. Under the same normalized mean square error index of -3.5dB, the method of this invention requires less pilot length, saving pilot overhead resources.

[0136] Figure 7 The horizontal axis in the figure represents the signal-to-noise ratio (SNR) for channel estimation in the distributed XL-MIMO system, in dB, and the vertical axis represents the normalized mean square error of channel estimation. Figure 7The black curve in the figure represents the relationship between normalized mean square error and signal-to-noise ratio obtained by simulation using prior art 1; the light blue curve represents the relationship between normalized mean square error and signal-to-noise ratio obtained by simulation using prior art 2; the dark blue curve represents the relationship between normalized mean square error and signal-to-noise ratio obtained by simulation using prior art 3; and the red curve represents the relationship between normalized mean square error and signal-to-noise ratio obtained by the method proposed in this invention.

[0137] from Figure 7 As can be seen, the method of the present invention is significantly better than the other three existing methods in terms of channel estimation normalized mean square error across the entire signal-to-noise ratio range. Furthermore, as the signal-to-noise ratio continues to increase, the method of the present invention still shows a significant performance improvement trend, indicating that the method of the present invention has better estimation performance and robustness than the existing technology.

Claims

1. A distributed XL-MIMO near-field channel estimation method based on a cooperative mechanism, characterized in that, In a distributed XL-MIMO near-field system consisting of multiple spatially distributed access points and user terminals, each access point is divided into one reference access point and multiple non-reference access points. Each access point is connected via a fronthaul link, and the coarse estimation and continuous domain correction of the LoS and NLoS components of the wireless channel of each access point are performed in segments. Under the cooperative mechanism, non-reference access points estimate their own channel parameters using the parameter estimation results of reference access points, based on the constructed channel parameter correlation mapping mechanism. The steps of this estimation method include the following: Step 1: In the distributed XL-MIMO near-field system, arbitrarily select one access point as the reference access point and the rest as non-reference access points; by characterizing the coupling relationship between angle parameters and distance parameters in the model building stage, construct a unified near-field parameterized expression model for LoS components and NLoS components under the same parameterized structure. Step 2: Based on the deployment topology relationship between access points in the distributed XL-MIMO system, establish a channel parameter association mapping mechanism between reference access points and non-reference access points; Step 3: Each access point constructs an observation signal based on the pilot signal sent by the received user terminal; the LoS component parameters are coarsely estimated using a uniformly quantized codebook, and the parameter continuous domain gradient is corrected using the coarse estimate from the discrete codebook as the initial value; the estimated LoS component parameters are shared to each non-reference access point through the fronthaul link to limit the parameter search space of the non-reference access points. Step 4: Based on the channel parameter association mapping mechanism between access points, each non-reference access point performs discrete codebook coarse estimation and continuous domain gradient correction of the necessary low-dimensional AoA parameters of the LosS component within the limited parameter search space. The parameters of the remaining LosS components are derived from the estimation results of the reference access point and the topology information. Step 5: After offsetting the influence of the LoS component, a sparse representation model is constructed for the NLoS component. A coarse estimate is obtained through a sparse reconstruction algorithm, and the angle parameter and distance parameter are optimized by continuous domain gradient correction. The reference access point shares the critical path parameters of the NLoS component with each non-reference access point to support them in completing NLoS component estimation and channel reconstruction within the dimensionality reduction search space. Step 6: Based on the parameter estimation results of the access point and the non-reference access point, reconstruct the LoS component and NLoS component of each access point, and output the channel estimation results of each access point in the distributed XL-MIMO near-field system.

2. The distributed XL-MIMO near-field channel estimation method according to claim 1, characterized in that, The coupling relationship between angle parameters and distance parameters described in step 1 during the model building phase is as follows: ； in, This represents the coupling relationship between the angle and distance parameters of the access point and the user terminal when establishing the near-field channel model between the access point and the user terminal. Indicates the signal carrier wavelength. In the LoS transmitter array representing the access point, the first... The spatial projection ratio of the Loss component on the origin side In the LoS receiver array of the access point, the first... Spatial projection ratio of the Loss component on the receiving end side This indicates the LoS propagation distance between the access point and the user terminal.

3. The distributed XL-MIMO near-field channel estimation method according to claim 2, characterized in that, The steps for constructing the unified near-field parameterized representation model described in step 1 are as follows: The first step is to model the LoS components of each access point and the user terminal into the following near-field parameterized expression based on the coupling relationship between the angle and distance parameters of each access point, according to the following formula: ； in, This indicates the Loss component channel of the access point. Indicates the path gain of the Loss component at the access point. The phase rotation factor representing the Loss component. This represents the near-field steering vector on the receiving end side. This represents the near-field steering vector at the origin. This indicates the transpose operation. This indicates an element-wise multiplication operation. Indicates the coupling relationship The angle and distance coupled modulation matrix is ​​composed of; The second step is to model the NLoS components of each access point and user terminal as follows: ； in, Indicates the first NLoS component of the access point Channels corresponding to each scatterer Indicates the first NLoS component of the access point The path gain corresponding to each scatterer Indicates the first NLoS component Phase rotation factor corresponding to each scatterer Indicates the distance from the center of the user terminal array to the... The propagation distance of each scatterer Indicates the distance from the center of the access point array to the... The propagation distance of each scatterer This indicates the center of the user terminal array and the center of the access point array.

4. The distributed XL-MIMO near-field channel estimation method according to claim 3, characterized in that, The steps for establishing the channel parameter association mapping mechanism between the reference access point and the non-reference access point in step 2 are as follows: The first step is to establish a channel angle parameter association mapping mechanism based on the LoS component angle parameters of the reference access point and the non-reference access point, as follows: ； in, , Representing the reference access point and the first The LoS components of a non-reference access point: departure angle and arrival angle. The radian system used to express 90°. , These represent the departure angle and arrival angle of the LoS components of the reference access point, respectively. , Both indicate that the system has known topological angle information; The second step involves establishing a mapping mechanism for the LosS component distance parameters between reference access points and non-reference access points based on the known topology distance information of the system, as follows: ； in, , These represent the channel propagation distances of the LoS components of the non-reference access point and the reference access point, respectively. This indicates that the system has known topological distance information. Represents the cosine function; The third step is to share the near-field steering vector of the NLoS component transmitter side of the reference access point with the non-reference access point; and to establish a channel distance parameter association mapping mechanism between the NLoS components of the reference access point and the non-reference access point in the same way as the first and second steps.

5. The distributed XL-MIMO near-field channel estimation method according to claim 4, characterized in that, The steps for coarsely estimating the LosS component parameters using a uniformly quantized codebook in step 3 are as follows: The first step is to establish the observation signals of each access point after the user terminal sends the pilot signal within a coherent time period: ； in, This represents the observed signal at the access point. This represents the radio frequency combining matrix at the receiving end. Indicates the total number of channels at the access point. This represents the pilot signal on the transmitting side of the user terminal. This represents zero-mean complex Gaussian white noise; The second step is to generate the LoS channel parameter quantization codebook for the reference access point: ； in, This represents the quantization codebook of the LosS component channel parameters of the reference access point. , , These represent the sampled values ​​of the Loss component distance, AoD, and AoA parameters of the reference access point, respectively. , These represent the maximum and minimum distance quantization values ​​of the LoS component of the reference access point, respectively. , These represent the maximum and minimum values ​​of the AoD quantization for the Loss component, respectively. , These represent the maximum and minimum values ​​of the AoA quantization for the Loss component, respectively. , , These represent the quantization step size of the LosS component distance, AoD, and AoA parameters, respectively. The third step is based on the least squares method. Coarse estimation of the LoS component parameters of the reference access point in the search.

6. The distributed XL-MIMO near-field channel estimation method according to claim 1, characterized in that, The steps for performing continuous domain gradient correction of parameters using the discrete codebook coarse estimate as the initial value in step 3 are as follows: The first step is to use the Frobenius norm squared error between the observed signal and the model reconstructed signal as the objective function, that is, to construct a continuous domain parameter optimization problem by minimizing the error between the observed signal and the reconstructed channel response; the optimization variables include the LosS component distance, AoA, and AoD of the reference access point. The second step is to use the coarse estimate of the LoS component parameters as the initial value for continuous domain gradient correction, and calculate the gradient of the objective function with respect to each optimization parameter to obtain the parameter update direction. The third step is to adaptively adjust the gradient descent step size using a simplified Armijo criterion. After setting the initial step size, it is determined whether the updated parameters make the objective function satisfy the descent condition that the current iteration objective function value is not greater than the previous iteration objective function value. If not, the step size is reduced by a preset ratio. Repeat the above process until the descent condition is met; The fourth step is to update each parameter along the gradient descent direction using a determined step size. When the normalization error of each LosS component parameter is less than the preset threshold or the maximum number of iterations is reached, the iteration is terminated and the fifth step is executed; otherwise, the process returns to the second step. The fifth step is to use the continuously corrected LoS component parameter estimation results as the final LoS component parameter estimation values ​​for the reference access point.

7. The distributed XL-MIMO near-field channel estimation method according to claim 1, characterized in that, The steps in step 4 for performing discrete codebook coarse estimation and continuous domain gradient correction of the necessary low-dimensional AoA parameters for the LosS components are as follows: The first step is to share the final estimation results of the LoS component parameters of the reference access point with each non-reference access point. After constructing the observation signals of the non-reference access points, the non-reference access points only need to estimate the AoA parameter in the LoS component, and the other LoS component parameters are determined by the estimation results of the reference access point and the known topology information of the system. The second step involves constructing a discrete angle search codebook for the AoA parameters only for the non-reference access point, and calculating the error between the observed signal and the reconstructed channel within this discrete search space, thereby obtaining a coarse estimate of the AoA parameters through a traversal search. The third step involves using the coarse discrete estimation result as the initial value and employing the same continuous domain gradient correction method as the reference access point to refine the AoA parameter. The optimization ends when the normalization error of the AoA parameter is less than the preset threshold or the maximum number of iterations is reached, thus obtaining the final AoA parameter estimation result for the non-reference access point.

8. The distributed XL-MIMO near-field channel estimation method according to claim 1, characterized in that, The step 4, which describes deriving the remaining LoS component parameters from the reference access point estimation results and topology information, refers to the process of obtaining the final estimation result by performing continuous domain correction optimization on the AoA parameters of the non-reference access point. Then, by combining the LoS component parameter estimation results of the reference access point with the established channel parameter correlation mapping mechanism between the reference access point and the non-reference access point, the final parameter estimation results of the AoD parameters and distance parameters of the non-reference access point can be derived, thus completing the LoS component estimation of the non-reference access point.

9. The distributed XL-MIMO near-field channel estimation method according to claim 1, characterized in that, The steps described in step 5, which involve obtaining a coarse estimate using the sparse reconstruction algorithm and then optimizing the angle and distance parameters using continuous domain gradient correction, are as follows: The first step is that after each access point completes the LoS component estimation, it removes the estimated LoS component from the observed signal to obtain a residual signal containing only NLoS component and noise, which is used for subsequent NLoS component parameter estimation. The second step involves constructing near-field polar domain discrete codebooks for each access point based on the system's preset angle and distance parameter sampling rules, and characterizing the near-field steering vector under different combinations of angle and distance parameters; and generating the corresponding sensing matrix based on the constructed near-field polar domain discrete codebooks. The third step involves using an orthogonal matching pursuit algorithm combined with compressed sensing to iteratively select several support components in the sensing matrix, determine the set of angle and distance parameter indices corresponding to the NLoS components, and thus obtain a coarse estimation result for the NLoS components. The fourth step involves using the coarse estimation result as the initial value to construct a continuous domain gradient correction optimization problem with the residual signal reconstruction error as the objective function, and further optimizing the angle and distance parameters of the NLoS component. During the parameter update process, the Armijo criterion is used to adaptively adjust the step size until the change in the objective function value between two consecutive iterations is less than a preset threshold or the maximum number of iterations is reached, at which point the optimization ends and the continuously corrected NLoS component parameter estimation result is obtained. The fifth step is to share the NLoS component critical path parameters estimated by the reference access point to the non-reference access point. The non-reference access point performs dimensionality reduction on the near-field polar domain discrete codebook of the receiving end based on the shared parameters of the transmitting end and the known topology information of the system, and completes the NLoS component estimation by adopting the sparse reconstruction and continuous domain gradient correction process consistent with the reference access point. The sixth step is to reconstruct the corresponding NLoS component channel matrix based on the optimized NLoS component parameters of each access point, and then superimpose it with the estimated LoS components to obtain the final near-field channel estimation results for each access point.

10. A distributed XL-MIMO near-field channel estimation system based on a cooperative mechanism, characterized in that, This is implemented based on the distributed XL-MIMO near-field channel estimation method according to any one of claims 1-9; Includes the following modules: The distributed near-field parameterized modeling module is used to characterize the coupling relationship between angle and distance parameters during the model building phase, constructing a unified near-field parameterized expression model for LoS and NLoS components. The access point parameter association mapping module, based on the constructed unified near-field parameterized expression model and combined with the deployment topology relationships between access points, establishes a channel parameter association mapping mechanism between reference access points and non-reference access points. The LoS component cooperative parameter estimation module is used to perform coarse estimation and continuous domain gradient correction of the reference access point's LoS component parameters, and shares the LoS component parameter estimation results of the reference access point with the non-reference access points, completing the non-... The system performs a coarse estimation of the low-dimensional AoA parameters of the reference access point and continuous domain gradient correction to obtain the LoS component parameter estimation results for each access point. The NLoS component cooperative parameter estimation module performs coarse estimation of the NLoS component parameters of the reference access point and continuous domain gradient correction, and shares critical path parameters with non-reference access points. This completes the coarse estimation of the NLoS component parameters of the non-reference access points in the reduced-dimensional search space and continuous domain gradient correction, obtaining the NLoS component parameter estimation results for each access point. The channel reconstruction module reconstructs the near-field channel matrix based on the LoS and NLoS component parameter estimation results of each access point and outputs the final channel estimation results.

Images