Channel estimation and positioning system and method based on CP-Mama architecture

Abstract

The invention relates to a channel estimation and positioning system based on a CP-Mama framework. The channel estimation and positioning system comprises a pilot signal sending and receiving module, a data preprocessing module, a CP-Mama network module, a positioning result calculation module and a channel reconstruction module. The pilot signal sending and receiving module is used for sending a pilot signal to the user equipment through the base station and receiving an echo of the pilot signal; the data preprocessing module is used for extracting features highly related to user space coordinates according to the received pilot signals and constructing a received signal matrix; the CP-Mama network module comprises an encoder, a decoder and a jump connection layer; and the positioning result calculation module is used for calculating the position of the user equipment and a sight distance channel. The invention also relates to a method using the channel estimation and positioning system based on the CP-Mama architecture. According to the invention, the positioning precision and the channel estimation effect are improved; the calculation complexity is reduced; the problem that the spatial correlation is reduced due to the increase of the space between sparse array units is solved.

Description

Technical Field

[0001] This invention relates to the field of integrated sensing and communication technology, and more specifically to a channel estimation and positioning system and method based on the CP-Mamba architecture. Background Technology

[0002] As a core objective of sixth-generation integrated sensing and communication (ISAC) technology, this technology aims to maximize energy efficiency while enhancing sensing and communication capabilities. To this end, multiple-input multiple-output (MIMO) systems must improve spatial resolution to achieve accurate perception of dense targets while enhancing spectral efficiency. A highly promising solution is ultra-large-scale MIMO (XL-MIMO), which extends the traditional large-scale MIMO architecture by deploying more antennas. However, traditional compact arrays (CAs) using half-wavelength element spacing suffer from prohibitively high hardware costs. To address this challenge, sparse MIMO technology has emerged—this novel architecture achieves larger apertures without increasing the number of array elements, providing a new approach to system optimization. Furthermore, when the element spacing of a sparse XL-MIMO system exceeds half a wavelength, even larger element spacing can be achieved, significantly improving spatial resolution. Since limited spatial diversity affects signal resolution, accurate channel estimation and localization are crucial in sparse XL-MIMO systems.

[0003] Currently, research on sparse XL-MIMO is still in its early stages, with typical studies as follows: In their paper "An Exploration of the Advantages of Sparse Arrays in Ultra-Large-Scale MIMO Systems: Do Half-Wavelength Arrays Still Have Advantages in the Near Field?", Chen Xianzhe, Ren Hong, Pan Cunhua, Wang Chengxiang, and Wang Jiangzhou derived a closed-form expression for the power distribution and main lobe size of a sparse uniform planar array around the near-field focus and proposed an effective degree of freedom estimation method, fully demonstrating the potential of sparse arrays (SA) in suppressing interference and improving the capacity of XLMIMO systems.

[0004] The book "Two-Dimensional Orientation Estimation of Coprime Planar Arrays Based on Coprime Array Tensors" by Zheng Hang, Zhou Chengwei, Gu Yujie, and Shi Zhiguo designs a direction of arrival (DOA) estimator based on tensor decomposition for coprime arrays. By utilizing higher-order singular value decomposition and the inherent low-rank characteristics of the array, this method can not only interpolate missing sensor data but also effectively suppress noise, thus exhibiting stronger robustness in multi-target, low signal-to-noise ratio (SNR) environments.

[0005] The book "Direct MUSIC Methods on Sparse Arrays" by P. Vaidyanathan and P. Pal proposes a direct multi-signal classification method applicable to nested coprime arrays, which successfully avoids the co-array ambiguity problem and improves resolution.

[0006] The book "Sparse Coprime Arrays for Azimuth Estimation" by A. Raza, W. Liu, and Q. Shen developed a simplified coprime array that significantly reduces the number of sensors and cost while maintaining virtual aperture and DOA performance.

[0007] The paper "Channel Estimation of Sparse Arrays in Subarray-Based Hybrid Beamforming Systems" by J. Eisenbeis, N. Kern, M. Tingulstad, LG de Oliveira, and T. Zwick proposes a channel estimation scheme based on sparse array assistance, applicable to subarray hybrid beamforming architectures.

[0008] The shortcomings of existing technology are: 1. Existing technologies still employ a decoupled architecture. In near-field spherical wave propagation scenarios, user localization and channel estimation are essentially coupled. This decoupled design not only reduces accuracy but also requires independent pilot design and algorithms, leading to increased resource consumption and complexity, which is particularly detrimental to resource-constrained XL-MIMO systems.

[0009] 2. Traditional methods are prone to performance bottlenecks in high-dimensional, nonlinear, and dynamic channel environments. Summary of the Invention

[0010] To address the aforementioned problems, this invention provides a channel estimation and positioning system and method based on the CP-Mamba architecture. The aim is to improve positioning accuracy and channel estimation performance, reduce computational complexity, and solve the problem of decreased spatial correlation due to increased spacing between sparse array (SA) cells.

[0011] To solve the above problems, the technical solution provided by the present invention is as follows: A channel estimation and localization system based on CP-Mamba architecture includes a pilot signal transmission and reception module, a data preprocessing module, a CP-Mamba network module, a localization result calculation module, and a channel reconstruction module, wherein: The pilot signal transmitting and receiving module includes a base station antenna array configured with a sparse array, used to transmit pilot signals to user equipment through the base station and receive their echoes; The data preprocessing module is used to extract features related to the user space coordinate height based on the received pilot signals and construct a received signal matrix. The CP-Mamba network module includes an encoder, a decoder, and a skip connection layer. The encoder comprises multiple alternately connected encoder convolutional layers and a Mamba module, and is used to extract local and long-range dependency features of the pilot signal. The decoder comprises an upsampling layer, a decoder convolutional layer, and the Mamba module, and is used with the encoder to reconstruct the local coordinates and channel matrix. The skip connection layer is used to fuse multi-scale features between the encoder and the decoder to prevent information loss. The positioning result calculation module is used to calculate the user equipment location and line-of-sight channel; The channel reconstruction module is used to reconstruct a complete channel based on the line-of-sight channel obtained from the positioning results and the non-line-of-sight channel output by the CP-Mamba network module.

[0012] Preferably, the received signal matrix is ​​expressed by the following formula: in: Used to characterize the received signal matrix; Used to characterize the stacked synthesizer matrix; Used to characterize the channel between the base station and the user equipment; Used to characterize the stacked noise vector; The stacked synthesizer matrix is ​​expressed by the following formula: in: Analog synthesizer used to characterize the base station, ; Elements used to characterize the analog synthesizer; Follows a uniform random variable ; Used to characterize the number of radio frequency chains; P is used to characterize the total number of subcarriers in the system, and N is used to characterize the total number of antennas in the array; The stacked noise vector is expressed by the following formula: in: Used to characterize noise terms; Used to characterize the original complex Gaussian noise vector, and follows... distributed.

[0013] Preferably, the channel between the base station and the user equipment includes line-of-sight components and non-line-of-sight components, expressed by the following formula: in: Used to characterize the line-of-sight component; Used to characterize the non-line-of-sight component; Used to characterize the number of clusters between user equipment and base stations; Used to characterize the The number of scatterers in each cluster; Used to characterize the near-field steering vector; Used to characterize the In the cluster, the first Angle of arrival of each scatterer between the base station array reference point and the UE; Used to characterize coordinates antenna To the l-th cluster The distance between the scatterers; Used to characterize user equipment and the first The th cluster The distance between the centers of the scatterer array; Used to characterize the speed of light; Used to characterize the unit response.

[0014] Preferably, the sparse array is a modular array or a nested array, and its total aperture satisfies: in: Used to characterize the total aperture of the modular array; Used to characterize the total aperture of the nested array; The number of modules used to characterize the modular array or the number of dense array antenna elements contained in the nested array; Used to characterize the number of antenna elements contained in each module; Used to characterize the spacing between each module unit; Used to characterize the center spacing of the defined modules; The number of sparse subarray antenna elements used to characterize the nested array.

[0015] Preferably, the Mamba module is constructed based on an improved selective state-space model, and its state updates and output calculations are expressed by the following formula: in: Used to characterize the state matrix; Used to characterize the input matrix; Used to characterize the output matrix; Used to characterize implicit latent states.

[0016] Preferably, the CP-Mamba network module uses mean squared error as the loss function, expressed as follows: in: Used to characterize the real channel; The estimated channel is used to characterize the CP-Mamba network module; Used to characterize the number of samples in a batch.

[0017] Preferably, the CP-Mamba network module adopts a U-shaped convolutional network, and the encoder and the decoder share feature maps through the skip connection layer to improve the channel matrix reconstruction accuracy.

[0018] A method utilizing the aforementioned channel estimation and localization system based on the CP-Mamba architecture includes the following steps: S1. The base station transmits pilot signals and receives echo pilot signals from user equipment, and then constructs the received signal matrix; S2. Input the received signal matrix into the encoder of the CP-Mamba network module, extract local features using the encoder's convolutional layer, and establish long-range dependencies through the Mamba module to obtain the output feature vector; S3. During the positioning phase, the output feature vector is output as user equipment coordinates via the decoder, and the channel between the base station and the user equipment is calculated based on geometric relationships; S4. In the channel estimation stage, the user equipment coordinates output in the positioning stage and the line-of-sight component of the channel determined at that location are used as prior inputs to the CP-Mamba network module, and then the non-line-of-sight component is reconstructed. S5. By superimposing the line-of-sight component and the non-line-of-sight component, a complete channel estimation result is obtained.

[0019] Compared with the prior art, the present invention has the following advantages: 1. The two-stage joint positioning and channel estimation method of the present invention. First, the coordinates of the user equipment are estimated, and then the line-of-sight channel is calculated; then, the non-line-of-sight channel is estimated using the line-of-sight channel and pilot signals, thereby completely reconstructing the channel; this staged method not only significantly improves the positioning accuracy and channel estimation effect, but also greatly reduces the computational complexity.

[0020] 2. In the design of the CP-Mamba joint channel estimation and user localization architecture proposed in this invention, U-Net provides a high-resolution, gradient-stable backbone network, which complements Mamba's global sequence modeling, thereby solving the problem of reduced spatial correlation due to increased spacing between sparse array (SA) cells.

[0021] 3. Numerical simulation results show that, compared with traditional dense arrays (CAs), the present invention exhibits superior performance in localization and channel estimation. Attached Figure Description

[0022] Figure 1 This is a schematic diagram of a sparse XL-MIMO system model according to a specific embodiment of the present invention; Figure 2 These are schematic diagrams illustrating different array architectures in specific embodiments of the present invention; Figure 3 This is a schematic diagram of the two-stage framework for channel estimation and localization in a specific embodiment of the present invention; Figure 4 This is a schematic diagram of the architecture of the Mamba module in a specific embodiment of the present invention; Figure 5 This is a schematic diagram of the CP-Mamba network backbone according to a specific embodiment of the present invention; Figure 6 This is a schematic diagram illustrating the MPE performance of different antenna arrays of CP-Mamba according to a specific embodiment of the present invention; Figure 7 This is a schematic diagram illustrating the MPE performance of nested arrays under different methods in specific embodiments of the present invention; Figure 8 This is a schematic diagram illustrating the NMSE performance of nested arrays under different methods in specific embodiments of the present invention. Detailed Implementation

[0023] The present invention will be further illustrated below with reference to specific embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. After reading the present invention, any modifications of the present invention in various equivalent forms by those skilled in the art will fall within the scope defined by the appended claims.

[0024] This invention application claims a channel estimation and positioning system based on the CP-Mamba architecture, comprising a pilot signal transmission and reception module, a data preprocessing module, a CP-Mamba network module, a positioning result calculation module, and a channel reconstruction module, wherein: The pilot signal transmission and reception module includes a base station antenna array configured with a sparse array, used to transmit pilot signals to user equipment through the base station and receive their echoes; The data preprocessing module is used to extract features related to the user space coordinate height based on the received pilot signals and to construct the received signal matrix; The CP-Mamba network module includes an encoder, a decoder, and a skip connection layer. The encoder contains multiple alternating encoder convolutional layers and Mamba modules. The encoder is used to extract local features and long-range dependency features of the pilot signal. The decoder contains an upsampling layer, a decoder convolutional layer, and a Mamba module. The decoder and encoder are used to reconstruct the local coordinates and channel matrix. The skip connection layer is used to fuse multi-scale features between the encoder and decoder to prevent information loss. The positioning result calculation module is used to calculate the location of user equipment and the line-of-sight channel; The channel reconstruction module is used to reconstruct a complete channel based on the line-of-sight channel obtained from the positioning results, combined with the non-line-of-sight channel output by the CP-Mamba network module.

[0025] It should be noted that, as Figure 1 As shown, this invention is based on a sparse XL-MIMO OFDM scenario: a base station (BS) is equipped with N sparse antenna elements (SAs) to provide services to a single-antenna user equipment (UE). The center of the base station array is located at the origin, with only N radio frequency links configured (NRF < N), and K OFDM subcarriers are allocated to the UE. Assuming there are a total of K orthogonal subcarriers, the frequency of the k-th subcarrier is expressed by Equation 1: (1) Where: f c B and B are used to characterize the carrier frequency and bandwidth, respectively; Used to characterize the carrier wavelength. User equipment is randomly distributed in the front half-plane of the base station, with a polar coordinate radius of r and an azimuth angle of r. This indicates the user's orientation relative to the base station. The coordinates of the user equipment are expressed according to Equation 2: (2).

[0026] It needs to be further explained that, such as Figure 2 The diagram shows typical array (CA), uniform sparse array (USA), and non-uniform array (NUSA) architectures, which also include modular arrays (MOA) and nested arrays (NA). Specifically: (1) Typical Array (CA): Traditional CA uses antenna elements arranged in a uniform linear pattern, with the spacing between each element being the operating wavelength. Half of it. Therefore, the total aperture of a typical array containing N elements is expressed by Equation 3: (3).

[0027] (2) Plain array (USA): All cell spacing is 0. sparse parameters It is important to note that when In this case, the typical array is a special case of the plain array.

[0028] (3) Modular Array (MOA): This array contains There are 1 module, and each module contains 1 module. Total antenna elements, Each module has one unit. The spacing between each module unit is... The module size is expressed according to Formula 4: (4).

[0029] Define the center spacing of the modules as ( The total aperture is expressed by Equation 5: (5).

[0030] (4) Nested Array (NA): A two-layer NA consists of inner layers spaced at intervals of 1 / 2. of A uniform linear subarray and the outer layer spacing of Composed of sparse subarrays. This hierarchical design improves spatial resolution while maintaining structural coherence. The physical dimensions of the inner and outer arrays are expressed by Equations 6 and 7, respectively: (6) (7) The total aperture is expressed according to Formula 8: (8).

[0031] It should be further noted that, due to the sparse array (SA) spacing exceeding This can produce grating lobes and angular blurring effects. In scenarios where communication reliability and beam control are paramount, typical arrays are generally more suitable for channel estimation tasks. However, larger spacing improves angular resolution, making sparse arrays (SAs) more suitable for positioning tasks. In integrated sensing and communication (ISAC) scenarios, sparse arrays (SAs) are more attractive than sparse arrays (CAs) due to their significant advantages in hardware cost, mutual coupling control, and spatial multiplexing.

[0032] In this specific embodiment, the sparse array is a modular array or a nested array, and its total aperture satisfies Equations 9 and 10: (9) (10) in: Used to characterize the total aperture of the modular array; Used to characterize the total aperture of the nested array; The number of modules used to characterize a modular array or the number of dense array antenna elements contained in a nested array. Used to characterize the number of antenna elements contained in each module; Used to characterize the spacing between each module unit; Used to characterize the center spacing of the defined modules; Used to characterize the number of sparse subarray antenna elements in a nested array.

[0033] It should be noted that as the array spacing increases, the Rayleigh distance and near-field region expand accordingly. In such scenarios, near-field channel characteristics are mainly affected by the angle of arrival (AoA) and distance. To comprehensively simulate the near-field propagation characteristics in a very large-scale array system, this invention uses a near-field channel model for modeling. It is assumed that a two-layer near-field array (NA) or a multi-layer near-field array (MOA) is used to serve a system with known polar coordinates. A single-antenna user equipment. In an NA configuration, the array element... The position coordinates are expressed according to Equation 11: (11) in: and ; In MOA configuration, array unit The position coordinates are expressed according to Equation 12: (12) in: and .

[0034] Because the aperture of the SA is relatively large, the Rayleigh distance increases accordingly, causing a large number of user equipment unit links to fall within the electromagnetic near-field range. This invention uses a non-uniform spherical wave model for modeling.

[0035] It should be further explained that the channel between the base station and the user equipment includes line-of-sight components and non-line-of-sight components, expressed by equations 13, 14, and 15: (13) (14) (15) in: Used to characterize the line-of-sight component; Used to characterize non-line-of-sight components; Used to characterize the number of clusters between user equipment and base stations; Used to characterize the The number of scatterers in each cluster; Used to characterize the near-field steering vector; Used to characterize the In the cluster, the first Angle of arrival of each scatterer between the base station array reference point and the UE; Used to characterize coordinates antenna To the l-th cluster The distance between the scatterers; Used to characterize user equipment and the first The th cluster The distance between the centers of the scatterer array; Used to characterize the speed of light; Used to characterize the unit response.

[0036] In this specific embodiment, only retain The line-of-sight (LoS) term is expressed as the channel gain according to Equation 16: (16) in: Used to characterize the distance from user equipment to base station antenna; Used to characterize the path loss factor. Small-scale complex channel gain used to characterize non-line-of-sight (NLoS) paths.

[0037] It should be noted that in the first During the first time slot, the first Uplink pilot signal received on each subcarrier Expressed according to Equation 17: (17) in: Pilot symbols used to characterize the transmissions of user equipment (UE). At that time, the received signal matrix is ​​expressed according to Equation 1: (18) in: Used to characterize the received signal matrix; Used to characterize the stacked synthesizer matrix; Used to characterize the channel between the base station and the user equipment; Used to characterize the stacked noise vector; The stacked synthesizer matrix is ​​expressed according to Equations 19 and 20: (19) (20) in: Analog synthesizer used to characterize the base station, ; Elements used to characterize analog synthesizers; Follows a uniform random variable ; Used to characterize the number of radio frequency chains; P is used to characterize the total number of subcarriers in the system, and N is used to characterize the total number of antennas in the array; The stacked noise vector is expressed by equations 21 and 22: (twenty one) (twenty two) in: Used to characterize noise terms; Used to characterize the original complex Gaussian noise vector, and follows... distributed.

[0038] It should be noted that under near-field spherical wave propagation conditions, user localization and channel estimation are essentially strongly coupled problems. Traditional methods that separate the two not only reduce accuracy but also require separate pilot and algorithm design for each task, increasing system resource overhead and implementation complexity. This drawback is particularly pronounced in resource-constrained ultra-large-scale MIMO systems. Meanwhile, while sparse arrays (SAs) can improve angular resolution by increasing array aperture, larger element spacing leads to reduced spatial correlation and effects such as grating lobes and angular ambiguity, resulting in decreased signal resolution and uncertain angle estimation. This means that in sparse XL-MIMO near-field scenarios, channel estimation faces challenges such as low spatial correlation, angular ambiguity, and signal gain fluctuations, requiring more advanced methods to overcome these challenges. Therefore, this invention proposes a two-stage method to reduce channel estimation errors caused by antenna sparsity. To address the weakened spatial correlation caused by increased element spacing in spatially adaptive arrays (SAs), we design a CP-Mamba architecture by leveraging the advantages of U-Net in local feature extraction and Mamba's ability to capture global dependencies.

[0039] It needs to be further explained that, such as Figure 3 As shown, this invention constructs a two-stage collaborative processing framework to simultaneously improve user positioning accuracy and channel estimation quality. In the first stage (positioning stage), the CP-Mamba network extracts features related to the user's spatial coordinate height from the echo pilot signal sent by the base station and reflected by the user equipment. Using the extracted features and combined with formula (15), the network can predict the three-dimensional coordinates of the user equipment. Because sparse arrays have a larger aperture and higher angular resolution, the localization phase primarily relies on the line-of-sight (LoS) path to determine the UE's position, extracting precise information from the LoS signal. Once the user's position estimate is obtained... Based on the geometric relationship, the corresponding LoS channel gain and phase distribution, i.e., the line-of-sight channel response, can be calculated using formula (13). Next, the second stage (channel estimation stage) begins, where the CP-Mamba model calculates the user location output from the positioning stage. and the LoS channel determined by that location. This prior knowledge is injected into the network input to reconstruct the non-line-of-sight (NLoS) channel components. Since the LosS component has already been determined by the localization results, the network only needs to estimate the remaining NLoS component at this stage, greatly reducing the dimensionality of unknown parameters and the required pilot overhead. By combining the reconstructed NLoS component with the known LosS channel, a complete downlink channel estimate can be obtained. + This two-stage process fully leverages the guiding role of location information in channel estimation, achieving synergistic gains for both tasks: the prior knowledge provided by location makes channel estimation more accurate, while accurate channel estimation, in turn, verifies the validity of the location results. This framework effectively reduces channel estimation errors caused by antenna sparsity and improves the reliability of location and channel estimation in near-field environments.

[0040] It should be noted that, in order to effectively address near-field channel interference and establish a long-range signal dependence model, the Mamba module of this invention is built based on an improved selective state-space model (SSM). This module improves upon the classic continuous-time SSM, and its state update and output calculation are expressed by equations 23 and 24: (twenty three) (twenty four) in: Used to characterize the state matrix; Used to characterize the input matrix; Used to characterize the output matrix; Used to characterize implicit latent states.

[0041] It needs to be further explained that, such as Figure 4 The specific design of the Mamba module is shown below. Assume the input tensor is... ,in and These represent the batch size, sequence length, and feature dimension, respectively. The model employs a two-branch dynamic architecture to achieve efficient long-distance dependency modeling. First, x is decomposed into parallel features through linear projection. .

[0042] In the main branch, After activation by depthwise convolution (CONV) and sigmoid weighted linear units (SiLU), time-varying parameters are generated through joint parameterization. and state matrix Time coefficient Discretization is performed using a projection layer derived from the Softplus map to generate ∆. Subsequently, selective scan operators are integrated. and Aggregating multi-scale spatiotemporal features into Meanwhile, the auxiliary branch is adjusted via sigmoid gating. ,generate Ultimately, linear projection restores the dimensions to [the original dimensions]. .

[0043] It should be noted that, based on the aforementioned Mamba module, this invention designs a CP-Mamba network architecture as the core model for joint localization and channel estimation. The CP-Mamba network module adopts a U-shaped convolutional network structure, and the encoder and decoder share feature maps through skip connection layers to improve the accuracy of channel matrix reconstruction.

[0044] It should be further explained that the network backbone of the CP-Mamba network module, such as... Figure 5 As shown. This architecture integrates the encoder-decoder topology and skip connection mechanism of the U-Net

[16] convolutional network, as well as the global sequence dependency modeling capability that the Mamba module excels at. In the CP-Mamba architecture, the encoder contains several cascaded convolutional layers and Mamba layers. In the encoder network, the first encoder module uses a convolutional layer to extract local features; the subsequent encoder modules downsample the spatial dimension and increase the number of feature channels through a convolutional layer with a stride of 2, while the Mamba module focuses on capturing fine local details and improving resolution recovery capability. In the decoder network, an upsampling layer is applied to restore the spatial resolution of the feature map, and finally another convolutional layer is used as the output layer of the decoder network. The skip connection between the encoder and decoder in the U-shaped structure effectively improves the feature reuse efficiency and avoids the loss of detailed information during high-level feature extraction, thereby achieving higher accuracy in both localization coordinates and channel matrix reconstruction.

[0045] A method utilizing a channel estimation and localization system based on the CP-Mamba architecture includes the following steps: S1. The base station sends pilot signals and receives echo pilot signals from user equipment, and then constructs a received signal matrix; S2. Input the received signal matrix into the encoder of the CP-Mamba network module, extract local features using the encoder's convolutional layer, and establish long-range dependencies through the Mamba module to obtain the output feature vector; S3. During the positioning phase, the output feature vector is decoded to output the coordinates of the user equipment, and the channel between the base station and the user equipment is calculated based on the geometric relationship; S4. In the channel estimation stage, the user equipment coordinates output in the positioning stage and the line-of-sight components of the channel determined at that location are used as prior inputs to the CP-Mamba network module, and then the non-line-of-sight components are reconstructed. S5. By superimposing the line-of-sight component and the non-line-of-sight component, the complete channel estimation result is obtained.

[0046] It should be noted that the goal of the positioning phase is to estimate the coordinates of the user equipment (UE) using pilot signals transmitted by the base station. During this phase, the encoder receives pilot input. And output feature vectors (Through Equation 24), the vector is then processed by the decoder and enters the pooling layer to finally predict the coordinates of the user device. .

[0047] The localization phase primarily relies on line-of-sight (LoS) paths to determine coordinates. Once the user equipment's location is obtained, the line-of-sight channel can be calculated using geometric relationships. In contrast, the channel estimation phase requires consideration of both line-of-sight and non-line-of-sight paths to accurately reconstruct the complete channel.

[0048] During the channel estimation phase, the encoder uses the predicted user equipment coordinates and line-of-sight channel As input, feature vectors are generated. This vector is then passed to the decoder to reconstruct the non-line-of-sight components. By combining the reconstructed non-line-of-sight components with the line-of-sight channel, the complete channel can be obtained. .

[0049] It should be noted that the CP-Mamba network module uses mean squared error as the loss function to learn the latent representation of high-dimensional data, expressed as Equation 25: (25) in: Used to characterize the real channel; Estimated channels used to characterize CP-Mamba network modules; Used to characterize the number of samples in a batch.

[0050] In this specific embodiment, the parameters are set as follows: , and MHz. Assume user equipment (UE) is arranged around the base station, with the following locations: and This invention evaluated four array configurations: CA, USA, NA, and MOA, each using 20,000 training samples. The NA array employs a two-level structure: the first level contains 4 units, and the second level contains 124 units. The MOA array consists of 16 modules, each containing 8 units and a specific size. This approach balances spatial sampling optimization with feasibility. Positioning performance is evaluated using mean square error (MPE), expressed as Equation 26: (26).

[0051] The channel estimation performance is expressed using the normalized mean square error (NMSE), as shown in Equation 27: (27).

[0052] like Figure 6 As shown in this specific embodiment, the performance of the proposed CP-Mamba-based method in line-of-sight (LoS) scenarios was evaluated. The test environment included two configurations: co-located (CA) and scattered antennas (SAs). Under the same signal-to-noise ratio conditions, the scattered antenna, with its large aperture advantage, significantly outperformed the co-located scheme in positioning accuracy. Figure 7 The results show a comparison of the CP-Mamba method with various benchmark schemes, including deep learning-based wireless networks and ultra-large-scale linear programming models, employing an asymmetric architecture. The results demonstrate that CP-Mamba maintains the highest positioning accuracy across all signal-to-noise ratio levels.

[0053] Compared to other benchmark methods, CP-Mamba not only has fewer parameters and lower FLOPs consumption, but also has superior channel estimation performance, fully demonstrating the efficiency and scalability of the method.

[0054] like Figure 8 As shown, this specific embodiment compares the Normalized Mean Square Error (NMSE) performance of CP-Mamba with the benchmark method. Under all signal-to-noise ratio conditions, CP-Mamba outperforms the benchmark scheme in channel estimation. This advantage stems from the positioning phase of our scheme—which provides relatively accurate user equipment (UE) location estimation, a capability not possessed by other methods, thereby significantly improving channel estimation accuracy by integrating UE location information.

[0055] It should be noted that this invention proposes a two-stage joint localization and channel estimation framework based on CP-Mamba. The first stage locates the user equipment coordinates, providing data support for channel estimation in the second stage, thereby achieving coordinated optimization of localization and channel estimation. To this end, this invention designs a CP-Mamba network architecture: using U-Net for local feature extraction and combining it with Mamba modules to model long-range dependencies. To evaluate the impact of diversity antennas (SA), three diversity configuration schemes are studied in this specific embodiment. Simulation results show that this framework outperforms existing baseline methods in both localization accuracy and channel estimation performance, and diversity antennas consistently outperform co-located antennas (CA) in both tasks.

[0056] In the above detailed description, various features are combined together in a single embodiment to simplify this disclosure. This approach to disclosure should not be construed as reflecting an intention that embodiments of the claimed subject matter require more features than are explicitly stated in each claim. Rather, as reflected in the appended claims, the invention is presented with fewer features than all of the features of the single disclosed embodiment. Therefore, the appended claims are hereby explicitly incorporated into the detailed description, wherein each claim stands alone as a preferred embodiment of the invention.

[0057] The disclosed embodiments have been described above to enable any person skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein can be applied to other embodiments without departing from the spirit and scope of this disclosure. Therefore, this disclosure is not limited to the embodiments given herein, but is consistent with the broadest scope of the principles and novel features disclosed in this application.

[0058] The foregoing description includes examples of one or more embodiments. It is certainly impossible to describe all possible combinations of components or methods in order to describe the above embodiments, but those skilled in the art will recognize that further combinations and arrangements of the various embodiments are possible. Therefore, the embodiments described herein are intended to cover all such changes, modifications, and variations that fall within the scope of the appended claims. Furthermore, the term "comprising" as used in the specification or claims is interpreted in a manner similar to the term "including," as interpreted when used as a conjunction in the claims. Additionally, the use of any term "or" in the specification of the claims is intended to mean "non-exclusive or."

[0059] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A channel estimation and localization system based on CP-Mamba architecture, characterized in that: It includes a pilot signal transmission and reception module, a data preprocessing module, a CP-Mamba network module, a positioning result calculation module, and a channel reconstruction module, wherein: The pilot signal transmitting and receiving module includes a base station antenna array configured with a sparse array, used to transmit pilot signals to user equipment through the base station and receive their echoes; The data preprocessing module is used to extract features related to the user space coordinate height based on the received pilot signals and construct a received signal matrix. The CP-Mamba network module includes an encoder, a decoder, and a skip connection layer. The encoder comprises multiple alternately connected encoder convolutional layers and a Mamba module, and is used to extract local and long-range dependency features of the pilot signal. The decoder comprises an upsampling layer, a decoder convolutional layer, and the Mamba module, and is used with the encoder to reconstruct the local coordinates and channel matrix. The skip connection layer is used to fuse multi-scale features between the encoder and the decoder to prevent information loss. The positioning result calculation module is used to calculate the user equipment location and line-of-sight channel; The channel reconstruction module is used to reconstruct a complete channel based on the line-of-sight channel obtained from the positioning results and the non-line-of-sight channel output by the CP-Mamba network module.

2. The channel estimation and localization system based on CP-Mamba architecture according to claim 1, characterized in that: The received signal matrix is ​​expressed by the following formula: in: Used to characterize the received signal matrix; Used to characterize the stacked synthesizer matrix; Used to characterize the channel between the base station and the user equipment; Used to characterize the stacked noise vector; The stacked synthesizer matrix is ​​expressed by the following formula: in: Analog synthesizer used to characterize the base station, ; Elements used to characterize the analog synthesizer; Follows a uniform random variable ; Used to characterize the number of radio frequency chains; P is used to characterize the total number of subcarriers in the system, and N is used to characterize the total number of antennas in the array; The stacked noise vector is expressed by the following formula: in: Used to characterize noise terms; Used to characterize the original complex Gaussian noise vector, and follows... distributed.

3. The channel estimation and localization system based on CP-Mamba architecture according to claim 2, characterized in that: The channel between the base station and the user equipment includes line-of-sight components and non-line-of-sight components, expressed by the following formula: in: Used to characterize the line-of-sight component; Used to characterize the non-line-of-sight component; Used to characterize the number of clusters between user equipment and base stations; Used to characterize the The number of scatterers in each cluster; Used to characterize the near-field steering vector; Used to characterize the In the cluster, the first Angle of arrival of each scatterer between the base station array reference point and the UE; Used to characterize coordinates antenna To the l-th cluster The distance between the scatterers; Used to characterize user equipment and the first The th cluster The distance between the centers of the scatterer array; Used to characterize the speed of light; Used to characterize the unit response.

4. The channel estimation and localization system based on CP-Mamba architecture according to claim 3, characterized in that: The sparse array is a modular array or a nested array, and its total aperture satisfies: in: Used to characterize the total aperture of the modular array; Used to characterize the total aperture of the nested array; The number of modules used to characterize the modular array or the number of dense array antenna elements contained in the nested array; Used to characterize the number of antenna elements contained in each module; Used to characterize the spacing between each module unit; Used to characterize the center spacing of the defined modules; The number of sparse subarray antenna elements used to characterize the nested array.

5. The channel estimation and localization system based on CP-Mamba architecture according to claim 4, characterized in that: The Mamba module is built based on an improved selective state-space model, and its state updates and output calculations are expressed by the following formula: in: Used to characterize the state matrix; Used to characterize the input matrix; Used to characterize the output matrix; Used to characterize implicit latent states.

6. The channel estimation and localization system based on CP-Mamba architecture according to claim 5, characterized in that: The CP-Mamba network module uses mean squared error as the loss function, expressed as follows: in: Used to characterize the real channel; The estimated channel is used to characterize the CP-Mamba network module; Used to characterize the number of samples in a batch.

7. The channel estimation and localization system based on CP-Mamba architecture according to claim 6, characterized in that: The CP-Mamba network module adopts a U-shaped convolutional network, and the encoder and the decoder share feature maps through the skip connection layer to improve the channel matrix reconstruction accuracy.

8. A method for channel estimation and localization system based on CP-Mamba architecture as described in any one of claims 1 to 7, characterized in that: Includes the following steps: S1. The base station transmits pilot signals and receives echo pilot signals from user equipment, and then constructs the received signal matrix; S2. Input the received signal matrix into the encoder of the CP-Mamba network module, extract local features using the encoder's convolutional layer, and establish long-range dependencies through the Mamba module to obtain the output feature vector; S3. During the positioning phase, the output feature vector is output as user equipment coordinates via the decoder, and the channel between the base station and the user equipment is calculated based on geometric relationships; S4. In the channel estimation stage, the user equipment coordinates output in the positioning stage and the line-of-sight component of the channel determined at that location are used as prior inputs to the CP-Mamba network module, and then the non-line-of-sight component is reconstructed. S5. By superimposing the line-of-sight component and the non-line-of-sight component, a complete channel estimation result is obtained.

Images