A near-field beam training method and apparatus for a single-user distributed MIMO system

Abstract

This invention discloses a near-field beam training method and apparatus for a single-user distributed MIMO system. Based on the far-field hierarchical codebook of each access point (AP) in the distributed MIMO system, far-field hierarchical beam training is performed on the corresponding APs to obtain the beam direction of each AP pointing towards the user. The user's position is then filtered based on several beam directions to obtain the user's position information relative to the distributed MIMO system. This invention uses far-field codebooks for beam training in distributed MIMO systems to estimate user angles, and then uses geometric relationships to determine the user's position based on these angles. This method only requires time resources for user angle estimation, significantly reducing beam training overhead compared to existing near-field beam training techniques.

Description

Technical Field

[0001] This invention belongs to the field of near-field beam training technology, and particularly relates to a near-field beam training method and apparatus for a single-user distributed MIMO system. Background Technology

[0002] To meet the demands of 6G communication for ultra-high speed, ultra-low latency, ultra-high reliability, and massive connectivity, distributed multiple-input multiple-output (MIMO) has attracted widespread attention. Compared to centralized MIMO systems, distributed MIMO systems offer significant advantages in terms of interference resistance and coverage. By deploying a large number of access points (APs) in different geographical locations for signal processing and collaboration, they provide stronger interference resistance and wider signal coverage, thereby improving the reliability and performance of the communication system.

[0003] Currently, with the significant increase in the number of antennas and carrier frequencies in 6G systems, the Rayleigh distance is significantly extended, and near-field characteristics become more pronounced. This renders traditional far-field beam training methods unsuitable for near-field MIMO channels. In centralized MIMO systems, inspired by the near-field beam focusing phenomenon, a polar-domain codebook is constructed to divide the near-field space into a two-dimensional grid of angles and distances. The beam training overhead is the number of these two-dimensional grids. Moreover, directly adopting the beam training methods of centralized MIMO systems in distributed MIMO systems would result in even greater training overhead. Summary of the Invention

[0004] The purpose of this invention is to provide a near-field beam training method and apparatus for a single-user distributed MIMO system, so as to reduce the near-field beam training overhead of the distributed MIMO system.

[0005] This invention adopts the following technical solution: a near-field beam training method for a single-user distributed MIMO system, comprising the following steps:

[0006] Based on the far-field hierarchical codebook of each AP in the distributed MIMO system, far-field hierarchical beam training is performed on the corresponding AP to obtain the beam direction of each AP pointing to the user.

[0007] The user's location is filtered based on several beam directions to obtain the user's location information relative to the distributed MIMO system.

[0008] Furthermore, after obtaining the user's location information relative to the distributed MIMO system, the process also includes:

[0009] The optimal near-field codeword for each AP is calculated based on its location information.

[0010] Furthermore, filtering the user's location based on several beam directions includes:

[0011] The user's location is filtered based on the geometric relationship of several beam directions.

[0012] Furthermore, filtering the user's location based on the geometric relationships of several beam directions includes:

[0013] Several intersection points are determined based on several beam directions;

[0014] Based on the location information of each intersection point in the service area of ​​the distributed MIMO system, intersection points in the location-dense area are selected to form an intersection point set.

[0015] Calculate the average location information based on the location information of the intersection points in the intersection point set;

[0016] The average location information is used as the user's location information relative to the distributed MIMO system.

[0017] Furthermore, before determining several intersection points based on several beam directions, the process also includes:

[0018] Generate a symmetrical beam direction that is symmetrical about the corresponding AP based on the beam direction;

[0019] Choose a beam direction or symmetrical beam direction for each AP as the final beam direction of the AP.

[0020] Furthermore, selecting a beam direction or symmetrical beam direction for each AP includes:

[0021] Generate 2 based on the beam direction or symmetrical beam direction of each AP. M A combination of beam directions;

[0022] In beam direction combination, the optimal beam direction combination is selected based on the average distance between the intersection points formed by the beam directions;

[0023] For each AP, a beam direction or a symmetrical beam direction is selected based on the optimal beam direction combination.

[0024] Furthermore, in beam direction combination, selecting the optimal beam direction combination based on the average distance between the intersection points formed by the beam directions includes:

[0025] The beam direction combination corresponding to the minimum average distance is selected as the optimal beam direction combination.

[0026] Furthermore, selecting the beam direction combination corresponding to the minimum average distance as the optimal beam direction combination includes:

[0027] The minimum value among the average distances is selected based on the KD-Tree.

[0028] Furthermore, a density-based clustering algorithm is used to filter out intersections in densely located regions.

[0029] Another technical solution of the present invention: a near-field beam training device for a single-user distributed MIMO system, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-described method.

[0030] The beneficial effects of this invention are: This invention uses far-field codebooks for beam training in distributed MIMO systems to estimate user angles, and then uses geometric relationships to determine the user's position based on the user angles; this method only requires time resources in the user angle estimation process, which reduces beam training overhead compared to the hierarchical near-field beam training in the prior art. Attached Figure Description

[0031] Figure 1 This is a downlink diagram of a single-user distributed MIMO system in an embodiment of the present invention;

[0032] Figure 2 This is a schematic diagram of the far-field beam training process based on the angular domain in an embodiment of the present invention;

[0033] Figure 3 This is a flowchart of the user location filtering process in an embodiment of the present invention;

[0034] Figure 4 This is a performance analysis diagram of the achievable rate versus signal-to-noise ratio for different beam training methods in embodiments of the present invention. Detailed Implementation

[0035] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments.

[0036] Existing research on near-field beam training mainly focuses on centralized MIMO systems. However, directly adopting the beam training method of centralized MIMO systems in distributed MIMO systems would result in a much larger training overhead and would not fully utilize the advantages of multiple APs distributed in space. Therefore, how to design a near-field beam training method for single-user distributed MIMO systems has become an urgent problem to be solved.

[0037] This invention discloses a near-field beam training method for a single-user distributed MIMO system, comprising the following steps: performing far-field layered beam training on the corresponding APs based on the far-field layered codebook of each AP in the distributed MIMO system to obtain the beam direction of each AP pointing to the user; filtering the user's position according to several beam directions to obtain the user's position information relative to the distributed MIMO system.

[0038] This invention addresses the use of far-field codebooks for beam training in distributed MIMO systems to estimate user angles, and then uses geometric relationships to determine the user's position based on those angles. This method only requires time resources for user angle estimation, reducing beam training overhead compared to hierarchical near-field beam training in existing technologies.

[0039] After obtaining the user's location information relative to the distributed MIMO system, this invention further includes: calculating the near-field optimal codeword for each AP based on the location information. Therefore, this invention uses a far-field codebook to perform beam training on the APs of the distributed MIMO system in the near-field environment. When the user's location is determined, the corresponding near-field optimal codeword is calculated based on the user's location information for data transmission, which can significantly reduce the beam training overhead during beam training.

[0040] In this invention, a single cell is used as the service area of ​​the distributed MIMO system, such as... Figure 1 As shown, consider a single-cell distributed MIMO near-field downlink system with M access points (APs). In the case of circular AP deployment, all APs are uniformly distributed along a circular trajectory of radius R. A service area coordinate system is established with the center of the service area as the origin, any direction as the x-axis, and the direction perpendicular to the x-axis as the y-axis. In this coordinate system, the polar coordinates of the m-th AP are (R, φ). m )=(R,2π(m-1) / M), rectangular coordinates are (Rcosφ) m ,Rsinφ m ), m=1,2,…,M,φ m This represents the deployment angle of m APs in a circular trajectory.

[0041] Each access point (AP) is deployed with a uniform linear array (ULA) consisting of N antennas. The total number of transmit antennas in the distributed MIMO system is N. t =MN, and the spacing between two adjacent antennas in each AP is equal to half a wavelength. In this embodiment of the invention, each ULA array is arranged along the y-axis in the service area coordinate system, then the coordinate of the nth antenna of the mth AP is (Rcosφ). m ,Rsinφ m +δ n d), where Let λ be half the wavelength, and λ be the wavelength. All access points (APs) simultaneously serve a single-antenna user. The signal received by the user is:

[0042]

[0043] Where x represents the transmission symbol of the AP, This represents the channel matrix between the user and the m-th AP. This represents the beam steering vector at the m-th AP, which is essentially a codeword selected from a predefined near-field codebook. The power is represented by σ. 2 Additive white Gaussian noise. Beam training involves measuring the power of the received signal y and obtaining the optimal codeword through a predefined codebook or corresponding calculations.

[0044] Since XL-MIMO channels are typically dominated by several main paths, it is only necessary to search for the physical location of the main paths through beam training, without needing to obtain explicit channel information. Therefore, this invention focuses on near-field LOS channels to find the optimal beam steering vector aligned with the main paths.

[0045] If the user's polar coordinates are (r, θ), where r represents the distance between the user and the origin of the service area coordinate system, and θ represents the angle of the line connecting the user and the origin relative to the x-axis, then the user's relative coordinates to the m-th AP are (x...). m ,y m )=(rcosθ-Rcosφ m ,rsinθ-Rsinφ m Then the polar coordinates of the user relative to the m-th AP can be represented as: Based on the spherical wavefront propagation model, the near-field Loss channel (i.e., channel matrix) from the m-th AP to the user can be modeled as follows:

[0046]

[0047] in, Let b be the complex path gain of the m-th AP. m (r,θ) represents the near-field steering vector of the m-th AP, i.e.:

[0048]

[0049] in, This represents the distance between the nth antenna of the mth AP and the user. Based on the above analysis, the channel matrix... It can be represented in the following form:

[0050]

[0051] Since near-field communication needs to consider not only the angle domain but also the distance domain, beam training methods used in far-field communication cannot be directly applied to near-field communication.

[0052] In this invention, the relative direction of the user is estimated first, that is, each AP estimates the relative direction of the user based on far-field beam training.

[0053] Specifically, the angular domain-based hierarchical beam training is as follows: Figure 2As shown, for the m-th AP, the codewords of each layer in the far-field codebook should cover a specific region. Based on the idea of ​​binary search, the coverage area of ​​the upper-layer codewords is the union of the coverage areas of the lower-layer codewords. For the l-th layer codeword, the N at the center of the activation array... l =2 l 1 antenna, N l The smaller the value, the greater the probability that the user is located in the far field of this AP. The l-th layer has a total of N values. l Each codeword points to a specific character. set up Let m represent the set of layer l codewords that call the m-th AP. Then:

[0054]

[0055] in, Let b(∞,θ) = a(θ) represent the i-th codeword of the l-th layer of the m-th AP, where b(∞,θ) = a(θ) represents the far-field steering vector.

[0056] Using a binary tree-based beam search, hierarchical beam training of the angular domain of the m-th AP can be achieved. Assume... Let i represent the optimal codeword of the m-th AP at layer l. * To obtain the optimal codeword index, the angle of the user relative to the m-th AP can be estimated as follows: The actual user is located at Within the angular range. Therefore, each layer of training requires 2 training overheads (i.e., the time resources occupied), so the total overhead for the m-th AP after L layers of training is 2L.

[0057] The beam direction of each AP to the user can be obtained by the above method. With the known position of the AP, the next step is to solve for the intersection point generated by the two beam directions. Finally, the steering vector is substituted to derive the optimal near-field codeword.

[0058] In one embodiment, since both the beam direction and the AP location are within the service area of ​​the distributed MIMO system, belonging to a straight line (i.e., beam direction) and a point (i.e., AP location and user location) in the service area coordinate system, the user's location can be filtered based on the geometric relationship of several beam directions.

[0059] Specifically, Figure 3Figure (a) shows the obtained beam direction of each AP, where k1, k2, k3, and k4 represent the slopes of the beam directions of the four APs, respectively. Since the array beam pattern of the APs is symmetrical in 360°, the identified direction has a certain probability of being symmetrical about the array. Therefore, to improve accuracy, each AP uses two slopes that are opposite to each other. Thus, two straight lines can be obtained at each AP, that is, a symmetrical beam direction symmetrical about the corresponding AP is generated based on the beam direction.

[0060] Next, a beam direction or symmetrical beam direction needs to be selected for each AP as the final beam direction of the AP. For example... Figure 3 As shown in (b), these lines, arranged in a certain way, have a total of 2 M There are several combinations, where -k1, -k2, -k3, and -k4 represent the slopes of the symmetrical beam directions of the four APs. In other words, two combinations are generated based on the beam direction or symmetrical beam direction of each AP. M A beam direction combination; where M represents the number of APs in the distributed MIMO system.

[0061] To select the set of lines that best meets the requirements, it's easy to see that this set of lines produces a denser network of intersections, such as... Figure 3 As shown in (c), the line combination that meets the conditions can be screened by minimizing the average distance between the intersection points. That is, in the beam direction combination, the optimal beam direction combination is selected according to the average distance between the intersection points formed by the beam directions; and the beam direction or symmetrical beam direction is selected for each AP according to the optimal beam direction combination.

[0062] More specifically, selecting the optimal beam direction combination based on the average distance between the intersection points formed by the beam directions in beam direction combination includes: selecting the beam direction combination corresponding to the minimum average distance as the optimal beam direction combination. Specifically, two different beam directions will intersect in the service area of ​​a distributed MIMO system, and there will be a distance between any two intersection points. The aforementioned average distance refers to the average distance obtained by averaging all distances. Next, each beam direction combination generates an average distance, and this invention selects the beam direction combination corresponding to the minimum average distance among multiple combinations as the final beam direction combination.

[0063] In this embodiment of the invention, the minimum value among the average distances is selected based on a KD-Tree. A KD-Tree is an efficient multidimensional spatial data structure, particularly suitable for nearest neighbor searches and range queries on low-dimensional data. By appropriately selecting the splitting axis and splitting points, a balanced tree structure can be constructed, thereby accelerating query operations.

[0064] Then, several intersection points are determined based on several beam directions; based on the location information of each intersection point in the service area of ​​the distributed MIMO system, intersection points in the location-dense area are selected to form an intersection point set; the average location information is calculated based on the location information of the intersection points in the intersection point set; and the average location information is used as the user's location information relative to the distributed MIMO system.

[0065] Because users are far from certain access points (APs), the intersection points they make with other straight lines have large errors. Therefore, this embodiment uses a density-based clustering algorithm to filter intersection points in densely populated areas. More specifically, the density-based DBSCAN clustering algorithm is selected to filter out a subset of intersection points to form an intersection point set, such as... Figure 3 As shown in (d).

[0066] Each intersection point in the intersection set has coordinate values ​​in the service area coordinate system. Therefore, in this embodiment, the average of these coordinate values ​​(r', θ') is taken as the final user location information. Substituting this into the near-field steering vector yields the user's optimal near-field codeword for each AP, which can be expressed as:

[0067]

[0068] Among them, v m =b(r',θ'), v m This represents the best near-field codeword for the m-th AP.

[0069] In summary, this invention mainly achieves user location estimation through a series of calculations, with an algorithm complexity of O(2^3). M-1 M 2 logM), where O(·) represents the complexity function. Assume the complexity of calculating the average distance between P intersection points in a KD-tree is PlogP, and the number of intersection points between M lines is M(M-1) / 2. Then there are 2 M Substituting P = M(M-1) / 2 into the given equation and taking an approximation, we can obtain the algorithm complexity described above. It can be observed that when M is small, the computational complexity is acceptable.

[0070] The effectiveness of the method of the present invention will be demonstrated through simulation experiments, that is, the effectiveness of the method of the present invention will be verified by MATLAB simulation.

[0071] Specifically, the cell radius is set to 100m, the number of APs is set to 4, and they are evenly distributed on a circle of radius R. Each AP has 128 ULA antennas, so the total number of antennas is MN = 512. The wavelength λ is set to 0.01m, and the Rayleigh distance of a single AP can reach 327m. Users are randomly distributed within the cell, and the achievable data rate is given by R = log2(1 + γ|h H v| 2Calculations show that h represents the channels for users and all APs, and v = [v1, ..., v2]. m ,…,v M ], v m Let γ represent the steering vector of the m-th AP, and γ be the signal-to-noise ratio. All numerical results are obtained through more than 10,000 random positions.

[0072] Table 1 compares the training overhead of different beam training methods. The far-field exhaustive beam training method requires searching MN angles; the near-field exhaustive beam training method has S=6 distance sampling points, and its training overhead is the product of the angle sampling points and the distance sampling points. Therefore, the beam training overheads of far-field exhaustive and near-field exhaustive are 512 and 3072, respectively.

[0073] The near-field two-stage beam training method first searches in the angle domain and then searches in the range domain, so its training cost is the sum of the angle sampling points and the range sampling points, reaching 518; for the near-field two-stage layered beam training, a layered approach is adopted in the angle domain and the range domain respectively, and the training cost is significantly reduced to 24.

[0074] In contrast, the method proposed in this invention first utilizes a far-field codebook to obtain the user's angle relative to different APs, with multiple APs searching simultaneously. The beam training overhead is only 2log2(N). Then, a two-stage filtering method is used to obtain the user's final estimated point without incurring additional training overhead. It can be seen that this invention fully leverages the architectural advantages of distributed MIMO systems, and the cooperation between multiple APs significantly reduces beam training overhead.

[0075] Table 1

[0076]

[0077] in addition, Figure 4 The performance of achievable rates relative to signal-to-noise ratio (SNR) is shown under different beam training methods, with SNR ranging from 5 dB to 15 dB and AP deployment radius set to 50 m. It can be observed that the beam training method proposed in this invention outperforms the two-stage near-field beam training method under centralized MIMO (i.e., the near-field two-stage beam training method in Table 1) and the two-stage hierarchical near-field beam training method (i.e., the near-field two-stage hierarchical beam training in Table 1), while approaching the performance of the near-field polar domain codebook with a beam training overhead of thousands under centralized MIMO (i.e., the near-field exhaustive search beam training method in Table 1), and far exceeding the performance of the far-field codebook under centralized MIMO (i.e., the far-field exhaustive search beam training method in Table 1).

[0078] Simulation results show that the beam training method of the present invention can achieve a near-near-field extreme domain exhaustive beam training rate while significantly reducing training overhead.

[0079] The present invention also discloses a near-field beam training device for a single-user distributed MIMO system, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-described method.

[0080] Based on this understanding, the implementation of all or part of the processes in the above-described embodiments of the present invention can be accomplished by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include at least: any entity or device capable of carrying computer program code to a storage device, a recording medium, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium. Examples include USB flash drives, portable hard drives, magnetic disks, or optical disks. In some jurisdictions, according to legislation and patent practice, computer-readable media cannot be electrical carrier signals or telecommunication signals.

[0081] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0082] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed in this invention can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0083] In the embodiments provided by this invention, it should be understood that the disclosed apparatus / device and method can be implemented in other ways. For example, the apparatus / device embodiments described above are merely illustrative.

[0084] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A near-field beam training method for a single-user distributed MIMO system, characterized in that, Includes the following steps: Based on the far-field hierarchical codebook of each AP in the distributed MIMO system, far-field hierarchical beam training is performed on the corresponding AP to obtain the beam direction of each AP pointing to the user. The user's position is filtered based on several beam directions to obtain the user's position information relative to the distributed MIMO system; Filtering the user's location based on several beam directions includes: Several intersection points are determined based on several beam directions; Based on the location information of each intersection point in the service area of ​​the distributed MIMO system, the intersection points in the location-dense area are selected to form an intersection point set. Calculate the average position information based on the position information of the intersection points in the intersection point set; The average location information is used as the user's location information relative to the distributed MIMO system.

2. The near-field beam training method for a single-user distributed MIMO system as described in claim 1, characterized in that, After obtaining the user's location information relative to the distributed MIMO system, the process also includes: Based on the location information, the optimal near-field codeword for each AP is calculated.

3. The near-field beam training method for a single-user distributed MIMO system as described in claim 2, characterized in that, Before determining several intersection points based on several beam directions, the process also includes: Based on the beam direction, a symmetrical beam direction symmetrical about the corresponding AP is generated; For each AP, select the beam direction or symmetrical beam direction as the final beam direction of the AP.

4. The near-field beam training method for a single-user distributed MIMO system as described in claim 3, characterized in that, Selecting the beam direction or symmetrical beam direction for each AP array includes: Generate 2 based on the beam direction or symmetrical beam direction of each AP. M A combination of beam directions; among which... M This indicates the number of access points (APs) in the distributed MIMO system. In beam direction combination, the optimal beam direction combination is selected based on the average distance between the intersection points formed by the beam directions; For each AP, the beam direction or symmetrical beam direction is selected based on the optimal beam direction combination.

5. The near-field beam training method for a single-user distributed MIMO system as described in claim 4, characterized in that, In beam direction combination, selecting the optimal beam direction combination based on the average distance between the intersection points formed by the beam directions includes: The beam direction combination corresponding to the minimum value among the average distances is selected as the optimal beam direction combination.

6. The near-field beam training method for a single-user distributed MIMO system as described in claim 5, characterized in that, Selecting the beam direction combination corresponding to the minimum value among the average distances as the optimal beam direction combination includes: The minimum value among the average distances is selected based on the KD-Tree.

7. The near-field beam training method for a single-user distributed MIMO system as described in claim 2, characterized in that, The intersection points in the densely located regions are selected using a density-based clustering algorithm.

8. A near-field beam training device for a single-user distributed MIMO system, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method as described in any one of claims 1-7.

Images