An array antenna hybrid far and near field beam training method and device

Abstract

The application provides an array antenna mixed far and near field beam training method and device, comprising: dividing a phased array array into a plurality of sub-arrays composed of adjacent array elements; sending a first training signal based on a same far field angle codebook to a user equipment through each sub-array; obtaining optimal angle indexes corresponding to each sub-array fed back by the user equipment; when it is determined that the user equipment is in a far field based on an angle consistency measurement between the optimal angle indexes, determining a downlink beam according to the optimal angle indexes; or when it is determined that the user equipment is in a near field based on the angle consistency measurement between the optimal angle indexes, generating an initial distance estimation according to the optimal angle indexes, performing near field beam training to obtain a near field focusing beam, and reducing implementation complexity and beam training overhead in a mixed field scene.

Description

Technical Field

[0001] This invention relates to the field of communication technology, and in particular to a method and apparatus for training hybrid far- and near-field beams of an array antenna. Background Technology

[0002] With the development of millimeter-wave and terahertz communication technologies, multi-antenna array technology, such as massive MIMO and VMIMO arrays, is commonly used in related communication systems to meet the demands of high data rates and large-capacity transmission. Due to the increased carrier frequency and array aperture, the propagation conditions between user equipment and base stations no longer satisfy the traditional far-field plane wave approximation in some scenarios, and the near-field spherical wave effect becomes increasingly significant. This can lead to mixed propagation scenarios involving both far-field and near-field users within the same system. In such scenarios, beam alignment is typically achieved through beam training to obtain array gain and establish a communication link. Because the channel characterization parameters differ between the far-field and near-field, the codebook construction and search strategies for beam training differ accordingly.

[0003] Existing technical solutions mainly fall into two categories: First, they perform far-field angle scanning or near-field angle-distance joint two-dimensional search separately. The former cannot adapt to near-field users, while the latter incurs huge training overhead. Second, they attempt to uniformly process mixed fields through equivalent parameters, but this still requires repeated scanning and evaluation of a large number of candidate parameters, or relies on complex multi-layered codebooks and inter-layer coordination mechanisms.

[0004] However, these methods all result in high beam training overhead and high implementation complexity in mixed field scenarios. Summary of the Invention

[0005] This invention provides a method and apparatus for training hybrid near-field and far-field beams of an array antenna, which solves the shortcomings of existing technologies in beam training under mixed-field scenarios, such as high overhead and high implementation complexity.

[0006] This invention provides a method for training hybrid far- and near-field beamforming of an array antenna, comprising: The phased array is divided into multiple sub-arrays composed of adjacent array elements; Each of the subarrays sends a first training signal based on the same far-field angle codebook to the user equipment. Obtain the optimal angle index corresponding to each subarray, as fed back by the user equipment; When it is determined that the user equipment is in the far field based on the angular consistency metric between the optimal angle indices, the downlink beam is determined according to the optimal angle index; or... When it is determined that the user equipment is in the near field based on the angular consistency metric between the optimal angle indices, an initial distance estimate is generated according to the optimal angle index, and near-field beam training is performed to obtain a near-field focusing beam.

[0007] According to the present invention, a hybrid near-field and far-field beamforming training method for array antennas, after obtaining the optimal angle index corresponding to each sub-array fed back by the user equipment, further includes: Calculate the statistical differences between the optimal angle indices of each subarray to obtain an angle consistency measure; Based on the comparison result between the angle consistency metric and the preset discrimination threshold, it is determined whether the user equipment is in the far field or near field.

[0008] According to the present invention, a hybrid far- and near-field beam training method for an array antenna is provided, wherein the statistical difference includes the degree of angular dispersion, and the calculation of the degree of angular dispersion includes: determining the statistical center value of each optimal angle index as a global reference direction, and calculating the degree of dispersion of each index relative to the global reference direction; The discrimination threshold is determined based on the Rayleigh distance defined by the aperture of the phased array and the operating wavelength, as well as the maximum center distance between each subarray.

[0009] According to the present invention, a hybrid far- and near-field beam training method for array antennas is provided, wherein the statistical difference includes phase curvature, and the calculation of the phase curvature includes: calculating the curvature of the phase profile on the array aperture based on the beam phase information corresponding to each sub-array; The discrimination threshold is a preset phase curvature threshold.

[0010] According to the present invention, a hybrid near-field and far-field beam training method for an array antenna is provided, wherein generating an initial distance estimate based on the optimal angle index includes: Based on the optimal angle index and array center position of each subarray, the deviation of the angle estimate of each subarray relative to the common reference direction is calculated. Substituting the deviation into an approximate linear model characterizing the relationship between angle deviation and distance, the initial distance estimate is obtained by solving the least squares fitting algorithm.

[0011] According to the present invention, a hybrid near-field and far-field beam training method for an array antenna is provided, wherein obtaining the initial distance estimate by a least-squares fitting algorithm includes: Construct an objective function with distance-related parameters as optimization variables; wherein, the objective function represents the sum of squares of the products of the deviation between the optimal angle index of each subarray and the common reference direction, the distance-related parameters, and the center position of the corresponding subarray. Find the distance-related parameter values ​​that minimize the objective function; The reciprocal of the distance-related parameter value is used as the initial distance estimate for the user equipment.

[0012] According to the present invention, a hybrid near-field and far-field beam training method for an array antenna is provided, wherein performing near-field beam training to obtain a near-field focused beam includes: The statistical center of the optimal angle index and the initial distance estimate are used as the starting point for the iterative search; Within the local area corresponding to the starting point, a second training signal is alternately sent in the angle and distance dimensions to perform a one-dimensional discrete search. The angle estimate and distance estimate are updated based on the received signal quality fed back by the user equipment until the convergence condition is met.

[0013] According to the present invention, a hybrid near-field and far-field beam training method for an array antenna is provided, wherein the step of alternately performing a one-dimensional discrete search in the angular and distance dimensions within the local region corresponding to the starting point includes: With the current distance estimate fixed, within the preset angle search interval, traverse multiple angle candidate points and send the corresponding near-field focusing training beams. Based on the user equipment feedback, select the angle candidate point with the best reception quality and update the angle estimate. The angle estimate is fixed and updated. Within the preset distance search interval, multiple distance candidate points are traversed and the corresponding near-field focusing training beams are sent. The distance candidate point with the best reception quality is selected based on the user equipment feedback to update the distance estimate.

[0014] According to the present invention, a hybrid near-field and far-field beamforming training method for an array antenna is provided, wherein dividing the phased array into multiple sub-arrays composed of adjacent array elements includes: dividing the phased array into multiple consecutive sub-arrays with equal apertures according to the geometric arrangement of the phased array; the geometric center positions of the array surfaces of each sub-array are different, and the geometric center of the array surfaces of each sub-array has a known positional offset in the direction of the array axis. The far-field angle codebook is a codebook constructed based on the discrete Fourier transform.

[0015] This invention protects a hybrid far- and near-field beam training device with an array antenna, comprising: The partitioning module is used to divide the phased array into multiple subarrays composed of adjacent array elements; The transmitting module is used to transmit a first training signal based on the same far-field angle codebook to the user equipment through each of the subarrays; The acquisition module is used to acquire the optimal angle index corresponding to each subarray, fed back by the user equipment; The determination module is configured to determine the downlink beam based on the optimal angle index when the user equipment is determined to be in the far field based on the angle consistency metric between the optimal angle indices; or, when the user equipment is determined to be in the near field based on the angle consistency metric between the optimal angle indices, generate an initial distance estimate based on the optimal angle index and perform near-field beam training to obtain a near-field focusing beam.

[0016] The present invention also provides an electronic device, 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 array antenna hybrid far- and near-field beam training method as described above.

[0017] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the array antenna hybrid near-field beam training method as described above.

[0018] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the array antenna hybrid near-field beam training method as described above.

[0019] The hybrid near-field and far-field beam training method and apparatus for array antennas provided by this invention utilizes the physical structure of a phased array. By dividing the array into geometric subarrays and reusing the far-field angle codebook for initial scanning, near-field and far-field discrimination can be achieved. This eliminates the need for designing complex codebooks specifically for mixed fields and for introducing additional RF hardware such as adjustable delay units, making it easy to deploy in engineering. Furthermore, a local fine-grained search based on the initial estimate is triggered only when the signal is determined to be near-field, thus avoiding the huge overhead of always performing a global traversal of the two-dimensional parameter space in traditional schemes, significantly improving training efficiency. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0021] Figure 1 This is a flowchart illustrating the hybrid near-field and far-field beam training method for array antennas provided by the present invention. Figure 2 This is a schematic diagram of the phased array architecture design based on geometric subarrays provided by the present invention; Figure 3 This is a schematic diagram of subarray partitioning; Figure 4This is a schematic diagram of the angular focusing characteristics of the user subarray in the far and near fields; Figure 5 This is a schematic diagram of the user subarray angle focusing characteristics; Figure 6 This is a schematic diagram of the focusing characteristics of subarrays SR1 and SR4; Figure 7 This is a schematic diagram comparing the achievable rates of the method of this invention with other methods under different SNRs; Figure 8 This is a schematic diagram comparing the root mean square error performance of the spatial angle estimation method of the present invention with other methods under different signal-to-noise ratios; Figure 9 This is a schematic diagram comparing the root mean square error performance of the method of this invention with other methods under different signal-to-noise ratios for user distance estimation. Figure 10 This is a schematic diagram comparing the array gain retention performance of the method of this invention with other methods under different signal-to-noise ratios; Figure 11 This is a schematic diagram of the structure of the array antenna hybrid near-field beam training device provided by the present invention; Figure 12 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0023] Figure 1 This is a flowchart illustrating the hybrid near-field and far-field beam training method for array antennas provided by the present invention.

[0024] like Figure 1 As shown in this embodiment, a hybrid near-field and far-field beam training method for array antennas is provided. This method is applicable to large-scale / ultra-large-scale phased array communication systems in millimeter-wave and terahertz communications, and can solve the problems of high beam training overhead and high implementation complexity in hybrid near-field scenarios. The method mainly includes the following steps: 101. Divide the phased array into multiple subarrays composed of adjacent array elements.

[0025] Specifically, based on the inherent geometric arrangement characteristics of the ULA uniform linear phased array, the entire large-scale phased array is divided into S continuous subarrays with different geometric center positions on each array surface, such as... Figure 2The diagram illustrates a phased array architecture design based on geometric subarrays. The partitioning process is implemented solely through array control logic, requiring no additional RF chains, adjustable delay units, or other hardware, and without altering the original phased array hardware architecture. Simultaneously, near-field and far-field codebooks of corresponding resolutions are pre-prepared to provide a foundation for training signal transmission. By creating geometric differences in the spatial positions of each subarray, a physical basis is provided for using the angular characteristics of the subarrays to distinguish between near and far fields. Furthermore, each subarray can independently perform beam scanning operations without affecting the overall array's operational logic.

[0026] 102. Each subarray sends a first training signal based on the same far-field angle codebook to the user equipment.

[0027] Specifically, the base station first sends a training trigger signal to the user equipment to initiate subarray training and instruct on the training configuration. Then, the i-th subarray (i=1,2,...,S) is selected, and each subarray, based on the same far-field angle codebook, sequentially sends multiple sets of first training signals to the user equipment. The codebook usage rules and signal transmission timing of each subarray remain consistent, eliminating the need to design a new codebook specifically for mixed-field scenarios. Reusing a mature far-field angle codebook avoids the design, storage, and scheduling complexity of mixed-field codebooks, significantly reducing the computational and hardware overhead of the initial preparation, while ensuring the comparability of the scanning results of each subarray.

[0028] 103. Obtain the optimal angle index corresponding to each subarray, as fed back by the user equipment.

[0029] Specifically, after receiving the first training signals transmitted by each subarray, the user equipment measures the reception quality of each group of signals in real time, and uses the beam index corresponding to the signal with the best reception quality as the optimal angle index of that subarray. This refers to the optimal far-field angle codeword index. After the user equipment completes the reception and measurement of the first training signals for all subarrays, the optimal angle index of all subarrays is determined. ,..., } The data is fed back to the base station in a unified manner. The angular reference features of each subarray are obtained through the actual measurement results of the user equipment, providing real and effective data support for near-field and far-field discrimination. Moreover, the index data of all subarrays can be obtained with only one scan, without the overhead of repeated training.

[0030] 104. When it is determined that the user equipment is in the far field based on the angle consistency metric between the optimal angle indices, the downlink beam is determined according to the optimal angle index.

[0031] Specifically, after the base station confirms that it has received the optimal angle indices of all subarrays, it aggregates these indices to obtain a coarse angle estimation result. If the user equipment is determined to be in the far field based on this coarse angle estimation result, the base station directly controls the entire phased array to generate the corresponding far-field downlink beam according to the statistical characteristics of the optimal angle indices. Beam alignment is directly completed in the far-field scenario, maximizing the simplification of the far-field beam training process and shortening the beam alignment latency.

[0032] 105. When it is determined that the user equipment is in the near field based on the angular consistency metric between the optimal angle indices, an initial distance estimate is generated according to the optimal angle index, and near-field beam training is performed to obtain the near-field focusing beam.

[0033] Specifically, if the user equipment is determined to be in the near field, the base station uses the optimal angle index of each subarray. ,..., } The algorithm uses the geometric position information of the subarrays to convert the angular features of the subarrays into distance estimates, thus obtaining the initial distance estimate. This is the coarse distance estimation. Subsequently, statistical characteristics of the optimal angle index and the initial distance estimation are used. Based on this, a near-field beam training process is initiated, and a near-field focusing beam adapted to the near-field spherical wave effect is obtained through refined parameter search. Refined training is only triggered when the field is determined to be near-field, avoiding performing a two-dimensional global search of angle and distance for all users, which significantly reduces the overall training overhead in mixed-field scenarios. At the same time, the near-field focusing beam can match the propagation characteristics of spherical waves, improving the array gain of near-field communication.

[0034] This embodiment achieves on-demand optimization of beam training in mixed-field scenarios through subarray partitioning, far-field codebook scanning, and far- and near-field branching processes. It simplifies the far-field process and provides fine-grained training on demand in the near-field, thereby reducing overall training overhead and implementation complexity. Furthermore, it requires no additional hardware and is easy to deploy in engineering.

[0035] Furthermore, based on the above embodiments, after obtaining the optimal angle index corresponding to each subarray fed back by the user equipment in this embodiment, it further includes: calculating the statistical difference between the optimal angle indices of each subarray to obtain an angle consistency measure; and determining whether the user equipment is in the far field or near field based on the comparison result of the angle consistency measure and the preset discrimination threshold.

[0036] Specifically, the base station indexes each subarray using its optimal angle. ,..., } Based on the basic data, the differences between each index are obtained through statistical calculations, which constitutes the angular consistency measure. Angular consistency measurement The magnitude of the value is inversely proportional to the degree of matching between the optimal angle indices of each subarray; that is, the smaller the value, the more consistent the angle indices of each subarray, and the larger the value, the more obvious the differences in the angle indices of each subarray. The abstract differences in the angle indices of each subarray are transformed into quantifiable and comparable specific indicators, providing clear numerical basis for near-field and far-field discrimination and avoiding errors in subjective judgment. The base station pre-sets a preset discrimination threshold based on the hardware parameters of the phased array and the geometric arrangement characteristics of the subarrays, and then applies the calculated angle consistency metric... The angular consistency metric is compared with a preset threshold. If the angular consistency metric is less than or equal to the preset threshold, it indicates that the angular index consistency of each subarray is high, and the user equipment is determined to be in the far field. If the angular consistency metric is greater than the preset threshold, it indicates that the angular index consistency of each subarray is low, and the user equipment is determined to be in the near field. This near-field discrimination without distance priors or near-field codebooks can be completed using only the existing data from the far-field scan of the subarrays, without adding extra training signals or scanning overhead. Furthermore, the discrimination rules are simple, easy to implement, and suitable for engineering applications.

[0037] Furthermore, based on the above embodiments, the statistical differences in this embodiment include the degree of angular dispersion. Calculating the degree of angular dispersion includes: determining the statistical center value of each optimal angle index as the global reference direction, and calculating the degree of dispersion of each index relative to the global reference direction; the discrimination threshold is determined based on the Rayleigh distance defined by the aperture and working wavelength of the phased array and the maximum center distance between each subarray.

[0038] Specifically, the calculation of the degree of angular dispersion includes: First, determine the statistical center value of each optimal angle index as the global reference direction. The calculation formula is (1): ,..., } (1) Choosing the statistical center value instead of the average value as the reference direction effectively avoids the influence of individual extreme value indices on the overall reference direction, ensuring the objectivity of the reference direction. Subsequently, the optimal angle index for each subarray is calculated. Relative to global reference direction The deviation values ​​are statistically calculated for all deviation values, such as variance and standard deviation. The resulting statistical result is the angular dispersion, which directly reflects the degree of dispersion of the angle indices of each subarray relative to the global reference direction. A clear execution standard for calculating angular dispersion ensures the consistency and comparability of calculation results across different scenarios, thereby improving the accuracy of the judgment results.

[0039] The preset discrimination threshold is determined by the Rayleigh distance defined by the aperture and working wavelength of the phased array and the maximum center distance between each subarray. The formula for calculating the Rayleigh distance is (2): = / (2) In the formula, D is the aperture of the phased array. Lambda This is the operating wavelength of the phased array.

[0040] The maximum center distance between each subarray is denoted as . , which is the distance between the phase centers of the two spatially farthest subarrays. The base station is based on Rayleigh distance. By combining the numerical characteristics of the maximum center distance of the subarrays with the requirements of actual communication scenarios, a preset discrimination threshold suitable for the current phased array architecture is calculated. This preset discrimination threshold is deeply bound to the hardware parameters and geometric features of the phased array, rather than being an empirical value detached from reality. This makes the threshold more closely match the working characteristics of the current array, significantly improving the accuracy of near-field and far-field discrimination and avoiding misjudgments caused by unreasonable thresholds.

[0041] By specifying the angular consistency metric as the degree of angular dispersion, the calculation rules and threshold determination criteria are clarified, making it more operable. At the same time, the threshold design based on hardware parameters improves the adaptability of the discrimination results.

[0042] Furthermore, based on the above embodiments, the statistical differences in this embodiment include phase curvature, and the calculation of phase curvature includes: calculating the curvature of the phase profile on the array aperture based on the beam phase information corresponding to each subarray; the discrimination threshold is a preset phase curvature threshold.

[0043] Specifically, firstly, the beam phase information corresponding to the transmission of the first training signal by each subarray is extracted. This beam phase information is inherent measurement data of the phased array RF link and does not require additional scanning. Then, based on the beam phase information, a phase profile across the entire phased array aperture is fitted, and the curvature value of the phase profile is calculated. The result is the phase curvature. In near-field scenarios, the curvature of the phase profile is larger due to the spherical wave effect, while in far-field scenarios, the curvature is smaller due to the plane wave effect, showing a significant numerical difference. Utilizing phase characteristics to distinguish between near and far fields forms a complementary dual-index discrimination mechanism with the degree of angular dispersion, improving the robustness of the discrimination results. Even if one index deviates due to noise, the other index can still guarantee the accuracy of the discrimination.

[0044] Angular consistency measurement and phase curvature Each corresponds to a specific preset judgment threshold, i.e., the angle consistency index. and phase curvature index The calculation formulas are (3) and (4) respectively: = (3) = (4) In the formula, The robustness coefficient of the angle consistency index. The robustness coefficient of the phase curvature index. For mesh compensation robustness coefficient, typically 、 、 All are set to 1; The distance between the centers of the two furthest subarrays; Rayleigh distance; The grid spacing of the far-field angle codebook; Lambda The operating wavelength of the phased array; a The sub-array half-aperture; This serves as the global reference direction.

[0045] The preset threshold for phase curvature is a preset phase curvature threshold, which is the aforementioned phase curvature index. The phased array is pre-set by the base station based on the working characteristics of the phased array and the phase change law of the near-field spherical wave, and can be flexibly adjusted according to the signal-to-noise ratio and array size of the actual communication scenario.

[0046] Angular dispersion and phase curvature are two complementary discrimination indicators. When the angular consistency metric is less than its corresponding preset threshold, or the phase curvature is less than its corresponding preset threshold, the user equipment is determined to be in the near field. Only when both the angular consistency metric and the phase curvature do not exceed their corresponding preset thresholds is the user equipment determined to be in the far field. The discrimination logic formula is as follows: like If the distance is close to the target area, it is considered near-field; otherwise, it is considered far-field. By using two indicators for joint discrimination, the probability of misclassification between near and far-field is significantly reduced, making the discrimination results more consistent with the actual propagation scenario. At the same time, both indicators are calculated based on existing data from the far-field scanning of the subarray, without adding any additional training and computational overhead.

[0047] Furthermore, based on the above embodiments, this embodiment generates an initial distance estimate based on the optimal angle index, including: calculating the deviation of the angle estimate of each subarray relative to the common reference direction based on the optimal angle index of each subarray and the array center position; substituting the deviation into an approximate linear model characterizing the relationship between angle deviation and distance, and solving the initial distance estimate by a least squares fitting algorithm.

[0048] Specifically, the global reference direction is first used as the common reference direction, combined with the optimal angle index of each subarray and the array center position of the subarray. The deviation of the angle estimate of each subarray from the common reference direction is calculated; this deviation is the angle deviation amount. The calculation process fully utilizes the spatial differences between each subarray, establishing a correlation between the angular deviation and the distance to the user equipment. The angular characteristics of the subarrays are transformed into distance-related observations, providing a data foundation for distance estimation. Furthermore, the data required for the calculation is all existing data from previous scans, incurring no additional overhead.

[0049] The base station pre-constructs an approximate linear model representing the relationship between angle deviation and distance. This approximate linear model is based on the propagation law of near-field spherical waves and utilizes the spatial position differences of the subarrays. Establish angle deviation The linear correlation between the distance to the user equipment and the distance to the user equipment is established. Then, the calculated angular deviations of all subarrays are substituted into the approximate linear model, and the model is solved using the least squares (LS) fitting algorithm. The solution obtained is the initial distance estimate for the user equipment. (Coarse distance estimation). It directly inverts the inconsistency of subarray angles into distance estimates without designing a new near-field codebook. It achieves effective near-field distance estimation without increasing any training overhead, providing reliable initial values ​​for subsequent fine-tuning of near-field training.

[0050] Furthermore, based on the above embodiments, this embodiment obtains the initial distance estimate by using a least squares fitting algorithm, including: constructing an objective function with distance-related parameters as optimization variables; wherein the objective function represents the sum of squares of the products of the deviation between the optimal angle index of each subarray and the common reference direction, the distance-related parameters, and the center position of the corresponding subarray; solving for the distance-related parameter value that minimizes the objective function; and using the reciprocal of the distance-related parameter value as the initial distance estimate of the user equipment.

[0051] Specifically, first, set the distance-related parameters. The least-squares fitting optimization variable is inversely proportional to the actual distance to the user equipment, and is subsequently based on the angle deviation. Subarray center position and global reference direction Construct the objective function, and the specific calculation formula for the objective function is (5): (5) In the formula, S The number of subarrays. For the first s The phase center of each subarray is offset relative to the common reference point. The objective function is essentially the sum of the squares of the products of the deviations between the optimal angle indices of each subarray and the common reference direction, along with the distance-related parameters and the corresponding subarray center positions. This gives the least-squares fitting a clear optimization objective, ensuring the directionality and accuracy of the fitting process. Simultaneously, the construction of the objective function fully incorporates the geometric characteristics of the subarrays, improving the reliability of the fitting results.

[0052] The base station uses numerical calculations to find the distance-related parameter values ​​that minimize the aforementioned objective function. The optimal distance correlation parameter represents the highest model fit, indicating that the correlation between the angle deviation and distance best reflects reality. By minimizing the sum of squares, the impact of noise on the fitting results is effectively reduced, ensuring the optimal distance correlation parameter. The accuracy.

[0053] Due to distance-related parameters The distance is inversely proportional to the actual distance to the user equipment; therefore, the reciprocal of the optimal distance-related parameter values ​​obtained from the solution is used to obtain the initial distance estimate. The calculation formula is (6): (6) The transformation from fitted parameters to actual distance estimation is completed, resulting in the initial distance estimate. It provides a precise starting point for near-field fine training, avoids blind searching in near-field training, and significantly shortens the convergence time.

[0054] Furthermore, based on the above embodiments, this embodiment performs near-field beam training to obtain a near-field focusing beam, including: using the statistical center of the optimal angle index and the initial distance estimate as the starting point of the iterative search; in the local area corresponding to the starting point, alternately sending a second training signal in the angle dimension and the distance dimension to perform a one-dimensional discrete search; updating the angle estimate and the distance estimate based on the received signal quality fed back by the user equipment until the convergence condition is met.

[0055] Specifically, firstly, the statistical center of the optimal angle index of each subarray is determined. As the starting point for iteration of the angle dimension, the initial distance estimate As the starting point for iteration in the distance dimension, we set the iteration count t=1, and let... 、 = Together, these two constitute the initial point for the near-field beam training iterative search. (Using a coarse angle...) and coarse distance Using the initial point as the starting point for iteration, near-field fine-tuning training only searches a local region near the initial point, avoiding global traversal and significantly reducing training overhead and computational complexity.

[0056] Base station at the starting point of iteration ( , ) Within the corresponding angle and distance local area, a preset search interval is defined. Then, an alternating one-dimensional discrete search method is adopted. That is, the parameters of one dimension are fixed first, and the other dimension is discretized and scanned. During the scanning process, a second training signal, namely the near-field focusing training beam, is sent to the user equipment through the entire array. After completing one dimension scan, the same operation is performed on the other dimension, and so on. This transforms the two-dimensional near-field parameter search problem into multiple one-dimensional discrete searches, which significantly reduces the computational complexity. At the same time, the second training signal is sent through the entire array to ensure the array gain of the near-field focusing beam.

[0057] After receiving the second training signal, the user equipment measures the reception quality of each group of signals and feeds it back to the base station. The base station uses the parameter corresponding to the signal with the best reception quality as the update value for that dimension, and updates the angle estimate accordingly. and distance estimates After each alternating scan of angle and distance, the base station determines whether the convergence condition is met. If not, it continues iterating; if so, training terminates. The convergence condition is one of the following two conditions; iteration terminates if either is met: ① The number of iterations reaches a preset upper limit. t = , This is the pre-set maximum number of iterations. ② The parameters updated in this iteration are unchanged from the previous iteration, i.e. = The parameters are updated by receiving actual measurement results of signal quality, ensuring the accuracy of the parameters. At the same time, the explicit convergence condition makes the training process controllable, avoiding the increased overhead caused by infinite iteration, and the convergence condition can be flexibly adjusted according to actual communication needs.

[0058] After the iteration terminates, the base station will obtain the final parameters. Substitute the near-field focusing training beam codeword into the formula to generate the final near-field focusing beam. The near-field focusing beam is used for near-field user data transmission and completes near-field beam alignment.

[0059] By transforming the two-dimensional search into an alternating one-dimensional search, the overhead and complexity of near-field training are significantly reduced, while the explicit convergence conditions make the training process more controllable.

[0060] Furthermore, based on the above embodiments, this embodiment performs a one-dimensional discrete search alternately in the angle and distance dimensions within the local area corresponding to the starting point, including: fixing the current distance estimate, traversing multiple angle candidate points and sending corresponding near-field focusing training beams within a preset angle search interval, and updating the angle estimate by selecting the angle candidate point with the best reception quality based on user equipment feedback; fixing the updated angle estimate, traversing multiple distance candidate points and sending corresponding near-field focusing training beams within a preset distance search interval, and updating the distance estimate by selecting the distance candidate point with the best reception quality based on user equipment feedback.

[0061] Specifically, first fix the current distance estimate. Within a preset angle search interval, the interval is truncated and a discrete set of candidate angle points is generated. The number of candidate points is The formula for calculating the angle search interval is (7): (7) In the formula, The step size for angle search is [-1, 1], which represents the valid range of angle values ​​to avoid invalid angle searches.

[0062] The base station then targeted candidate points at each angle. Omega ∈ Generate the corresponding near-field focusing training beam codeword The system then transmits a second training signal to the user equipment via the full array. The user equipment measures the reception quality metrics for each group of signals. The data is then fed back to the base station, which selects the candidate angle with the best reception quality as the updated angle estimate. The updated formula is (8): (8) Under the premise of a fixed distance, it achieves fine-grained scanning of the angular dimension, gradually converges to the optimal angular parameters, and the search method of local intervals avoids unnecessary overhead.

[0063] The updated angle estimate was then fixed. Within a preset distance search interval, the distance search interval is truncated and a discrete set of distance candidate points is generated. The number of candidate points is The formula for calculating the distance to the search interval is (9): (9) In the formula, This is the distance search scaling factor. This is the minimum value for the near-field distance estimate. This represents the maximum value of the near-field distance estimate. This represents the valid range of distance values.

[0064] Base stations target each distance candidate point r ∈ Generate the corresponding near-field focusing training beam codeword The system then transmits a second training signal to the user equipment via the entire array, and the user equipment measures the reception quality index of each group of signals. The data is then fed back to the base station, which selects the candidate distance point with the best reception quality as the updated distance estimate. The updated formula is (10): (10) Under the premise of a fixed optimal angle, a refined scan of the distance dimension is achieved, allowing the distance parameter to gradually converge to the optimal value. This alternating optimization with angle scanning ensures the performance of the near-field focusing beam. The alternating one-dimensional discrete search is broken down into specific methods of fixed-distance scanning of the angle and fixed-angle scanning of the distance. This clarifies the rules for defining the search interval and the handling of candidate points, making near-field refined training more feasible. Simultaneously, the search method within local intervals ensures low training overhead and high accuracy.

[0065] Furthermore, based on the above embodiments, this embodiment divides the phased array into multiple sub-arrays composed of adjacent array elements, including: dividing the phased array into multiple consecutive sub-arrays with equal apertures according to the geometric arrangement of the phased array; the geometric center positions of the array surfaces of each sub-array are different, and the geometric center of the array surfaces of each sub-array has a known positional offset in the direction of the array axis; the far-field angle codebook is a codebook constructed based on discrete Fourier transform.

[0066] Specifically, when dividing the phased array into subarrays, the geometric arrangement of the phased array (preferably a ULA uniform linear phased array) is strictly followed, dividing it into multiple continuous subarrays with equal apertures. This ensures consistent hardware characteristics across subarrays and avoids scanning result deviations caused by differences in array apertures. Simultaneously, the geometric center positions of the array surfaces of each subarray are distinct, and the phase center of each subarray has a known positional offset along the array axis. This positional offset is a pre-measured fixed value, providing clear geometric data for calculating angular deviations and least-squares fitting. Examples of subarray division include dividing the phased array into four continuous subarrays (ULA) or four quadrant subarrays (UPA), with a division quantity of S. The division method can be flexibly adjusted according to the array size, and all division methods require no additional hardware, implemented only through control logic. This ensures the comparability of scanning results across subarrays, while the known phase center positional offset provides a reliable geometric basis for algorithm calculations, improving accuracy and feasibility.

[0067] The far-field angle codebook used in this invention is a codebook constructed based on Discrete Fourier Transform (DFT). It is a mature codebook for far-field beam training and does not require redesign for mixed-field scenarios. The base station can directly reuse existing codebook resources. By utilizing the mature DFT codebook, the overhead of codebook redesign, storage, and debugging is avoided, and it can be directly adapted to the codebook system of existing phased array communication systems.

[0068] This invention utilizes the geometric differences in the center positions of subarray surfaces as a source of near-field observability. A large-scale phased array antenna array is divided into multiple subarrays composed of adjacent elements, each with a different geometric center position and a known relative position. Simultaneously, each subarray has a different array center position in physical space, with its center distributed geometrically along the array's main axis or in a two-dimensional plane. Each subarray can independently perform beam scanning, channel measurement, or angle estimation while sharing or aligning to the same set of far-field angle codebooks. The subarray structure requires no additional RF chains or hardware; it is implemented solely through array partitioning and control logic.

[0069] This invention determines the two thresholds (angle consistency index and phase curvature index) for near and far-field users. It does not rely on distance priors or introduce near-field codebooks. It only uses the subarray scanning results themselves and completes the near and far-field determination through two complementary physical indices.

[0070] The subarray LS fitting method of this invention does not design a new near-field codebook or increase the scanning dimension; it directly inverts the "subarray angle inconsistency" into distance estimation. This method utilizes the difference in the array center positions formed by the spatially separated arrangement of multiple subarrays to map the angle estimation results of each subarray into observations related to user distance. Under preset master angle estimation conditions, it constructs an approximately linear relationship model between subarray angles and distance parameters. The base station obtains a coarse estimate of the user distance by performing least-squares fitting on the linear relationship model. The distance estimation process does not require additional near-field training signals or two-dimensional parameter search; it can be completed solely based on existing measurement results from the subarray angle scanning phase, thus achieving effective estimation of near-field user distance parameters without increasing training overhead.

[0071] This invention employs an angle-range alternating one-dimensional scanning method. Building upon coarse distance and coarse angle measurements, it rapidly converges to the optimal near-field parameters within a local region by alternately fixing one dimension and scanning another. In each training iteration, the method separates the angle and distance parameters. First, under a fixed distance estimate, an angle scan is performed within a preset angle search window, updating the angle estimate. Then, under a fixed updated angle estimate, a distance scan is performed within a preset distance search window, updating the distance estimate. These angle and distance scans are performed alternately, and the iteration is terminated based on changes in the received signal quality index or parameter convergence conditions. By transforming the two-dimensional near-field parameter search problem into multiple alternating one-dimensional scanning processes, the computational complexity is significantly reduced while maintaining near-field beam performance, improving the algorithm's engineering feasibility in large-scale array systems.

[0072] To verify the effectiveness of the present invention, the following tests were conducted: (1) Performance verification parameter settings Center frequency: 100GHz Phased array element count: 256 (ULA) Array spacing: 1 / 2 wavelength Array partitioning: S=4 Far-field codebook size: U=256 Near-field codebook size: UL=256 40 Near-field training iterations: Tmax = 3 Figure 3 This is a schematic diagram of subarray partitioning. Figure 4 This is a schematic diagram of the angular focusing characteristics of the user subarray in the far and near fields. Figure 5 This is a schematic diagram of the angular focusing characteristics of the user subarray. Left: far field, right: near field. Figure 6 This is a schematic diagram of the focusing characteristics of subarrays SR1 and SR4. Left: far field, right: near field.

[0073] Table 1 compares the training complexity of different algorithms.

[0074] Figure 7 This is a schematic diagram comparing the achievable rates of the method of this invention with other methods under different SNRs. Figure 8 This is a schematic diagram comparing the root mean square error performance of the spatial angle estimation method of the present invention with other methods under different signal-to-noise ratios. Figure 9 This is a schematic diagram comparing the root mean square error performance of the method of this invention with other methods under different signal-to-noise ratios for user distance estimation. Figure 10 This is a schematic diagram comparing the array gain retention performance of the method of this invention with other methods under different signal-to-noise ratios.

[0075] The method of the present invention has the following advantages: 1. No need to construct additional mixed field codebooks or add RF chains / adjustable delay devices; the near and far field discrimination and alignment path selection can be achieved directly using the subarray set structure.

[0076] Compared to existing technologies that "unify the mixed field using angle-distance equivalent parameters and construct codebooks for candidate parameters separately," this application does not rely on additional polarization domain / equivalent domain codebook construction, nor does it require the introduction of new RF link devices such as adjustable time delay (TTD) circuits to achieve near-field focusing. Instead, this application, based on the geometric subarray partitioning of the array, uses the inherent "difference in the geometric center position of the subarray surface" of the ultra-large-scale array as an observable dimension. Through the consistency / inconsistency features between multiple sets of optimal angle indices obtained from subarray scanning, as well as the phase curvature features, it directly completes the near-field and far-field discrimination, and selects either angle-only scanning (far-field path) or angle-distance fine search (near-field path) accordingly. Therefore, this solution can be implemented without changing the existing hardware architecture and codebook system, and has stronger engineering compatibility and deployment feasibility.

[0077] 2. The training and computation costs are significantly lower, avoiding the "angle × distance" candidate scanning and repeated training caused by the unified modeling of mixed fields.

[0078] Compared to hybrid near-field training methods, existing solutions typically require constructing or calling corresponding codebooks for multiple candidate equivalent parameters one by one, and then filtering based on matching degree / sparseness metrics. This process essentially still involves traversing or evaluating multiple rounds of the "angle-distance" two-dimensional candidate set. The training overhead and computational complexity increase linearly or approximately linearly with the size of the candidate set, and exploratory training of the near-field dimension is often unavoidable before determining near-field, leading to additional pilot signal consumption and latency. This application proposes a "discrimination first, alignment later" approach. It utilizes subarray far-field angle codebook scanning to generate statistics for discrimination, pre-determining the user's propagation area. When determined to be far-field, it directly reverts to full-array far-field beam training without introducing the distance dimension. Only when determined to be near-field is a subsequent alternating angle-distance one-dimensional scan triggered to refine the parameters. Thus, the two-dimensional joint search that is common in mixed fields is transformed into local one-dimensional alternating optimization that is mainly based on subarray scanning and triggered on demand. This significantly reduces the overall training overhead and online computation, and reduces performance loss and resource waste caused by misselection of candidates or repeated training.

[0079] The hybrid far- and near-field beam training device for array antennas provided by the present invention is described below. The hybrid far- and near-field beam training device for array antennas described below can be referred to in correspondence with the hybrid far- and near-field beam training method for array antennas described above.

[0080] Figure 11 This is a schematic diagram of the structure of the array antenna hybrid near-field beam training device provided by the present invention.

[0081] like Figure 11 As shown, the present invention provides a hybrid far- and near-field beam training device for array antennas, comprising: The partitioning module 1101 is used to divide the phased array array into multiple sub-arrays composed of adjacent array elements; The transmitting module 1102 is used to transmit a first training signal based on the same far-field angle codebook to the user equipment through each subarray; The acquisition module 1103 is used to acquire the optimal angle index corresponding to each subarray, which is fed back by the user equipment; The determination module 1104 is used to determine the downlink beam based on the optimal angle index when the user equipment is determined to be in the far field based on the angle consistency metric between the optimal angle indices; or, when the user equipment is determined to be in the near field based on the angle consistency metric between the optimal angle indices, generate an initial distance estimate based on the optimal angle index and perform near-field beam training to obtain a near-field focusing beam.

[0082] Figure 12 This is a schematic diagram of the structure of the electronic device provided by the present invention.

[0083] like Figure 12As shown, the electronic device may include a processor 1210, a communications interface 1220, a memory 1230, and a communication bus 1240, wherein the processor 1210, the communications interface 1220, and the memory 1230 communicate with each other via the communication bus 1240. The processor 1210 can call logic instructions in the memory 1230 to execute a hybrid far- and near-field beamforming training method for an array antenna. This method includes: dividing the phased array into multiple subarrays composed of adjacent array elements; sending a first training signal based on the same far-field angle codebook to the user equipment through each of the subarrays; obtaining the optimal angle index corresponding to each subarray fed back by the user equipment; when it is determined that the user equipment is in the far field based on the angle consistency metric between the optimal angle indices, determining the downlink beam according to the optimal angle index; or, when it is determined that the user equipment is in the near field based on the angle consistency metric between the optimal angle indices, generating an initial range estimate according to the optimal angle index, and performing near-field beamforming training to obtain a near-field focused beam.

[0084] Furthermore, the logical instructions in the aforementioned memory 1230 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0085] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the array antenna hybrid far-field and near-field beam training method provided by the above methods. The method includes: dividing a phased array into multiple sub-arrays composed of adjacent array elements; sending a first training signal based on the same far-field angle codebook to a user equipment through each of the sub-arrays; obtaining the optimal angle index corresponding to each sub-array fed back by the user equipment; when it is determined that the user equipment is in the far field based on the angle consistency metric between the optimal angle indices, determining the downlink beam according to the optimal angle index; or, when it is determined that the user equipment is in the near field based on the angle consistency metric between the optimal angle indices, generating an initial range estimate according to the optimal angle index, and performing near-field beam training to obtain a near-field focused beam.

[0086] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements a method for training hybrid far- and near-field beamforming of an array antenna provided by the methods described above. This method includes: dividing a phased array into multiple sub-arrays composed of adjacent array elements; sending a first training signal based on the same far-field angle codebook to a user equipment through each of the sub-arrays; obtaining the optimal angle index corresponding to each sub-array fed back by the user equipment; determining a downlink beam according to the optimal angle index when the user equipment is determined to be in the far field based on an angle consistency metric between the optimal angle indices; or, generating an initial range estimate according to the optimal angle index when the user equipment is determined to be in the near field based on an angle consistency metric between the optimal angle indices, and performing near-field beamforming training to obtain a near-field focused beam.

[0087] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0088] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0089] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; 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; and these 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.

Claims

1. A method for training hybrid near-field and far-field beamforming of an array antenna, characterized in that, include: The phased array is divided into multiple sub-arrays composed of adjacent array elements; Each of the subarrays sends a first training signal based on the same far-field angle codebook to the user equipment. Obtain the optimal angle index corresponding to each subarray, as fed back by the user equipment; When it is determined that the user equipment is in the far field based on the angle consistency metric between the optimal angle indices, the downlink beam is determined according to the optimal angle index; or, When it is determined that the user equipment is in the near field based on the angular consistency metric between the optimal angle indices, an initial distance estimate is generated according to the optimal angle index, and near-field beam training is performed to obtain a near-field focusing beam.

2. The array antenna hybrid near-field beam training method according to claim 1, characterized in that, After obtaining the optimal angle index corresponding to each subarray fed back by the user equipment, the process further includes: Calculate the statistical differences between the optimal angle indices of each subarray to obtain an angle consistency measure; Based on the comparison result between the angle consistency metric and the preset discrimination threshold, it is determined whether the user equipment is in the far field or near field.

3. The array antenna hybrid near-field beam training method according to claim 2, characterized in that, The statistical difference includes the degree of angular dispersion. Calculating the degree of angular dispersion includes: determining the statistical center value of each optimal angle index as the global reference direction, and calculating the degree of dispersion of each index relative to the global reference direction. The discrimination threshold is determined based on the Rayleigh distance defined by the aperture of the phased array and the operating wavelength, as well as the maximum center distance between each subarray.

4. The array antenna hybrid near-field and far-field beam training method according to claim 2, characterized in that, The statistical differences include phase curvature, and the calculation of the phase curvature includes: calculating the curvature of the phase profile on the array aperture based on the beam phase information corresponding to each subarray; The discrimination threshold is a preset phase curvature threshold.

5. The array antenna hybrid near-field beam training method according to claim 1, characterized in that, The step of generating an initial distance estimate based on the optimal angle index includes: Based on the optimal angle index and array center position of each subarray, the deviation of the angle estimate of each subarray relative to the common reference direction is calculated. Substituting the deviation into an approximate linear model characterizing the relationship between angle deviation and distance, the initial distance estimate is obtained by solving the least squares fitting algorithm.

6. The array antenna hybrid near-field and far-field beam training method according to claim 5, characterized in that, The initial distance estimate obtained by solving the least squares fitting algorithm includes: Construct an objective function with distance-related parameters as optimization variables; wherein, the objective function represents the sum of squares of the products of the deviation between the optimal angle index of each subarray and the common reference direction, the distance-related parameters, and the center position of the corresponding subarray. Find the distance-related parameter values ​​that minimize the objective function; The reciprocal of the distance-related parameter value is used as the initial distance estimate for the user equipment.

7. The array antenna hybrid near-field beam training method according to claim 1, characterized in that, The near-field focusing beam obtained by performing near-field beam training includes: The statistical center of the optimal angle index and the initial distance estimate are used as the starting point for the iterative search; Within the local region corresponding to the starting point, a second training signal is alternately sent in the angle and distance dimensions to perform a one-dimensional discrete search. The angle estimate and distance estimate are updated based on the received signal quality fed back by the user equipment until the convergence condition is met.

8. The array antenna hybrid near-field and far-field beam training method according to claim 7, characterized in that, The step of alternately performing a one-dimensional discrete search in the angle and distance dimensions within the local region corresponding to the starting point includes: With the current distance estimate fixed, within the preset angle search interval, traverse multiple angle candidate points and send the corresponding near-field focusing training beams. Based on the user equipment feedback, select the angle candidate point with the best reception quality and update the angle estimate. The angle estimate is fixed and updated. Within the preset distance search interval, multiple distance candidate points are traversed and the corresponding near-field focusing training beams are sent. The distance candidate point with the best reception quality is selected based on the user equipment feedback to update the distance estimate.

9. The method for training hybrid far- and near-field beamforming of an array antenna according to any one of claims 1-8, characterized in that, The step of dividing the phased array into multiple subarrays composed of adjacent array elements includes: dividing the phased array into multiple consecutive subarrays with equal apertures according to the geometric arrangement of the phased array; the geometric center positions of the array surfaces of each subarray are different, and the geometric center of the array surfaces of each subarray has a known positional offset in the direction of the array axis. The far-field angle codebook is a codebook constructed based on the discrete Fourier transform.

10. A hybrid near-field and far-field beamforming training device with an array antenna, characterized in that, include: The partitioning module is used to divide the phased array into multiple subarrays composed of adjacent array elements; The transmitting module is used to transmit a first training signal based on the same far-field angle codebook to the user equipment through each of the subarrays; The acquisition module is used to acquire the optimal angle index corresponding to each subarray, fed back by the user equipment; The determination module is used to determine the downlink beam based on the optimal angle index when it is determined that the user equipment is in the far field based on the angle consistency metric between the optimal angle indices. Alternatively, when it is determined that the user equipment is in the near field based on the angular consistency metric between the optimal angle indices, an initial distance estimate is generated according to the optimal angle index, and near-field beam training is performed to obtain a near-field focusing beam.

Images