A beam tracking method, device

Abstract

The application relates to the technical field of millimeter wave communication, and discloses a beam tracking method and device, wherein the method comprises the following steps: acquiring an initial beam pair, performing beam tracking based on the initial beam pair, the initial beam pair being obtained by performing beam training on out-of-band information at an initial moment, when the beam tracking reaches a beam prediction moment, acquiring historical tracking information in a historical first period, predicting whether beam training needs to be performed again in a future second period based on the historical tracking information, if the prediction result indicates that the beam training needs to be performed again, determining a beam training moment based on the prediction result, so that, when the beam tracking reaches the beam training moment, the beam training is performed again to obtain an updated beam pair, and the beam tracking is continued based on the updated beam pair. The technical scheme provided by the application can reduce the training cost of beam tracking, ensure the link quality of beam tracking, and achieve the balance between low training cost and high link quality.

Description

Technical Field

[0001] This application relates to the field of millimeter-wave communication technology, and in particular to a beam tracking method and device. Background Technology

[0002] Currently, millimeter-wave signal transmission often relies on waveform shaping technology, using directional beams formed by base stations and terminals equipped with large-scale antenna arrays for communication. This beamforming technology depends on precise beam alignment and continuous beam tracking to maintain the link quality of millimeter-wave communication.

[0003] In related technologies, beam tracking generally employs fixed-period beam training, i.e., beam alignment is achieved by periodically updating beam pairs. However, when the motion state of the terminal changes dynamically, frequent beam training generates a large amount of redundant training overhead, while untimely beam training leads to severe beam misalignment and link quality degradation, making it impossible to achieve a balance between low training overhead and high link quality.

[0004] Therefore, how to reduce the training overhead of beam tracking while ensuring the link quality of beam tracking has become a key research focus in the field of millimeter-wave communication. Summary of the Invention

[0005] This application provides a beam tracking method and device that can reduce the training overhead of beam tracking while ensuring the link quality of beam tracking, achieving a balance between low training overhead and high link quality.

[0006] This application provides a beam tracking method, the method comprising: acquiring an initial beam pair; performing beam tracking based on the initial beam pair, wherein the initial beam pair is obtained by beam training on out-of-band information at an initial time; when beam tracking reaches a beam prediction time, acquiring historical tracking information within a first historical time period; predicting, based on the historical tracking information, whether beam training needs to be re-performed in a future second time period; if the prediction result indicates that beam training needs to be re-performed, determining a beam training time based on the prediction result, such that when beam tracking reaches the beam training time, beam training is re-executed to obtain an updated beam pair, and beam tracking continues based on the updated beam pair.

[0007] In one implementation, after predicting whether beam training is needed in the second future time period, the method further includes: if the prediction result indicates that beam training is not needed, determining the next beam prediction time based on the current beam prediction time and the second time period; if the prediction result indicates that beam training is needed, determining the next beam prediction time based on the beam training time and the first time period.

[0008] In one implementation, predicting whether beam training is needed in a future second time period based on the historical tracking information includes: using a pre-trained time series prediction network to perform time series prediction based on the historical tracking information to obtain the mean signal-to-noise ratio (SNR) sequence and the SNR variance sequence for the future second time period; determining the SNR confidence interval for each moment in the second time period based on the mean SNR sequence and the SNR variance sequence; comparing the SNR confidence interval for each moment with a preset boundary threshold; and determining whether beam training is needed based on the comparison results for each moment.

[0009] In one implementation, determining whether beam training is needed based on the comparison results at each time step includes: if the comparison result at any time step indicates that the lower interval value of the signal-to-noise ratio confidence interval is less than the preset boundary threshold, it is determined that beam training is needed; if the comparison results at each time step indicate that the lower interval value of the signal-to-noise ratio confidence interval is greater than or equal to the preset boundary threshold, it is determined that beam training is not needed.

[0010] In one embodiment, the time series prediction network includes a mean prediction network and a variance prediction network; based on the historical tracking information, performing time series prediction using the pre-trained time series prediction network includes: inputting the historical tracking information into the mean prediction network, performing mean prediction processing using the mean prediction network to obtain the signal-to-noise ratio mean sequence for a future second time period; inputting the historical tracking information into the variance prediction network, performing variance prediction processing using the variance prediction network to obtain the signal-to-noise ratio variance sequence for a future second time period; wherein, the historical tracking information includes the historical signal-to-noise ratio sequence for a historical first time period and historical out-of-band information.

[0011] In one implementation, the beam training is performed as follows: acquiring out-of-band information at the current moment, determining a first line-of-sight center and a second line-of-sight center based on the out-of-band information; performing beam detection based on the first line-of-sight center and the second line-of-sight center respectively, determining multiple candidate beam pairs based on the detection results, and acquiring the signal-to-noise ratio (SNR) of each candidate beam pair; determining a target beam pair from the multiple candidate beam pairs based on the SNR of each candidate beam pair, and using the target beam pair as an initial beam pair or an updated beam pair.

[0012] In one embodiment, beam detection is performed based on the first line-of-sight center and the second line-of-sight center, respectively. Determining multiple candidate beam pairs based on the detection results includes: performing beam detection based on the first line-of-sight center to obtain a first candidate beam set, and performing beam detection based on the second line-of-sight center to obtain a second candidate beam set; for any first candidate beam in the first candidate beam set, traversing multiple second candidate beams in the second candidate set, and combining the first candidate beam with any second candidate beam to form a candidate beam pair, thereby obtaining multiple candidate beam pairs.

[0013] In one implementation, beam detection based on the first line-of-sight center to obtain a first candidate beam set includes: determining multiple first initial beams located within a first search radius of the first line-of-sight center based on a pre-trained transmission network, and determining the conditional likelihood value of each first initial beam; for any first initial beam, if the conditional likelihood value of the first initial beam is determined to be greater than a preset threshold, the first initial beam is added to the first candidate beam set.

[0014] In one embodiment, beam detection based on the second line-of-sight center to obtain a second candidate beam set includes: determining a plurality of second initial beams located within a second search radius of the second line-of-sight center based on a pre-trained receiving network, and determining the conditional likelihood value of each second initial beam; for any second initial beam, if it is determined that the conditional likelihood value of the second initial beam is greater than a preset threshold, the second initial beam is added to the second candidate beam set.

[0015] A second aspect of this application provides a computer device, including: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the computer instructions to perform the beam tracking method as described in the first aspect.

[0016] The technical solution provided in one or more embodiments of this application performs advance prediction and adaptive triggering of beam training based on historical tracking information, achieving low-overhead and high-reliability beam tracking. Specifically, during beam tracking, it is determined whether beam training should be performed based on historical tracking information. If the beam prediction result indicates that beam training is not required, the tracking process continues to advance according to the initial beam pair and prediction period. If the prediction indicates that beam training is required, the beam training time is further determined, and beam training is re-executed when that time actually arrives. The subsequent communication quality is maintained by updating the beam pair, thereby enabling the beam training time to be adaptively adjusted according to changes in channel state and terminal motion state.

[0017] Compared to the forced beam training based on a fixed period in traditional technologies, this technical solution can reduce the training overhead of beam tracking while ensuring the link quality of beam tracking, so as to achieve a balance between low training overhead and high link quality, thereby avoiding the problem of large training overhead or beam misalignment caused by the changing motion state of the terminal. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the specific embodiments of this application or the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0019] Figure 1 A schematic diagram illustrating the steps of a beam tracking method provided in this application embodiment; Figure 2 A schematic diagram illustrating the distribution of a signal-to-noise ratio confidence interval, provided for one embodiment of this application; Figure 3 A schematic diagram illustrating the beam training steps provided in one embodiment of this application; Figure 4 A schematic diagram of beam training search provided in one embodiment of this application; Figure 5 A timing diagram of conventional beam tracking provided for one embodiment of this application; Figure 6 A timing diagram illustrating a beam tracking method provided in one embodiment of this application; Figure 7 A schematic diagram of the structure of a beam tracking device provided in one embodiment of this application; Figure 8 This is a schematic diagram of the structure of a computer device provided in one embodiment of this application. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0021] Furthermore, the use of terms such as "first," "second," etc., in this application is for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of embodiments in this application, unless otherwise stated, "multiple" means two or more. Additionally, the use of "based on" or "according to" implies openness and inclusiveness, because processes, steps, calculations, or other actions "based on" or "according to" one or more of the stated conditions or values ​​may in practice be based on additional conditions or beyond the stated values.

[0022] With the continuous development of the wireless communication industry, especially in 5G and 6G wireless communication, the millimeter-wave band is considered crucial for meeting the explosive growth of data traffic due to its abundant spectrum resources. Millimeter-wave communication typically relies on waveform shaping technology, which uses base stations and terminals equipped with massive MIMO antenna arrays to continuously perform beam alignment and beam tracking to form the required high-gain directional beams for communication. Traditional beam tracking schemes often employ fixed-period beam training to update beam pairs and maintain communication link quality. However, this approach lacks the ability to detect dynamic changes in the channel and cannot achieve an adaptive balance between training overhead and link stability.

[0023] For example, in beam communication scenarios based on UAV communication, millimeter-wave links typically rely on line-of-sight propagation, requiring high-gain directional beams to compensate for path loss. However, the high-speed mobility, frequently changing flight attitudes, and dynamic channel conditions of UAVs pose significant challenges to beam tracking. Using a fixed-period beam tracking scheme cannot adapt to the dynamic changes in UAV motion. For instance, when the UAV is relatively stationary or moving at low speed, frequent beam training and tracking, while maintaining beam alignment, incurs unnecessary training and signaling overhead, which is particularly detrimental in energy-constrained UAV scenarios. For example, when the UAV's attitude changes abruptly or it performs high-speed maneuvers, a fixed period may fail to track beam changes in time, leading to a sharp drop in link quality or even link interruption, thus failing to achieve an adaptive balance between training overhead and link stability. The above description is merely a scenario example provided in the specification and is not intended to limit the invention. Any modifications or equivalent substitutions made within the spirit and principles of this invention should be included within the scope of protection of this invention.

[0024] In view of the above, this application provides one or more embodiments of a beam tracking method or device that can solve the above problems, reduce the training overhead of beam tracking while ensuring the link quality of beam tracking, and achieve a balance between low training overhead and high link quality.

[0025] Firstly, please refer toFigure 1 One embodiment of this application provides a beam tracking method, which may include the following steps: S1: Obtain an initial beam pair and perform beam tracking based on the initial beam pair, wherein the initial beam pair is obtained by beam training on out-of-band information at the initial time. In millimeter-wave communication scenarios, both base stations and terminals are equipped with large-scale wireless arrays, forming directional electromagnetic beams by adjusting the phase and amplitude of antenna elements. A beam pair, as the combination of the base station's transmit beam and the terminal's receive beam, is essential for establishing a high-quality communication link; only when the beam pairs are aligned can a high-quality communication link be established. The aforementioned initial beam pair is the beam pair determined through beam training at the initial moment of beam tracking, and communication transmission is based on this initial beam pair until the beam training moment arrives. The out-of-band information may include position and attitude information characterizing the terminal's motion state, transmitted through a separate navigation channel. For example, in a millimeter-wave communication scenario based on UAV communication, the out-of-band information includes three-dimensional position coordinates (e.g., x, y, z) and three-dimensional attitude angles (e.g., pitch angle α, roll angle β, and yaw angle γ), acquired by the UAV's GPS positioning system and INS inertial navigation system, respectively.

[0026] The beam tracking operation described above can be understood as the process of continuously monitoring channel changes and updating beam pairs as necessary after initial beam alignment to maintain communication link quality. The beam training operation described above can be understood as the process of finding the optimal beam pair by measuring and comparing the signal quality (e.g., represented by signal-to-noise ratio) of different beam pair combinations. The beam pair combinations described above are possible combinations of any transmit beam from the base station and any receive beam from the terminal. At the initial moment of beam tracking initiation, the communication link has not yet been established; therefore, it is necessary to first determine the initial beam pair based on the out-of-band information described above to establish a usable communication link. For example, the beam training operation described above can simply adopt an exhaustive search method, that is, by obtaining the signal-to-noise ratio of each beam pair combination, the beam pair combination with the highest signal-to-noise ratio is determined as the required beam pair (initial beam pair or updated beam pair).

[0027] For example, a local search method based on the line-of-sight center is used to determine the required beam pair. This involves using the out-of-band information to determine the line-of-sight direction between the base station and the terminal, and then quickly locating a possible range among the possible beam pair combinations based on the line-of-sight direction. The beam pair combination with the highest signal-to-noise ratio within this possible range is then selected as the required beam pair. Optionally, after defining the possible range based on the line-of-sight direction, the possible range can be further narrowed to further reduce training overhead. For example, beam pair combinations that do not meet a preset threshold can be pre-filtered from the possible range using methods such as conditional likelihood filtering or confidence filtering.

[0028] S3: When beam tracking reaches the beam prediction time, obtain historical tracking information within the first historical time period, and based on the historical tracking information, predict whether beam training needs to be performed again in the second future time period. The aforementioned beam prediction time can be understood as the prediction time node determined during beam tracking to determine whether beam training is required. The first beam prediction time after the initial moment in beam tracking is a pre-set time node, and subsequent beam prediction times are adaptively adjusted time nodes based on historical tracking information. The aforementioned first time period is a continuous time window preceding the beam prediction time, using historical tracking information from the first time period as reference information for beam prediction. This historical tracking information may include historical signal-to-noise ratio (SNR) sequences and historical out-of-band information. The historical SNR sequence is a sequence formed by arranging the SNR of each moment within the historical first time period in chronological order, and the historical out-of-band information includes the out-of-band information of each moment within the historical first time period. The aforementioned second time period is a pre-set complete beam prediction cycle. Under ideal conditions where the channel state and terminal motion state do not change (i.e., when beam training is not required for prediction representation), beam training is not required during the second time period, and the next moment after the end of the second time period is taken as the next beam prediction time.

[0029] Traditional beam tracking schemes typically use a fixed period for beam prediction. This beam prediction determines whether beam training is needed within the fixed period to periodically update the beam pairs. Due to the volatile motion of the terminal, this can easily lead to significant training overhead or beam inaccuracies. Therefore, by predicting the need for beam training and the timing of beam training in advance based on historical tracking information, the timing of beam training and beam prediction can be adaptively adjusted when there are significant changes in state information. Optionally, a first time period can be used as the next prediction interval after beam training, and a second time period can be used when the terminal motion is smooth, the channel is stable, and beam training is not triggered, thereby achieving low-overhead, high-reliability beam tracking. Specifically, predicting whether beam training is needed in the second future time period based on historical tracking information can be achieved using a pre-trained neural network to predict the signal-to-noise ratio (SNR). The predicted SNR serves as a representation of the channel state and terminal motion state at various future times. For example, historical tracking information is processed by a pre-trained neural network to output relevant prediction information of the signal-to-noise ratio at each time point in the future second time period, such as the mean and variance of the signal-to-noise ratio. Then, whether beam training is needed is determined by triggering a preset prediction information threshold or a preset reliability threshold.

[0030] S5: If the prediction result indicates that beam training needs to be re-performed, determine the beam training time based on the prediction result, so that when beam tracking reaches the beam training time, beam training is re-executed to obtain an updated beam pair, and beam tracking continues based on the updated beam pair.

[0031] The above prediction results include beam training triggering results at each moment within the second time period. Based on these prediction results, the moment when beam training needs to be triggered, i.e., the beam training moment, can be determined. It should be noted that there is at most one beam training moment within the second time period. During the prediction process, beam training is sequentially inferred according to the temporal order of each moment within the second time period. When it is inferred that beam training is required at a certain moment, that moment is designated as the beam training moment, and the prediction result indicates that beam training needs to be performed again. Subsequent moments are not inferred. When the beam tracking process actually progresses to the beam training moment, the beam training operation is immediately triggered to re-determine the beam pair, i.e., the updated beam pair obtained above, to replace the initial beam pair as the beam configuration for maintaining millimeter-wave communication for subsequent beam tracking. For reference, if the current beam training moment is not the first beam training moment determined through beam training, the updated beam pair obtained at the current beam training moment replaces the updated beam pair obtained in the previous round of beam training.

[0032] In this embodiment, if the prediction result indicates that beam training is not required, the beam tracking process continues normally until the end of the next second time period. At the end of the next second time period, beam prediction is restarted for the next second time period, i.e., historical tracking information from the first time period before the end of the second time period is obtained, predicting whether beam training needs to be re-performed in the second time period after the end of the second time period. Ideally, in the case where beam training is not required throughout the entire beam tracking process, the beam tracking process should continuously repeat with the second time period as the prediction cycle.

[0033] For example, beam tracking based on updated beam pairs includes: determining the next beam prediction time based on the prediction result; when the beam tracking process progresses to the next beam prediction time, obtaining historical tracking information within a historical first time period before the next beam prediction time; predicting whether beam training needs to be re-performed within a future second time period after the next beam prediction time based on the historical tracking information; if the prediction result indicates that beam training needs to be re-performed, determining the next beam training time based on the prediction result, so that when beam tracking progresses to the next beam training time, beam training is re-executed to obtain a new updated beam pair, and beam tracking continues based on the newly obtained updated beam pair.

[0034] Based on the above ideas, the technical solution provided in this embodiment of the application performs advance prediction and adaptive triggering of beam training based on historical tracking information, achieving low-overhead and high-reliability beam tracking. Specifically, during beam tracking, the system determines whether beam training should be performed based on historical tracking information. If the beam prediction result indicates that beam training is not required, the tracking process continues according to the initial beam pair and prediction period (i.e., the second time period). If the prediction indicates that beam training is required, the beam training time is further determined, and beam training is re-executed when that time actually arrives. The subsequent communication quality is maintained by updating the beam pair, thereby enabling the beam training time to be adaptively adjusted according to changes in channel state and terminal motion state. Compared with the forced beam training based on a fixed period in traditional technologies, this technical solution can reduce the training overhead of beam tracking and ensure the link quality of beam tracking, achieving a balance between low training overhead and high link quality, thus avoiding the problem of large training overhead or beam inaccuracy caused by the changing motion state of the terminal.

[0035] In some implementations, after predicting whether beam training is needed in the second future time period, the next beam prediction time needs to be determined based on the prediction results. The beam prediction time period and the beam prediction time are adaptively adjusted according to different prediction results, thereby avoiding redundant training overhead. A reasonable beam prediction time can also avoid misjudgments or omissions in beam training timing, thus improving the reliability of beam tracking. Specifically, if the prediction results indicate that beam training is not needed, the next beam prediction time is determined based on the current beam prediction time and the second time period; if the prediction results indicate that beam training is needed, the next beam prediction time is determined based on the beam training time and the first time period.

[0036] In one implementation, when the prediction result indicates that beam training is not required in the second time period, it means that the current prediction determines that the communication link is stable in the second time period and there is no need to make predictions again in this period. The next beam prediction time is determined to be the current prediction time plus the length of the second time period, so that a new round of prediction can be started after the current prediction ends, thereby avoiding overly frequent prediction calculations and ensuring that there are no misjudgments or omissions in beam training time.

[0037] In one implementation, when the prediction result indicates that beam training is needed in the second future time period, the beam training time is first determined, and then the next beam prediction time is determined as the beam training time plus the length of the first time period. Since the beam pair changes after beam training, historical signal-to-noise ratio (SNR) data is no longer compatible with the updated beam pair. Setting the next beam prediction time after the beam training time, within the first time period, ensures that sufficient historical SNR data for the updated beam pair has been accumulated when the next round of prediction begins, thereby ensuring the accuracy of beam prediction and further improving the link quality of beam tracking.

[0038] In one embodiment, the first time period and the second time period are pre-adjusted. Specifically, if the first time period is too short, it will lead to a decrease in communication link quality due to inaccurate prediction; if it is too long, it will lead to a decrease in communication link quality due to low prediction frequency. If the second time period is too short, it will lead to a waste of computing resources due to excessive prediction; if the second time period is too long, it will lead to a decrease in prediction accuracy due to increased training difficulty. Preferably, the duration of the first time period is 7 CSI cycles, and the duration of the second time period is 5 CSI cycles, wherein the time interval between adjacent moments is one CSI cycle, typically 10ms, equivalent to the first time period being 70ms and the second time period being 50ms.

[0039] The technical solutions provided by these implementation methods dynamically determine the next beam prediction time based on the prediction results, so as to adaptively adjust the beam prediction period, thereby further reducing training overhead and improving link quality. Specifically, if the prediction indicates that no training is needed, the next beam prediction time is set to the beam prediction time plus the length of the second time segment, avoiding redundant computational overhead or misjudgment and omission of the beam training time; if the prediction indicates that training is needed, the next beam prediction time is set to the beam training time plus the length of the first time segment, thereby ensuring that there is enough historical signal-to-noise ratio data for the next round of prediction, thus ensuring the accuracy of beam prediction and further improving the link quality of beam tracking.

[0040] In some implementations, signal-to-noise ratio (SNR) related information for each moment within a future second time period is inferred based on historical tracking information, such as the SNR confidence interval for each moment. This information is then used to predict whether beam training is needed within the future second time period. Specifically, based on historical tracking information, a pre-trained time-series prediction network is used to perform time-series prediction to obtain the mean SNR sequence and variance SNR sequence for the future second time period. Based on the mean SNR sequence and variance SNR sequence, the SNR confidence interval for each moment within the second time period is determined. The SNR confidence interval for each moment is compared with a preset boundary threshold, and based on the comparison results for each moment, it is determined whether beam training is needed. For example, according to the temporal order of each moment within the second time period, the SNR confidence interval corresponding to each moment is compared with the preset boundary threshold. If the SNR confidence interval for any moment exceeds the preset boundary threshold, it is determined that beam training is needed within the second time period.

[0041] The aforementioned time-series prediction network is a pre-trained neural network for processing time-series data, such as a recurrent neural network, convolutional neural network, support vector machine, and random forest. The historical tracking information includes historical signal-to-noise ratio (SNR) sequences and historical out-of-band information. The aforementioned time-series prediction network can be divided into two sub-networks to predict the mean SNR sequence and the variance SNR sequence, respectively. The mean SNR sequence reflects the central position of the expected SNR distribution at each time point, the variance SNR sequence reflects the degree of SNR fluctuation at each time point, and the SNR confidence interval is a numerical range calculated based on the mean and variance SNR at the corresponding time point according to a specific confidence level, for example, to reflect the uncertainty of the prediction.

[0042] In one implementation, determining whether beam training is needed based on the comparison results at each time point includes: if the comparison result at any time point indicates that the lower interval value of the signal-to-noise ratio confidence interval is less than a preset boundary threshold, beam training is determined to be needed; if the comparison results at all times point indicate that the lower interval value of the signal-to-noise ratio confidence interval is greater than or equal to the preset boundary threshold, beam training is determined not to be needed. Specifically, according to the temporal order of the times within the second time period, the first time point at which the corresponding signal-to-noise ratio confidence interval exceeds the preset boundary threshold is taken as the beam training time within the second time period. Traditional techniques only perform beam training after a decline in communication quality is detected, which has a certain lag and may lead to communication interruption. This application, by setting a reasonable preset boundary threshold, can trigger beam training before the actual decline in communication quality, further improving the link reliability of beam tracking.

[0043] In one implementation, the aforementioned time series prediction network includes a pre-trained mean prediction network and a variance prediction network. Time series prediction is performed using the pre-trained mean prediction network and variance prediction network based on historical tracking information. Specifically, historical tracking information is input into the mean prediction network, which performs mean prediction processing to obtain the signal-to-noise ratio (SNR) mean sequence for the next future time period. The mean prediction processing can be understood as the mean prediction network performing forward propagation calculations based on the historical tracking information to obtain the SNR mean at each future time point. Similarly, historical tracking information is input into the variance prediction network, which performs variance prediction processing to obtain the SNR variance sequence for the next future time period. The variance prediction processing can be understood as the variance prediction network performing forward propagation calculations based on the historical tracking information to obtain the SNR variance at each future time point. The aforementioned historical tracking information includes the historical SNR sequence for the first historical time period and historical out-of-band information.

[0044] In one embodiment, future time signal-to-noise ratio It approximately follows a Gaussian distribution At any moment Mean prediction and variance prediction can be performed in the following ways: , ,in, Indicates time The average signal-to-noise ratio, Indicates time The signal-to-noise ratio variance For the first period, This is the second time period. For a moment The out-of-band information can be represented as This includes the terminal's location information (three-dimensional coordinates). 3D coordinates 3D coordinates ) and attitude information (pitch angle) Roll angle Yaw angle ).in, and Let these represent the objective functions of the mean prediction network and the variance prediction network, respectively. These are the learning parameters for the mean prediction network. For the training parameters of the variance prediction network, the objective function described above aims to maximize the log-likelihood of the predictions, and is trained as follows: ,in, During the multi-round training of the time series prediction network, the learning parameters of the mean prediction network and the variance prediction network are adjusted based on the training results of any round.

[0045] In one embodiment, see Figure 2 Future moments The signal-to-noise ratio confidence interval is expressed as ,Right now .in, Indicates time signal-to-noise ratio, Indicates time The average signal-to-noise ratio, Indicates time The signal-to-noise ratio variance The critical value of the standard normal distribution, for example hour, The solid blue line represents the actual measured signal-to-noise ratio and its change over time, the shaded area represents the signal-to-noise ratio confidence interval (prediction confidence interval), and the prediction start point represents the current beam prediction time, from which the prediction of the second future time period begins.

[0046] These implementations provide a technical solution that uses a time-series prediction network to infer from historical tracking information to determine whether beam training is needed and the precise timing for beam training. Specifically, the time-series prediction network includes a mean prediction network and a variance prediction network. During the training phase, these mean and variance prediction networks aim to maximize the predicted log-likelihood, enabling the network to accurately fit the conditional distribution characteristics of the signal-to-noise ratio (SNR), resulting in more accurate subsequent predictions and ensuring the link quality of beam tracking. Simultaneously, threshold condition triggering based on the SNR confidence interval allows for the early decision-making of beam training before channel quality deteriorates, avoiding link interruptions caused by beam training lag and further improving the link reliability of beam tracking.

[0047] In some implementations, please refer to Figure 3 Before beam tracking, beam training is required to determine the initial beam pair. During beam tracking, it is necessary to determine whether beam training is needed again based on the prediction results to obtain an updated beam pair. The beam training described above is based on out-of-band information at the initial time or the predicted beam training time. Specifically, the beam training is performed according to the following steps: S11: Obtain the out-of-band information at the current moment, and determine the first line-of-sight center and the second line-of-sight center based on the out-of-band information; S13: Perform beam detection based on the first line-of-sight center and the second line-of-sight center respectively, determine multiple candidate beam pairs based on the detection results, and obtain the signal-to-noise ratio of each candidate beam pair; S15: Based on the signal-to-noise ratio of each candidate beam pair, determine the target beam pair among multiple candidate beam pairs, and use the target beam pair as the initial beam pair or the updated beam pair.

[0048] The first line-of-sight center is the beam pointing along the ideal line-of-sight propagation direction determined by the base station, and the second line-of-sight center is the beam pointing along the ideal line-of-sight propagation direction determined by the terminal. By determining the line-of-sight center, the beam search range is significantly reduced, thus lowering training overhead. In step S11, the first and second line-of-sight centers are determined based on the line-of-sight center in the local coordinate system, which is determined based on out-of-band information. Specifically, based on the terminal's attitude and coordinate information, the beam indices and relative azimuth angles of the transmitter and receiver arrays are determined. The line-of-sight center in the local coordinate system is then determined based on the beam indices and relative azimuth angles, and beam index mapping is performed to obtain the line-of-sight centers of the base station and the terminal.

[0049] For example, utilizing out-of-band information Obtain the line-of-sight center in the local coordinate system. The line-of-sight centers are mapped to the codebooks of the base station and the terminal, respectively, to obtain the beam indices of the first line-of-sight center and the second line-of-sight center, so as to determine the first line-of-sight center and the second line-of-sight center in the channel model according to their respective beam indices.

[0050] In one embodiment, in step S13 above, beam detection is performed based on a first line-of-sight center and a second line-of-sight center, respectively, and multiple candidate beam pairs are determined based on the detection results. The beam detection operation described above can be understood as evaluating the communication performance of a specific beam or beam pair by actually transmitting test signals and measuring the received quality within a beam space defined by the line-of-sight center.

[0051] Specifically, beam detection is performed based on a first line-of-sight center to obtain a first candidate beam set. This involves defining a first beam search space within a preset search radius on the base station side using the first line-of-sight center, selecting each beam within the first beam search space as a first candidate beam, and evaluating the communication performance of each beam within the first beam search space. Similarly, beam detection is performed based on a second line-of-sight center to obtain a second candidate beam set. This involves defining a second beam search space within a preset search radius on the terminal side using the second line-of-sight center, selecting each beam within the second beam search space as a second candidate beam, and evaluating the communication performance of each beam within the second beam search space. Further, for any first candidate beam in the first candidate beam set, multiple second candidate beams in the second candidate beam set are traversed, and each first candidate beam is combined with any second candidate beam to form a candidate beam pair, resulting in multiple candidate beam pairs.

[0052] In one implementation, a pre-trained transmission network is used for beam detection based on a first line-of-sight center to obtain a first candidate beam set. The transmission network serves as a pre-trained base station-side neural network, used to predict the probability distribution, i.e., the conditional likelihood value, of each beam at the base station becoming a target transmission beam. Specifically, based on the pre-trained transmission network, out-of-band information and the first line-of-sight center are processed to determine multiple first initial beams located within a first search radius of the first line-of-sight center, and to determine the conditional likelihood value of each first initial beam. For any first initial beam, if the conditional likelihood value of the first initial beam is determined to be greater than a preset threshold, the first initial beam is added to the first candidate beam set.

[0053] In one implementation, a pre-trained receiving network is used for beam detection based on a second line-of-sight center to obtain a second candidate beam set. The aforementioned transmitting network serves as a pre-trained terminal neural network, used to predict the probability distribution, i.e., the conditional likelihood value, of each beam of the terminal becoming the target receiving beam. Specifically, based on the pre-trained receiving network, out-of-band information and the second line-of-sight center are processed to determine multiple second initial beams located within a second search radius of the second line-of-sight center, and to determine the conditional likelihood value of each second initial beam. For any second initial beam, if the conditional likelihood value of the second initial beam is determined to be greater than a preset threshold, the second initial beam is added to the second candidate beam set. The aforementioned first search radius and second search radius are preset, preferably the same search radius.

[0054] In one embodiment, see Figure 4 Taking drones as terminal devices as an example, through the transmission network and receiving network The search space of the first beam is determined respectively. Second beam search space and output the first search radius Second search radius Conditional likelihood of each beam within the beam and In the search space and Within each case, the conditional likelihood exceeds the threshold. The beam was selected as the corresponding first candidate beam set. Second candidate beam set After obtaining the candidate beam set, the base station transmits the beams sequentially. Meanwhile, the terminal traverses the collection. To select the appropriate beam This forms a beam pair for communication. Select the beam pair with the highest signal-to-noise ratio as the target beam pair, and use the target beam pair as the initial beam pair or the updated beam pair.

[0055] In one embodiment, the beam training process described above can be represented as follows: ,in, , , .in, Indicates the selected beam pair. This represents the signal-to-noise ratio value obtained for the corresponding beam pair. Indicates the first candidate beam set. Indicates the second candidate beam set. Indicates the transmission network. Indicates the receiving network. Indicates the preset threshold. This represents the original probability distribution output by the transmitting network. This represents the original probability distribution of the network output.

[0056] The technical solutions provided by these implementation methods perform local searches based on the line-of-sight center and narrow the search range through probabilistic filtering to further reduce training overhead. Specifically, a first line-of-sight center on the base station side and a second line-of-sight center on the terminal side are determined respectively, thereby significantly narrowing the initial search range to the vicinity of the line-of-sight direction. Subsequently, the line-of-sight center and out-of-band information are processed through pre-trained transmit and receive networks to compress the beam detection range to a local area. Furthermore, candidate beam combinations are determined within the local area defined by the line-of-sight center, thereby selecting the beam pair with the highest signal-to-noise ratio as the target beam pair. While accurately determining the required initial or updated beam pairs, this significantly reduces the number of beam pairs that need to be measured during beam training, thereby effectively reducing training overhead and signaling burden, and significantly improving the robustness of beam alignment and the stability of link quality.

[0057] Please see Figure 5 This application also provides a scenario example of beam tracking in conventional techniques. Figure 5 This is an example diagram of beam tracking based on a fixed period in traditional techniques. Specifically, a fixed period is set. , Fixed at 20 CSI cycles, or 200 ms, each time a Upon timing, regardless of the current link quality, a beam training operation is forcibly triggered, including: traversing all beams within a preset search radius R, centered on the beam directions of both the base station and the terminal; the base station sequentially transmitting the reference signal for each beam; and the terminal sequentially receiving and measuring the signal quality of each beam pair, selecting the beam pair with the best signal quality from all measured beam pairs. As a new communication beam, the base station and terminal switch to the new beam to continue communication, the timer resets, and they wait for the next fixed cycle. This fixed-cycle beam tracking method is not suitable for scenarios where the terminal state changes frequently. For example, if the terminal device is a drone, when the drone is moving smoothly, the fixed cycle will generate a lot of redundant training overhead, wasting spectrum and energy resources. However, when the drone is maneuvering at high speed or its attitude changes drastically, the fixed cycle may cause serious beam misalignment due to untimely updates, resulting in a decrease in communication rate or link interruption. In other words, it is impossible to achieve an adaptive balance between low overhead and high reliability.

[0058] Please see Figure 6 This application also provides a specific scenario example of applying the beam tracking method described above, where a drone is used as the terminal. Specifically, at the initial moment... Perform the first beam training to determine the initial beam pair. The base station and terminal communicate based on the initial beam pair and continuously perform beam tracking. As beam tracking progresses... At a given time, beam prediction is performed, which involves determining whether beam training is needed, and if so, determining the timing for beam training. In this example, The representation of the prediction results of time-based beam prediction requires beam training, and the beam training time is... Then when beam tracking advances to Beam training operations are performed at specific times to determine the updated beam pairs. Furthermore, the next beam prediction time is determined as... , When beam tracking advances to At time 10, the beam prediction operation is performed again. In this example, If the result of the time-based beam prediction does not require beam training, then the next beam prediction time is determined as... , And so on, until the entire beam tracking process is complete. In this example, As a first time period, it is set to 7 CSI cycles. This is a second time period, set as 5 CSI cycles.

[0059] Compared to traditional techniques, this example uses dynamic training triggered by the lower bound of the prediction confidence interval, reducing total training overhead by 43.37%. It also uses conditional likelihood probability to select high-probability beams for training, reducing the average outage probability by 98.47% and improving beam alignment by 3.75%. The technical solution provided in this example can adaptively adjust tracking behavior based on the UAV's motion state and channel changes, proactively mitigating the risk of link quality degradation through an active prediction mechanism, achieving an adaptive balance between low overhead and high reliability in beam tracking.

[0060] Please see Figure 7 This application also provides a beam tracking device, the device comprising: A beam initialization unit 100 is used to acquire an initial beam pair and perform beam tracking based on the initial beam pair, wherein the initial beam pair is obtained by beam training on out-of-band information at the initial time. The beam prediction unit 200 is used to acquire historical tracking information within the first historical time period when beam tracking reaches the beam prediction time, and based on the historical tracking information, predict whether beam training needs to be performed again in the second future time period. The beam training unit 300 is used to determine the beam training time based on the prediction result if the prediction result indicates that beam training needs to be re-performed, so that when beam tracking reaches the beam training time, beam training is re-executed to obtain an updated beam pair, and beam tracking continues based on the updated beam pair.

[0061] in, In one embodiment, the beam prediction unit 200 is specifically used to acquire historical tracking information within a first historical time period when beam tracking reaches the beam prediction time; based on the historical tracking information, perform time series prediction using a pre-trained time series prediction network to obtain the mean signal-to-noise ratio (SNR) sequence and the SNR variance sequence within a future second time period; based on the mean SNR sequence and the SNR variance sequence, determine the SNR confidence interval for each moment within the second time period; compare the SNR confidence interval for each moment with a preset boundary threshold; and based on the comparison results for each moment, determine whether beam training is required.

[0062] In one embodiment, the beam prediction unit 200 is further configured to determine the next beam prediction time based on the current beam prediction time and the second time period when the prediction result indicates that beam training is not required, and to determine the next beam prediction time based on the beam training time and the first time period when the prediction result indicates that beam training is required.

[0063] In one embodiment, the beam training unit 300 is specifically configured to, if the prediction result indicates that beam training needs to be re-performed, determine the beam training time based on the prediction result, so that when beam tracking reaches the beam training time, it acquires out-of-band information at the current time, determines a first line-of-sight center and a second line-of-sight center based on the out-of-band information, performs beam detection based on the first line-of-sight center and the second line-of-sight center respectively, determines multiple candidate beam pairs based on the detection results, acquires the signal-to-noise ratio of each candidate beam pair, determines a target beam pair among the multiple candidate beam pairs based on the signal-to-noise ratio of each candidate beam pair, uses the target beam pair as the updated beam pair, and continues beam tracking based on the updated beam pair.

[0064] Further functional descriptions of the above modules and units are the same as those in the corresponding embodiments described above, and will not be repeated here.

[0065] One beam tracking device in this application embodiment is presented in the form of a functional unit. Here, a unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that execute one or more software or fixed programs, or other devices that can provide the above functions.

[0066] Please see Figure 8 , Figure 8 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application, such as... Figure 8 As shown, the computer device includes one or more processors 10, memory 20, and interfaces for connecting the components, including high-speed interfaces and low-speed interfaces. The components communicate with each other via different buses and can be mounted on a common motherboard or otherwise installed as needed. The processors can process instructions executed within the computer device, including instructions stored in or on memory to display graphical information of a GUI on external input / output devices (such as display devices coupled to the interfaces). In some alternative implementations, multiple processors and / or multiple buses can be used with multiple memories and multiple memory modules, if desired. Similarly, multiple computer devices can be connected, each providing some of the necessary operations (e.g., as a server array, a group of blade servers, or a multiprocessor system). Figure 8 Take a processor 10 as an example.

[0067] Processor 10 may be a central processing unit, a network processor, or a combination thereof. Processor 10 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. The programmable logic device may be a complex programmable logic device (CAMP), a field-programmable gate array (FPGA), a general-purpose array logic (GDA), or any combination thereof.

[0068] The memory 20 stores instructions executable by at least one processor 10 to cause the at least one processor 10 to perform the method shown in the above embodiments.

[0069] The memory 20 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on the use of the computer device. Furthermore, the memory 20 may include high-speed random access memory and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some alternative embodiments, the memory 20 may optionally include memory remotely located relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0070] The memory 20 may include volatile memory, such as random access memory; the memory may also include non-volatile memory, such as flash memory, hard disk or solid-state drive; the memory 20 may also include a combination of the above types of memory.

[0071] The computer device also includes a communication interface 30 for communicating with other devices or communication networks.

[0072] This application also provides a computer-readable storage medium. The methods described in this application can be implemented in hardware or firmware, or implemented as recordable on a storage medium, or implemented as computer code downloaded over a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and subsequently stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code. When the software or computer code is accessed and executed by the computer, processor, or hardware, the methods shown in the above embodiments are implemented.

[0073] The apparatus, module, or unit described in the above embodiments can be implemented by a computer chip or entity, or by a product having a certain function. A typical implementation device is a computer. Specifically, the computer can be, for example, a personal computer, laptop computer, cellular phone, camera phone, smartphone, personal digital assistant, media player, navigation device, email device, game console, tablet computer, wearable device, or any combination of these devices.

[0074] For ease of description, the above devices are described separately by function as various units. Of course, in implementing this application, the functions of each unit can be implemented in one or more software and / or hardware.

[0075] Those skilled in the art will understand that embodiments of this application can be provided as methods, apparatus, or computer devices. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0076] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus, and computer devices according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0077] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0078] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0079] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0080] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the device and apparatus embodiments are basically similar to the method embodiments, so the descriptions are relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0081] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

[0082] Although embodiments of this application have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of this application, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A beam tracking method, characterized in that, The method includes: Obtain an initial beam pair and perform beam tracking based on the initial beam pair, wherein the initial beam pair is obtained by beam training on out-of-band information at the initial time. When beam tracking reaches the beam prediction time, historical tracking information within the first historical time period is obtained. Based on the historical tracking information, it is predicted whether beam training needs to be performed again in the second historical time period. If the prediction result indicates that beam training needs to be re-performed, the beam training time is determined based on the prediction result, so that when beam tracking reaches the beam training time, beam training is re-executed to obtain an updated beam pair, and beam tracking continues based on the updated beam pair.

2. The method according to claim 1, characterized in that, After predicting whether beam training is needed in the second future time period, the method further includes: If the prediction result indicates that the beam training is not required, the next beam prediction time is determined based on the current beam prediction time and the second time period. If the prediction result indicates that beam training is required, the next beam prediction time is determined based on the beam training time and the first time period.

3. The method according to claim 1, characterized in that, Based on the historical tracking information, predicting whether beam training is needed in the second future time period includes: Based on the historical tracking information, a pre-trained time series prediction network is used to perform time series prediction in order to obtain the mean signal-to-noise ratio sequence and the variance signal-to-noise ratio sequence for the second future time period. Based on the signal-to-noise ratio mean sequence and the signal-to-noise ratio variance sequence, the signal-to-noise ratio confidence interval for each time point within the second time period is determined; The signal-to-noise ratio confidence interval at each time point is compared with a preset boundary threshold. Based on the comparison results at each time point, it is determined whether beam training is required.

4. The method according to claim 3, characterized in that, Based on the comparison results at each time point, determining whether beam training is needed includes: If the comparison result at any time indicates that the lower interval value of the signal-to-noise ratio confidence interval is less than the preset boundary threshold, it is determined that beam training is required. If the comparison results at each time point all indicate that the lower interval value of the signal-to-noise ratio confidence interval is greater than or equal to the preset boundary threshold, then it is determined that the beam training is not required.

5. The method according to claim 3, characterized in that, The time series prediction network includes a mean prediction network and a variance prediction network; Based on the historical tracking information, time series prediction using a pre-trained time series prediction network includes: The historical tracking information is input into the mean prediction network, and the mean prediction network is used to perform mean prediction processing to obtain the signal-to-noise ratio mean sequence in the second future time period. The historical tracking information is input into the variance prediction network, and the variance prediction network is used to perform variance prediction processing to obtain the signal-to-noise ratio variance sequence in the second future time period. The historical tracking information includes the historical signal-to-noise ratio sequence and historical out-of-band information within the first historical time period.

6. The method according to claim 1, characterized in that, The beam training is performed in the following manner: Obtain the out-of-band information at the current moment, and determine the first line-of-sight center and the second line-of-sight center based on the out-of-band information; Beam detection is performed based on the first line-of-sight center and the second line-of-sight center respectively. Based on the detection results, multiple candidate beam pairs are determined, and the signal-to-noise ratio of each candidate beam pair is obtained. Based on the signal-to-noise ratio of each candidate beam pair, a target beam pair is determined from the plurality of candidate beam pairs, and the target beam pair is used as the initial beam pair or the updated beam pair.

7. The method according to claim 6, characterized in that, Beam detection is performed based on the first line-of-sight center and the second line-of-sight center, respectively. Based on the detection results, multiple candidate beam pairs are determined, including: Beam detection is performed based on the first line-of-sight center to obtain a first candidate beam set, and beam detection is performed based on the second line-of-sight center to obtain a second candidate beam set. For any first candidate beam in the first candidate beam set, traverse multiple second candidate beams in the second candidate set, and combine the first candidate beam with any second candidate beam to form a candidate beam pair, so as to obtain multiple candidate beam pairs.

8. The method according to claim 7, characterized in that, Beam detection is performed based on the first line-of-sight center to obtain the first candidate beam set, which includes: Based on a pre-trained transmission network, a plurality of first initial beams located within a first search radius of the first line-of-sight center are determined, and the conditional likelihood value of each of the first initial beams is determined. For any of the first initial beams, if the conditional likelihood value of the first initial beam is determined to be greater than a preset threshold, the first initial beam is added to the first candidate beam set.

9. The method according to claim 7, characterized in that, Beam detection is performed based on the second line-of-sight center to obtain the second candidate beam set, which includes: Based on a pre-trained receiving network, a plurality of second initial beams located within the second search radius of the second line-of-sight center are determined, and the conditional likelihood value of each second initial beam is determined. For any of the second initial beams, if the conditional likelihood value of the second initial beam is determined to be greater than a preset threshold, the second initial beam is added to the second candidate beam set.

10. A computer device, characterized in that, include: A memory and a processor are communicatively connected, the memory storing computer instructions, and the processor executing the computer instructions to perform the beam tracking method according to any one of claims 1 to 9.

Images