Beam alignment method, apparatus, network device, and storage medium

Abstract

The embodiment of the application discloses a beam alignment method and device, a network device and a storage medium. The method comprises the following steps: determining a predicted beam forming vector corresponding to a current time slot according to an AOA posterior probability vector corresponding to a previous time slot; measuring a received signal measurement value of the current time slot according to the predicted beam forming vector; updating the AOA posterior probability vector of the previous time slot according to the received signal measurement value of the current time slot and the predicted beam forming vector, and obtaining an AOA posterior probability vector corresponding to the current time slot; performing cyclic iteration until the current time slot is the last time slot; and determining a target AOA angle according to the AOA posterior probability vector corresponding to the last time slot. The embodiment can improve the speed and accuracy of beam alignment.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of signal processing, in particular to a beam alignment method and device, a network device and a storage medium. BACKGROUND

[0002] The communication connection between the network device (such as a base station) and the terminal device needs to be transmitted. When the terminal device sends a signal to the network device, the network device needs to determine the angle of arrival (AOA) of the signal between the terminal device and the network device, that is, to perform a beam alignment operation. However, the transmission of the signal in free space will be affected by the environment. According to the free space propagation model, as the frequency of electromagnetic waves increases, the path loss experienced by the signal during transmission will increase exponentially. Although the use of large-scale antenna arrays and beamforming technology can overcome severe path loss and achieve deep signal propagation, it sacrifices the breadth of the signal during transmission. The network device can only obtain the maximum array gain when the transmit and receive beams are aligned. Once the beam alignment deviates, the network device can only obtain limited array gain, resulting in poor communication quality and even communication interruption. Moreover, the directivity of the beam requires the network device to perform beam alignment when initially establishing a communication connection. Slow beam alignment will also affect the communication instructions. Therefore, how to improve the speed and accuracy of beam alignment has become a problem to be solved. SUMMARY

[0003] The embodiments of the present application disclose a beam alignment method, device, network device and storage medium, which can improve the speed and accuracy of beam alignment.

[0004] The embodiments of the present application disclose a beam alignment method applied to a network device, the method comprising:

[0005] determining a predicted beamforming vector corresponding to a current time slot according to an AOA posterior probability vector corresponding to a previous time slot; the AOA posterior probability vector comprises AOA posterior probabilities corresponding to a plurality of first angle intervals in a preset angle range;

[0006] measuring a received signal measurement value of the current time slot according to the predicted beamforming vector;

[0007] updating the AOA posterior probability vector of the previous time slot according to the received signal measurement value of the current time slot and the predicted beamforming vector, to obtain an AOA posterior probability vector corresponding to the current time slot;

[0008] The AOA posterior probability vector corresponding to the current time slot is taken as the AOA posterior probability vector corresponding to the last time slot, and the step of determining the predicted beamforming vector corresponding to the current time slot according to the AOA posterior probability vector corresponding to the last time slot is re-executed until the current time slot is the last time slot.

[0009] The target AOA angle is determined according to the AOA posterior probability vector corresponding to the last time slot.

[0010] In an embodiment, a signal model of the received signal is constructed based on the predicted beamforming vector and the channel path;

[0011] The AOA posterior probability vector corresponding to the last time slot is taken as the AOA posterior probability vector corresponding to the last time slot, and the step of determining the predicted beamforming vector corresponding to the current time slot according to the AOA posterior probability vector corresponding to the last time slot is re-executed until the current time slot is the last time slot.

[0012] The fading coefficient mean and the fading coefficient variance corresponding to the current time slot are obtained by updating the fading coefficient mean and the fading coefficient variance of the channel path corresponding to the last time slot according to the received signal measurement value of the current time slot and the predicted beamforming vector.

[0013] The conditional density function corresponding to the received signal measurement value of the current time slot is determined based on the signal model, the predicted beamforming vector, the fading coefficient mean and the fading coefficient variance corresponding to the current time slot.

[0014] The AOA posterior probability vector corresponding to the current time slot is determined based on the conditional density function corresponding to the current time slot and the AOA posterior probability vector of the last time slot according to the Bayesian formula.

[0015] In an embodiment, before the step of determining the predicted beamforming vector corresponding to the current time slot according to the AOA posterior probability vector corresponding to the last time slot, the method further comprises:

[0016] The predicted beamforming vector corresponding to the first time slot is determined according to the initial AOA posterior probability vector.

[0017] The received signal measurement value of the first time slot is measured according to the predicted beamforming vector corresponding to the first time slot.

[0018] The initial fading coefficient mean and the initial fading coefficient variance of the channel path are updated according to the received signal measurement value of the first time slot and the predicted beamforming vector, to obtain the fading coefficient mean and the fading coefficient variance corresponding to the first time slot.

[0019] determining a conditional density function corresponding to the received signal measurement value of the first time slot according to the predicted beamforming vector, the mean and variance of the fading coefficient corresponding to the first time slot based on the signal model;

[0020] determining the AOA posterior probability vector corresponding to the first time slot based on the conditional density function corresponding to the first time slot and the initial AOA posterior probability vector according to the Bayes formula.

[0021] In one embodiment, the updating of the mean and variance of the fading coefficient of the channel path corresponding to the previous time slot to obtain the mean and variance of the fading coefficient corresponding to the current time slot according to the received signal measurement value of the current time slot and the predicted beamforming vector comprises:

[0022] obtaining the received signal prediction value of the current time slot according to the mean and variance of the fading coefficient of the channel path corresponding to the previous time slot and the predicted beamforming vector, and the signal model;

[0023] determining the signal observation error of the current time slot according to the received signal prediction value of the current time slot and the received signal measurement value of the current time slot;

[0024] updating the mean and variance of the fading coefficient of the channel path corresponding to the previous time slot to obtain the mean and variance of the fading coefficient corresponding to the current time slot according to the signal observation error of the current time slot and the filter gain.

[0025] In one embodiment, the determining of the AOA posterior probability vector corresponding to the current time slot based on the conditional density function and the AOA posterior probability vector of the previous time slot according to the Bayes formula comprises:

[0026] determining the conditional probability corresponding to each of the plurality of first angle intervals according to the conditional density function;

[0027] determining the standard likelihood corresponding to each of the plurality of first angle intervals according to the conditional probability corresponding to each of the plurality of first angle intervals and the AOA posterior probability vector of the previous time slot;

[0028] determining the AOA posterior probability corresponding to each of the plurality of first angle intervals in the current time slot based on the standard likelihood corresponding to each of the plurality of first angle intervals and the AOA posterior probability corresponding to each of the plurality of first angle intervals in the previous time slot according to the Bayes formula.

[0029] In an embodiment, the method further comprises:

[0030] The beamforming prediction neural network performs beamforming vector prediction according to the AOA posterior probability vector corresponding to the previous time slot, a time slot number of the previous time slot, and a transmit power of the terminal, to determine the predicted beamforming vector of the current time slot.

[0031] The method further comprises:

[0032] The angle of arrival prediction neural network performs angle of arrival prediction according to the AOA posterior probability vector corresponding to the last time slot, to determine the target AOA angle.

[0033] The beamforming prediction neural network and the angle of arrival prediction neural network are trained together, and a training set of the beamforming prediction neural network and the angle of arrival prediction neural network comprises a plurality of sample AOA angles.

[0034] In an embodiment, each of the first angle intervals comprises a plurality of second angle intervals; and the method further comprises:

[0035] The target second angle interval is any second angle interval within the preset angle range.

[0036] The AOA posterior probability of each second angle interval included in each first angle interval in the current time slot is added to obtain the AOA posterior probability of each first angle interval in the current time slot.

[0037] Embodiments of the present application disclose a beam alignment device applied to a network device, the device comprising:

[0038] A vector determination module is configured to determine a predicted beamforming vector corresponding to a current time slot according to an angle of arrival (AOA) posterior probability vector corresponding to a previous time slot; the AOA posterior probability vector comprises AOA posterior probabilities corresponding to a plurality of first angle intervals within a preset angle range.

[0039] a signal measurement module, configured to measure a received signal measurement value of the current time slot according to the predicted beamforming vector;

[0040] a probability updating module, configured to update the AOA posterior probability vector of the previous time slot according to the received signal measurement value of the current time slot and the predicted beamforming vector, to determine an AOA posterior probability vector corresponding to the current time slot;

[0041] a step execution module, configured to take the AOA posterior probability vector corresponding to the current time slot as a new AOA posterior probability vector corresponding to the previous time slot, and re-execute the step of determining the predicted beamforming vector corresponding to the current time slot according to the AOA posterior probability vector corresponding to the previous time slot, until the current time slot is the last time slot;

[0042] an angle determination module, configured to determine a target AOA angle according to the AOA posterior probability vector corresponding to the last time slot.

[0043] Embodiments of the present application disclose a network device, comprising:

[0044] a memory in which executable program codes are stored;

[0045] a processor coupled with the memory;

[0046] The processor invokes the executable program codes stored in the memory to execute the method in any of the above embodiments.

[0047] Embodiments of the present application disclose a computer readable storage medium, which stores a computer program, wherein the computer program, when executed by a processor, causes the processor to execute the method in any of the above embodiments.

[0048] By the beam alignment method, apparatus, network device and storage medium disclosed in the embodiments of the present application, the network device can determine a predicted beam forming vector corresponding to a current time slot according to an AOA posterior probability vector corresponding to a previous time slot, measure a received signal measurement value of the current time slot according to the predicted beam forming vector, and update the AOA posterior probability vector of the previous time slot according to the received signal measurement value of the current time slot and the predicted beam forming vector, to obtain an AOA posterior probability vector corresponding to the current time slot, so that the AOA posterior probability vector corresponding to the current time slot can be taken as a new AOA posterior probability vector corresponding to the previous time slot, and the step of determining the predicted beam forming vector corresponding to the current time slot according to the AOA posterior probability vector corresponding to the previous time slot is re-executed until the current time slot is the last time slot, and the network device can determine a target AOA angle according to the AOA posterior probability vector corresponding to the last time slot. By adaptively determining the predicted beam forming vector according to the AOA posterior probability vector, and then updating the AOA posterior probability vector according to the received signal measurement value measured by the predicted beam forming vector, the network device can learn the current channel information, so that after the iteration is completed, the AOA posterior probability vector corresponding to the last time slot can be directly used, not only the beam alignment is completed when the last time slot is reached, but also the direction of the AOA is directly predicted on the continuous space, i.e., the preset angle range, so that the interference of the complex channel environment on the AOA prediction is avoided, and the accuracy of the beam alignment is improved. BRIEF DESCRIPTION OF DRAWINGS

[0049] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiments will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0050] Figure 1 is an application scenario diagram of a beam alignment method disclosed in an embodiment of the present application;

[0051] Figure 2 is a flow diagram of a beam alignment method disclosed in an embodiment of the present application;

[0052] Figure 3 is a flow diagram of another beam alignment method disclosed in an embodiment of the present application;

[0053] Figure 4 is a diagram for interval division of a preset angle range disclosed in an embodiment of the present application;

[0054] Figure 5is a flowchart of a method for updating an AOA posterior probability vector according to an embodiment of the present application;

[0055] Figure 6 is a flowchart of a beam alignment process according to an embodiment of the present application;

[0056] Figure 7 is a comparison diagram of the accuracy of a beam alignment method under different signal-to-noise ratios and the accuracy of other existing methods according to an embodiment of the present application;

[0057] Figure 8 is a modular diagram of a beam alignment apparatus according to an embodiment of the present application;

[0058] Figure 9 is an electronic block diagram of a network device according to an embodiment of the present application. DETAILED DESCRIPTION

[0059] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of protection of the present application.

[0060] It should be noted that the terms "comprising" and "having" and any variations thereof in the embodiments of the present application are intended to cover the inclusion of not exclusive, for example, a process, method, system, product or device comprising a series of steps or units does not have to be limited to those steps or units clearly listed, but can include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0061] It can be understood that the terms "first", "second" and the like used in the present application can be used herein to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish the first element from another element. For example, without departing from the scope of the present application, a first angle interval can be referred to as a second angle interval, and similarly, a second angle interval can be referred to as a first angle interval. The first angle interval and the second angle interval are both angle intervals, but they are not the same angle interval.

[0062] In the related art, the beam alignment method is based on a pre-designed codebook, the beam directions in the codebook are scanned by beam training, and the optimal beam is selected according to the strength of the received signal. The beam alignment method mainly includes exhaustive search and hierarchical search. Exhaustive search is to search each beam direction by using a narrow beam, and find the transceiving beam that best matches the channel main path direction. Because exhaustive search needs to search all beams in the codebook, the beam alignment time is long, and the spectrum resource consumption is high. Hierarchical search is to use a wide beam to search the codebook to determine the angle interval of the channel main path, and then reduce the beam width in the determined angle interval and continue to search. By reducing the angle interval and beam width multiple times, the best narrow beam that matches the communication main path is found. Hierarchical search is superior to exhaustive search under high signal-to-noise ratio conditions, and can effectively shorten the beam alignment time, but under low signal-to-noise ratio conditions, it needs longer search time than the exhaustive search method to achieve similar beam alignment accuracy.

[0063] Therefore, the two related beam alignment methods of exhaustive search and hierarchical search can only obtain high beam alignment accuracy under certain signal-to-noise ratio conditions, and when complex mobile communication scenarios are involved, neither of the two beam alignment methods can guarantee the speed and accuracy of beam alignment. For example, in a terahertz wave-based communication scenario, the beam of the terahertz wave is extremely narrow, the channel is sparse and complex, and the beam alignment is difficult.

[0064] To adapt to the complex and variable terahertz wave communication scenario, in the related art, the generalized likelihood ratio is used to judge the unknown channel signal-to-noise ratio in real time, to realize the switching between the wide beam codebook of hierarchical search and the narrow beam codebook of exhaustive search, and to adaptively select the appropriate beam search codebook under high and low signal-to-noise ratios, which reduces the time and frequency resource consumption of beam search and ensures the accuracy of beam alignment. However, with the increase of antenna array size and communication frequency, the beam width is continuously reduced, and the time and frequency resource consumption of the beam alignment based on spatial scanning is still high.

[0065] In addition, in related technologies, based on multi-perspective images of the environment within the terminal's visual perception range, scattering object vector features corresponding to each scattering object within the terminal's visual perception range are obtained. Based on the scattering object vector features and the target terminal's location information, the optimal transmit and receive beam pair corresponding to the target terminal is determined. This method characterizes the electromagnetic propagation characteristics of the channel between the terminal and the base station using the scattering object vector features and the terminal's location information, thereby determining the optimal transmit and receive beam pair sequence number corresponding to the target terminal. This overcomes the drawback of existing beam alignment methods, which have high time-frequency resource overhead, and achieves rapid beam alignment. However, this method requires the use of an additional photography system to obtain multi-perspective images, as well as an additional information processing system with strong image processing capabilities. The target detection algorithm is used to extract feature information of surrounding objects, further increasing the cost of the system. Furthermore, the beam alignment methods in the above-mentioned related technologies all select the optimal beam from a pre-designed codebook. The beam effect is subject to the quality of the preset codebook, lacking flexibility and affecting the accuracy of beam alignment.

[0066] The embodiments of the present application disclose a beam alignment method, apparatus, network device, and storage medium, which can improve the speed and accuracy of beam alignment.

[0067] The following is a detailed description with reference to the accompanying drawings.

[0068] like Figure 1 As shown, Figure 1 1 is a schematic diagram of an application scenario of a beam alignment method disclosed in an embodiment of the present application. The application scenario may include a network device 110 and a terminal device 120. The network device 110 may include a base station and any other signal receiver capable of receiving signals, and the terminal device 120 may include a mobile phone, a computer, and any other signal transmitter capable of sending signals. The embodiment of the present application does not limit the types of the network device 110 and the terminal device 120. Figure 1 The graphic of the network device 110 is set as a base station, and the graphic of the terminal device 120 is set as a mobile phone. This is only an example and does not mean that the types of the network device 110 and the terminal device 120 are limited.

[0069] In order to establish a communication connection with the network device 110, the terminal device 120 can send an uplink pilot signal to the network device 110. The network device 110 can generate a corresponding beam for reception based on the predicted beamforming vector and measure the received signal measurement value. Optionally, the terminal device 120 can send an uplink pilot signal in τ time slots respectively. The network device 110 completes beam training in the τ time slots, determines the AOA (Angle-of-Arrival) from the terminal device 120 to the network device 110, and obtains the target AOA angle. The size of τ is pre-set.

[0070] For the predicted beamforming vector and target AOA angle that need to be predicted, the predicted beamforming vector of the current time slot can be a mapping of the AOA posterior probability vector corresponding to the previous time slot, and the target AOA angle can be a mapping of the AOA posterior probability vector corresponding to the last time slot. The AOA posterior probability vector may include AOA posterior probabilities corresponding to multiple first angle intervals within a preset angle range, where the preset angle range refers to a range within which the signal arrival angle AOA from the terminal device 120 to the network device 110 may fall. The multiple first angle intervals may constitute the preset angle range, and the first angle intervals do not overlap.

[0071] In one embodiment, the network device 110 can determine the predicted beamforming vector corresponding to the current time slot based on the AOA posterior probability vector corresponding to the previous time slot, and then measure the received signal measurement value of the current time slot based on the predicted beamforming vector. The AOA posterior probability vector of the previous time slot is updated based on the received signal measurement value and the predicted beamforming vector of the current time slot to determine the AOA posterior probability vector corresponding to the current time slot. The AOA posterior probability vector corresponding to the current time slot can be used as the new AOA posterior probability vector corresponding to the previous time slot, and the step of determining the predicted beamforming vector corresponding to the current time slot based on the signal arrival angle AOA posterior probability vector corresponding to the previous time slot is re-executed until the current time slot is the last time slot. The network device can determine the target AOA angle based on the AOA posterior probability vector corresponding to the last time slot.

[0072] like Figure 2 As shown, Figure 2 : is a flowchart of a beam alignment method disclosed in an embodiment of the present application. The beam alignment method can be applied to the network device in the above embodiment. The beam alignment method may include the following steps:

[0073] Step 210: Determine a predicted beamforming vector corresponding to the current time slot based on the AOA posterior probability vector corresponding to the previous time slot.

[0074] The network device inputs the AOA posterior probability vector corresponding to the previous time slot into the adaptive beamforming strategy function to obtain the predicted beamforming vector corresponding to the current time slot. The adaptive beamforming strategy function corresponding to each time slot can be different. Specifically, as shown in formula (1), formula (1) is used to represent that the network device determines the predicted beamforming vector corresponding to the current time slot based on the AOA posterior probability vector corresponding to the previous time slot.

[0075]

[0076] wherein, t is the time slot sequence number of the previous time slot, t+1 is the time slot sequence number of the current time slot, G t (·) is the adaptive beamforming strategy function corresponding to the time slot sequence number of the previous time slot, π t (φ) is the AOA posterior probability vector corresponding to the previous time slot.

[0077] Step 220, measuring the received signal measurement value of the current time slot according to the predicted beamforming vector.

[0078] Since the network device includes a uniform linear antenna array composed of M antennas, the M antennas can obtain an array response vector when receiving a signal, in order to enhance the signal in the AOA direction and eliminate or suppress the interference signals and noise in other directions, it is necessary to adjust the array response vector by emitting a beam, specifically, the network device can generate a beam corresponding to the predicted beamforming vector, measure the received signal measurement value of the current time slot based on the beamforming technology, and the received signal measurement value can include power, amplitude, phase and other data.

[0079] Step 230, updating the AOA posterior probability vector of the previous time slot according to the received signal measurement value of the current time slot and the predicted beamforming vector, to obtain the AOA posterior probability vector corresponding to the current time slot.

[0080] Since the AOA angle must fall within the preset angle range, i.e. the event that the AOA angle falls within each first angle interval constitutes a complete event group, in the Bayesian theory, in order to determine the AOA posterior probability of the target first angle interval in the current time slot, the target first angle interval being any one of the plurality of first angle intervals, the prior probability corresponding to the target first angle interval and the conditional probability corresponding to the plurality of first angle intervals are needed, wherein the prior probability refers to the probability estimated before obtaining the received signal measurement value of the current time slot, for example, if there are 3 first angle intervals, the probability of estimating that the AOA falls within each first angle interval is the same, then the prior probability corresponding to each first angle interval is one third, the posterior probability refers to the probability determined according to the received signal measurement value after obtaining the received signal measurement value of the current time slot, the posterior probability can be a modification of the prior probability according to the received signal measurement value of the current time slot, and the conditional probability corresponding to the target first angle interval refers to the probability of measuring the received signal measurement value of the current time slot under the condition that the AOA angle falls within the target first angle interval and the network device receives the signal according to the predicted beamforming vector of the current time slot.

[0081] Therefore, the network device can take the posterior probability of the target first angle interval corresponding to the previous time slot as the prior probability of the target first angle interval corresponding to the current time slot, and determine the conditional probability corresponding to each of the plurality of first angle intervals according to the received signal measurement value and the predicted beamforming vector of the current time slot, so as to obtain the AOA posterior probability corresponding to the target first angle interval in the current time slot.

[0082] In step 240, the AOA posterior probability vector corresponding to the current time slot is taken as the AOA posterior probability vector corresponding to the previous time slot, and step 210 is re-executed until the current time slot is the last time slot, and step 250 is executed.

[0083] In step 250, the target AOA angle is determined according to the AOA posterior probability vector corresponding to the last time slot.

[0084] The network device inputs the AOA posterior probability vector corresponding to the last time slot into the AOA estimation strategy function, and the target AOA angle can be obtained. Specifically, as shown in formula (2), formula (2) is used to represent that the network device determines the target AOA angle according to the AOA posterior probability vector corresponding to the last time slot,

[0085]

[0086] wherein, is the target AOA angle, F(·) is the AOA estimation strategy function, π τ (φ) is the AOA posterior probability vector corresponding to the last time slot.

[0087] In the embodiments of the present application, the network device can determine the predicted beamforming vector corresponding to the current time slot according to the AOA posterior probability vector corresponding to the previous time slot, measure the received signal measurement value of the current time slot according to the predicted beamforming vector, and update the AOA posterior probability vector of the previous time slot according to the received signal measurement value of the current time slot and the predicted beamforming vector, to obtain the AOA posterior probability vector corresponding to the current time slot, so that the AOA posterior probability vector corresponding to the current time slot can be taken as the AOA posterior probability vector corresponding to the new previous time slot, and the step of determining the predicted beamforming vector corresponding to the current time slot according to the AOA posterior probability vector corresponding to the previous time slot is re-executed until the current time slot is the last time slot, and the network device can determine the target AOA angle according to the AOA posterior probability vector corresponding to the last time slot. Through the method of adaptively determining the predicted beamforming vector according to the AOA posterior probability vector, and then updating the AOA posterior probability vector by the received signal measurement value measured by the predicted beamforming vector, the network device can learn the current channel information, so that after the iteration is completed, the AOA posterior probability vector corresponding to the last time slot can be directly used, not only the beam alignment is completed when the last time slot is reached, but also the direction of the AOA is directly predicted on the continuous space, i.e., the preset angle range, thereby avoiding the interference of the complex channel environment on the AOA prediction, and improving the accuracy of the beam alignment.

[0088] As shown in Figure 3 , Figure 3 is a flow diagram of another beam alignment method disclosed in the embodiments of the present application. The beam alignment method can be applied to the network device in the above embodiments. The beam alignment method can include the following steps:

[0089] In step 310, the beamforming vector prediction is performed by the beamforming prediction neural network according to the AOA posterior probability vector corresponding to the previous time slot, the time slot sequence number of the previous time slot, and the transmit power of the terminal, to determine the predicted beamforming vector of the current time slot.

[0090] It is relatively difficult to directly solve the adaptive beamforming strategy function in formula (1). The beamforming prediction neural network is designed in the embodiments of the present application to perform beamforming vector prediction to obtain the predicted beamforming vector. The beamforming prediction neural network can be a deep neural network. The network device can form an input vector by taking the AOA posterior probability vector corresponding to the previous time slot, the time slot sequence number of the previous time slot, and the transmit power of the terminal as inputs and input the input vector into the beamforming prediction neural network to predict the predicted beamforming vector w t+1 of the current time slot. Wherein, π tis the AOA posterior probability vector corresponding to the previous time slot, P is the transmit power of the terminal, and t is the time slot number of the previous time slot. It should be noted that, since the beamforming prediction neural network is used as an adaptive beamforming strategy function, continuous variables cannot be used as inputs of the neural network, and therefore the preset angle range is divided into a plurality of first angle intervals in the foregoing embodiment.

[0091] Specifically, the structure of the beamforming prediction neural network can be as shown in Table (1),

[0092]

[0093]

[0094] Table (1)

[0095] In step 320, a received signal measurement value of the current time slot is measured according to the predicted beamforming vector.

[0096] In step 330, the AOA posterior probability of the target second angle interval in the previous time slot is updated according to the received signal measurement value of the current time slot and the predicted beamforming vector, to obtain the AOA posterior probability of the target second angle interval in the current time slot.

[0097] To improve the accuracy of updating the AOA posterior probability vector, the network device can further divide each first angle interval, and each first angle interval can include a plurality of second angle intervals. Optionally, the preset angle range can be [φ min ,φ max ], the network device can first divide the preset angle range into N c first angle intervals, including For each first angle interval, the network device can further divide it into N s second angle intervals, that is, each φ i is divided into N s second angle intervals, including i = 1, …, N c ,φ i,j is the jth second angle interval in the ith first angle interval, j = 1, …, N s . As shown in Figure 4 Figure 4 is a schematic diagram of interval division of a preset angle range according to an embodiment of the present application, wherein N c and N s are both equal to 5, Figure 4 (a) in (a) divides the preset angle range into 5 first angle intervals, including φ1, φ2, …, φ5, and each first angle interval corresponds to an AOA posterior probability, Figure 4 ​(b) in the (b) in the above formula (2) is taken as an example, 5 second angle intervals are divided, including φ 2,1 ,φ 2,2 ,…,φ 2,5 .

[0098] Step 340, the AOA posterior probability of each first angle interval in the current time slot is obtained by adding the AOA posterior probability of each second angle interval included in the first angle interval in the current time slot.

[0099] Although the smaller second angle interval unit can improve the accuracy when updating the AOA posterior probability vector, compared with the AOA posterior probability vector containing the AOA posterior probability corresponding to each second angle interval, the AOA posterior probability vector containing the AOA posterior probability corresponding to each first angle interval can reduce the complexity of the calculation when determining the predicted beamforming vector, and will not affect the accuracy of the predicted beamforming vector, so after updating the AOA posterior probability, the network device can add the AOA posterior probability of each second angle interval included in each first angle interval in the current time slot to obtain the AOA posterior probability of each first angle interval in the current time slot. Specifically, as shown in formula (3),

[0100]

[0101] wherein, is the AOA posterior probability of the i-th first angle interval in the current time slot, is the AOA posterior probability of the j-th second angle interval in the i-th first angle interval in the current time slot.

[0102] Step 350, it is judged whether the current time slot is the last time slot, if not, step 360 is executed, and if yes, step 370 is executed.

[0103] Step 360, the AOA posterior probability vector corresponding to the current time slot is taken as the AOA posterior probability vector corresponding to the new last time slot, and step 310 is re-executed.

[0104] Step 370, the angle of arrival prediction neural network determines the target AOA angle according to the AOA posterior probability vector corresponding to the last time slot.

[0105] Similarly, it is difficult to solve the AOA estimation strategy function in formula (2). Embodiments of the present application design an angle of arrival prediction neural network. The network device can input the AOA posterior probability corresponding to the last time slot of each second angle interval to the angle of arrival prediction neural network to determine the target AOA angle. The angle of arrival prediction neural network can be a single-layer neural network. That is, the number of network layers of the angle of arrival prediction neural network can be less than the number of network layers of the beamforming prediction neural network. In the case that the accuracy of the AOA posterior probability vector corresponding to the last time slot obtained by the beamforming prediction neural network after multiple loop iterations is high, the angle of arrival prediction neural network does not need to design a large number of network layers to determine a relatively accurate target AOA angle, thereby improving the efficiency of beam alignment and reducing the complexity of the network structure.

[0106] Specifically, the structure of the angle of arrival prediction neural network can be as shown in table (2),

[0107]

[0108]

[0109] Table (2)

[0110] Optionally, the beamforming prediction neural network and the angle of arrival prediction neural network are trained together. The training set of the beamforming prediction neural network and the angle of arrival prediction neural network includes a plurality of sample AOA angles. The network device can randomly select a plurality of sample AOA angles in a preset angle range and generate a τ sample AOA posterior probability vector corresponding to each sample AOA angle, thereby inputting the τ sample AOA posterior probability vectors to the beamforming prediction neural network to generate a predicted beamforming vector corresponding to each time slot, inputting the predicted beamforming vector of the last time slot to the angle of arrival prediction neural network to obtain the target AOA angle, calculating the loss according to the target AOA angle and the corresponding sample AOA angle, updating the parameters of the beamforming prediction neural network and the angle of arrival prediction neural network, thereby completing the training of the beamforming prediction neural network and the angle of arrival prediction neural network.

[0111] In an embodiment of the present application, the network device can predict the beamforming vector based on the AOA posterior probability vector corresponding to the previous time slot, the time slot number of the current time slot, and the transmit power of the terminal through a beamforming prediction neural network to determine the predicted beamforming vector of the current time slot. The network device can also update the AOA posterior probability corresponding to the target second angle interval in the previous time slot based on the received signal measurement value of the current time slot and the predicted beamforming vector to obtain the AOA posterior probability corresponding to the target second angle interval in the current time slot, and then add the AOA posterior probabilities corresponding to the current time slot of each second angle interval included in each first angle interval to obtain the AOA posterior probability corresponding to the current time slot of each first angle interval. When the current time slot is the last time slot, the network device can also predict the arrival angle based on the AOA posterior probability vector corresponding to the last time slot through an arrival angle prediction neural network to determine the target AOA angle.

[0112] Among them, the network device uses a beamforming prediction neural network to learn the current channel information and predict the optimal beam, avoiding the complicated beam training process, improving the accuracy of the predicted beamforming vector, and also improving the speed of beam alignment. The network device only needs to have a certain matrix calculation capability and does not require additional auxiliary equipment. The network device uses a small amount of pilot resources sent by the terminal device and a neural network that has been trained in advance. It can directly complete the beam alignment through simple matrix addition and matrix multiplication operations, thereby improving the convenience of beam alignment. In addition, compared with the related art in which beam selection is limited to a pre-designed codebook with a limited number of candidate beams, the network device in the embodiment of the present application uses a neural network to directly obtain the arrival angle estimate within a preset angle range, thereby avoiding the influence of the quality of the codebook on the beam alignment effect and improving the accuracy of beam alignment.

[0113] like Figure 5 As shown, Figure 5 : This is a flow chart of a method for updating an AOA posterior probability vector disclosed in an embodiment of the present application. The method for updating an AOA posterior probability vector can be applied to the network device in the above embodiment. The method for updating an AOA posterior probability vector can include the following steps:

[0114] Step 510: Update the fading coefficient mean and fading coefficient variance of the channel path corresponding to the previous time slot based on the received signal measurement value and the predicted beamforming vector of the current time slot to obtain the fading coefficient mean and fading coefficient variance corresponding to the current time slot.

[0115] To predict the received signal, the network device further includes a signal model of the received signal, which is used to obtain a received signal prediction value for comparison with a received signal measurement value. The signal model of the received signal is constructed based on the predicted beamforming vector and the channel path. The network device can obtain a plurality of conditional probabilities corresponding to a plurality of first angle intervals according to the signal model of the received signal. The network device determines the predicted beamforming vector included in the signal model in step 210, and updates all parameters in the signal model to improve the accuracy of obtaining the plurality of conditional probabilities corresponding to the plurality of first angle intervals. The parameters of the channel path can include a fading coefficient, and the network device can update the mean and variance of the fading coefficient included in the channel path corresponding to the previous time slot according to the current received signal measurement value and the predicted beamforming vector to obtain the mean and variance of the fading coefficient corresponding to the current time slot.

[0116] Specifically, as shown in equation (4), equation (4) is a formula of the channel path,

[0117] h = a (f) equation (4);

[0118] where h is the channel path, a is the fading coefficient, a (f) is an array response vector of a uniform linear antenna array composed of M antennas, and f is the AOA angle, d is the antenna spacing, and l is the wavelength. Optionally, a can be subject to a complex Gaussian distribution, and therefore, the network device can update the mean and variance of the fading coefficient corresponding to the previous time slot to obtain the mean and variance of the fading coefficient corresponding to the current time slot, that is, complete the parameter update of the signal model. The mean of the fading coefficient can include a plurality of means corresponding to a plurality of first angle intervals, and the variance of the fading coefficient can include a plurality of variances corresponding to a plurality of first angle intervals.

[0119] In one embodiment, the network device can obtain a received signal prediction value of the current time slot according to the mean and variance of the fading coefficient of the channel path corresponding to the previous time slot, the predicted beamforming vector, and the signal model, determine a signal observation error of the current time slot according to the received signal prediction value of the current time slot and the received signal measurement value of the current time slot, and update the mean and variance of the fading coefficient of the channel path corresponding to the previous time slot according to the signal observation error of the current time slot and the filter gain to obtain the mean and variance of the fading coefficient corresponding to the current time slot.

[0120] The filter gain is used to optimize the signal observation error of the current time slot in the updating process of the fading coefficient. The filter can include a Kalman filter, and the network device can obtain the filter gain according to an observation error covariance matrix and a noise error covariance matrix. Specifically, the manner of updating the fading coefficient mean and the fading coefficient variance is shown in formula (5),

[0121]

[0122] wherein, is the fading coefficient mean corresponding to the current time slot of the jth second angle interval in the ith first angle interval, is the fading coefficient mean corresponding to the last time slot of the jth second angle interval in the ith first angle interval, is the fading coefficient variance corresponding to the current time slot of the jth second angle interval in the ith first angle interval, is the fading coefficient variance corresponding to the last time slot of the jth second angle interval in the ith first angle interval, i,j,t+1 is obtained according to the channel model, and specifically, y t+1 is the received signal measurement value of the current time slot.

[0123] In step 520, based on the signal model, the conditional density function corresponding to the received signal measurement value of the current time slot is determined according to the predicted beamforming vector, the fading coefficient mean and the fading coefficient variance corresponding to the current time slot.

[0124] wherein the signal model can be shown in formula (6),

[0125]

[0126] wherein x t is the uplink pilot signal when the time slot sequence number is t, w t is the predicted beamforming vector corresponding to the time slot sequence number t, z t is the additive white Gaussian noise when the time slot sequence number is t, x t satisfies the power constraint |x t | 2 = P, the predicted beamforming vector w t satisfies the constant modulus constraint wherein is the ith component of w t .

[0127] Therefore, the network device determines the signal path h according to the fading coefficient mean value and the fading coefficient variance corresponding to the current time slot, so that the conditional density function of the received signal measurement value at the current time can be obtained by substituting the signal path h and the predicted beamforming vector of the current time slot into formula (6), and the conditional density function is shown in formula (7),

[0128]

[0129] wherein, denotes the beamforming prediction neural network corresponding to the previous time slot, π t is the AOA posterior probability vector corresponding to the previous time slot, φ = φ i,j is used to represent that the AOA angle falls into the jth second angle interval in the ith first angle interval, is used to represent that the predicted beamforming vector of the current time slot is determined by the beamforming prediction neural network corresponding to the previous time slot according to the AOA posterior probability vector corresponding to the previous time slot.

[0130] In step 530, the AOA posterior probability vector corresponding to the current time slot is determined according to the conditional density function corresponding to the current time slot and the AOA posterior probability vector of the previous time slot based on the Bayes formula.

[0131] In one embodiment, the network device can determine the conditional probability corresponding to each of the plurality of first angle intervals according to the conditional density function, determine the standard likelihood corresponding to each of the plurality of angle intervals according to the conditional probability corresponding to each of the plurality of first angle intervals and the AOA posterior probability vector of the previous time slot, and determine the AOA posterior probability of each of the plurality of first angle intervals corresponding to the current time slot based on the standard likelihood corresponding to each of the plurality of first angle intervals and the AOA posterior probability corresponding to each of the plurality of first angle intervals in the previous time slot. By implementing this embodiment, the AOA posterior probability vector can be updated by the idea of Bayesian estimation, and the convenience of vector updating can be improved.

[0132] Specifically, as shown in formula (8), formula (8) is a formula for determining the AOA posterior probability vector corresponding to the current time slot based on the Bayes formula,

[0133]

[0134] wherein, is the AOA posterior probability of the jth second angle interval in the ith first angle interval corresponding to the current time slot, is the total probability, is the standard likelihood corresponding to the jth second angle interval in the ith first angle interval, is the AOA posterior probability vector corresponding to the previous time slot.

[0135] Since the network device needs to determine the predicted beamforming vector corresponding to the current time slot according to the AOA posterior probability vector corresponding to the previous time slot, and needs to update the mean and variance of the fading coefficients included in the channel path corresponding to the previous time slot, when the current time slot is the first time slot, i.e., there is no previous time slot, the network device can update or determine the data by using the preset initial quantity.

[0136] In one embodiment, before determining the predicted beamforming vector corresponding to the current time slot according to the AOA posterior probability vector corresponding to the previous time slot, the network device can determine the predicted beamforming vector corresponding to the first time slot according to the initial AOA posterior probability vector, measure the received signal measurement value of the first time slot according to the predicted beamforming vector corresponding to the first time slot, update the initial mean and variance of the fading coefficients of the channel path according to the received signal measurement value of the first time slot and the predicted beamforming vector, obtain the mean and variance of the fading coefficients corresponding to the first time slot, determine the conditional density function corresponding to the received signal measurement value of the first time slot based on the signal model and according to the predicted beamforming vector, the mean and variance of the fading coefficients corresponding to the first time slot, and determine the AOA posterior probability vector corresponding to the first time slot based on the Bayes formula and according to the conditional density function corresponding to the first time slot and the initial AOA posterior probability vector.

[0137] The initial AOA posterior probability vector can be a uniform distribution vector, i.e., the posterior probabilities corresponding to each first angle interval can be equal, i.e., the probability of the AOA angle falling into each first angle interval is considered to be equal at the first time slot, and the probability of the AOA angle falling into each first angle interval is updated through subsequent loop iterations. The initial mean of the fading coefficients can be 0, and the initial variance of the fading coefficients can be 1. Specifically, when t = 0, 1 is an all-1 vector, and for i,j , where Implementing this embodiment improves the practicality and completeness of the beam alignment method.

[0138] As shown in Figure 6 , Figure 6 is a flowchart of a beam alignment process disclosed in an embodiment of the present application, wherein when t = 0, the initial AOA posterior probability vector P and t are input into the beamforming prediction neural network, and the beamforming prediction neural network outputs the predicted beamforming vector w t+1, the network device updates the predicted beamforming vector w t+1 , the network device measures the received signal measurement value y t+1 , the network device updates the fading coefficient mean value and the fading coefficient variance , the network device further updates the AOA posterior probability vector , the network device adds the AOA posterior probability of each second angle interval included in each first angle interval in the current time slot, to obtain the AOA posterior probability of each first angle interval in the current time slot, that is, to determine the AOA posterior probability vector t+1 , the network device further updates the AOA posterior probability vector t+1 , P, and t, and iterates the beamforming prediction neural network until the last time slot, and when the current time slot is the last time slot, that is, t = τ, the target AOA angle is determined by the angle of arrival prediction neural network

[0139] In the embodiments of the present application, the network device can update the fading coefficient mean value and the fading coefficient variance of the channel path corresponding to the previous time slot according to the received signal measurement value of the current time slot and the predicted beamforming vector, obtain the fading coefficient mean value and the fading coefficient variance corresponding to the current time slot, and determine the conditional density function corresponding to the received signal measurement value of the current time slot based on the signal model, the predicted beamforming vector, the fading coefficient mean value and the fading coefficient variance corresponding to the current time slot. The network device further determines the AOA posterior probability vector corresponding to the current time slot based on the Bayes formula, the conditional density function corresponding to the current time slot and the AOA posterior probability vector of the previous time slot, updates the fading coefficient mean value and the fading coefficient variance, improves the accuracy of updating the AOA posterior probability vector, and improves the convenience of updating the AOA posterior probability vector based on the idea of the Bayes formula.

[0140] In order to more clearly show the effect of the beam alignment method disclosed in the embodiments of the present application, the present application is simulated and compared with three existing algorithms, including a compressed sensing algorithm (CS), a halved hierarchical search algorithm (hieBS), and a hierarchical search algorithm based on posterior distribution (hiePM). In the simulation experiment, the network device includes M = 64 antennas, N c = 128, and the number of time slots for uplink pilot transmission of the terminal device is set to τ = 2log2N c= 14. The network device trains the beamforming prediction neural network and the angle of arrival prediction neural network for 10 batches in each training period, each batch containing at most 2^12 samples, and the loss function uses the mean square error function. The performance of the neural network in the training process is monitored by calculating the loss function of the verification data set of 10^5 samples, and every 10 training periods, if the loss function of the neural network on the verification data set is less than the current minimum loss, the new model parameters are saved and the minimum loss is updated.

[0141] As shown in Figure 7 , Figure 7 is a comparison diagram of the accuracy of the beam alignment method disclosed in the embodiments of the present application under different signal-to-noise ratios and the accuracy of other existing methods, wherein "training period: 100 times" refers to the beam alignment method corresponding to the neural network with a training period of 100 times in the embodiments of the present application, "training period: 500 times" refers to the beam alignment method corresponding to the neural network with a training period of 500 times in the embodiments of the present application, "training period: 1000 times" refers to the beam alignment method corresponding to the neural network with a training period of 1000 times in the embodiments of the present application, and "training period: 10000 times" refers to the beam alignment method corresponding to the neural network with a training period of 10000 times in the embodiments of the present application. With the increase of the training period of the neural network, the mean square error obtained by the beam alignment method disclosed in the embodiments of the present application becomes smaller and smaller, and when the training period of the neural network reaches 10000, the mean square error of the beam alignment method disclosed in the embodiments of the present application on the test set is smaller than that of the three existing methods.

[0142] As shown in Figure 8 , Figure 8 is a modular diagram of a beam alignment device disclosed in the embodiments of the present application. The beam alignment device 800 can include a vector determination module 810, a signal measurement module 820, a probability updating module 830, a step execution module 840, and an angle determination module 850, wherein:

[0143] The vector determination module 810 is configured to determine a predicted beamforming vector corresponding to a current time slot according to an angle of arrival (AOA) posterior probability vector corresponding to a previous time slot; the AOA posterior probability vector includes AOA posterior probabilities corresponding to a plurality of first angle intervals in a preset angle range;

[0144] The signal measurement module 820 is configured to measure a received signal measurement value of the current time slot according to the predicted beamforming vector;

[0145] The probability updating module 830 is configured to update the AOA posterior probability vector of the previous time slot according to the received signal measurement value of the current time slot and the predicted beamforming vector, to obtain an AOA posterior probability vector corresponding to the current time slot.

[0146] The step execution module 840 is configured to take the AOA posterior probability vector corresponding to the current time slot as the AOA posterior probability vector corresponding to the new previous time slot, and re-execute the step of determining the predicted beamforming vector corresponding to the current time slot according to the AOA posterior probability vector corresponding to the signal arriving angle of the previous time slot, until the current time slot is the last time slot.

[0147] The angle determination module 850 is configured to determine the target AOA angle according to the AOA posterior probability vector corresponding to the last time slot.

[0148] In an embodiment, a signal model of the received signal is constructed based on the predicted beamforming vector and the channel path; the probability updating module 830 is further configured to update the mean and variance of the fading coefficients included in the channel path corresponding to the previous time slot according to the received signal measurement value of the current time slot and the predicted beamforming vector, to obtain the mean and variance of the fading coefficients corresponding to the current time slot; determine the conditional density function corresponding to the received signal measurement value of the current time slot based on the signal model, the predicted beamforming vector, the mean and variance of the fading coefficients corresponding to the current time slot; and determine the AOA posterior probability vector corresponding to the current time slot based on the Bayes formula, the conditional density function corresponding to the current time slot and the AOA posterior probability vector of the previous time slot.

[0149] In an embodiment, the vector determination module 810 is further configured to determine the predicted beamforming vector corresponding to the first time slot according to the initial AOA posterior probability vector; the signal measurement module 820 is further configured to measure the received signal measurement value of the first time slot according to the predicted beamforming vector corresponding to the first time slot; and the probability updating module 830 is further configured to update the initial mean and variance of the fading coefficients of the channel path according to the received signal measurement value of the first time slot and the predicted beamforming vector, to obtain the mean and variance of the fading coefficients corresponding to the first time slot; determine the conditional density function corresponding to the received signal measurement value of the first time slot based on the signal model, the predicted beamforming vector, the mean and variance of the fading coefficients corresponding to the first time slot; and determine the AOA posterior probability vector corresponding to the first time slot based on the Bayes formula, the conditional density function corresponding to the first time slot and the initial AOA posterior probability vector.

[0150] In an embodiment, the probability updating module 830 is further configured to obtain a predicted value of a received signal of a current time slot according to a mean value and a variance of a fading coefficient of a channel path corresponding to a previous time slot, a predicted beamforming vector, and a signal model; determine a signal observation error of the current time slot according to the predicted value of the received signal of the current time slot and a measured value of the received signal of the current time slot; and update the mean value and the variance of the fading coefficient of the channel path corresponding to the previous time slot according to the signal observation error of the current time slot and a filter gain, to obtain a mean value and a variance of a fading coefficient corresponding to the current time slot.

[0151] In an embodiment, the probability updating module 830 is further configured to determine a conditional probability corresponding to each of a plurality of first angle intervals according to a conditional density function; determine a standard likelihood corresponding to each of the plurality of first angle intervals according to the conditional probability corresponding to each of the plurality of first angle intervals and an AOA posterior probability vector of a previous time slot; and determine an AOA posterior probability corresponding to each of the plurality of first angle intervals in a current time slot according to the standard likelihood corresponding to each of the plurality of first angle intervals and the AOA posterior probability corresponding to each of the plurality of first angle intervals in the previous time slot based on a Bayes formula.

[0152] In an embodiment, the vector determining module 810 is further configured to determine a predicted beamforming vector of a current time slot by performing beamforming vector prediction on an AOA posterior probability vector corresponding to a previous time slot, a time slot serial number of the previous time slot, and a transmit power of a terminal through a beamforming prediction neural network; and the angle determining module 850 is further configured to determine a target AOA angle by performing AOA prediction on the AOA posterior probability vector corresponding to the last time slot through an AOA prediction neural network; wherein the beamforming prediction neural network and the AOA prediction neural network are trained together, and a training set of the beamforming prediction neural network and the AOA prediction neural network includes a plurality of sample AOA angles.

[0153] In an embodiment, each of the plurality of first angle intervals includes a plurality of second angle intervals; the probability updating module 830 is further configured to update an AOA posterior probability corresponding to a target second angle interval in a previous time slot according to a measured value of a received signal of a current time slot and a predicted beamforming vector, to obtain an AOA posterior probability corresponding to the target second angle interval in the current time slot; wherein the target second angle interval is any second angle interval within a preset angle range; and add the AOA posterior probabilities corresponding to the second angle intervals included in each of the plurality of first angle intervals in the current time slot to obtain an AOA posterior probability corresponding to each of the plurality of first angle intervals in the current time slot.

[0154] In the embodiments of the present application, the network device can determine the predicted beamforming vector corresponding to the current time slot according to the AOA posterior probability vector corresponding to the previous time slot, measure the received signal measurement value of the current time slot according to the predicted beamforming vector, and update the AOA posterior probability vector of the previous time slot according to the received signal measurement value of the current time slot and the predicted beamforming vector, to obtain the AOA posterior probability vector corresponding to the current time slot, so that the AOA posterior probability vector corresponding to the current time slot can be taken as the AOA posterior probability vector corresponding to the new previous time slot, and the step of determining the predicted beamforming vector corresponding to the current time slot according to the AOA posterior probability vector corresponding to the previous time slot is re-executed until the current time slot is the last time slot, and the network device can determine the target AOA angle according to the AOA posterior probability vector corresponding to the last time slot. Through the method of adaptively determining the predicted beamforming vector according to the AOA posterior probability vector, and updating the AOA posterior probability vector through the received signal measurement value measured by the predicted beamforming vector to perform cyclic iteration on the AOA posterior probability vector, the network device can learn the current channel information, so that after the iteration is completed, the AOA posterior probability vector corresponding to the last time slot can be directly used, not only the beam alignment is completed when the last time slot is reached, but also the direction of the AOA is directly predicted on the continuous space, that is, the preset angle range, thereby avoiding the interference of the complex channel environment on the AOA prediction, and improving the accuracy of the beam alignment.

[0155] As shown in Figure 9 In one embodiment, a network device is provided, which can include:

[0156] a memory 910 storing executable program codes;

[0157] a processor 920 coupled with the memory 910;

[0158] The processor 920 invokes the executable program codes stored in the memory 910 to implement the beam alignment method provided in the above embodiments.

[0159] The memory 910 can include a random access memory (RAM) and can also include a read-only memory (ROM). The memory 910 can be used to store instructions, programs, codes, code sets or instruction sets. The memory 910 can include a program storage area and a data storage area, wherein the program storage area can store instructions for implementing an operating system, instructions for implementing at least one function (such as a touch function, a sound playing function, an image playing function, etc.), instructions for implementing the above various method embodiments, etc. The data storage area can also store data created by the electronic device during use, etc.

[0160] The processor 920 can include one or more processing cores. The processor 920 connects various parts within the entire electronic device by various interfaces and lines, performs various functions of the electronic device and processes data by running or executing instructions, programs, code sets or instruction sets stored in the memory 910, and calling data stored in the memory 910. Alternatively, the processor 920 can be implemented in at least one of a hardware form of a digital signal processing (DSP), a field-programmable gate array (FPGA), a programmable logic array (PLA). The processor 920 can integrate a combination of one or several of a central processing unit (CPU), a graphics processor (GPU), and a modem. Among them, the CPU mainly processes an operating system, a user interface, and an application program; the GPU is responsible for rendering and drawing display content; and the modem is used for processing wireless communication. It can be understood that the above-mentioned modem can also not be integrated into the processor 920, but can be implemented by a separate communication chip.

[0161] It can be understood that the electronic device can include more or less structural elements than those in the above structural block diagram, for example, including a power module, a physical key, a WiFi (Wireless Fidelity) module, a speaker, a Bluetooth module, a sensor, etc., which can also not be limited herein.

[0162] The embodiments of the present application disclose a computer readable storage medium storing a computer program, wherein the computer program causes a computer to execute the method described in each of the above embodiments.

[0163] In addition, the embodiments of the present application further disclose a computer program product, when the computer program product runs on a computer, causes the computer to execute all or part of the steps of any one of the beam alignment methods described in the above embodiments.

[0164] Those skilled in the art can understand that all or part of the steps of various methods in the above embodiments can be completed by instructing the relevant hardware through a program, and the program can be stored in a computer readable storage medium, including Read-Only Memory (ROM), Random Access Memory (RAM), Programmable Read-Only Memory (PROM), Erasable Programmable Read-Only Memory (EPROM), One-time Programmable Read-Only Memory (OTPROM), Electrically-Erasable Programmable Read-Only Memory (EEPROM), Compact Disc Read-Only Memory (CD-ROM) or other optical disk storage, magnetic disk storage, magnetic tape storage, or any other medium that can be used to carry or store data in a computer readable manner.

[0165] The above describes in detail a beam alignment method, device, network equipment and storage medium disclosed by the embodiments of the present application. The principles and implementation manners of the present application are described by applying specific examples. The above embodiment description is only used to help understand the method and core idea of the present application. Meanwhile, for those skilled in the art, according to the idea of the present application, the specific implementation manner and application range can be changed. In summary, the content of the specification should not be understood as a limitation of the present application.

Claims

1. A beam alignment method, characterized in that: Applied to a network device, the method includes: Determining a predicted beamforming vector corresponding to a current time slot based on an a posteriori probability vector of the angle of arrival (AOA) of the signal corresponding to the previous time slot; wherein the a posteriori probability vector of the AOA includes a posteriori probabilities of the AOA corresponding to a plurality of first angle intervals within a preset angle range; Measuring and obtaining a received signal measurement value of the current time slot according to the predicted beamforming vector; updating the AOA posterior probability vector of the previous time slot according to the received signal measurement value of the current time slot and the predicted beamforming vector to obtain the AOA posterior probability vector corresponding to the current time slot; Using the AOA posterior probability vector corresponding to the current time slot as the AOA posterior probability vector corresponding to the new previous time slot, and re-performing the step of determining the predicted beamforming vector corresponding to the current time slot based on the signal arrival angle AOA posterior probability vector corresponding to the previous time slot until the current time slot is the last time slot; Determining a target AOA angle according to the AOA posterior probability vector corresponding to the last time slot; The updating, based on the received signal measurement value of the current time slot and the predicted beamforming vector, of the AOA posterior probability vector of the previous time slot to obtain the AOA posterior probability vector corresponding to the current time slot includes: updating, based on the received signal measurement value of the current time slot and the predicted beamforming vector, a fading coefficient mean and a fading coefficient variance of the channel path corresponding to the previous time slot to obtain a fading coefficient mean and a fading coefficient variance corresponding to the current time slot; Determining, based on a signal model, a conditional density function corresponding to a received signal measurement value of the current time slot according to the predicted beamforming vector, a mean value of a fading coefficient corresponding to the current time slot, and a variance of a fading coefficient; Determining, according to the conditional density function, the conditional probabilities corresponding to the plurality of first angle intervals respectively; Determining the standard likelihoods corresponding to the multiple first angle intervals respectively according to the conditional probabilities respectively corresponding to the multiple first angle intervals and the AOA posterior probability vector of the previous time slot; Based on the Bayesian formula, the AOA posterior probability corresponding to each first angle interval in the current time slot is determined according to the standard likelihood corresponding to each first angle interval and the AOA posterior probability corresponding to each first angle interval in the previous time slot.

2. The method according to claim 1, characterized in that Before determining the predicted beamforming vector corresponding to the current time slot based on the signal angle of arrival AOA posterior probability vector corresponding to the previous time slot, the method further includes: Determine the predicted beamforming vector corresponding to the first time slot based on the initial AOA posterior probability vector; Measuring and obtaining a received signal measurement value of the first time slot according to the predicted beamforming vector corresponding to the first time slot; updating an initial fading coefficient mean and an initial fading coefficient variance of the channel path according to a received signal measurement value of the first time slot and the predicted beamforming vector, to obtain a fading coefficient mean and a fading coefficient variance corresponding to the first time slot; Determining, based on the signal model, a conditional density function corresponding to a received signal measurement value of the first time slot according to the predicted beamforming vector, a fading coefficient mean value, and a fading coefficient variance corresponding to the first time slot; Based on the Bayesian formula, the AOA posterior probability vector corresponding to the first time slot is determined according to the conditional density function corresponding to the first time slot and the initial AOA posterior probability vector.

3. The method according to claim 1, characterized in that The updating, based on the received signal measurement value of the current time slot and the predicted beamforming vector, of the fading coefficient mean and the fading coefficient variance of the channel path corresponding to the previous time slot to obtain the fading coefficient mean and the fading coefficient variance corresponding to the current time slot, includes: Obtaining a predicted value of a received signal for the current time slot according to a fading coefficient mean and a fading coefficient variance of the channel path corresponding to the previous time slot, the predicted beamforming vector, and the signal model; Determining a signal observation error of the current time slot based on a predicted value of the received signal of the current time slot and a measured value of the received signal of the current time slot; According to the signal observation error and filter gain of the current time slot, the fading coefficient mean and fading coefficient variance of the channel path corresponding to the previous time slot are updated to obtain the fading coefficient mean and fading coefficient variance corresponding to the current time slot.

4. The method according to claim 1, wherein The determining, based on the signal arrival angle AOA posterior probability vector corresponding to the previous time slot, a predicted beamforming vector corresponding to the current time slot includes: Predicting a beamforming vector for the current time slot by using a beamforming prediction neural network based on the AOA posterior probability vector corresponding to the previous time slot, the time slot number of the previous time slot, and the transmit power of the terminal; The determining the target AOA angle according to the AOA posterior probability vector corresponding to the last time slot includes: Performing an angle of arrival prediction based on the AOA posterior probability vector corresponding to the last time slot using an angle of arrival prediction neural network to determine the target AOA angle; The beamforming prediction neural network and the arrival angle prediction neural network are trained together, and the training sets of the beamforming prediction neural network and the arrival angle prediction neural network include multiple sample AOA angles.

5. The method according to any one of claims 1 to 4, characterized in that: Each of the first angle intervals includes a plurality of second angle intervals; and updating the AOA posterior probability vector of the previous time slot according to the received signal measurement value of the current time slot and the predicted beamforming vector to obtain the AOA posterior probability vector corresponding to the current time slot includes: updating, based on the received signal measurement value of the current time slot and the predicted beamforming vector, the AOA posterior probability corresponding to the target second angle interval in the previous time slot to obtain the AOA posterior probability corresponding to the target second angle interval in the current time slot; wherein the target second angle interval is any second angle interval within the preset angle range; The AOA posterior probabilities corresponding to the second angle intervals included in each first angle interval in the current time slot are added together to obtain the AOA posterior probability corresponding to each first angle interval in the current time slot.

6. A beam alignment device, characterized in that: Applied to network equipment, the device includes: A vector determination module is configured to determine a predicted beamforming vector corresponding to a current time slot based on an a posteriori probability vector of the angle of arrival (AOA) of the signal corresponding to the previous time slot; the AOA a posteriori probability vector includes AOA a posteriori probabilities corresponding to a plurality of first angle intervals within a preset angle range; a signal measurement module, configured to measure and obtain a received signal measurement value of the current time slot according to the predicted beamforming vector; a probability updating module, configured to update the AOA posterior probability vector of the previous time slot according to the received signal measurement value of the current time slot and the predicted beamforming vector, to obtain the AOA posterior probability vector corresponding to the current time slot; a step execution module, configured to use the AOA posterior probability vector corresponding to the current time slot as the AOA posterior probability vector corresponding to the new previous time slot, and re-execute the step of determining the predicted beamforming vector corresponding to the current time slot based on the signal arrival angle AOA posterior probability vector corresponding to the previous time slot, until the current time slot is the last time slot; An angle determination module, configured to determine a target AOA angle based on an AOA posterior probability vector corresponding to the last time slot; The probability updating module is further configured to update the fading coefficient mean and the fading coefficient variance of the channel path corresponding to the previous time slot based on the received signal measurement value of the current time slot and the predicted beamforming vector, to obtain the fading coefficient mean and the fading coefficient variance corresponding to the current time slot; and determine, based on a signal model, the conditional density function corresponding to the received signal measurement value of the current time slot based on the predicted beamforming vector and the fading coefficient mean and the fading coefficient variance corresponding to the current time slot; The probability update module is also used to determine the conditional probabilities corresponding to the multiple first angle intervals according to the conditional density function; determine the standard likelihoods corresponding to the multiple first angle intervals according to the conditional probabilities corresponding to the multiple first angle intervals and the AOA posterior probability vector of the previous time slot; based on the Bayesian formula, determine the AOA posterior probability corresponding to each first angle interval in the current time slot according to the standard likelihood corresponding to each first angle interval and the AOA posterior probability corresponding to each first angle interval in the previous time slot.

7. A network device, characterized in that: include: a memory storing executable program code; a processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the method according to any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, wherein when the computer program is executed by a processor, the processor is caused to perform the method according to any one of claims 1 to 5.

Images