Method, apparatus and electronic device for determining optimal beam

Abstract

The application discloses a method and device for determining an optimal beam and electronic equipment. The method comprises: obtaining first state information of a target device at a first time point, wherein the target device is a base station or a drone; predicting the first state information by using a prediction model to obtain an initial beam at a second time point, wherein the second time point is a next time point of the first time point; determining a similar beam of the initial beam from a codebook of the target device, and determining an initial beam pair according to the initial beam and the similar beam; and determining an optimal beam of the target device on each radio frequency chain according to the initial beam pair, wherein the optimal beam is a beam configuration with the best communication performance. The application solves the technical problem that the initial communication quality after each beam switching is unstable due to random initialization of a traditional heuristic algorithm, which is not conducive to beam tracking and stable communication.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of wireless communication, in particular to a method and device for determining optimal beam and electronic equipment. BACKGROUND

[0002] In the low-altitude communication scenarios such as unmanned aerial vehicle (UAV) inspection and live broadcast, the picture needs to be transmitted back in real time. When multiple UAVs perform cluster tasks, not only the communication between the UAVs and the base station needs to be ensured, but also the data interaction between the UAVs needs to be ensured, which puts high requirements on the communication rate and stability of the network. Since the UAVs fly at a high speed, the base station needs to track and update the communication beam in real time to avoid communication lag or interruption caused by high latency. The initial communication quality after beam switching is unstable due to random initialization of the traditional heuristic algorithm, which is not conducive to beam tracking and stable communication.

[0003] At present, there is no effective solution to the above problems. SUMMARY

[0004] The embodiments of the present application provide a method and device for determining optimal beam and electronic equipment to at least solve the technical problem that the initial communication quality after beam switching is unstable due to random initialization of the traditional heuristic algorithm, which is not conducive to beam tracking and stable communication.

[0005] According to an aspect of an embodiment of the present application, a method for determining optimal beam is provided, comprising: obtaining first state information of a target device at a first time, wherein the target device is a base station or a UAV; predicting the first state information by using a prediction model to obtain an initial beam at a second time, wherein the second time is the next time of the first time; determining a similar beam of the initial beam from a codebook of the target device, and determining an initial beam pair according to the initial beam and the similar beam; determining an optimal beam of the target device on each radio frequency link according to the initial beam pair, wherein the optimal beam is the best beam configuration in terms of communication performance.

[0006] Optionally, the prediction model is obtained by training in the following manner: obtaining a training set for training the prediction model, wherein the training set includes first historical state information of the base station, second historical state information of the UAV, and historical communication quality data between the base station and the UAV; predicting the training set by using an initial prediction model to obtain a historical prediction result; determining a total loss function according to the historical prediction result, the first historical state information and the historical communication quality data; and iteratively training the initial prediction model according to the total loss function to obtain the prediction model.

[0007] Optionally, the second historical state information comprises a first longitude, a first latitude, a first height of the UAV, a pitch angle, a roll angle, a yaw angle of the UAV, and a relative azimuth angle between the UAV and the base station, wherein the relative azimuth angle is determined by: obtaining a second longitude, a second latitude and a second height corresponding to the base station, and obtaining an equatorial radius and an ellipsoid flattening; determining a first value according to the first latitude, the equatorial radius and the ellipsoid flattening, and determining a second value according to the second latitude, the equatorial radius and the ellipsoid flattening; determining a first geocentric coordinate corresponding to the UAV according to the first value, the first longitude, the first latitude and the first height, and determining a second geocentric coordinate corresponding to the base station according to the second value, the second longitude, the second latitude and the second height, wherein the first geocentric coordinate comprises coordinates of the UAV in X-axis, Y-axis and Z-axis directions, and the second geocentric coordinate comprises coordinates of the base station in the X-axis, Y-axis and Z-axis directions; determining a coordinate difference value according to the first geocentric coordinate and the second geocentric coordinate, wherein the coordinate difference value comprises coordinate difference values of the UAV and the base station in the X-axis, Y-axis and Z-axis directions; determining a conversion coordinate of the UAV according to the coordinate difference value and a base station rotation matrix, wherein the base station rotation matrix is determined by the second longitude and the second latitude; and determining the relative azimuth angle according to the conversion coordinate and a coordinate origin.

[0008] Optionally, the total loss function is determined according to the historical prediction result, the first historical state information and the historical communication quality data, comprising: obtaining a first predicted azimuth angle, a first predicted downtilt angle and a first predicted speed of the UAV at t moment in the historical prediction result, obtaining a real azimuth angle and a real downtilt angle in the first historical state information, and obtaining a real speed in the historical communication quality data, wherein t is a positive integer; determining a main loss function according to the first predicted azimuth angle, the first predicted downtilt angle, the real azimuth angle and the real downtilt angle; determining a speed loss function according to the predicted speed and the real speed; obtaining a second predicted azimuth angle and a second predicted downtilt angle of the UAV at t-1 moment in the historical prediction result; determining a beam smoothness constraint loss function according to the first predicted azimuth angle, the second predicted azimuth angle, the first predicted downtilt angle and the second predicted downtilt angle; determining an effective coverage constraint loss function according to the first predicted azimuth angle, a preset maximum azimuth angle, a preset minimum azimuth angle, the first predicted downtilt angle, a preset maximum downtilt angle and a preset minimum downtilt angle; and determining the total loss function according to the main loss function, the speed loss function, the beam smoothness constraint loss function and the effective coverage constraint loss function.

[0009] Optionally, the initial beam at the second moment is obtained by predicting the first state information by using the prediction model, comprising: predicting the first state information by using the prediction model to obtain second state information at the second moment; and determining the initial beam at the second moment according to the second state information.

[0010] Optionally, determining the similar beam of the initial beam from the codebook of the target device comprises: determining the similarity between the initial beam and each beam in the codebook of the target device to obtain a set of similarities; and determining the beam with the highest similarity in the set of similarities as the similar beam of the initial beam.

[0011] Optionally, determining the optimal beam of the target device on each radio frequency chain according to the initial beam pair comprises: step 1, obtaining a first beam pair corresponding to a current time and obtaining a neighboring beam pair of the first beam pair; step 2, determining a first spectral efficiency corresponding to the first beam pair and determining a second spectral efficiency corresponding to the neighboring beam pair; step 3, in the case where the difference between the second spectral efficiency and the first spectral efficiency is greater than or equal to a preset threshold, determining a beam pair change rate according to the first spectral efficiency, the second spectral efficiency and the initial beam pair; step 4, determining an intermediate value according to the beam pair change rate, a decay coefficient and a zero prevention coefficient; step 5, determining a second beam pair corresponding to a next time of the current time according to the intermediate value, the beam pair change rate and the first beam pair, and determining the second beam pair as the first beam pair; and step 6, repeating steps 1 to 5 until the iteration is stopped when the difference between the second spectral efficiency and the first spectral efficiency is less than the preset threshold, and determining the second beam pair obtained last as the optimal beam.

[0012] According to another aspect of the embodiments of the present application, a determination device of an optimal beam is also provided, comprising: an obtaining module, configured to obtain first state information of a target device at a first time, wherein the target device is a base station or a UAV; a prediction module, configured to predict the first state information by using a prediction model to obtain an initial beam at a second time, wherein the second time is a next time of the first time; a first determination module, configured to determine a similar beam of the initial beam from a codebook of the target device, and determine an initial beam pair according to the initial beam and the similar beam; and a second determination module, configured to determine an optimal beam of the target device on each radio frequency chain according to the initial beam pair, wherein the optimal beam is a beam configuration with the best communication performance.

[0013] According to still another aspect of the embodiments of the present application, an electronic device is also provided, comprising: a memory, configured to store program instructions; and a processor, connected with the memory, configured to execute the program instructions to realize the following functions: obtaining first state information of a target device at a first time, wherein the target device is a base station or a UAV; predicting the first state information by using a prediction model to obtain an initial beam at a second time, wherein the second time is a next time of the first time; determining a similar beam of the initial beam from a codebook of the target device, and determining an initial beam pair according to the initial beam and the similar beam; and determining an optimal beam of the target device on each radio frequency chain according to the initial beam pair, wherein the optimal beam is a beam configuration with the best communication performance.

[0014] According to another aspect of the embodiments of the present application, a nonvolatile storage medium is also provided, which comprises a stored computer program, wherein a device in which the nonvolatile storage medium is located performs the above-mentioned optimal beam determination method by running the computer program.

[0015] According to another aspect of the embodiments of the present application, a computer program product is also provided, which comprises computer instructions, and the computer instructions are executed by a processor to implement the above-mentioned optimal beam determination method.

[0016] In the embodiments of the present application, the first state information of the target device at the first time is obtained, wherein the target device is a base station or a drone; the first state information is predicted by using a prediction model to obtain an initial beam at a second time, wherein the second time is the next time of the first time; a similar beam of the initial beam is determined from a codebook of the target device, and an initial beam pair is determined according to the initial beam and the similar beam; and the optimal beam of the target device at each radio frequency link is determined according to the initial beam pair, wherein the optimal beam is the best beam configuration in terms of communication performance, thereby achieving the purpose of fast and accurate beam tracking, and achieving the technical effect of reducing communication delay, and further solving the technical problems that the initial communication quality after each beam switching is unstable due to random initialization of the traditional heuristic algorithm, which is not conducive to beam tracking and stable communication. BRIEF DESCRIPTION OF DRAWINGS

[0017] The accompanying drawings, which are included to provide a further understanding of the present application, constitute a part of the present application and illustrate the exemplary embodiments of the present application and their descriptions serve to explain the present application, but do not constitute improper limitations on the present application. In the drawings:

[0018] Figure 1 FIG. 1 is a hardware structure block diagram of a computer terminal for implementing an optimal beam determination method according to an embodiment of the present application;

[0019] Figure 2 FIG. 2 is a flowchart of an optimal beam determination method according to an embodiment of the present application;

[0020] Figure 3 FIG. 3 is a memory unit structure diagram of a ConvLSTM according to an embodiment of the present application;

[0021] Figure 4 FIG. 4 is a heuristic beam search schematic diagram according to an embodiment of the present application;

[0022] Figure 5 FIG. 5 is a search number comparison diagram of the present application and the exhaustive search;

[0023] Figure 6 FIG. 6 is a performance curve comparison diagram according to an embodiment of the present application;

[0024] Figure 7 is a structural diagram of a determination device of an optimal beam according to an embodiment of the present application. DETAILED DESCRIPTION

[0025] In order for those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor should fall within the scope of protection of the present application.

[0026] It should be noted that the terms "first", "second", and the like in the specification and claims of the present application and the above-described drawings are used to distinguish similar objects, and do not necessarily have to describe a specific order or a chronological sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device that includes a series of steps or units does not have to be limited to only 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.

[0027] The information collected by the embodiments of the present application is information and data authorized by the user or authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure and application of related data comply with relevant laws, regulations and standards in the relevant region, necessary security measures are taken, public order and good customs are not violated, and appropriate operation portals are provided for users to choose authorization or refuse automatic decision results; if the user chooses to refuse, the expert decision process is entered.

[0028] First, some of the nouns or terms that appear in the process of explaining and describing the embodiments of the present application are applicable to the following explanations:

[0029] Discrete Fourier Transform (DFT): used to analyze the frequency components of a discrete signal in time or space. DFT converts a discrete, periodic or finite length signal into a series of complex coefficients, which describe the amplitude and phase of different frequency components in the signal.

[0030] DFT codebook: In the field of wireless communication, DFT codebook is a set of beam weighting coefficients designed based on DFT principle, used to guide the beamforming of antenna array. Each weighting coefficient in DFT codebook corresponds to a specific beam direction, by adjusting these coefficients, the pointing of the beam can be controlled to maximize the direction towards the target receiver while suppressing interference in other directions. In this application, DFT codebook is used as an efficient beam search library, combined with prediction and search algorithms to quickly find the best beam configuration suitable for communication between UAV and base station.

[0031] Long Short-Term Memory (LSTM): A special type of recurrent neural network designed to overcome the gradient vanishing problem of traditional RNN, suitable for processing and predicting time series data. LSTM introduces memory cells and three gating mechanisms - input gate, forget gate and output gate, which can selectively remember or forget information, effectively learning long-term dependencies, suitable for tasks such as speech recognition, time series prediction, etc. that require understanding the temporal context of data.

[0032] Convolutional Neural Networks (CNN): A model of deep learning designed for processing data with grid structure such as images. Through convolutional layers to capture local features in space, pooling layers to reduce data dimension, and fully connected layers for classification or regression.

[0033] Mean-Square Error (MSE): A common metric for evaluating the accuracy of model predictions, used to measure the average squared difference between predicted and true values, sensitive to the size of errors, suitable for scenarios where data needs to be accurately aligned.

[0034] Mean Absolute Error (MAE): A common metric for evaluating the accuracy of model predictions, used to calculate the average absolute difference between predicted and true values, less sensitive to outliers, providing a direct understanding of errors.

[0035] In related technologies, the road loss of the base station deployment frequency band is high, and the millimeter wave frequency band site has been deployed. In the later 5G and 6G period, the base station needs to rely on a super large-scale antenna array to obtain beamforming gain to resist road loss. The more the number of antennas, the more the number of beams and the narrower the beam, the base station needs to spend more time tracking and searching for the optimal beam to maintain high-speed and low-latency communication. Related technologies cannot meet the fast beam tracking between the base station and the UAV.

[0036] In order to realize the low-latency fast beam tracking of the base station to the UAV, and make the base station and the UAV maintain stable high-speed communication, the embodiments of the present application provide a method for determining an optimal beam, which can be run inFigure 1 The computer terminal shown below is described.

[0037] The method for determining the optimal beam provided by the embodiments of the present application can be executed in a mobile terminal, a computer terminal or similar computing device. Figure 1 A hardware structure block diagram of a computer terminal for implementing the method for determining the optimal beam is shown. As shown in the figure, Figure 1 The computer terminal 10 can include one or more processors (the processor can include but not limited to a microprocessor MCU or a programmable logic device FPGA processing device, etc.), a memory 104 for storing data, and a transmission module 106 for communication function through wired and / or wireless network connection. In addition, it can also include a display, a keyboard, a cursor control device, an input / output interface (I / O interface), a universal serial bus (USB) port (which can be included as one of the ports of the I / O interface), a network interface, a BUS bus. Those skilled in the art can understand, Figure 1 The structure shown is only a schematic, which does not limit the structure of the above-mentioned electronic device. For example, the computer terminal 10 can include more or less components than Figure 1 shown, or have a different configuration than Figure 1 shown.

[0038] It should be noted that the one or more processors and / or other data processing circuits described above can be referred to herein as "data processing circuits" in general. The data processing circuit can be embodied in whole or in part as software, hardware, firmware or any other combination. In addition, the data processing circuit can be a single independent processing module, or any one of the other elements combined into the computer terminal 10 in whole or in part. As referred to in the embodiments of the present application, the data processing circuit as a kind of processor control (for example, the selection of the variable resistance terminal path connected with the interface).

[0039] The memory 104 can be used to store software programs of application software and modules, such as program instructions / data storage devices corresponding to the method for determining an optimal beam in the embodiments of the present application, and the processor executes various functional applications and data processing by running the software programs and modules stored in the memory 104, that is, implements the method for determining an optimal beam described above. The memory 104 can include a high-speed random access memory, and can also include a non-volatile memory, such as one or more magnetic storage devices, flash memories, or other non-volatile solid-state memories. In some examples, the memory 104 can further include a memory remotely arranged with respect to the processor, and these remote memories can be connected to the computer terminal 10 through a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0040] The transmission module 106 is used to receive or send data via a network. Specific examples of the above-mentioned network can include a wireless network provided by a communication provider of the computer terminal 10. In one example, the transmission module 106 includes a network adapter (Network Interface Controller, NIC) which can be connected to other network devices through a base station so as to be able to communicate with the Internet. In one example, the transmission module 106 can be a radio frequency (Radio Frequency, RF) module which is used to communicate with the Internet in a wireless manner.

[0041] The display can be, for example, a touch screen type liquid crystal display (LCD) which can enable a user to interact with the user interface of the computer terminal 10.

[0042] It should be noted that, in some optional embodiments, the above-mentioned Figure 1 The computer terminal shown can include hardware elements (including circuitry), software elements (including computer code stored on a computer-readable medium), or a combination of both hardware and software elements. It should be noted that, Figure 1 is merely one example of a particular implementation, and is intended to illustrate the types of components that can be present in the above-described computer terminal.

[0043] Under the above-mentioned operating environment, the embodiments of the present application provide a method for determining an optimal beam, and it should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a group of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described herein can be executed in an order different from that shown herein.

[0044] Figure 2 is a flowchart of a method for determining an optimal beam according to the embodiments of the present application, and Figure 2As shown, the method comprises the following steps:

[0045] In step S202, first state information of the target device at a first time is obtained, wherein the target device is a base station or a UAV.

[0046] In the above step S202, the first state information can include position information, attitude and direction, and communication state. In the case of a UAV as the target device, the position information in the first state information includes the current latitude, longitude and height of the UAV. The attitude and direction include the pitch angle, roll angle and yaw angle of the UAV. The communication state includes key indicators of the health of the base station and UAV communication link, i.e., parameters related to communication quality, such as reference signal received power (RSRP), signal-to-interference-plus-noise ratio (SINR) and data transmission rate (or uplink rate). In the case of a base station as the target device, the position information in the first state information includes the latitude, longitude and height of the base station, and the attitude and direction include the azimuth angle and downtilt angle of the antenna, indicating the direction of the base station radiation. The communication state includes parameters related to communication quality.

[0047] The first state information is obtained at the first time, so that the subsequent beam tracking algorithm can be based on the latest and most accurate data for prediction and search, ensuring that the communication between the UAV and the base station is always in the best state.

[0048] In step S204, a prediction model is used to predict the first state information to obtain an initial beam at a second time, wherein the second time is the next time of the first time.

[0049] In the above step S204, the prediction model is a trained ConvLSTM model, which can predict a relatively ideal beam direction based on time and spatial characteristics, serving as a starting point for subsequent beam search. Specifically, the prediction model understands the dynamic relationship between device state and beam direction through learning from historical data. The prediction model receives the first state information of the target device at the first time as input, and predicts the communication condition at the second time through its internal computer mechanism (such as convolution operation to capture spatial features and recurrent mechanism to understand time series dependence). The result of the prediction is the initial beam (representing the theoretically optimal initial beam direction). The predicted initial beam serves as the starting point for heuristic search, and the algorithm will further adjust the beam direction based on this, until the best beam is found, ensuring that the communication between the base station and the UAV is always efficient and stable.

[0050] In step S206, a similar beam of the initial beam is determined from a codebook of the target device, and an initial beam pair is determined based on the initial beam and the similar beam.

[0051] In step S206, the codebook is a predefined set of beam weighting coefficients used to generate different beam directions. In wireless communication systems, especially in scenarios using large-scale antenna arrays, codebooks provide an efficient, mathematically-based beamforming approach.

[0052] The codebook described above can be a DFT codebook, which is a set of beam weighting coefficients designed based on the DFT principle. The weighting coefficient of the jth antenna in the ith beam is as follows:

[0053]

[0054] where j represents an imaginary number, is the number of beams, is the number of antennas, the base station has beams, and the UAV has beams.

[0055] Due to the complexity of actual application scenarios and the limitations of model prediction, the initial beam predicted by the model may not accurately match the beam in the DFT codebook. By matching similar beams, the beam configuration closest to the prediction result can be found, the similar beam is obtained, and the initial beam pair is formed according to the initial beam and the similar beam.

[0056] It should be noted that when determining the similar beam of the initial beam, the codebook used can be the DFT codebook of the target device, or other forms of codebooks corresponding to the target device. The DFT codebook described above is only used as an example of one of the codebooks and does not limit the scheme. In actual use, other codebooks can also be used to determine the similar beam of the initial beam.

[0057] Step S208, according to the initial beam pair, determine the optimal beam of the target device on each radio frequency link, wherein the optimal beam is the beam configuration with the best communication performance.

[0058] In step S208, according to the initial beam pair, a heuristic search algorithm can be used to further optimize the beam configuration and determine the optimal beam of the target device on each radio frequency link. The optimal beam refers to the beam configuration that can provide the best communication performance (such as the highest spectral efficiency, the lowest latency, the maximum data transmission rate, etc.) under given communication conditions. This process is essentially based on the model prediction result, through local search and iterative optimization, to find the beam that can truly maximize the communication quality.

[0059] ​​​Through the steps S202 to S208, the purpose of fast and accurate beam tracking is achieved. In the scenario of fast movement of the unmanned aerial vehicle, this method can quickly respond to the change of the position of the unmanned aerial vehicle, reduce the time of beam search, thereby significantly reducing the communication delay and improving the stability of communication. Not only the communication between the unmanned aerial vehicle and the base station is optimized, but also the foundation for efficient communication of multiple radio frequency links is laid, avoiding the problem of communication interruption or quality decline caused by slow beam search process. The unstable communication performance problem caused by random initialization of the traditional heuristic algorithm is effectively solved, and the communication quality after each beam switching is ensured through prediction and optimization of the beam configuration, providing a more stable and high-speed communication experience for mobile devices such as unmanned aerial vehicles. The following is described.

[0060] In the above method for determining the optimal beam, the prediction model is trained in the following manner: a training set for training the prediction model is obtained, wherein the training set includes: first historical state information of the base station, second historical state information of the unmanned aerial vehicle, and historical communication quality data between the base station and the unmanned aerial vehicle; an initial prediction model is used to predict the training set to obtain historical prediction results; a total loss function is determined according to the historical prediction results, the first historical state information, and the historical communication quality data; and the initial prediction model is iteratively trained according to the total loss function to obtain the prediction model.

[0061] In some embodiments of the present application, LSTM, as a commonly used time series prediction model, can better process time series data, but the local features of spatial data are strong, and LSTM cannot depict this feature, so it is not suitable for scenarios where data has spatial characteristics. Similarly, CNN, as a commonly used model for processing spatial data, can better extract the local features of spatial data, but cannot effectively capture the dependency of time series data, so it is not suitable for scenarios where data has time characteristics. In the scenario where the communication data between the base station and the unmanned aerial vehicle has both time and space characteristics, ConvLSTM combining LSTM and CNN architecture is used for beam prediction. ConvLSTM discards the fully connected layer, introduces convolution in space and introduces cycle in time, while retaining the memory unit and gating mechanism of LSTM, can capture time and space characteristics, consider both long-term dependency of time series data and local features of spatial structure, and can effectively process spatio-temporal data. Figure 3 The memory unit structure diagram of ConvLSTM is as follows, and the data is processed in the following order:

[0062] (1)

[0063] (2)

[0064] (3)

[0065] (4)

[0066] (5)

[0067] (6)

[0068] in, It's convolution. It refers to the convolution kernel. Different subscripts in W indicate the convolution kernel used in different processes. This is the bias vector; different subscripts in 'b' represent the bias vectors used in different processes. / It is the hidden state of the previous / current time step. This is the current input data. It is the sigmoid function. It is the hyperbolic tangent function. It is a temporary cell state. It represents the cell state at the current time t. It represents the cell state at the previous time step t-1. , , , , All are calculations The intermediate value. Equations (1), (2) to (4), and (5) to (6) are the calculation processes of the forget gate, input gate, and output gate, respectively. Among them, the forget gate determines the information to be retained at the previous moment, the input gate determines the information to be added, and the output gate determines the information to be output.

[0069] In training the prediction model, a training set is first obtained, where the first historical state information of the base station in the training set can include its position (longitude, latitude, height), configuration parameters of the antenna (such as azimuth and downtilt angle, beamwidth), the second historical state information of the UAV can include its position (longitude, latitude, height), flight attitude information (such as pitch angle, roll angle, yaw angle, relative azimuth angle with the base station), and the historical communication quality data includes reference signal received power (RSRP), signal to interference plus noise ratio (SINR), and data transmission rate, etc. An untrained initial prediction model is used to predict the data in the above training set, and the model attempts to predict the communication quality or beam direction between the base station and the UAV according to the given historical state information of the base station and the UAV, to obtain historical prediction results. Based on the difference between the historical prediction results and the actual historical communication quality data and historical state information, a total loss function is defined. The training of the model is guided by minimizing the total loss function, so as to make it more close to the actual data. With the aid of the back propagation algorithm and the optimizer, the parameters of the model are iteratively adjusted according to the gradient of the total loss function, so that the model is continuously learned and optimized. This process continues until the prediction ability of the model reaches a satisfactory level, or the loss function converges to a stable value. In this way, the initial prediction model is trained for multiple iterations to obtain the prediction model. For example, after 35 rounds of training using 3 million data, the total loss function value of the initial prediction model gradually decreases, and the prediction accuracy is finally improved to 86.7%.

[0070] The whole training process aims to make the prediction model understand the complex interaction between the base station and the UAV through a large amount of historical data, predict the results that conform to the physical laws and environmental characteristics, so that in actual application, when receiving new state information of the base station and the UAV, the model can quickly give a more accurate prediction, which is used to guide the rapid tracking of the beam and the optimization of the communication quality.

[0071] In the above-mentioned method for determining the optimal beam, the second historical state information includes the UAV's first longitude, first latitude, first altitude, UAV's pitch angle, roll angle, yaw angle, and relative azimuth angle between the UAV and the base station. The relative azimuth angle is determined as follows: obtaining the second longitude, second latitude, and second altitude corresponding to the base station, and obtaining the equatorial radius and ellipsoidal flattening; determining a first value based on the first latitude, equatorial radius, and ellipsoidal flattening; determining a second value based on the second latitude, equatorial radius, and ellipsoidal flattening; determining the first geocentric coordinates of the UAV based on the first value, first longitude, first latitude, and first altitude; and determining the second value based on the second value. The second geocentric coordinates of the base station are determined using the first geocentric coordinates (including the coordinates of the UAV in the X, Y, and Z axes) and the second geocentric coordinates (including the coordinates of the base station in the X, Y, and Z axes). Based on these coordinates, the coordinate difference between the UAV and the base station is determined, including the coordinate differences in the X, Y, and Z axes. The transformed coordinates of the UAV are determined based on the coordinate difference and the base station rotation matrix, where the base station rotation matrix is ​​determined using the second longitude and second latitude. Finally, the relative azimuth is determined based on the transformed coordinates and the origin.

[0072] In some embodiments of this application, to facilitate the extraction of spatial features, the UAV coordinates need to be transformed to a local coordinate system with the base station as the origin. Specifically, let the second longitude, second latitude, and second altitude of the base station be... , , The first longitude, first latitude, and first altitude of the drone are , , First, convert the latitude, longitude, and altitude (i.e., longitude, latitude, and altitude, abbreviated as latitude, longitude, and altitude) coordinates of the base station and the UAV to geocentric coordinates, and then convert them to a station-centered coordinate system with the base station as the origin. In the station-centered coordinate system, the East (E) axis extends eastward along the latitude circle, the North (N) axis extends northward along the longitude circle, and the U-axis is perpendicular to the ground and pointing upwards. Calculate the radius of curvature of the circumpolar coordinate system using the following formula. :

[0073] (7)

[0074] (8)

[0075] in, It is the equatorial radius. It refers to the ellipsoidal flattening. This is used in calculating the flattening of the ellipsoid corresponding to When, in formula (7) This represents the first latitude, and what we get at this point is... For the above first value, in the calculation of the base station corresponding to the second latitude, the first obtained for the above second value.

[0076] The base station and the unmanned aerial vehicle respectively calculate the respective geocentric coordinates according to formula (9) to formula (11):

[0077] (9)

[0078] (10)

[0079] (11)

[0080] In the calculation of the first geocentric coordinates corresponding to the unmanned aerial vehicle, the first value of in formula (9)-(11) represents the first value, the first height, the first latitude, the first longitude, and the X, Y and Z obtained by substituting the data corresponding to each letter into formula (9)-(11) are the above first geocentric coordinates. In the calculation of the second geocentric coordinates corresponding to the base station, the second value of in formula (9)-(11) represents the second value, the second height, the second latitude, the second longitude, and the X, Y and Z obtained by substituting the data corresponding to each letter into formula (9)-(11) are the above second geocentric coordinates.

[0081] According to the first geocentric coordinates and the second geocentric coordinates, the coordinate difference value is determined, specifically, let , , be the geocentric coordinate difference value of the unmanned aerial vehicle and the base station, , , respectively represent the coordinates in the X-axis, Y-axis and Z-axis directions of the first geocentric coordinates corresponding to the unmanned aerial vehicle, , , respectively represent the coordinates in the X-axis, Y-axis and Z-axis directions of the second geocentric coordinates corresponding to the base station. After constructing the base station rotation matrix (i.e. the first matrix on the right side of formula 12), the coordinate difference value can be processed according to formula (12) to obtain the ENU coordinates of the unmanned aerial vehicle in the station-centered coordinate system:

[0082] (12) ​

[0083] After obtaining the ENU coordinates of the UAV in the station-centered coordinate system (i.e., the above-mentioned converted coordinates), the base station takes the ENU coordinates of the station-centered coordinate system as the coordinate origin, and according to the converted coordinates and the coordinate origin, the relative azimuth angle between the UAV and the base station can be obtained.

[0084] In the above step, the total loss function is determined according to the historical prediction result, the first historical state information and the historical communication quality data, including: obtaining the first predicted azimuth angle, the first predicted downtilt angle and the first predicted rate of the historical prediction result at time t, obtaining the real azimuth angle and the real downtilt angle in the first historical state information, and obtaining the real rate in the historical communication quality data, wherein t is a positive integer; determining a main loss function according to the first predicted azimuth angle, the first predicted downtilt angle, the real azimuth angle and the real downtilt angle; determining a rate loss function according to the predicted rate and the real rate; obtaining the second predicted azimuth angle and the second predicted downtilt angle of the historical prediction result at time t-1; determining a beam smoothness constraint loss function according to the first predicted azimuth angle, the second predicted azimuth angle, the first predicted downtilt angle and the second predicted downtilt angle; determining an effective coverage constraint loss function according to the first predicted azimuth angle, a preset maximum azimuth angle, a preset minimum azimuth angle, the first predicted downtilt angle, a preset maximum downtilt angle and a preset minimum downtilt angle; and determining the total loss function according to the main loss function, the rate loss function, the beam smoothness constraint loss function and the effective coverage constraint loss function.

[0085] In some embodiments of the present application, in order to improve the beam prediction accuracy of ConvLSTM, the total loss function used in the prediction model includes a main loss function, a rate loss function, a physical constraint loss function and an environment adaptive weight. The main loss function includes an angle smoothness loss function and a cosine similarity loss function. The physical constraint loss function includes a beam smoothness constraint loss function and an effective coverage constraint loss function. Specifically, the total loss function is as follows:

[0086] (13)

[0087] (14)

[0088] (15)

[0089] (16)

[0090] (17) ​​​​

[0091] (18)

[0092] (19)

[0093] (20)

[0094] wherein, is a real azimuth angle / real downtilt angle, is a predicted azimuth angle / predicted downtilt angle, which can be a first predicted azimuth angle / first predicted downtilt angle corresponding to the t time point, is a difference between the real azimuth angle and the predicted azimuth angle, is a difference between the real downtilt angle and the predicted downtilt angle, is an error threshold value, is a cosine balance coefficient, is an environment weight, is an environment threshold value of an angle prediction weight, and SINR represents a signal to interference plus noise ratio, is a rate coefficient for preventing a main loss from being disturbed, is a real / predicted rate, which can be a first predicted rate corresponding to the t time point, and respectively are a first predicted azimuth angle / first predicted downtilt angle at the t time point and a second predicted azimuth angle / second predicted downtilt angle at the t-1 time point, is a smoothing coefficient for suppressing beam hopping to ensure stable communication, is a penalty coefficient for exceeding a physical limit, represents a preset azimuth angle minimum value, represents a preset azimuth angle maximum value, represents a preset downtilt angle minimum value, represents a preset downtilt angle maximum value, and RELU represents an activation function. Wherein, MSE and MAE are fused, small errors use MSE, and large errors use MAE, which can reduce the influence of abnormal values; The angle prediction weight is adjusted according to the environment, which can reduce the influence of interference on the model.

[0095] ​​​​​​​In step S204 of the above-mentioned method for determining the optimal beam, the prediction model is used to predict the first state information to obtain the initial beam at the second time, including: using the prediction model to predict the first state information to obtain the second state information at the second time; and determining the initial beam at the second time based on the second state information.

[0096] In some embodiments of this application, the first state information is input into a pre-trained prediction model (such as a ConvLSTM model). Based on its internal learning mechanism and prior training data experience, the model predicts the possible state information of the target device at a second time step, i.e., the second state information. According to the second state information, the model calculates or selects an initial beam direction, which is considered to be most likely to provide optimal communication quality in the current prediction environment, thereby determining the initial beam at the second time step.

[0097] In step S206 of the above-mentioned method for determining the optimal beam, determining the similar beams of the initial beam from the codebook of the target device includes: determining the similarity between the initial beam and each beam in the codebook of the target device to obtain a similarity set; and determining the beam with the highest similarity in the similarity set as the similar beam of the initial beam.

[0098] In some embodiments of this application, the base station and the drone communicate using their respective codebooks (such as DFT codebooks). Assuming the target device is the base station, let the model predict the initial beam at the second time step as... The beam in the DFT codebook is .because Limited accuracy makes it impossible to match A perfect match requires calculating beam similarity. , use and Most similar As a similar beam, when the target device is a drone, the processing method is the same as that on the base station side. The closer the similarity is to 1, the higher the beam similarity.

[0099] (twenty one)

[0100] The similarity between the initial beam and each beam in the codebook (such as the DFT codebook) of the target device is determined according to formula (21), and a similarity set is obtained. The beam with the highest similarity in the similarity set is determined as the similar beam of the initial beam.

[0101] In step S208 of the method for determining the optimal beam, the optimal beam of the target device on each radio frequency chain is determined according to the initial beam pair, including: step 1: obtaining a first beam pair corresponding to a current time, and obtaining a neighboring beam pair of the first beam pair; step 2: determining a first spectral efficiency corresponding to the first beam pair, and determining a second spectral efficiency corresponding to the neighboring beam pair; step 3: in the case where a difference between the second spectral efficiency and the first spectral efficiency is greater than or equal to a preset threshold, determining a beam pair change rate according to the first spectral efficiency, the second spectral efficiency and the initial beam pair; step 4: determining an intermediate value according to the beam pair change rate, a decay coefficient and a zero-prevention coefficient; step 5: determining a second beam pair corresponding to a next time of the current time according to the intermediate value, the beam pair change rate and the first beam pair, and determining the second beam pair as the first beam pair; and step 6: repeating steps 1 to 5 until the iteration is stopped when the difference between the second spectral efficiency and the first spectral efficiency is less than the preset threshold, and determining the second beam pair obtained in the last time as the optimal beam.

[0102] In some embodiments of the present application, the base station opens the antenna of the first radio frequency chain with the unmanned aerial vehicle, the base station transmits the reference signal using the sending beam of the predicted beam pair, and the unmanned aerial vehicle receives using the receiving beam of the predicted beam pair. As shown in FIG. 2, each point is a beam pair, and the red point beam search area in the square formed around the blue point predicted beam pair. The unmanned aerial vehicle uses the heuristic algorithm to determine the beam pair for the next communication and feeds back to the base station from the blue point. After receiving the information, the base station transmits the reference signal using the beam, and the unmanned aerial vehicle uses the heuristic algorithm to determine the beam pair for the next communication and feeds back to the base station again, and the above process is repeated until the optimal beam of the first radio frequency chain between the base station and the unmanned aerial vehicle is determined, that is, the optimal beam of the first radio frequency chain between the base station and the unmanned aerial vehicle is determined. Figure 4 The code word used by the first subarray, wherein, is the analog beam matrix of the base station / unmanned aerial vehicle. The base station opens the second radio frequency chain with the unmanned aerial vehicle, and determines the code word used by the first subarray according to the above steps. The code word used by the second subarray, and the process is repeated until the code word used by all subarrays in is determined. The code word used by the second subarray, and the process is repeated until the code word used by all subarrays in is determined.

[0103] (22)

[0104] (23)

[0105] (24)

[0106] (25)

[0107] ​​​​​wherein, is the first beam pair corresponding to the current moment, is the beam pair change rate, is the adjacent beam pair of the first beam pair, , is the initial beam pair, represents the first spectral efficiency, represents the second spectral efficiency, is the attenuation coefficient, and the value is 0.85, is the anti-zero coefficient, and the value is , , , represents the intermediate value, represents the second beam pair. Iteration is stopped when the beam pair spectral efficiency is less than a preset threshold (such as ), and the last obtained second beam pair is determined as the optimal beam, and the optimal beam pair is used for communication.

[0108] Spectral efficiency (abbreviation: spectral efficiency) is used to evaluate the communication quality, and the spectral efficiency is calculated as follows:

[0109] (26)

[0110] wherein, / is the number of radio frequency links (total) of the base station / unmanned aerial vehicle, is the signal power, is the noise variance, / is the analog beam matrix of the base station / unmanned aerial vehicle, H represents the channel matrix, and I represents the unit matrix, and the dimension of the unit matrix is determined according to .

[0111] Since the scheme in the embodiment of the present application has no complex closed-form expression, the algorithm complexity of the actual simulation is given, Figure 5 is a search frequency comparison diagram of the present application and the traversal search, and the search frequency is the average of 5000 independent searches. The simulation configuration is: 256 antennas of the base station, 4 radio frequency links, and full connection structure; 16 antennas of the unmanned aerial vehicle, 2 links, and full connection structure. The traversal search frequency in the diagram is about 110 billion times, and the search frequency of the scheme in the present application is only 15 times. Compared with the optimal performance of the traversal search, the complexity of the beam switching scheme proposed in the embodiment of the present application is reduced by 99.9999%. Figure 6 is a performance curve comparison diagram according to the embodiment of the present application, from Figure 6 It can be seen that the performance of the scheme of the present application is about 99% of the optimal performance, which has obvious advantages.

[0112] Figure 7 is a structural diagram of a determination device of an optimal beam according to an embodiment of the present application, as shown in the figure, the device comprises: Figure 7

[0113] An acquisition module 40 is configured to acquire first state information of a target device at a first time point, wherein the target device is a base station or a UAV.

[0114] A prediction module 42 is configured to predict the first state information by using a prediction model to obtain an initial beam at a second time point, wherein the second time point is a next time point of the first time point.

[0115] A first determination module 44 is configured to determine a similar beam of the initial beam from a codebook of the target device, and determine an initial beam pair according to the initial beam and the similar beam.

[0116] A second determination module 46 is configured to determine an optimal beam of the target device on each radio frequency chain according to the initial beam pair, wherein the optimal beam is a beam configuration with the best communication performance.

[0117] Through the acquisition module, the prediction module, the first determination module and the second determination module in the determination device of the optimal beam, the purpose of fast and accurate beam tracking is achieved, thereby realizing the technical effect of reducing the communication delay, and further solving the technical problem that the initial communication quality after each beam switching is unstable due to random initialization of the traditional heuristic algorithm, which is not conducive to beam tracking and stable communication.

[0118] In the determination device of the optimal beam, a training module 48 is further included, which is configured to train the prediction model. Specifically, the prediction model is obtained by training in the following manner: a training set for training the prediction model is acquired, wherein the training set comprises first historical state information of a base station, second historical state information of a UAV, and historical communication quality data between the base station and the UAV; an initial prediction model is used to predict the training set to obtain a historical prediction result; a total loss function is determined according to the historical prediction result, the first historical state information and the historical communication quality data; and the initial prediction model is iteratively trained according to the total loss function to obtain the prediction model.

[0119] ​In the training module in the determination apparatus of the optimal beam, the second historical state information includes a first longitude, a first latitude, a first altitude of the unmanned aerial vehicle, a pitch angle, a roll angle, a yaw angle of the unmanned aerial vehicle, and a relative azimuth angle between the unmanned aerial vehicle and the base station, wherein the training module is further configured to determine the relative azimuth angle, and specifically, the relative azimuth angle is determined by the following manner: obtaining a second longitude, a second latitude and a second altitude corresponding to the base station, and obtaining an equatorial radius and an ellipsoid flattening; determining a first value according to the first latitude, the equatorial radius and the ellipsoid flattening, and determining a second value according to the second latitude, the equatorial radius and the ellipsoid flattening; determining a first geocentric coordinate corresponding to the unmanned aerial vehicle according to the first value, the first longitude, the first latitude and the first altitude, and determining a second geocentric coordinate corresponding to the base station according to the second value, the second longitude, the second latitude and the second altitude, wherein the first geocentric coordinate includes coordinates of the unmanned aerial vehicle in X-axis, Y-axis and Z-axis directions, and the second geocentric coordinate includes coordinates of the base station in the X-axis, the Y-axis and the Z-axis directions; determining a coordinate difference value according to the first geocentric coordinate and the second geocentric coordinate, wherein the coordinate difference value includes coordinate difference values of the unmanned aerial vehicle and the base station in the X-axis, the Y-axis and the Z-axis directions; determining a conversion coordinate of the unmanned aerial vehicle according to the coordinate difference value and a base station rotation matrix, wherein the base station rotation matrix is determined by the second longitude and the second latitude; and determining the relative azimuth angle according to the conversion coordinate and a coordinate origin.

[0120] In the training module in the determination apparatus of the optimal beam, the training module is further configured to obtain a first predicted azimuth angle, a first predicted downtilt angle and a first predicted rate in a t-th moment in the historical prediction result, obtain a real azimuth angle and a real downtilt angle in the first historical state information, and obtain a real rate in the historical communication quality data, wherein t is a positive integer; determine a main loss function according to the first predicted azimuth angle, the first predicted downtilt angle, the real azimuth angle and the real downtilt angle; determine a rate loss function according to the predicted rate and the real rate; obtain a second predicted azimuth angle and a second predicted downtilt angle in a (t-1)-th moment in the historical prediction result; determine a beam smoothness constraint loss function according to the first predicted azimuth angle, the second predicted azimuth angle, the first predicted downtilt angle and the second predicted downtilt angle; determine an effective coverage constraint loss function according to the first predicted azimuth angle, a preset maximum azimuth angle, a preset minimum azimuth angle, the first predicted downtilt angle, a preset maximum downtilt angle and a preset minimum downtilt angle; and determine a total loss function according to the main loss function, the rate loss function, the beam smoothness constraint loss function and the effective coverage constraint loss function.

[0121] In the prediction module in the determination apparatus of the optimal beam, the prediction module is further configured to predict the first state information by using a prediction model to obtain second state information in a second moment; and determine an initial beam in the second moment according to the second state information.

[0122] In the first determining module in the apparatus for determining the optimal beam, the first determining module is further configured to determine a similarity between the initial beam and each beam in a codebook of the target device, to obtain a set of similarities; and determine the beam with the highest similarity in the set of similarities as the similar beam of the initial beam.

[0123] In the second determining module in the apparatus for determining the optimal beam, the second determining module is further configured to implement the following steps: Step 1: obtaining a first beam pair corresponding to a current time and obtaining a neighboring beam pair of the first beam pair; Step 2: determining a first spectral efficiency corresponding to the first beam pair and determining a second spectral efficiency corresponding to the neighboring beam pair; Step 3: in a case where a difference between the second spectral efficiency and the first spectral efficiency is greater than or equal to a preset threshold, determining a beam pair change rate based on the first spectral efficiency, the second spectral efficiency, and the initial beam pair; Step 4: determining an intermediate value based on the beam pair change rate, a decay coefficient, and a zero-prevention coefficient; Step 5: determining a second beam pair corresponding to a next time of the current time based on the intermediate value, the beam pair change rate, and the first beam pair, and determining the second beam pair as the first beam pair; and Step 6: repeating Steps 1 to 5 until the iteration is stopped when the difference between the second spectral efficiency and the first spectral efficiency is less than the preset threshold, and determining the second beam pair obtained in the last iteration as the optimal beam.

[0124] It should be noted that, Figure 7 The apparatus for determining the optimal beam is configured to implement the method for determining the optimal beam as shown in Figure 2 Therefore, the related explanations and descriptions in the method for determining the optimal beam are also applicable to the apparatus for determining the optimal beam, and will not be repeated here.

[0125] The embodiments of the present application also provide an electronic device, which comprises a memory and a processor, wherein the memory is configured to store program instructions; the processor is connected with the memory and is configured to execute the program instructions to realize the following functions: obtaining first state information of a target device at a first time, wherein the target device is a base station or a drone; predicting the first state information by using a prediction model to obtain an initial beam at a second time, wherein the second time is a next time of the first time; determining a similar beam of the initial beam from a codebook of the target device, and determining an initial beam pair based on the initial beam and the similar beam; and determining an optimal beam of the target device at each radio frequency chain based on the initial beam pair, wherein the optimal beam is a beam configuration with the best communication performance.

[0126] It should be noted that the electronic device is configured to implement the method for determining the optimal beam as shown in Figure 2 Therefore, the related explanations and descriptions in the method for determining the optimal beam are also applicable to the electronic device, and will not be repeated here.

[0127] The embodiment of the present application further provides a nonvolatile storage medium comprising a stored computer program, wherein a device in which the nonvolatile storage medium is located executes the following optimal beam determination method by running the computer program: obtaining first state information of a target device at a first time, wherein the target device is a base station or a drone; predicting the first state information by using a prediction model to obtain an initial beam at a second time, wherein the second time is a next time of the first time; determining a similar beam of the initial beam from a codebook of the target device, and determining an initial beam pair according to the initial beam and the similar beam; and determining an optimal beam of the target device at each radio frequency chain according to the initial beam pair, wherein the optimal beam is a beam configuration with the best communication performance.

[0128] It should be noted that the nonvolatile storage medium is used to execute the optimal beam determination method shown in the above description, and thus the related explanations in the optimal beam determination method are also applicable to the nonvolatile storage medium, which will not be described here again. Figure 2

[0129] The embodiment of the present application further provides a computer program product comprising computer instructions, which, when executed by a processor, implement the steps of the optimal beam determination method in the various embodiments of the present application.

[0130] The embodiment of the present application further provides a computer program, which, when executed by a processor, implements the steps of the optimal beam determination method in the various embodiments of the present application.

[0131] The serial numbers of the embodiments of the present application are only for description, and do not represent the advantages or disadvantages of the embodiments.

[0132] In the above-described embodiments of the present application, the description of each embodiment has its own focus, and the parts not described in detail in a certain embodiment can be referred to the related description of other embodiments.

[0133] In the several embodiments provided by the present application, it should be understood that the disclosed technology can be implemented in other ways. Of course, the unit described as the division is only a logical function division, and there can be another division manner during actual implementation, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units shown or discussed can be indirect coupling or communication connection through some interfaces, units or modules, and can be electrical or other forms.

[0134] ​The units described as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units, that is, may be located in one place, or may be distributed to multiple units. Part or all of the units can be selected according to actual needs to achieve the purpose of the embodiment scheme.

[0135] In addition, the functional units in each embodiment of the present application can be integrated in one processing unit, or each unit can be physically present alone, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.

[0136] The integrated unit, if realized in the form of a software functional unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application essentially or the part of the prior art that contributes to the technical solutions or all or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The foregoing storage medium includes: a U disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a mobile hard disk, a magnetic disk or an optical disk, and various program code storage media.

[0137] The above is only the preferred embodiment of the present application, and it should be pointed out that for those skilled in the art, without departing from the principles of the present application, a number of improvements and refinements can be made, which should be considered as the protection scope of the present application.

Claims

1. A method for determining an optimal beam, the method comprising: receiving a plurality of beams; and determining an optimal beam from the plurality of beams. The method comprises the following steps: acquiring first state information of a target device at a first time, wherein the target device is a base station or a drone; using a prediction model to predict the first state information to obtain an initial beam at a second time, wherein the second time is the next time of the first time; determining a similar beam of the initial beam from a codebook of the target device, and determining an initial beam pair according to the initial beam and the similar beam; determining an optimal beam of the target device on each radio frequency link according to the initial beam pair, wherein the optimal beam is a beam configuration with the best communication performance; determining an optimal beam of the target device on each radio frequency link according to the initial beam pair, comprising: step 1: acquiring a first beam pair corresponding to a current time, and acquiring a neighboring beam pair of the first beam pair; step 2: determining a first spectral efficiency corresponding to the first beam pair, and determining a second spectral efficiency corresponding to the neighboring beam pair; step 3: in the case where the difference between the second spectral efficiency and the first spectral efficiency is greater than or equal to a preset threshold, determining a beam pair change rate according to the first spectral efficiency, the second spectral efficiency and the initial beam pair; step 4: determining an intermediate value according to the beam pair change rate, a decay coefficient and a zero prevention coefficient; step 5: determining a second beam pair corresponding to the next time of the current time according to the intermediate value, the beam pair change rate and the first beam pair, and determining the second beam pair as the first beam pair; step 6: repeating steps 1 to 5 until the difference between the second spectral efficiency and the first spectral efficiency is less than the preset threshold, and stopping iteration, and determining the last obtained second beam pair as the optimal beam.

2. The method of claim 1, wherein, The prediction model is trained by the following method: acquiring a training set for training the prediction model, wherein the training set comprises first historical state information of the base station, second historical state information of the drone, and historical communication quality data between the base station and the drone; using an initial prediction model to predict the training set to obtain a historical prediction result; determining a total loss function according to the historical prediction result, the first historical state information and the historical communication quality data; iteratively training the initial prediction model according to the total loss function to obtain the prediction model.

3. The method of claim 2, wherein, The second historical state information comprises a first longitude, a first latitude, a first height of the drone, a pitch angle, a roll angle and a yaw angle of the drone, and a relative azimuth angle between the drone and the base station, wherein the relative azimuth angle is determined by the following method: acquiring a second longitude, a second latitude and a second height corresponding to the base station, and acquiring an equatorial radius and an ellipsoid flattening; determining a first value according to the first latitude, the equatorial radius and the ellipsoid flattening, and determining a second value according to the second latitude, the equatorial radius and the ellipsoid flattening; Determine a first geocentric coordinate corresponding to the UAV according to the first value, the first longitude, the first latitude, and the first height, and determine a second geocentric coordinate corresponding to the base station according to the second value, the second longitude, the second latitude, and the second height, wherein the first geocentric coordinate comprises coordinates of the UAV in X-axis, Y-axis, and Z-axis directions, and the second geocentric coordinate comprises coordinates of the base station in the X-axis, Y-axis, and Z-axis directions; Determine a coordinate difference value according to the first geocentric coordinate and the second geocentric coordinate, wherein the coordinate difference value comprises coordinate difference values of the UAV and the base station in the X-axis, Y-axis, and Z-axis directions; Determine a conversion coordinate of the UAV according to the coordinate difference value and a base station rotation matrix, wherein the base station rotation matrix is determined by the second longitude and the second latitude; Determine the relative azimuth angle according to the conversion coordinate and a coordinate origin.

4. The method of claim 2, wherein, Determine a total loss function according to the historical prediction result, the first historical state information, and the historical communication quality data, including: Obtain a first predicted azimuth angle, a first predicted downtilt angle, and a first predicted rate of the historical prediction result at time t, obtain a real azimuth angle and a real downtilt angle of the first historical state information, and obtain a real rate of the historical communication quality data, wherein t is a positive integer; Determine a main loss function according to the first predicted azimuth angle, the first predicted downtilt angle, the real azimuth angle, and the real downtilt angle; Determine a rate loss function according to the predicted rate and the real rate; Obtain a second predicted azimuth angle and a second predicted downtilt angle of the historical prediction result at time t-1; Determine a beam smoothness constraint loss function according to the first predicted azimuth angle, the second predicted azimuth angle, the first predicted downtilt angle, and the second predicted downtilt angle; Determine an effective coverage constraint loss function according to the first predicted azimuth angle, a preset maximum azimuth angle, a preset minimum azimuth angle, the first predicted downtilt angle, a preset maximum downtilt angle, and a preset minimum downtilt angle; Determine the total loss function according to the main loss function, the rate loss function, the beam smoothness constraint loss function, and the effective coverage constraint loss function.

5. The method of claim 1, wherein, Predict the first state information by using a prediction model to obtain an initial beam at a second time, including: Predict the first state information by using the prediction model to obtain second state information at the second time; Determine the initial beam at the second time according to the second state information.

6. The method of claim 1, wherein, Determine a similar beam of the initial beam from a codebook of the target device, including: Determine a similarity between the initial beam and each beam in the codebook of the target device to obtain a similarity set; Determine the beam with the highest similarity in the similarity set as the similar beam of the initial beam.

7. An optimal beam determination device, characterized in that, Including: An acquisition module is configured to acquire first state information of a target device at a first time, wherein the target device is a base station or a UAV; a prediction module, configured to predict the first state information by using a prediction model to obtain an initial beam at a second time, wherein the second time is a next time of the first time; a first determination module, configured to determine a similar beam of the initial beam from a codebook of the target device, and determine an initial beam pair according to the initial beam and the similar beam; a second determination module, configured to determine an optimal beam of the target device on each radio frequency chain according to the initial beam pair, wherein the optimal beam is a beam configuration with the best communication performance; the determination of the optimal beam of the target device on each radio frequency chain according to the initial beam pair comprises the following steps: step 1: obtaining a first beam pair corresponding to a current time and obtaining a neighboring beam pair of the first beam pair; step 2: determining a first spectrum efficiency corresponding to the first beam pair and determining a second spectrum efficiency corresponding to the neighboring beam pair; step 3: in a case where a difference between the second spectrum efficiency and the first spectrum efficiency is greater than or equal to a preset threshold, determining a beam pair change rate according to the first spectrum efficiency, the second spectrum efficiency and the initial beam pair; step 4: determining an intermediate value according to the beam pair change rate, an attenuation coefficient and a zero-prevention coefficient; step 5: determining a second beam pair corresponding to a next time of the current time according to the intermediate value, the beam pair change rate and the first beam pair, and determining the second beam pair as the first beam pair; step 6: repeating steps 1 to 5 until the difference between the second spectrum efficiency and the first spectrum efficiency is less than the preset threshold, and stopping iteration, and determining the last obtained second beam pair as the optimal beam.

8. An electronic device, comprising: comprise: a memory, configured to store program instructions; The processor is connected with the memory and used for executing program instructions to realize the following functions: obtaining first state information of a target device at a first time, wherein the target device is a base station or a drone; predicting the first state information by using a prediction model to obtain an initial beam at a second time, wherein the second time is a next time of the first time; determining a similar beam of the initial beam from a codebook of the target device, and determining an initial beam pair according to the initial beam and the similar beam; determining an optimal beam of the target device at each radio frequency link according to the initial beam pair, wherein the optimal beam is a beam configuration with the best communication performance; determining the optimal beam of the target device at each radio frequency link according to the initial beam pair, comprising: step 1: obtaining a first beam pair corresponding to a current time, and obtaining a neighboring beam pair of the first beam pair; step 2: determining a first spectral efficiency corresponding to the first beam pair, and determining a second spectral efficiency corresponding to the neighboring beam pair; step 3: in the case that a difference between the second spectral efficiency and the first spectral efficiency is greater than or equal to a preset threshold, determining a beam pair change rate according to the first spectral efficiency, the second spectral efficiency and the initial beam pair; step 4: determining an intermediate value according to the beam pair change rate, an attenuation coefficient and an anti-zero coefficient; step 5: determining a second beam pair corresponding to a next time of the current time according to the intermediate value, the beam pair change rate and the first beam pair, and determining the second beam pair as the first beam pair; step 6: repeating steps 1 to 5 until the difference between the second spectral efficiency and the first spectral efficiency is less than the preset threshold, and stopping iteration, and determining the last obtained second beam pair as the optimal beam.

9. A non-volatile storage medium, characterized by The non-volatile storage medium comprises a stored computer program, wherein a device where the non-volatile storage medium is located executes the optimal beam determination method in any one of claims 1 to 6 by running the computer program.

10. A computer program product comprising computer instructions, characterized in that, The computer instructions are executed by the processor to realize the optimal beam determination method in any one of claims 1 to 6.

Images