Beam determination method, nonvolatile storage medium and computer equipment

Abstract

The invention discloses a beam determination method, a nonvolatile storage medium and computer equipment. The method comprises the following steps: acquiring various beam selection data fed back by target equipment; the multiple kinds of beam selection data are input into a neural network model, the neural network model outputs respective preset search values of multiple beams, the neural network model comprises multiple sub-networks corresponding to the multiple kinds of beam selection data in a one-to-one mode, and the multiple sub-networks are used for searching the multiple beams according to the correspondingly input beam selection data. Determining a group of overall beam initial values and a group of offset beam initial values; determining a plurality of candidate beams in the plurality of beams according to the respective search values of the plurality of beams; and measuring the plurality of candidate beams to obtain measurement results of the plurality of candidate beams, and selecting a target beam communicating with the target device from the plurality of candidate beams. The technical problem that the accuracy of beam prediction is not high due to the fact that information which can be obtained is not fully utilized for beam prediction is solved.

Description

Technical Field

[0001] The present invention relates to the field of communication technology, and in particular to a beam determination method, a non-volatile storage medium, and a computer device. Background Art

[0002] The optimal communication beam between a base station and a user depends on many factors, including the relative position of the base station and the user, the location of obstructions in the channel environment, and the positions of scatterers and reflective surfaces. Related technologies typically use only traditional channel-related parameters for beam prediction, such as channel state information and Doppler shift, resulting in low beam prediction accuracy.

[0003] To address the above-mentioned problems, no effective solutions have been proposed so far. Summary of the Invention

[0004] Embodiments of the present invention provide a beam determination method, a non-volatile storage medium, and a computer device to at least solve the technical problem that beam prediction is performed only using traditional channel-related parameters, resulting in low beam prediction accuracy.

[0005] According to one aspect of an embodiment of the present invention, a beam determination method is provided, comprising: obtaining a plurality of beam selection data for beam selection fed back by a target device, wherein the plurality of beam selection data comprises communication state data for describing a communication channel, and perception state data for describing a position and motion state of the target device, wherein the target device comprises a terminal and a base station; inputting the plurality of beam selection data into a neural network model, and the neural network model outputting search values ​​of respective predetermined plurality of beams, wherein the neural network model comprises a plurality of sub-networks corresponding one-to-one to the plurality of beam selection data, and the plurality of sub-networks respectively determine a set of overall beam initial values ​​and a set of overall beam initial values ​​according to the corresponding input beam selection data. The offset beam initial value, the search value of each of the multiple beams is jointly determined by multiple groups of overall beam initial values ​​and multiple groups of offset beam initial values ​​output by multiple sub-networks, wherein the overall beam initial value is the initial search value of each of the multiple beams, and the offset beam initial value is used to characterize the degree of influence of the corresponding beam selection data on the search value of the multiple offset beams; according to the respective search values ​​of the multiple beams, multiple candidate beams are determined from the multiple beams; the multiple candidate beams are measured to obtain measurement results of the multiple candidate beams, wherein the measurement results of the multiple candidate beams are used to characterize the communication quality of the multiple candidate beams; according to the measurement results, a target beam for communicating with the target device is selected from the multiple candidate beams.

[0006] Optionally, a plurality of beam selection data are input into the neural network model, and the neural network model outputs the search values ​​of the predetermined plurality of beams, including: inputting the plurality of beam selection data into the plurality of sub-networks one by one, and the plurality of sub-networks respectively outputting a set of overall beam initial values ​​and a set of offset beam initial values; inputting the plurality of beam selection data into the weight selection network included in the neural network model, and the weight selection network outputting the overall weights and offset weights of the plurality of sub-networks; and determining the search values ​​of the plurality of beams according to the initial search values ​​of the plurality of beams and the overall weights and offset weights of the plurality of sub-networks.

[0007] Optionally, the method further includes: when the measurement results of multiple candidate beams meet the first predetermined condition, multiple beam selection data, the search values ​​of each of the multiple beams and the measurement results of the multiple candidate beams are combined into training samples and added to the online training sample set.

[0008] Optionally, the method also includes: when the number of samples included in the online training sample set meets the second predetermined condition, using the online training sample set to perform online training on the neural network model to obtain an updated neural network model, and determining the beam used for communication based on the updated neural network model.

[0009] Optionally, the neural network model is trained online using an online training sample set to obtain an updated neural network model, including: using the online training sample set, with the goal of minimizing the gap between the search value of each of the multiple beams output by the neural network model and the target search value, training the multiple sub-networks and weight selection network included in the neural network model to obtain multiple trained sub-networks and trained weight selection networks, wherein the target search value is determined based on the measurement results of the multiple candidate beams; and constructing an updated neural network model based on the multiple trained sub-networks and the trained weight selection network.

[0010] Optionally, the first predetermined condition is that the beam with the highest search value among the multiple beams is not the same beam as the beam with the best communication quality represented by the measurement result.

[0011] Optionally, the method further includes: among the multiple sub-networks, marking the sub-networks whose sum of the overall weight and the offset weight assigned within a predetermined time period is less than a predetermined threshold as disabled, thereby obtaining an updated neural network model.

[0012] According to another aspect of an embodiment of the present invention, a non-volatile storage medium is provided. The non-volatile storage medium includes a stored program, wherein when the program is running, the device where the non-volatile storage medium is located is controlled to execute any one of the above-mentioned beam determination methods.

[0013] According to yet another aspect of an embodiment of the present invention, a computer device is provided. The computer device includes a processor, and the processor is configured to run a program. When the program is run, any one of the above-mentioned beam determination methods is executed.

[0014] According to yet another aspect of an embodiment of the present invention, a computer program product is provided, including a computer program. When the computer program is executed by a processor, any one of the above-mentioned beam determination methods is implemented.

[0015] In an embodiment of the present invention, a plurality of beam selection data for beam selection fed back by a target device are obtained, wherein the plurality of beam selection data include communication status data for describing a communication channel, and perception status data for describing a position and motion status of a target device, wherein the target device includes a terminal and a base station; the plurality of beam selection data are input into a neural network model, and the neural network model outputs search values ​​of a plurality of predetermined beams, wherein the neural network model includes a plurality of sub-networks corresponding one-to-one to the plurality of beam selection data, and the plurality of sub-networks respectively determine a set of overall beam initial values ​​and a set of offset beam initial values ​​according to the corresponding input beam selection data, and the search values ​​of the plurality of beams are jointly determined by the plurality of sets of overall beam initial values ​​and the plurality of sets of offset beam initial values ​​output by the plurality of sub-networks, wherein the overall beam initial value is the initial search value of the plurality of beams, and the offset beam initial value is the initial search value of the plurality of beams. The initial value of the shifted beam is used to characterize the degree of influence of the corresponding beam selection data on the search value of multiple offset beams; based on the respective search values ​​of the multiple beams, multiple candidate beams are determined from the multiple beams; the multiple candidate beams are measured to obtain measurement results of the multiple candidate beams, wherein the measurement results of the multiple candidate beams are used to characterize the communication quality of the multiple candidate beams; based on the measurement results, a target beam for communicating with the target device is selected from the multiple candidate beams, thereby achieving the purpose of being able to utilize the communication status data for describing the channel and the perception status data for describing the position and motion status of the target device, and utilizing multiple types of data for beam prediction, so that the neural network model can more accurately select the optimal beam in the current environment, thereby improving the accuracy of beam prediction, and thus solving the technical problem of not fully utilizing the available information for beam prediction, resulting in low accuracy of beam prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0017] Figure 1 A hardware structure block diagram of a computer terminal for implementing a beam determination method is shown;

[0018] Figure 2 is a schematic flow chart of a beam determination method according to an embodiment of the present invention;

[0019] Figure 3 is a schematic structural diagram of a neural network model provided according to an optional embodiment of the present invention;

[0020] Figure 4 is a schematic diagram of parallel threads provided according to an optional embodiment of the present invention;

[0021] Figure 5 is a schematic diagram of the effect provided by an optional embodiment of the present invention;

[0022] Figure 6 4 is a structural block diagram of a beam determination device provided according to an optional embodiment of the present invention. DETAILED DESCRIPTION

[0023] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

[0024] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0025] According to an embodiment of the present invention, an embodiment of a beam determination method is provided. 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 set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0026] The method embodiment provided in the first embodiment of the present application can be executed in a mobile terminal, a computer terminal or a similar computing device. Figure 1FIG1 shows a hardware structure block diagram of a computer terminal for implementing a beam determination method. Figure 1 As shown, the computer terminal 10 may include one or more (illustrated as 102a, 102b, ..., 102n in the figure) processors (the processor may include but is not limited to a microprocessor MCU or a programmable logic device FPGA and other processing devices), a memory 104 for storing data. In addition, it may also include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of the BUS bus), a network interface, a power supply and / or a camera. It will be understood by those skilled in the art that Figure 1 The structure shown is only for illustration and does not limit the structure of the above electronic device. Figure 1 More or fewer components than shown, or with Figure 1 Different configurations shown.

[0027] It should be noted that the one or more processors and / or other data processing circuits described above may generally be referred to herein as "data processing circuitry." The data processing circuitry may be embodied in whole or in part as software, hardware, firmware, or any other combination thereof. Furthermore, the data processing circuitry may be a single, independent processing module, or may be incorporated in whole or in part into any of the other components of the computer terminal 10. As described in the embodiments of the present application, the data processing circuitry serves as a processor control (e.g., selection of a variable resistor terminal path connected to an interface).

[0028] Memory 104 can be used to store software programs and modules for application software, such as the program instructions / data storage device corresponding to the beam determination method in the embodiments of the present invention. The processor executes the software programs and modules stored in memory 104 to execute various functional applications and data processing, thereby implementing the beam tracking method of the aforementioned application. Memory 104 can include high-speed random access memory (RAM) and non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, memory 104 may further include memory remotely located from the processor, which can be connected to computer terminal 10 via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0029] The display may be, for example, a touch screen liquid crystal display (LCD) that enables a user to interact with a user interface of the computer terminal 10 .

[0030] Figure 2 FIG. 1 is a flow chart of a beam determination method according to an embodiment of the present invention. Figure 2 As shown, the method includes the following steps:

[0031] Step S202: Obtain a plurality of beam selection data for beam selection fed back by the target device, wherein the plurality of beam selection data includes communication status data for describing the communication channel, and perception status data for describing the position and motion status of the target device, wherein the target device includes a terminal and a base station.

[0032] In this step, the execution entity may be a mobile communication system, and in particular, a system using Multiple Input Multiple Output (MIMO) technology. The multiple beam selection data are all data related to the target device that can be obtained by the communication system and used for beam selection. Specifically, they may include communication status data related to communication signal quality, such as reference signal signal-to-noise ratio, data channel block error rate, channel state information, constellation diagram error vector magnitude, etc. They may also include perception status data of the target device's position and motion state, such as the terminal's coordinates in three-dimensional space, the terminal's azimuth angle, and speed. The target device may be a terminal or a base station.

[0033] In step S204, a plurality of beam selection data are input into the neural network model, and the neural network model outputs the search value of each of the predetermined plurality of beams, wherein the neural network model includes a plurality of sub-networks corresponding one-to-one to the plurality of beam selection data, and the plurality of sub-networks respectively determine a set of overall beam initial values ​​and a set of offset beam initial values ​​according to the corresponding input beam selection data, and the search value of each of the plurality of beams is jointly determined by the plurality of sets of overall beam initial values ​​and the plurality of sets of offset beam initial values ​​output by the plurality of sub-networks, wherein the overall beam initial value is the initial search value of each of the plurality of beams, and the offset beam initial value is used to characterize the degree of influence of the corresponding beam selection data on the search value of the plurality of offset beams.

[0034] In this step, a neural network model can be used to predict the search value of each of the multiple beams based on the various beam selection data. The search value can be a quantitative metric used to evaluate and select beams. By using the neural network model to predict and compare the search values ​​of different beams, the beam with the highest search value can be effectively selected for communication.

[0035] To fully utilize multiple beam selection data, multiple subnetworks can be set up within the neural network model. Each subnetwork is responsible for beam prediction based on a type of beam selection data, and the output results include two groups: one group is the overall beam initial value, that is, based on the corresponding input data, multiple predetermined beams are predicted and the initial search value of each of the multiple beams is determined; the other group is the offset beam initial value, that is, based on the corresponding input data, multiple offset beams around the currently used beam are predicted and the initial search value of each of the multiple offset beams is determined. It should be noted that the motion state of the device, such as the position and speed, may cause more drastic changes in the values ​​of the multiple offset beams around the current beam. The offset beam initial value can reflect this change to improve the accuracy of beam prediction. Among them, multiple offset beams refer to beams obtained by moving multiple steps with the current beam as the center. The specific range of the step size here can be determined based on experience. For example, the farthest position that the device can move to within one operation cycle and the beam used at that position can be determined based on experience. Then, the range of the step size of the multiple beams can be determined based on the step size between the beam used by the device at the farthest position and the current beam.

[0036] The neural network model can combine a set of overall beam initial values ​​and a set of offset beam initial values ​​output by multiple sub-networks, that is, multiple sets of overall beam initial values ​​and multiple sets of offset beam initial values, to jointly determine the search values ​​of the multiple beams finally output.

[0037] Step S206 : determining a plurality of candidate beams from the plurality of beams according to the respective search values ​​of the plurality of beams.

[0038] In this step, based on the search values ​​of the multiple beams output by the neural network model, the beams with the highest values ​​can be selected as candidate beams. The candidate beams are the beams that are most likely to provide the best communication performance as predicted by the neural network model.

[0039] Step S208: measuring the multiple candidate beams to obtain measurement results of the multiple candidate beams, wherein the measurement results of the multiple candidate beams are used to characterize the communication quality of the multiple candidate beams.

[0040] In this step, actual measurements may be performed on multiple candidate beams to obtain measurement results of each of the multiple candidate beams, wherein the measurement results may include indicators such as signal-to-noise ratio, power, bit error rate, etc. for measuring the communication quality of the beams.

[0041] The available information is not fully utilized for beam prediction, resulting in low accuracy of beam prediction. In step S210, a target beam for communicating with a target device is selected from multiple candidate beams based on the measurement results.

[0042] In this step, after obtaining the actual measurement results, the optimal beam from multiple candidate beams is selected as the target beam for communication with the target device. It should be noted that in this step, the beam with the highest search value is not directly selected as the target beam for communication. Instead, considering that the prediction of the neural network model reflects the probability distribution of the target beam position in a statistical sense and has uncertainty, multiple candidate beams are selected. The beam with the best actual performance is then selected for communication.

[0043] Through the above steps, the purpose of utilizing the communication status data used to describe the channel and the perception status data used to describe the position and motion status of the target device can be achieved. By using multiple types of data for beam prediction, the neural network model can more accurately select the optimal beam in the current environment, thereby improving the accuracy of beam prediction, and thus solving the technical problem of not fully utilizing the available information for beam prediction, resulting in low accuracy of beam prediction.

[0044] As an optional embodiment, a plurality of beam selection data are input into a neural network model, and the neural network model outputs the search values ​​of a plurality of predetermined beams, including: inputting a plurality of beam selection data into a plurality of sub-networks one by one, and having the plurality of sub-networks output a set of overall beam initial values ​​and a set of offset beam initial values ​​respectively; inputting a plurality of beam selection data into a weight selection network included in the neural network model, and having the weight selection network output the overall weights and offset weights of the plurality of sub-networks; and determining the search values ​​of the plurality of beams according to the initial search values ​​of the plurality of beams and the overall weights and offset weights of the plurality of sub-networks.

[0045] Optionally, the neural network model includes a beam selection network, specifically comprising multiple sub-networks, capable of outputting search values ​​for each of the multiple beams. Each of the sub-networks can make predictions based on different types of beam selection data. Specifically, each sub-network is responsible for a specific type of state data, and can predict the overall beam initial value and the offset beam initial value based solely on its own input state data.

[0046] In addition to the beam selection network, the neural network model also includes a weight selection network, which can output the overall weights and offset weights of each of the multiple sub-networks. The overall weight is a measure of the initial value of the overall beam, and the offset weight is a measure of the initial value of the offset beam. Since each sub-network outputs a set of overall beam initial values ​​and a set of offset beam initial values, when these search values ​​are combined, the weight selection network can determine the weights of the values ​​output by the multiple sub-networks. In other words, the initial search values ​​of the multiple beams output by the multiple sub-networks can be combined based on the overall weights and offset weights of the multiple sub-networks to obtain the final search values ​​of the multiple beams.

[0047] Optionally, each sub-network can be a dual-head output structure, that is, each sub-network has two output ports, which respectively output the initial value of the overall beam, that is, the search value of each beam, and the initial value of the offset beam, that is, the search value of each offset beam. It should be noted that because each sub-network processes a corresponding beam selection data, and different types of information sources have different correlations with the problem of optimal beam selection, for example, the three-dimensional coordinates of the terminal are directly related to the angle of the beam, but have nothing to do with the offset of the beam per unit time, while the rotation speed of the terminal is independent of the angle of the beam, but is strongly related to the offset of the beam per unit time. Therefore, it is very critical for each sub-network to adopt a dual-head output structure. For each type of state information, a network with a dual-head output structure is required to respectively give the search value of each beam and the search value of each offset beam, so that all the information contained in the state information can be fully utilized. In addition, under the premise that the sub-network has a dual-head output structure, the weight selection network can assign corresponding weight values ​​to each output port of each sub-network, that is, assign overall weight and offset weight to each sub-network, and the weight of each sub-network can be the average of the weights of the two output ports.

[0048] The optional embodiments of the present invention do not limit the specific structure of the sub-network, and can be common network structures such as a fully connected network, a partially connected network, a convolutional network, a recurrent network, and an attention mechanism.

[0049] It should be noted that the offset beam is the index of the communication beam in the current time slot n [b x,n b y,n ], the beam is moved several steps in the x-axis or y-axis direction, so the output of the second output port is (2R+1) 2 The search value of the offset beam is specifically the beam [b x,n +r x b y,n +r y ],r x and r yare integers ranging from -R to R, where R is the maximum moving step, x is the x-axis of the beam array, y is the y-axis of the beam array, and the index of each beam is expressed as [b x,n b y,n ].

[0050] Since the offset beam does not correspond to all beams one by one, it is necessary to output (2R+1) 2 The search value of the offset beam is mapped to the search value of all beams, and then combined with the search value of all beams output by the first output port. Specifically, there are two mapping methods. For the search value of the offset beam included in the above (2R+1) 2 The search value of the beams within the beams is the corresponding output value. For the beams not included in the above (2R+1) 2 The search value of the beam within the beam is set to 0.

[0051] As an optional embodiment, the method further includes: among multiple sub-networks, marking the sub-networks whose sum of the overall weight and the offset weight assigned within a predetermined time period is less than a predetermined threshold as disabled, to obtain an updated neural network model.

[0052] Optionally, the overall weight and offset weight assigned to each sub-network by the weight selection network over a long period of time in the past can be summed or averaged. For sub-networks whose summed value or average weight is lower than a preset threshold, it is considered that the state information corresponding to the sub-network has a weak correlation with the beam tracking problem, and the state information is deleted, and the beam selection sub-network corresponding to the state information is shut down. In this process, the input and output dimensions of the weight selection network remain unchanged, the inputs of the deleted state information are all set to zero, and the weights of the sub-networks corresponding to the deleted state information are no longer used. Automatically identifying and deleting state information that has a weak correlation with the beam tracking problem can effectively filter out key information from a large number of information sources, thereby reducing the overhead of state information collection, simplifying the neural network model, and reducing the system's computing resource consumption.

[0053] The advantage of this optional embodiment is that it can adaptively filter state information and exclude state information that is less relevant to the optimal beam selection problem, thereby reducing the overhead of state information collection, simplifying the neural network model, and reducing the system's computing resource consumption.

[0054] In addition, due to the complex nature of the external environment, when collecting communication and perception state information, it is not always possible to search for all the state information as expected, and some state information may be missing. The sub-network corresponding to the missing state information does not output the corresponding search value, and the state information is set to zero in the input of the weight selection network, and the corresponding weight of the output sub-network is not used.

[0055] As an optional embodiment, the method also includes: when the measurement results of multiple candidate beams meet the first predetermined condition, multiple beam selection data, the search values ​​of multiple beams and the measurement results of multiple candidate beams are combined into training samples and added to the online training sample set.

[0056] Optionally, after measuring the beam and obtaining the actual measurement result, if the measurement result meets the first predetermined condition, the actual measurement result, as well as the input data (i.e., multiple beam selection data) and output data (i.e., the search value of each of the multiple beams) of the neural network model can be recorded and added to the online training sample set as training samples. In this case, when the neural network model is trained using the online training sample set, the neural network model can adjust its own parameters based on the gap between the actual measurement result and the search value output by itself, so that the output of the neural network model matches the actual situation as much as possible, thereby achieving the purpose of being able to adjust the parameters of the neural network model in a timely manner according to the beam change law during the application process, thereby achieving the technical effect of enhancing the adaptability of the beam determination method, enabling the neural network model to quickly adapt to the current environment and more accurately select the optimal beam under the current environment.

[0057] In addition, when determining online training samples, since each sub-network will output two sets of initial search values, two sets of training samples can be generated for adjustment. The labels of one set of training samples are the search values ​​that multiple beams should have based on the beam measurement results, and the labels of the other set of training samples are the search values ​​that multiple offset beams should have based on the search values ​​that multiple beams should have and the beam currently used for communication.

[0058] It should be noted that the neural network model in this step adopts an online training method in actual application, that is, during the application of the model, the parameters in the neural network model are further adjusted according to the data of the application process to adapt to the complex and changeable application environment. In the application process, the various beam selection data that can be obtained by the communication system are input into the neural network model, the network outputs the search value of all beams, and measures the measurement results of all beams, and the communication beam with the best measurement results can be selected. During the training process, the network parameters can be continuously updated based on the real-time feedback data during the application process, with the goal of narrowing the gap between the network output search value and the actual measurement results, and the updated parameters are fed back to the neural network model in the application process, so that the neural network model can provide a more reasonable search value. Through the continuous interaction and iteration between the above two processes, the neural network can gradually adapt to the current communication environment and provide more and more accurate beam search values.

[0059] As an optional embodiment, the first predetermined condition is that the beam with the highest search value among the multiple beams is not the same beam as the beam with the best communication quality represented by the measurement result.

[0060] Optionally, if the beam with the highest search value among the multiple beams is found to be the same beam with the best communication quality as indicated by the measurement results, this indicates that the neural network model's prediction results are relatively accurate, and the data does not need to be input into the training sample. Conversely, if the beam with the highest search value among the multiple beams is not the same beam with the best communication quality as indicated by the measurement results, this indicates that the neural network model's prediction results are not accurate enough, and the current data is needed to adjust the neural network model's parameters. Therefore, if the first predetermined condition is met, the corresponding data can be used as a training sample and added to the online training sample set.

[0061] As an optional embodiment, the method also includes: when the number of samples included in the online training sample set meets the second predetermined condition, using the online training sample set to perform online training on the neural network model to obtain an updated neural network model, and determining the beam used for communication based on the updated neural network model.

[0062] Optionally, since each training session requires consuming a certain amount of computing resources to adjust the model, a second condition can be pre-set to limit the frequency of training the neural network model. Only when the number of samples in the online training sample set reaches a certain threshold is the online training sample set used to train the neural network model online to obtain an updated neural network model. The samples in the online training sample set are acquired and continuously updated in real time during the application process. Furthermore, online training is only started when the number of samples reaches a certain level, ensuring that the sample set is large enough to represent the true distribution of the data, thereby making online training effective. During online training, the parameters of the neural network model are adjusted based on the new sample set to better reflect the current data characteristics. After the model is updated, the new model will be immediately used to determine the beam used for communication. This updated model will be more suitable for the current external environment than the original model.

[0063] As an optional embodiment, an online training sample set is used to perform online training on the neural network model to obtain an updated neural network model, including: using the online training sample set, with the goal of minimizing the gap between the search value of each of the multiple beams output by the neural network model and the target search value, training the multiple sub-networks and weight selection network included in the neural network model to obtain multiple trained sub-networks and trained weight selection networks, wherein the target search value is determined based on the measurement results of multiple candidate beams; and constructing an updated neural network model based on the multiple trained sub-networks and the trained weight selection network.

[0064] Optionally, when using an online training sample set to train the neural network model, the "search value of each of the multiple beams" in the training sample set is the predicted data output by the neural network model based on previously learned content, and the "target search value" is the value that should actually be assigned to the beam determined based on the measurement results after actual measurement of multiple candidate beams. Therefore, the neural network model can adjust its own parameters with the goal of minimizing the gap between the search value of each of the multiple beams and the target search value, so that the output of the neural network model matches the actual situation as much as possible.

[0065] During training, the neural network model can be trained based on the actual measurement results of multiple candidate beams. To facilitate model operation, the measurement results can be processed and converted into the benefits obtained by using the corresponding beam. Specifically, for measurement results positively correlated with the benefits, such as signal-to-noise ratio, the benefit can be defined as t′(i) = t(i) / max(t). Conversely, for measurement results negatively correlated with the benefits, such as bit error rate, the benefit can be defined as t′(i) = min(t) / t(i), where t(i) represents the i-th element in the measurement result vector t, and max(t) and min(t) represent the largest and smallest elements in the measurement result vector t, respectively. If the measurement result t(i) is 0, making the formula t′(i) = min(t) / t(i) incomprehensible, the benefit of the beam can be set to 1. The benefits of unmeasured beams are 0. After the benefits are defined, they can be used in the model training process instead of the measurement results.

[0066] First, multiple sub-networks can be trained using online training sample sets, and the search value x output by the revenue and neural network model can be used to train the sub-networks. m Construct the target search value of the mth sample in the online training sample set β is a predetermined constant satisfying 0≤β≤1, t′ m is the profit of the mth sample. After obtaining the target search value of all M samples, the minimum mean square error loss function is used All sub-networks and weight selection networks are trained. It should be noted that during training, if the valid sub-network set of the mth sample does not contain a sub-network (that is, when the sample is collected, the state information corresponding to the network cannot be obtained, or when the sample is collected, although the state information corresponding to the network is obtained, but during training, the network has been closed because its corresponding state information is weakly related to the beam tracking problem), then the loss value of the sample is Will not be used to train this subnetwork.

[0067] After the training is completed, samples with an older storage time in the online training sample set are also deleted. For example, a validity period threshold is set to delete samples that have been stored in the sample pool for longer than the threshold. In this way, newer samples are used to train the neural network in each training, thereby ensuring that the neural network can learn and adapt to the latest changes in the communication environment in real time.

[0068] As an optional embodiment, an online training sample set is used to perform online training on the neural network model to obtain an updated neural network model, including: setting parallel training threads and application threads; based on the application thread, using the neural network model to perform beam prediction; based on the training thread, using the online training sample set to perform online training on the neural network model to obtain an updated neural network model.

[0069] Optionally, during the application process, parallel training threads and application threads can be started to perform beam prediction (i.e., the application process) and model training, respectively. In the application thread, the neural network model is used to perform the beam prediction task. Specifically, the method of the embodiment of the present invention can be used to determine the search value of each of the multiple beams based on the collected beam selection data.

[0070] A training thread can be a dedicated thread for online training, which involves training the model using new data samples after the model has been deployed and begins processing data. During online training, the parameters of the neural network model are adjusted based on the new sample data. This update can be incremental, meaning that only small adjustments are made to the model during each training session to maintain model stability.

[0071] The advantage of this alternative embodiment is that it enables continuous model learning and adaptation, which is particularly important for applications that need to handle constantly changing data and environments. Through parallel processing, training and application can proceed simultaneously, thereby improving the system's responsiveness and efficiency. In addition, this approach can reduce performance degradation caused by model staleness because the model can learn from new data in real time.

[0072] After training the latest neural network model, the training thread can feed the new network parameters back to the application thread, which then uses the new network to determine beam search values. Through multiple iterations of interaction between the application and training threads, the training thread, through repeated learning from the measurement results fed back by the application thread, can gradually adjust the neural network parameters and determine beam search values ​​that align with actual measurement results. The application thread, by leveraging the new neural network parameters fed back by the training thread, can gradually accurately predict beam measurement results in advance, thereby gradually reducing the number of beams in the candidate beam set, reducing beam determination overhead, and freeing up more time-frequency resources for data transmission, thereby improving overall system throughput.

[0073] As a specific embodiment, the method provided by the present invention may include the following steps.

[0074] Determine all communication and perception status information (multiple beam selection data) available to the communication system. Any communication system contains at least two types of status information: the current beam index value and the current beam index differential sequence. The current beam index value refers to the sequence number of the current beam in the set of all optional beams. The current beam index differential sequence refers to the difference between the current beam index and the beam index at the previous moment, the difference between the beam index at the previous moment and the beam index at the moment before that, and so on, to obtain a sequence consisting of index differences at several consecutive moments.

[0075] Assume that there are P types of communication and perception state information {u1 u2…u P The above information is combined and expressed as the beam tracking state vector s = [u1…u P ].

[0076] For ease of description, an example is given here. For example, a base station antenna array has a total of 128 beams, and the average coverage angle of each beam is 7.5°. There are 16 beam angles on the x-axis and 8 beam angles on the y-axis. Then, all 128 beams can cover a range of 120° in the horizontal direction and 60° in the vertical direction. The index of each beam is represented by 0≤b x <16, 0≤b y <8. Similarly, the user's beam index is expressed as Assuming that the beam tracking period is T, at the time nT~(n+1)T, the base station and the user use beam and At (n+1)T, the base station needs to update the beam again because the user's position may have changed. The current beam index value is expressed as The current beam index difference sequence is expressed as

[0077]

[0078] It should be further explained that the types of communication and perception status information can change dynamically. For example, a communication terminal can accurately perceive 3D positioning information in an open outdoor area. However, when encountering strong interference or entering an obstructed environment, the quality of the perception signal will significantly deteriorate, making it impossible to obtain positioning information. In this case, the number of status information types will decrease by 1. When the perception signal quality recovers, the number of status information types returns to the original value.

[0079] like Figure 3As shown in the figure, a beam tracking neural network (neural network model) is constructed based on the types of state information that the system can obtain. Due to the different usage environments of base stations and users, for example, the base station antenna is fixed, while the user terminal may be handheld and its posture is always changing. Or the base station antenna array is large and contains more array elements, while the user terminal array is small and has fewer array elements. Independent beam tracking neural networks should be constructed for beam tracking of the base station and user, respectively. Each beam tracking neural network contains P sub-networks f1, ..., f P And 1 weight network w. Among them, the p-th sub-network f p The input is the p-th state information u p The first output head (output port) outputs the search value x of all beams 2p-1 , the search value of all beams of the second output head (output port) is x 2p The mapping of the sub-network from input to output is not limited to a specific structure in the present invention; it can be a fully connected network, a partially connected network, a convolutional network, a recurrent network, an attention mechanism, or other common network structures. Assuming there are N layers in total, the mapping from the input layer to the N-1th layer and from the N-1th layer to the output layer is described below.

[0080] Since the second output head gives the index of the current communication beam [b x,n b y,n ] Based on the search value of the beam after moving a number of steps in the x-axis or y-axis direction, the output of the N-1 layer of the second output head is (2R+1) 2 The search value of a beam, specifically the beam [b x,n +r x b y,n +r y ],r x and r y All are integers ranging from -R to R. From the N-1 layer to the output layer, the above (2R+1) 2 The search value of each beam is mapped to the search value of all beams. All beams can be divided into two cases. For the beams included in the above (2R+1) 2 The search value of the beam within the beam is the output value corresponding to the N-1 layer. For the beam not included in the above (2R+1) 2 The search value of the beam within the beam is set to 0.

[0081] Get the search value x1,...,x of all sub-networks 2P Finally, all search values ​​are weighted averaged to obtain the final search value x. The weight of each output head of all sub-networks is expressed as [α1α2…α 2P ], The weights are given by the weight network w (weight selection network), and the input of the weight network w is the beam tracking state vector s = [u1…u P ], the output is weight [α1α2…α 2P ]. For the weight network, the present invention does not limit its specific structure. Since the weight is based on the beam tracking state vector s = [u1…u P ] is obtained, so for different beam tracking state vectors, the weights of the sub-networks are dynamically changing. When the user is in different communication environments, each sub-network has different weights, which makes the algorithm very flexible and can adapt to different scenarios.

[0082] When a data source is temporarily unavailable, for example, when the perception state information mentioned above cannot be obtained, the corresponding beam selection subnetwork will have no input and will no longer calculate the output value. The weighted average stage will no longer consider the output of the network. Therefore, in this case, the final search value can also be expressed as in Represents the set of all valid subnetworks.

[0083] Average the weights of the two output heads of all sub-networks from the current moment to the previous K moments to obtain the average weight of each sub-network For subnetworks whose average weight is less than a preset threshold, the state information corresponding to the subnetwork is considered to be weakly relevant to the beam tracking problem. This state information is deleted and the subnetwork is shut down. During this process, the input and output dimensions of the weight network remain unchanged. In the input of the weight network, the inputs of the input nodes corresponding to the deleted state information are all reset to zero, and the weights of the subnetwork corresponding to the deleted state information are no longer used or trained.

[0084] In the specific embodiment provided by the present invention, two threads can be run in parallel, namely, a beam tracking thread (application thread) and a beam training thread (training thread), such as Figure 4As shown. In the beam tracking thread, the relevant state information that can be obtained by the communication system is input into the above-mentioned beam tracking network. The network outputs the search value of all beams, selects beams with search values ​​greater than the threshold to form a candidate beam set, measures the beams in the candidate beam set one by one, selects the communication beam with the best measurement result, and feeds the measurement results and the relevant state information input by the neural network back to the beam training thread. In the beam training thread, based on the information fed back by the beam tracking thread, with the goal of narrowing the gap between the network output search value and the actual measurement results, the network parameters are continuously updated, and the updated parameters are fed back to the beam tracking thread, so that the beam tracking thread can give a more reasonable search value. Through the continuous interaction and iteration between the above two threads, the neural network can gradually adapt to the current communication environment and give more and more accurate beam search values.

[0085] Specifically, in the beam tracking thread, the beam tracking network and the beam tracking state vector s = [u1…u P After obtaining the search value x of all beams, sort the search values ​​of all beams in descending order and select the beams whose search value is greater than the threshold x. th The beams of For the candidate beam set The beams in are measured one by one to obtain the measurement result t. The communication quality of the beam can be measured by indicators such as signal-to-interference-noise ratio, power, bit error rate, and spectrum efficiency, which are not limited in the present invention. The beam with the best communication quality is selected based on the measurement results as the communication beam between the base station and the user. The threshold x th It can be determined by multiplying the average value of all beam search values ​​by a constant factor, or by multiplying the maximum value of all beam search values ​​by a constant factor. After determining the communication beam, the beam tracking state vector s = [u1…u P ], candidate beam set Valid subnetwork set The measurement results t of all beams are fed back to the beam training thread, where t(i) represents the i-th element of vector t, i.e., the candidate beam set. The measurement result of the i-th beam in . The effective sub-network set It refers to the subnetwork corresponding to all valid state information obtained when collecting the sample, and the subnetwork is not closed because its corresponding state information is weakly related to the beam tracking problem.

[0086] In the beam training thread, the information fed back by the above beam tracking thread is first saved as a sample. The present invention uses reinforcement learning to train the neural network. The advantage of reinforcement learning is that there is no need to pre-train the neural network offline, and it can be tracked and trained online synchronously, which has good adaptability to situations where the communication environment changes dynamically. The main principle of reinforcement learning for training neural networks is to associate the search value of each beam output by the neural network with the "benefit" that each beam can obtain. In the present invention, the benefit is characterized by the measurement result t. For example, the benefit of a beam with a high signal-to-noise ratio should be greater than the benefit of a beam with a low signal-to-noise ratio, or the benefit of a beam with a low bit error rate should be greater than the benefit of a beam with a high bit error rate. When the beam training thread receives a new sample, it is necessary to determine whether the beam with the greatest search value in the sample (that is, the first beam in the candidate beam set) is the same as the beam with the best measurement result. If they are the same, it means that for the beam tracking state vector s of the sample, the neural network can give a beam search value that is consistent with the actual measurement result, so the sample is no longer saved; if they are different, it means that for the beam tracking state vector s of the sample, the neural network cannot correctly give a beam search value that is consistent with the actual measurement result, so the sample is saved to the sample pool.

[0087] When the number of samples in the sample pool exceeds a certain threshold, a number of samples are randomly drawn from the sample pool to train the beam-tracking neural network. During training, the benefit of each beam in the sample must be calculated. For measurement results positively correlated with benefit, such as signal-to-noise ratio, the benefit is defined as t′(i) = t(i) / max(t). Conversely, for measurement results negatively correlated with benefit, such as bit error rate, the benefit is defined as t′(i) = min(t) / t(i), where t(i) represents the i-th element of vector t, and max(t) and min(t) represent the largest and smallest elements in vector t, respectively. If the measurement result t(i) is 0, making the formula t′(i) = min(t) / t(i) impossible to calculate, the benefit of that beam can be simply set to 1. The benefit of unmeasured beams is 0.

[0088] Assume that M samples are drawn from the sample pool During training, for the mth sample, firstly according to the measurement result t m Get the benefit t′ of each beam corresponding to the sample m , then the sample state s m Input the neural network to get the value x of the neural network output m And construct the mth sample target search value β is a constant that satisfies 0≤β≤1. After obtaining the target search value of all M samples, the minimum mean square error loss function is used All sub-networks and weight selection networks are trained in sequence. It should be noted that during training, if the effective sub-network set of the mth sample If a sub-network is not included, the loss value of the sample is Will not be used to train this subnetwork.

[0089] After training the latest neural network, the new network parameters are fed back to the beam tracking thread, which uses the new network to determine beam search values. Through multiple iterations of interaction between the beam tracking thread and the beam training thread, the beam training thread repeatedly learns from the measurement results fed back by the beam tracking thread, gradually adjusting the neural network parameters to determine beam search values ​​that align with actual measurement results. The beam tracking thread, in turn, utilizes the new neural network parameters fed back by the beam training thread to accurately predict beam measurement results in advance, thereby gradually reducing the number of beams in the candidate beam set, reducing beam tracking overhead, and utilizing more time-frequency resources for data transmission, thereby improving overall system throughput.

[0090] After the training is completed, samples with an older storage time in the sample pool are deleted. For example, a validity period threshold is set to delete samples that have been stored in the sample pool for longer than the threshold. In this way, newer samples are used to train the neural network each time, ensuring that the neural network can learn and adapt to the latest changes in the communication environment in real time.

[0091] In one embodiment, a method for applying the present invention to an actual system is given. Assume a UAV communication system, where the base station side is located at a ground fixed station, with a total of 128 beams, and the UAV is equipped with a user terminal, with a total of 64 beams. First, it is necessary to determine the type of data source that the system can provide. Assume that in the above communication system, the types of status information that can be provided include the UAV GPS coordinates (x, y, z axes), attitude angles (pitch angle, roll angle, azimuth angle), and the power strength of the current received signal. Based on this, 4 subnetworks can be set for the base station, and the current beam index of the base station, the UAV GPS coordinates, the current beam index differential sequence of the base station, and the power strength of the current received signal are input respectively. 5 subnetworks can be set for the user, and the current beam index of the user, the UAV GPS coordinates, the UAV attitude angle, the current beam index differential sequence of the user, and the power strength of the current received signal are input respectively.

[0092] Initially, the weights of the base station and user beam-tracking neural networks are randomly initialized, resulting in a virtually random set of candidate beams. However, as measurement and training progress, the neural network gradually learns which beams are more likely to have high received power and which are less likely to have high received power, thereby gradually converging to a relatively stable tracking state. If the communication environment changes and the original rules no longer apply, the neural network can quickly detect this through measurement results and adjust its weights based on the new measurement data, rapidly adapting to the changing environment. Figure 5 The performance of the proposed algorithm was compared with that of a comparable algorithm in the related art. The horizontal axis represents time, and the vertical axis represents the probability that the tracking beam generated by the neural network is identical to the true optimal beam, referred to as the beam alignment rate. In simulations, the two algorithms were controlled to search the same number of beams. The results show that, in the simulation environment, the proposed algorithm's beam alignment rate remains stable at around 90%, with minimal fluctuations. The comparative method, which only searches a few beams around the current beam, cannot handle complex scenarios such as beam hopping. The beam alignment rate fluctuates dramatically, making it impossible to maintain a stable communication link.

[0093] It should be noted that for the aforementioned method embodiments, for simplicity of description, they are all expressed as a series of action combinations. However, those skilled in the art should be aware that the present invention is not limited by the order of the actions described, because according to the present invention, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in this specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present invention.

[0094] Through the description of the above embodiments, those skilled in the art can clearly understand that the beam determination method according to the above embodiments can be implemented by software plus the necessary general hardware platform. Of course, it can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the existing technology, can be embodied in the form of a software product. The computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for enabling a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods described in various embodiments of the present invention.

[0095] According to an embodiment of the present invention, a beam determination device for implementing the above beam determination method is also provided. Figure 6 FIG. 1 is a structural block diagram of a beam determination device according to an embodiment of the present invention. Figure 6As shown, the beam determination device includes: an acquisition module 61, a prediction module 62, a determination module 63, a measurement module 64 and a selection module 65. The beam determination device is described below.

[0096] An acquisition module 61 is used to obtain a variety of beam selection data for beam selection fed back by a target device, wherein the various beam selection data include communication status data for describing a communication channel, and perception status data for describing a position and motion status of a target device, wherein the target device includes a terminal and a base station.

[0097] The prediction module 62 is connected to the acquisition module 61, and is used to input multiple beam selection data into the neural network model, and the neural network model outputs the search value of each of the predetermined multiple beams, wherein the neural network model includes multiple sub-networks corresponding one-to-one to the multiple beam selection data, and the multiple sub-networks respectively determine a set of overall beam initial values ​​and a set of offset beam initial values ​​according to the corresponding input beam selection data. The search value of each of the multiple beams is jointly determined by the multiple sets of overall beam initial values ​​and the multiple sets of offset beam initial values ​​output by the multiple sub-networks, wherein the overall beam initial value is the initial search value of each of the multiple beams, and the offset beam initial value is used to characterize the degree of influence of the corresponding beam selection data on the search value of the multiple offset beams.

[0098] The determination module 63 is connected to the prediction module 62 and is configured to determine a plurality of candidate beams from the plurality of beams according to the respective search values ​​of the plurality of beams.

[0099] The measurement module 64 is connected to the determination module 63 and is used to measure the multiple candidate beams to obtain measurement results of the multiple candidate beams, wherein the measurement results of the multiple candidate beams are used to characterize the communication quality of the multiple candidate beams.

[0100] The selection module 65 is connected to the measurement module 64 and is configured to select a target beam for communicating with the target device from a plurality of candidate beams according to the measurement result.

[0101] It should be noted that the acquisition module 61, prediction module 62, determination module 63, measurement module 64, and selection module 65 described above correspond to steps S202 to S210 in the embodiment. The examples and application scenarios implemented by these modules and corresponding steps are the same, but are not limited to those disclosed in the above embodiment. It should be noted that the above modules, as part of the device, can be run in the computer terminal 10 provided in the embodiment.

[0102] An embodiment of the present invention may provide a computer device. Optionally, in this embodiment, the computer device may be located in at least one of a plurality of network devices in a computer network. The computer device includes a memory and a processor.

[0103] The memory can be used to store software programs and modules, such as the program instructions / modules corresponding to the beam determination method and apparatus in the embodiments of the present invention. The processor executes the software programs and modules stored in the memory to execute various functional applications and data processing, thereby implementing the above-mentioned beam determination method. The memory may include high-speed random access memory and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory may further include memory remotely located relative to the processor, and these remote memories may be connected to the computer terminal via 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.

[0104] The processor can call the information and application stored in the memory through the transmission device to perform the following steps: obtain a variety of beam selection data for beam selection fed back by the target device, wherein the multiple beam selection data include communication status data for describing the communication channel, and perception status data for describing the position and motion status of the target device, wherein the target device includes a terminal and a base station; input the multiple beam selection data into the neural network model, and the neural network model outputs the search value of each of the predetermined multiple beams, wherein the neural network model includes multiple sub-networks corresponding to the multiple beam selection data one by one, and the multiple sub-networks respectively determine a set of overall beam initial values ​​according to the corresponding input beam selection data and a group of offset beam initial values, the search values ​​of the multiple beams are jointly determined by the multiple groups of overall beam initial values ​​and the multiple groups of offset beam initial values ​​output by the multiple sub-networks, wherein the overall beam initial value is the initial search value of the multiple beams, and the offset beam initial value is used to characterize the degree of influence of the corresponding beam selection data on the search values ​​of the multiple offset beams; according to the respective search values ​​of the multiple beams, multiple candidate beams are determined from the multiple beams; the multiple candidate beams are measured to obtain measurement results of the multiple candidate beams, wherein the measurement results of the multiple candidate beams are used to characterize the communication quality of the multiple candidate beams; according to the measurement results, a target beam for communicating with the target device is selected from the multiple candidate beams.

[0105] According to an embodiment of the present invention, a beam determination scheme is provided. A plurality of beam selection data for beam selection fed back by a target device is obtained, wherein the plurality of beam selection data include communication status data for describing a communication channel and perception status data for describing the position and motion status of the target device, wherein the target device includes a terminal and a base station; the plurality of beam selection data are input into a neural network model, and the neural network model outputs a predetermined search value of each of the plurality of beams, wherein the neural network model includes a plurality of sub-networks corresponding one-to-one to the plurality of beam selection data, and the plurality of sub-networks respectively determine a set of overall beam initial values ​​and a set of offset beam initial values ​​according to the corresponding input beam selection data, and the search values ​​of the plurality of beams are jointly determined by the plurality of sets of overall beam initial values ​​and the plurality of sets of offset beam initial values ​​output by the plurality of sub-networks, wherein the overall beam initial value is the initial search value of the plurality of beams, and the offset beam initial value is the initial search value of the plurality of beams. The value is used to characterize the degree of influence of the corresponding beam selection data on the search value of multiple offset beams; according to the respective search values ​​of the multiple beams, multiple candidate beams are determined in the multiple beams; the multiple candidate beams are measured to obtain measurement results of the multiple candidate beams, wherein the measurement results of the multiple candidate beams are used to characterize the communication quality of the multiple candidate beams; according to the measurement results, a target beam for communicating with the target device is selected from the multiple candidate beams, thereby achieving the purpose of being able to utilize the communication status data for describing the channel and the perception status data for describing the position and motion status of the target device, making full use of all available data for beam prediction, so that the neural network model can more accurately select the optimal beam in the current environment, improving the accuracy of beam prediction, and thus solving the technical problem of not fully utilizing the available information for beam prediction, resulting in low accuracy of beam prediction.

[0106] A person skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing the hardware related to the terminal device through a program, and the program can be stored in a non-volatile storage medium, which may include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0107] An embodiment of the present invention further provides a non-volatile storage medium. Optionally, in this embodiment, the non-volatile storage medium may be used to store program codes executed by the beam determination method provided in the above embodiment.

[0108] Optionally, in this embodiment, the non-volatile storage medium may be located in any computer terminal in a computer terminal group in a computer network, or in any mobile terminal in a mobile terminal group.

[0109] Optionally, in this embodiment, the non-volatile storage medium is configured to store program codes for executing the following steps: obtaining a plurality of beam selection data for beam selection fed back by a target device, wherein the plurality of beam selection data includes communication status data for describing a communication channel, and perception status data for describing a position and motion status of a target device, wherein the target device includes a terminal and a base station; inputting the plurality of beam selection data into a neural network model, and the neural network model outputs search values ​​of respective predetermined plurality of beams, wherein the neural network model includes a plurality of sub-networks corresponding one-to-one to the plurality of beam selection data, and the plurality of sub-networks respectively determine a set of overall beam initialization values ​​according to the corresponding input beam selection data. value and a group of offset beam initial values, the search values ​​of each of the multiple beams are jointly determined by multiple groups of overall beam initial values ​​and multiple groups of offset beam initial values ​​output by multiple sub-networks, wherein the overall beam initial value is the initial search value of each of the multiple beams, and the offset beam initial value is used to characterize the degree of influence of the corresponding beam selection data on the search values ​​of the multiple offset beams; according to the respective search values ​​of the multiple beams, multiple candidate beams are determined from the multiple beams; the multiple candidate beams are measured to obtain measurement results of the multiple candidate beams, wherein the measurement results of the multiple candidate beams are used to characterize the communication quality of the multiple candidate beams; according to the measurement results, a target beam for communicating with the target device is selected from the multiple candidate beams.

[0110] An embodiment of the present invention further provides a computer program product, including a computer program, which implements the steps of the beam determination method in each embodiment of the present application when the computer program is executed by a processor.

[0111] The serial numbers of the above embodiments of the present invention are for description only and do not represent the advantages or disadvantages of the embodiments.

[0112] In the above embodiments of the present invention, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0113] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only exemplary. For example, the division of the units can be a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0114] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0115] In addition, the functional units in the various embodiments of the present invention may be integrated into a single processing unit, each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0116] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a non-volatile storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk, and other media that can store program code.

[0117] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. A beam determination method, characterized in that: include: Acquire multiple beam selection data for beam selection fed back by a target device, wherein the multiple beam selection data include communication state data for describing a communication channel and perception state data for describing a position and a motion state of the target device, wherein the target device includes a terminal and a base station; Input the multiple beam selection data into a neural network model, and the neural network model outputs the search values ​​of the respective predetermined multiple beams, wherein the neural network model includes multiple sub-networks corresponding to the multiple beam selection data one by one, and the multiple sub-networks respectively determine a group of overall beam initial values ​​and a group of offset beam initial values ​​according to the corresponding input beam selection data, and the search values ​​of the respective multiple beams are jointly determined by the multiple groups of overall beam initial values ​​and the multiple groups of offset beam initial values ​​output by the multiple sub-networks, wherein the overall beam initial value is the initial search value of the respective multiple beams, and the offset beam initial value is used to characterize the influence of the corresponding beam selection data on the search values ​​of the multiple offset beams; Determining a plurality of candidate beams from the plurality of beams according to respective search values ​​of the plurality of beams; Measuring the multiple candidate beams to obtain measurement results of the multiple candidate beams, wherein the measurement results of the multiple candidate beams are used to characterize the communication qualities of the multiple candidate beams; A target beam for communicating with the target device is selected from the multiple candidate beams according to the measurement result.

2. The method according to claim 1, characterized in that The inputting of the plurality of beam selection data into a neural network model, and the neural network model outputting the search value of each of the plurality of predetermined beams, comprises: Inputting the plurality of beam selection data into the plurality of sub-networks one by one, and the plurality of sub-networks respectively outputting a corresponding set of overall beam initial values ​​and a corresponding set of offset beam initial values; Inputting the plurality of beam selection data into a weight selection network included in the neural network model, and having the weight selection network output the overall weights and offset weights of the plurality of sub-networks; The search values ​​of each of the plurality of beams are determined according to the initial search values ​​of each of the plurality of beams and the overall weights and offset weights of each of the plurality of sub-networks.

3. The method according to claim 1, characterized in that Also includes: When the measurement results of the multiple candidate beams meet the first predetermined condition, the multiple beam selection data, the search values ​​of the multiple beams and the measurement results of the multiple candidate beams are combined into training samples and added into an online training sample set.

4. The method according to claim 3, characterized in that: Also includes: When the number of samples included in the online training sample set meets the second predetermined condition, the neural network model is trained online using the online training sample set to obtain an updated neural network model, and the beam used for communication is determined based on the updated neural network model.

5. The method according to claim 4, characterized in that The online training sample set is used to perform online training on the neural network model to obtain an updated neural network model, including: Using the online training sample set, with the goal of minimizing the gap between the search value of each of the multiple beams output by the neural network model and the target search value, the multiple sub-networks and the weight selection network included in the neural network model are trained to obtain the trained multiple sub-networks and the trained weight selection network, wherein the target search value is determined according to the measurement results of the multiple candidate beams; The updated neural network model is constructed based on the trained multiple sub-networks and the trained weight selection network.

6. The method according to claim 3, characterized in that Also includes: Among the multiple sub-networks, the sub-networks for which the sum of the overall weight and the offset weight assigned within a predetermined time period is less than a predetermined threshold are marked as disabled, thereby obtaining an updated neural network model.

7. The method according to any one of claims 1 to 6, characterized in that: The first predetermined condition is that the beam with the highest search value among the multiple beams is not the same beam as the beam with the best communication quality represented by the measurement result.

8. A non-volatile storage medium, characterized in that: The non-volatile storage medium includes a stored program, wherein when the program is executed, the device where the non-volatile storage medium is located is controlled to execute the beam determination method according to any one of claims 1 to 7.

9. A computer device, characterized in that: include: Memory and processor, The memory stores a computer program; The processor is used to execute the computer program stored in the memory, and when the computer program is run, the processor executes the beam determination method according to any one of claims 1 to 7.

10. A computer program product comprising computer instructions, characterized in that The computer instructions are executed by a processor to perform the beam determination method according to any one of claims 1 to 7.

Images