Beam management method, terminal equipment and network equipment

Abstract

The invention provides a beam management method, terminal equipment and network equipment. The method comprises the steps that the terminal equipment determines the state of a first model; wherein the first model is used for carrying out beam prediction on a target beam set, and the state of the first model comprises one or more of an activated state, a non-activated state and a monitoring state; wherein the monitoring state is used for determining a monitoring result of the first model, and the monitoring result is used for indicating the accuracy of a prediction result of the first model. In the embodiment of the invention, the first model can be managed based on the state of the first model, for example, the accuracy of the prediction result of the first model is monitored in the monitoring state, so that the success rate of communication between the terminal equipment and the network equipment is improved. The problem that in a traditional scheme, the wave beams used for communication between the terminal equipment and the network equipment are determined by using the prediction result output by the first model all the time, and when the prediction result of the first model is inaccurate, communication between the terminal equipment and the network equipment cannot be achieved is avoided.

Description

Method, terminal device and network device for beam management Technical Field

[0001] The present application relates to the field of communication technology, and more specifically, to a method, terminal device, and network device for beam management. Background Art

[0002] At present, although the beam management method based on artificial intelligence (AI) model has been discussed and studied, how to monitor and manage the operating status of the AI ​​model has not been discussed. It is possible that the target beam in the prediction result output by the AI ​​model is not the truly optimal choice, but the terminal device and network equipment are still using the prediction result. When the link quality corresponding to the target beam is too poor, communication based on the target beam may be impossible.

[0003] Summary of the Invention

[0004] The present application provides a method, terminal device, and network device for beam management. The following describes various aspects of the present application.

[0005] In a first aspect, a method for beam management is provided, comprising: a terminal device determining a state of a first model; wherein the first model is used to perform beam prediction for a target beam set, and the state of the first model includes one or more of an activated state, an inactivated state, and a monitoring state; wherein the monitoring state is used to determine a monitoring result of the first model, and the monitoring result is used to indicate the accuracy of the prediction result of the first model.

[0006] In a second aspect, a method for beam management is provided, comprising: a network device determining a state of a first model; wherein the first model is used to perform beam prediction for a target beam set, and the state of the first model includes one or more of an activated state, an inactivated state, and a monitoring state; wherein the monitoring state is used to determine a monitoring result of the first model, and the monitoring result is used to indicate the accuracy of the prediction result of the first model.

[0007] In a third aspect, a terminal device is provided, including: a processing unit for determining the state of a first model; wherein the first model is used to perform beam prediction for a target beam set, and the state of the first model includes one or more of an activated state, an inactivated state, and a monitoring state; wherein the monitoring state is used to determine the monitoring result of the first model, and the monitoring result is used to indicate the accuracy of the prediction result of the first model.

[0008] In a fourth aspect, a network device is provided, including: a processing unit for determining the state of a first model; wherein the first model is used to perform beam prediction for a target beam set, and the state of the first model includes one or more of an activated state, an inactivated state, and a monitoring state; wherein the monitoring state is used to determine the monitoring result of the first model, and the monitoring result is used to indicate the accuracy of the prediction result of the first model.

[0009] In a fifth aspect, a terminal device is provided, comprising a processor, a memory, and a communication interface, wherein the memory is used to store one or more computer programs, and the processor is used to call the computer program in the memory so that the terminal device executes part or all of the steps in the method of the first aspect.

[0010] In a sixth aspect, a network device is provided, comprising a processor, a memory, and a transceiver, wherein the memory is used to store one or more computer programs, and the processor is used to call the computer program in the memory so that the network device executes part or all of the steps in the method of the second aspect.

[0011] In a seventh aspect, an embodiment of the present application provides a communication system, which includes the above-mentioned terminal device and / or network device. In another possible design, the system may also include other devices that interact with the terminal device or network device in the solution provided in the embodiment of the present application.

[0012] In an eighth aspect, an embodiment of the present application provides a computer-readable storage medium, which stores a computer program, and the computer program enables a communication device (for example, a terminal device or a network device) to execute part or all of the steps in the methods of the above aspects.

[0013] In a ninth aspect, embodiments of the present application provide a computer program product, wherein the computer program product includes a non-transitory computer-readable storage medium storing a computer program, wherein the computer program is operable to cause a communication device (e.g., a terminal device or a network device) to perform some or all of the steps of the methods described in each of the above aspects. In some implementations, the computer program product can be a software installation package.

[0014] In the tenth aspect, an embodiment of the present application provides a chip, which includes a memory and a processor. The processor can call and run a computer program from the memory to implement some or all of the steps described in the methods of the above aspects.

[0015] In an embodiment of the present application, the first model can be managed based on its state. For example, in a monitoring state, the accuracy of the prediction results of the first model can be monitored, which helps improve the success rate of communication between the terminal device and the network device. This avoids the traditional solution of always using the prediction results output by the first model to determine the beam used for communication between the terminal device and the network device. When the prediction results of the first model are inaccurate, it will cause the terminal device and the network device to be unable to communicate. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] FIG1 is a wireless communication system 100 used in an embodiment of the present application.

[0017] FIG2 is a schematic diagram of a multi-beam system applicable to an embodiment of the present application.

[0018] FIG3 is a schematic diagram of a multi-beam system applicable to another embodiment of the present application.

[0019] FIG4 is a schematic diagram of a basic process of downlink beam selection applicable to an embodiment of the present application.

[0020] FIG5 is a schematic diagram of a basic process of downlink beam selection applicable to another embodiment of the present application.

[0021] FIG6 is a schematic diagram of a basic process of downlink beam selection applicable to another embodiment of the present application.

[0022] FIG7 is a schematic diagram of a neuron to which embodiments of the present application are applicable.

[0023] FIG8 is a schematic diagram of a neural network applicable to an embodiment of the present application.

[0024] Figure 9 is a schematic diagram of a convolutional neural network applicable to an embodiment of the present application.

[0025] FIG10 is a schematic diagram of a recurrent neural network applicable to an embodiment of the present application.

[0026] FIG11 is a schematic diagram of a long short-term memory model applicable to an embodiment of the present application.

[0027] FIG12 is a schematic flowchart of a method for beam management according to an embodiment of the present application.

[0028] FIG13 is a schematic diagram of periodically entering a monitoring state according to an embodiment of the present application.

[0029] FIG14 is a schematic diagram of periodically entering a monitoring state according to another embodiment of the present application.

[0030] FIG15 is a schematic diagram of periodically entering a monitoring state according to another embodiment of the present application.

[0031] FIG16 is a schematic diagram of non-periodic entry into a monitoring state according to an embodiment of the present application.

[0032] FIG17 is a schematic diagram of non-periodic entry into a monitoring state according to another embodiment of the present application.

[0033] FIG18 is a schematic diagram of non-periodic entry into a monitoring state according to another embodiment of the present application.

[0034] FIG19 is a schematic diagram of non-periodic entry into a monitoring state according to another embodiment of the present application.

[0035] Figure 20 is a schematic diagram of a terminal device according to an embodiment of the present application.

[0036] Figure 21 is a schematic diagram of a network device according to an embodiment of the present application.

[0037] Figure 22 is a schematic structural diagram of a communication device according to an embodiment of the present application. DETAILED DESCRIPTION

[0038] The technical solutions of the present application will be described below with reference to the accompanying drawings. For ease of understanding, the following first introduces the communication system applicable to the embodiments of the present application, as well as the terminology and communication process involved, with reference to Figures 1 to 11.

[0039] Figure 1 illustrates a wireless communication system 100 used in an embodiment of the present application. The wireless communication system 100 may include a network device 110 and a terminal device 120. The network device 110 may be a device that communicates with the terminal device 120. The network device 110 may provide communication coverage for a specific geographic area and may communicate with the terminal device 120 within the coverage area.

[0040] FIG1 exemplarily shows a network device and two terminals. Optionally, the wireless communication system 100 may include multiple network devices and each network device may include other numbers of terminal devices within its coverage area, which is not limited in the embodiments of the present application.

[0041] Optionally, the wireless communication system 100 may further include other network entities such as a network controller and a mobility management entity, which is not limited in the embodiment of the present application.

[0042] It should be understood that the technical solutions of the embodiments of the present application can be applied to various communication systems, such as: fifth generation (5G) system or new radio (NR), long term evolution (LTE) system, LTE frequency division duplex (FDD) system, LTE time division duplex (TDD), etc. The technical solutions provided in this application can also be applied to future communication systems, such as the sixth generation mobile communication system, satellite communication system, etc.

[0043] The terminal device in the embodiments of the present application may also be referred to as user equipment (UE), access terminal, user unit, user station, mobile station, mobile station (MS), mobile terminal (MT), remote station, remote terminal, mobile device, user terminal, terminal, wireless communication device, user agent or user device. The terminal device in the embodiments of the present application may refer to a device that provides voice and / or data connectivity to a user and can be used to connect people, objects and machines, such as a handheld device with wireless connection function, a vehicle-mounted device, etc. The terminal device in the embodiments of the present application can be a mobile phone, a tablet computer, a laptop computer, a PDA, a mobile internet device (MID), a wearable device, a virtual reality (VR) device, an augmented reality (AR) device, a wireless terminal in industrial control, a wireless terminal in self-driving, a wireless terminal in remote medical surgery, a wireless terminal in a smart grid, a wireless terminal in transportation safety, a wireless terminal in a smart city, a wireless terminal in a smart home, etc. Optionally, the UE can be used to act as a base station. For example, the UE can act as a scheduling entity that provides sidelink signals between UEs in V2X or D2D, etc. For example, a cellular phone and a car communicate with each other using sidelink signals. The cellular phone and smart home devices communicate without relaying the communication signal through the base station.

[0044] The network device in the embodiments of the present application may be a device for communicating with a terminal device, and may also be referred to as an access network device or a radio access network device. For example, the network device may be a base station. The network device in the embodiments of the present application may refer to a radio access network (RAN) node (or device) that connects a terminal device to a wireless network. A base station can broadly cover various names as follows, or be replaced with the following names, such as: NodeB, evolved NodeB (eNB), next generation NodeB (gNB), relay station, access point, transmission point (TRP), transmission point (TP), master station MeNB, secondary station SeNB, multi-standard radio (MSR) node, home base station, network controller, access node, wireless node, access point (AP), transmission node, transceiver node, baseband unit (BBU), remote radio unit (RRU), active antenna unit (AAU), remote radio head (RRH), central unit (CU), distributed unit (DU), positioning node, etc. A base station can be a macro base station, a micro base station, a relay node, a donor node or the like, or a combination thereof. A base station can also refer to a communication module, a modem or a chip used to be set in the aforementioned device or apparatus. The base station can also be a mobile switching center and a device that performs base station functions in device-to-device D2D, vehicle-to-everything (V2X), and machine-to-machine (M2M) communications, a network-side device in a 6G network, or a device that performs base station functions in future communication systems. The base station can support networks with the same or different access technologies. The embodiments of this application do not limit the specific technology and specific device form used by the network equipment.

[0045] Base stations can be fixed or mobile. For example, a helicopter or drone can be configured to act as a mobile base station, and one or more cells can move based on the location of the mobile base station. In other examples, a helicopter or drone can be configured to act as a device that communicates with another base station.

[0046] In some deployments, the network device in the embodiments of the present application may refer to a CU or a DU, or the network device includes a CU and a DU. The gNB may also include an AAU.

[0047] The network equipment and terminal devices can be deployed on land, including indoors or outdoors, handheld or vehicle-mounted; they can also be deployed on water; they can also be deployed in the air on aircraft, balloons, and satellites. The embodiments of this application do not limit the scenarios in which the network equipment and terminal devices are located.

[0048] It should be understood that all or part of the functions of the communication device in this application can also be implemented through software functions running on hardware, or through virtualization functions instantiated on a platform (such as a cloud platform).

[0049] Multi-beam system

[0050] Communication systems (e.g., NR) are designed to provide wide-bandwidth communications in high-frequency bands (e.g., bands above 6 GHz). As the operating frequency increases, path loss during transmission increases, impacting the coverage capabilities of high-frequency systems. Therefore, to effectively ensure high-frequency coverage, an effective technical solution is to use massive multiple-in, multiple-out (MIMO) antenna arrays (MMIMO). This creates shaped beams with greater gain, overcomes propagation loss, and ensures the coverage of the communication system.

[0051] Currently, the most common large-scale antenna array is the millimeter-wave antenna array. Since the wavelength emitted by the millimeter-wave antenna array is shorter, the spacing between antenna elements of the antenna array can be shorter and the aperture of the antenna array can be smaller, so that more physical antenna elements can be integrated into a two-dimensional antenna array of limited size.

[0052] In addition, due to the limited size of the millimeter-wave antenna array, digital beamforming cannot be used due to factors such as hardware complexity, cost overhead, and power consumption. Instead, analog beamforming is usually used. While enhancing network coverage, it can also reduce the complexity of device implementation.

[0053] To facilitate understanding of the multi-beam system, the following text introduces the communication process based on beam communication by taking the scenario of communication between a network device and a terminal as an example with reference to Figures 2 and 3.

[0054] Referring to Figure 2 , in a traditional communication system (e.g., an LTE communication system), a relatively wide beam 210 is typically used to cover an entire cell (or "sector"). Thus, at each moment, terminals within the cell (e.g., terminals 211-215) can communicate with network devices via this relatively wide beam, for example, to obtain transmission resources allocated by the network devices.

[0055] Referring to Figure 3, in newer communication systems (e.g., NR), a multi-beam system 310 can be used to cover the entire cell. That is, each beam in the multi-beam system (e.g., beams 311 to 314) covers a smaller range in the cell, and beam sweeping is used to achieve the effect of multiple beams covering the entire cell.

[0056] During beam scanning, different beams are used at different times to cover different areas within the cell. For example, at time 1, the communication system can use beam 311 to cover the area where terminal 321 is located. At time 2, the communication system can use beam 312 to cover the area where terminal 322 is located. At time 3, the communication system can use beam 313 to cover the areas where terminals 323 and 324 are located. At time 4, the communication system can use beam 314 to cover the area where terminal 325 is located.

[0057] Multi-beam systems use narrower beams, allowing for more concentrated transmission energy and thus greater coverage. However, precisely because the beams are narrow, each beam can only cover a portion of a cell. Therefore, multi-beam systems can be understood as "trading time for space."

[0058] Generally, the beam used by the transmitter to transmit signals is called a “transmit beam,” and the beam used by the receiver to receive signals is called a “receive beam.”

[0059] In some cases, the transmit beam may also be referred to as a spatial domain transmission filter, and accordingly, the receive beam may also be referred to as a spatial domain reception filter. In other cases, the transmit beam may also be referred to as a spatial domain transmission parameter, and accordingly, the receive beam may also be referred to as a spatial domain reception parameter. For ease of understanding, the embodiments of the present application are mainly described using beams as an example.

[0060] In the communication scenario between a network device and a terminal, if the network device and the terminal support multi-beam transmission, then before the network device and the terminal communicate, both the network device and the terminal need to select appropriate transmit beams and receive beams through a beam management process (for example, which may include beam selection, beam measurement, measurement reporting, etc.). For example, in the process of transmit beam selection, the network device can use different transmit beams to send multiple reference signals in turn, and the resources corresponding to the multiple reference signals are different. Correspondingly, the terminal device also uses multiple receive beams to receive the above-mentioned multiple reference signals respectively, and measures the detected reference signals to obtain beam measurement results (also called "measurement results"). Then, the terminal device selects some reference signals from the multiple reference signals detected, and feeds back the resource identifiers of the above-mentioned part of the reference signals and their corresponding beam measurement results to the network device, so that the network device can select an appropriate transmit beam as the transmit beam for subsequent communication with the terminal.

[0061] Typically, after a network device selects a suitable transmit beam, the terminal device must select a receive beam that matches the transmit beam to communicate with the network device. For ease of understanding, the following describes the beam management process using downlink beam management as an example. Generally, beam management can be divided into beam selection, beam measurement, and measurement reporting.

[0062] Beam selection

[0063] For ease of understanding, the following text uses Figures 4 to 6 as an example to explain the basic process of downlink beam selection. The process of matching transmit and receive beams in downlink transmission can be roughly divided into three main processes, designated P1, P2, and P3. P1 is the coarse matching of downlink transmit and receive beams. P2 is the fine adjustment of the downlink transmit beam on the network side. P3 is the fine adjustment of the downlink receive beam on the terminal side.

[0064] As shown in Figure 4, it is assumed that the network device has four downlink transmit beams: transmit beam 0 to transmit beam 3, and the terminal device has four downlink receive beams: receive beam A to receive beam D. During initial access, coarse pairing can be achieved through the random access process. That is, after completing the initial access, a beam pairing with relatively good link quality can be established between the network device and the terminal device to support subsequent data transmission. At this time, if both the transmit beam and the receive beam are narrow, it will take a long time to complete the mutual alignment, which will introduce a large delay to the system. Therefore, in order to quickly complete the coarse pairing between the beams, the transmit beam and the receive beam may be relatively wide. The resulting beam pairing can achieve better performance, but may not be the optimal pairing.

[0065] On the basis of the P1 coarse pairing, fine adjustments can be made to the transmit beam and receive beam (corresponding to the P2 and P3 processes respectively), and finer beams can be used to further improve transmission performance. Referring to Figure 5, through the P1 process, coarse pairing is completed between the downlink transmit beam 2 and the downlink receive A. Then, the transmit beam 2 can be finely adjusted. Accordingly, the network device can send three narrower beams: beam 2-1, beam 2-1, and beam 2-3. Accordingly, the terminal device uses receive beam A to receive the signals transmitted on beam 2-1, beam 2-1, and beam 2-3 respectively, and performs layer 1-reference signal receiving power (layer1-reference signal receiving power, L1-RSRP) measurement. Then, based on the beam measurement results, the terminal device reports the selected beam or beams to the network device.

[0066] As shown in Figure 5, through the P1 process, coarse synchronization is achieved between downlink transmit beam 2 and downlink receive beam A. To fine-tune the receive beam, the network device can send measurement signals multiple times on beam 2. Accordingly, the terminal device can use three narrower beams: beam A-1, beam A-2, and beam A-3 to receive the signal transmitted on beam 2 and perform measurements. The terminal device can then determine which narrow beam is best for transmit beam 2 based on the beam measurement results. In this process, the terminal device does not need to report to the network device which narrow beam it has selected for receiving transmit beam 2.

[0067] It should be noted that the P1 to P3 processes introduced above indicate an exemplary process of beam selection. In the embodiment of the present application, the beam selection process may also be implemented in other ways, which is not limited in the embodiment of the present application.

[0068] The above describes the exemplary process of beam selection in conjunction with Figures 4 to 6, and the following still describes the beam measurement process and reporting process.

[0069] In addition, the process of matching the transmit and receive beams in uplink transmission can be roughly divided into three main processes, denoted as U1, U2, and U3. U1 is the coarse matching of the uplink transmit and receive beams. U2 is the fine adjustment of the uplink receive beam on the network side. U3 is the fine adjustment of the uplink transmit beam on the terminal side. The above processes U1 to U3 are similar to processes P1 to P3 and are not repeated below for the sake of brevity.

[0070] Beam measurement

[0071] As described above, the basic beam selection process requires corresponding beam measurements. Currently, beam measurements can be achieved by measuring the reference signal transmitted on the beam.

[0072] In some implementations, the measurement of the downlink beam may be achieved by measuring the CSI-RS or SSB transmitted on the downlink beam. In other implementations, the measurement of the uplink beam may be achieved by measuring the SRS or SSB transmitted on the downlink beam.

[0073] In some implementations, for beam measurements (uplink and downlink beam measurements), layer 1 (L1) measurements can be used. L1 measurements can be processed directly at the physical layer, resulting in shorter processing latency. Currently, L1 measurements used for beam measurements include Layer 1 Reference Signal Received Power (L1-RSRP) and Layer 1 Signal to Interference and Noise Ratio (L1-SINR).

[0074] In other implementations, for uplink beam measurements, the measurement quantity may be a layer 1 (L1) measurement quantity. The L1 measurement can be processed directly at the physical layer, with a shorter processing delay. Currently, L1 measurement quantities used for beam measurements may include: layer 1-reference signal received power (L1-RSRP), layer 1-signal to interference plus noise ratio (L1-SINR), and layer 1-reference signal received quality (L1-RSRQ).

[0075] It should be noted that in the embodiment of the present application, in addition to the L1 measurement quantity described above, other measurement quantities may also be used, such as the L3 measurement quantity. Of course, the measurement quantities applicable to the embodiment of the present application may also be measurement quantities newly introduced in future communication systems.

[0076] Measurement reporting

[0077] In some implementations, the terminal device can report one or more information to the network device based on the beam measurement results, each information including beam indication information (for example, the identification of the reference signal, the number of the reference signal, etc.) and the corresponding measurement quantity.

[0078] Beam management based on artificial intelligence (AI)

[0079] In the beam selection process described above with reference to Figures 4 to 6, it is usually necessary to traverse all combinations of receive and transmit beams before selecting a suitable beam. However, traversing all combinations takes a long time, resulting in low beam selection efficiency.

[0080] For example, suppose the network equipment deploys 64 different downlink transmission directions in FR2 (carried by up to 64 SSBs). Accordingly, the terminal device uses one or more antenna panels to simultaneously scan the receiving beams when receiving, and each antenna panel has 4 receiving beams. Then the terminal device needs to measure at least 256 beam pairs, which means that 256 resources of downlink resource overhead are required. From a time perspective, each SSB cycle is approximately 20ms, and 4 SSB cycles are required to complete the measurement of 4 receiving beams. Assuming that multiple receiving antenna panels can perform beam scanning simultaneously, it will take at least 80ms.

[0081] As the number of beams in future massive MIMO systems increases, using beam scanning-based beam management solutions to match optimal beam pairs will only increase reference signal transmission overhead and beam scanning latency. Therefore, to avoid these issues, Release 18 proposes AI-based beam management. The following describes this AI-based beam management solution, combining the training and prediction processes of the AI ​​model.

[0082] Assume that the AI ​​model is used to predict the available beams in beam set A. Accordingly, during the training phase, the beam measurement results of beam set B can be used as AI model training data. That is, the AI ​​model is trained based on the beam measurement results of beam set B so that the AI ​​model can predict the available beams from beam set A.

[0083] It should be noted that the beam measurement results of the above-mentioned beam set B may include the measurement results corresponding to the L1 measurement quantity, and / or the indication information of the selected beam in beam set B (for example, the transmitting beam identifier, the receiving beam identifier or the beam pair identifier, etc.).

[0084] In some implementations, the training data may also include label information of beam set A, and the label information is used to indicate one or more of the following beams in beam set A: optimal transmit beam, optimal receive beam, optimal beam pair, better multiple transmit beams, better multiple receive beams, better beam pair, etc.

[0085] In the prediction stage, the input of the AI ​​model may include the link quality measurement results (for example, L1 measurement quantity) corresponding to the beams in beam set A, and the prediction results output by the AI ​​model may include the target beam selected from beam set A and the link quality corresponding to the target beam.

[0086] In some implementations, the target beam may be one or more beams. For example, if the target beam is a single beam, the target beam may be the optimal beam or a relatively optimal beam in beam set A. For example, if the target beam is multiple beams, the target beam may be multiple beams in beam set A that meet the requirements. "Meeting the requirements" may be understood as meaning that the link quality corresponding to the beam meets the requirements, for example, the link quality corresponding to the beam is greater than or equal to a threshold.

[0087] In other implementations, the target beam may refer to one or more beam pairs, each of which may include a receive beam or a transmit beam. For example, if the target beam is a single beam pair, the target beam may be the optimal beam pair or a relatively optimal beam pair in beam set A. For example, if the target beam is multiple beam pairs, the target beam may be multiple beam pairs in beam set A that meet the requirements. "Meeting the requirements" may be understood as meaning that the link quality corresponding to the beam pair meets the requirements, for example, the link quality corresponding to the beam pair is greater than or equal to a threshold.

[0088] It should be noted that the link quality in the embodiment of the present application can be determined by one or more measurement quantities described above. Of course, the link quality in the embodiment of the present application can also be determined based on other measurement quantities in future communication systems, and the embodiment of the present application is not limited to this.

[0089] In addition, the link quality is determined based on one or more measurement quantities, which can be understood as the link quality being obtained by processing one or more measurement quantities. Of course, the link quality can also be a measurement quantity, which is not limited in the present embodiment.

[0090] It should also be noted that if the prediction result only indicates one beam in the beam pair, the other beam in the beam pair can be determined by other means. For example, it can be determined by one or some of the processes P1 to P3 introduced above. Of course, it can also be determined by one or some of the processes U1 to U3. This is not limited to the embodiments of the present application.

[0091] In some implementations, the beam set B may be a different beam set from the beam set A. In some implementations, the beam set B may be a subset of the beam set A. Accordingly, by measuring fewer beams (beams in the beam set B), predictions for more beams (beams in the beam set A) may be achieved. Compared with the above-mentioned scheme of selecting beams based on traversing all combinations, this helps to reduce the time required to execute the beam selection process. Of course, in the embodiment of the present application, the beams in the beam set B and the beams in the beam set A may be completely different beams. For example, there is no intersection between the beam set B and the beam set A, but the beam direction corresponding to the beam set B may be similar to the beam direction corresponding to the beam set A.

[0092] In some other implementations, the beam set B may be exactly the same as the beam set A.

[0093] For ease of understanding, the following describes the AI ​​model applicable to the embodiment of the present application in conjunction with Figures 7 to 10. Of course, the AI ​​model of the embodiment of the present application can also be other models, and the embodiment of the present application is not limited to this.

[0094] Neural Networks

[0095] In recent years, artificial intelligence research, exemplified by neural networks, has achieved remarkable success in many fields, and will continue to play a vital role in people's lives and production for a long time to come. A neural network can be understood as a computational model consisting of multiple interconnected neuron nodes. The connections between these nodes represent the weighted values ​​from input signals to output signals, often referred to as weights. Each node performs a weighted summation of different input signals and outputs the result through a specific activation function.

[0096] As shown in Figure 7, neurons can rely on activation functions to implement nonlinear mapping, where the input of the neuron can be recorded as A, and each dimension of the input is recorded as a j , the corresponding weight is recorded as w j , together with the summation units (SU), the input is strengthened or weakened. In addition, the output of SU can be input into the activation function f to obtain the output t, where the value of j is 1, 2, ..., n.

[0097] Common neural networks include convolutional neural network (CNN), recurrent neural network (RNN), deep neural network (DNN), etc.

[0098] The following describes a neural network applicable to embodiments of the present application in conjunction with FIG8 . The neural network shown in FIG8 can be divided into three categories based on the location of different layers: input layer 810 , hidden layer 820 , and output layer 830 . Generally speaking, the first layer is the input layer 810 , the last layer is the output layer 830 , and the intermediate layers between the first and last layers are all hidden layers 820 .

[0099] The input layer 810 is used to input data, where the input data can be, for example, a received signal received by a receiver. The hidden layer 820 is used to process the input data, for example, decompress the received signal. The output layer 830 is used to output processed output data, for example, a decompressed signal.

[0100] As shown in Figure 8, a neural network consists of multiple layers, each layer contains multiple neurons. The neurons between layers can be fully connected or partially connected. For connected neurons, the output of the neurons in the previous layer can serve as the input of the neurons in the next layer.

[0101] With the continuous advancement of neural network research, deep learning algorithms have been proposed in recent years. These algorithms introduce a large number of hidden layers into neural networks, forming DNNs. More hidden layers allow DNNs to better capture complex real-world situations. Theoretically, a model with more parameters has higher complexity and a greater "capacity," meaning it can handle more complex learning tasks. These neural network models are widely used in pattern recognition, signal processing, optimization and combination, anomaly detection, and other fields.

[0102] CNN is a deep neural network with a convolutional structure, and its structure is shown in FIG9 , which may include an input layer 910 , a convolutional layer 920 , a pooling layer 930 , a fully connected layer 940 , and an output layer 950 .

[0103] Each convolution layer 920 may include a plurality of convolution operators, which are also called kernels. The convolution operator can be regarded as a filter for extracting specific information from the input signal. The convolution operator can essentially be a weight matrix, which is usually predefined.

[0104] The weight values ​​in these weight matrices need to be obtained through a lot of training in practical applications. The weight matrices formed by the weight values ​​obtained through training can extract information from the input signal, thereby helping CNN to make correct predictions.

[0105] When CNN has multiple convolutional layers, the initial convolutional layer tends to extract more general features, which can also be called low-level features. As the depth of CNN increases, the features extracted by the subsequent convolutional layers become more and more complex.

[0106] Pooling layer 930 is often needed to reduce the number of training parameters. Therefore, it is often necessary to periodically introduce pooling layers after convolutional layers. For example, as shown in Figure 9, a single convolutional layer can be followed by a pooling layer, or multiple convolutional layers can be followed by one or more pooling layers. In signal processing, the sole purpose of the pooling layer is to reduce the spatial size of the extracted information.

[0107] The fully connected layer 940, after being processed by the convolution layer 920 and the pooling layer 930, is not sufficient for CNN to output the required output information. Because as mentioned above, the convolution layer 920 and the pooling layer 930 only extract features and reduce the parameters brought by the input data. However, in order to generate the final output information (for example, the bit stream of the original information transmitted by the transmitter), CNN also needs to use the fully connected layer 940. Generally, the fully connected layer 940 may include multiple hidden layers, and the parameters contained in the multiple hidden layers may be pre-trained based on relevant training data of a specific task type. For example, the task type may include decoding a data signal received by a receiver. For another example, the task type may also include channel estimation based on a pilot signal received by the receiver.

[0108] Following the multiple hidden layers in the fully connected layer 940, the final layer of the CNN is the output layer 950, which is used to output the results. Typically, this output layer 950 is configured with a loss function (e.g., a loss function similar to categorical cross entropy) to calculate the prediction error, or to evaluate the degree of difference between the output of the CNN model (also known as the predicted value) and the ideal result (also known as the true value).

[0109] To minimize the loss function, the CNN model needs to be trained. In some implementations, the backpropagation algorithm (BP) can be used to train the CNN model. The BP training process consists of a forward propagation process and a backward propagation process. During the forward propagation process (e.g., the propagation from 910 to 950 in Figure 9 is forward propagation), the input data is fed into the aforementioned layers of the CNN model, processed layer by layer, and transmitted to the output layer. If the output result of the output layer differs significantly from the ideal result, the minimization of the aforementioned loss function is used as the optimization goal, and the backpropagation process is switched to (e.g., the propagation from 950 to 910 in Figure 9 is backward propagation). The partial derivatives of the optimization goal with respect to each neuron weight are calculated layer by layer, forming the gradient of the optimization goal with respect to the weight vector, which serves as the basis for modifying the model weights. The CNN training process is completed during the weight modification process. When the aforementioned error reaches the desired value, the CNN training process ends.

[0110] It should be noted that the CNN shown in Figure 9 is only an example of a convolutional neural network. In specific applications, the convolutional neural network can also exist in the form of other network models, and the embodiments of the present application are not limited to this.

[0111] RNNs are designed to process sequential data. In traditional neural network models (for example, CNN models), the layers are fully connected, from the input layer to the hidden layer to the output layer, and the nodes within each layer are disconnected. However, these ordinary neural networks are inadequate for many problems. For example, if you want to predict the next word in a sentence, you generally need to use the previous word, because the previous and next words in a sentence are not independent. RNNs are called recurrent neural networks because the current output of a sequence is also related to the previous output. Specifically, the network remembers the previous information and applies it to the calculation of the current output. That is, the nodes between hidden layers are no longer disconnected but connected, and the input of the hidden layer includes not only the output of the input layer but also the output of the hidden layer at the previous moment. In theory, RNNs can process sequence data of any length.

[0112] Training an RNN is similar to training a traditional ANN (artificial neural network). The same backpropagation error algorithm is used, but there is a slight difference. If the RNN is expanded, the parameters W, U, and V are shared, while traditional neural networks are not. Furthermore, when using the gradient descent algorithm, the output of each step depends not only on the network state at the current step, but also on the state of the network at the previous steps. For example, at t = 4, the output must be propagated back three steps, and the gradients of the three subsequent steps must be added. This learning algorithm is called backpropagation through time (BPTT).

[0113] Given the existence of artificial neural networks and convolutional neural networks, why do we still need recurrent neural networks? The reason is simple. Both convolutional and artificial neural networks assume that elements are independent of each other, and that inputs and outputs are also independent, like cats and dogs. However, in the real world, many elements are interconnected, such as the changes in stock prices over time. For example, someone said, "I love traveling, and my favorite place is Yunnan. I must visit __ someday." Everyone knows to fill in the blank with "Yunnan." This is because we infer this information based on the context, but achieving this is quite difficult. Therefore, recurrent neural networks were developed. Their essence is that they possess memory, just like humans. Therefore, their output depends on the current input and memory.

[0114] Figure 10 shows the structure of an RNN. Each circle can be considered a unit, and since each unit performs the same function, it can be folded into the left half of the diagram. To explain RNN in one sentence, it's a unit structure that is reused.

[0115] At present, in order to solve the problem of gradient explosion or disappearance of RNN, a deformation is made on the basis of RNN to obtain the long short-term memory (LSTM) model.

[0116] As shown in Figure 11, LSTM introduces a new memory unit c t (also called "cell state"), which is used for linear cyclic information transmission and outputs information to the external state h of the hidden layer. t At each moment t, c t It records historical information up to the current moment. Unlike RNNs, which only consider the most recent state, memory cells determine which states should be retained and which should be forgotten, addressing the shortcomings of traditional RNNs in long-term memory.

[0117] Continuing to refer to FIG11, in order to achieve the above state selection, the memory unit introduces a gate control mechanism to control the path of information transmission, similar to the gate in the data circuit, "0" means closed, and "1" means open. The memory unit includes a forget gate 1110, an input gate 1120, and an output gate 1130. Among them, the forget gate is used to control the memory unit c at the previous moment. t-1 How much information needs to be forgotten? The input gate is used to control the candidate state at the current moment. How much information needs to be stored? The output gate is used to control the memory unit c at the current moment. t How much information needs to be output to the external state h t .

[0118] At present, although R18 discusses and studies the beam management method based on the first model (for example, the AI ​​model), it does not discuss how to monitor and manage the operating status of the first model. It is possible that the target beam (for example, the optimal transmit beam, the optimal receive beam, or the optimal beam pair) in the prediction result output by the first model is not the truly optimal choice, but the terminal device and the network device are still using the prediction result. When the link quality corresponding to the target beam is too poor, communication based on the target beam may be impossible.

[0119] Taking the first model as an AI model as an example, with the complex changes in the channel environment during the movement of the terminal device, the wireless communication environment (for example, channel state) corresponding to the training data of the AI ​​model may not match the wireless communication environment actually deployed, resulting in the AI ​​model being unable to accurately predict the beam used for communication, that is, the generalization of the trained AI model is limited. At this time, the target beam (for example, the optimal transmit beam, the optimal receive beam, or the optimal beam pair) in the prediction result output by the AI ​​model is not the truly optimal choice, and it may even be impossible to communicate based on the target beam due to the poor link quality corresponding to the target beam. Therefore, there is an urgent need for a method for managing the operating status of the AI ​​model to help improve the success rate of communication. The following describes the method for beam management in an embodiment of the present application in conjunction with Figure 12.

[0120] It should be noted that, for ease of understanding, the following description takes the first model as an AI model as an example. In an embodiment of the present application, the first model may also be other models used for beam management.

[0121] Figure 12 is a schematic flow chart of a method for beam management according to an embodiment of the present application. The method shown in Figure 12 includes step S1210.

[0122] In step S1210 , the terminal device determines the status of the AI ​​model.

[0123] The above-mentioned AI model is used to perform beam prediction for a target beam set, and the state of the AI ​​model (or "operating state") includes one or more of an activated state, an inactivated state, and a monitoring state.

[0124] The activation state indicates that the AI ​​model is in operation. At this point, the target beam indicated by the AI ​​model's prediction results can be used for communication. In other words, the AI ​​model in the activation state outputs a highly accurate prediction result, allowing communication based on the target beam indicated by the prediction results.

[0125] The inactive state indicates that the AI ​​model is not operating. At this point, the target beam indicated by the AI ​​model's predictions cannot be used for communication. In other words, the predictions output by the inactive AI model are less accurate, making communication impossible based on the target beam indicated by the predictions.

[0126] In some implementations, if the AI ​​model is in an inactive state, at this time, in order to ensure the continuity of communication between the terminal device and the network device, beam management can be performed on the target beam set based on a beam scanning method. Among them, beam management based on beam scanning can adopt the beam management method described above in conjunction with Figures 4 to 6. Of course, in the embodiment of the present application, if the continuity of communication between the terminal device and the network device is not considered, if the AI ​​model is in an inactive state, beam management can also be stopped.

[0127] In some implementations, if the AI ​​model is in an inactive state, it can be updated online to improve the accuracy of the AI ​​model's predictions. In the embodiments of the present application, whether to update the AI ​​model online can be determined based on the scope of application of the AI ​​model. This will be described below and will not be repeated here for the sake of brevity.

[0128] It may be because the AI ​​model itself cannot provide beam prediction. Therefore, the inactive AI model can be updated online. Of course, if the AI ​​model is in an inactive state, it may also be due to changes in the wireless communication environment of the terminal device, not an error in the AI ​​model. Therefore, it is also possible not to update the inactive AI model online. The specific online update solution will be introduced below and is not limited here for the sake of brevity.

[0129] The monitoring state, also known as the "model monitoring state (MMS)," is used to determine the AI ​​model's monitoring results, which in turn indicate the accuracy of the AI ​​model's predictions. The following sections will detail the methods for determining the accuracy of predictions, combining methods 1 and 2. For the sake of brevity, these methods will not be detailed here.

[0130] In some implementations, the monitoring result may be obtained by continuously monitoring the state of the AI ​​model within a period of time (hereinafter referred to as the "first time period"). Therefore, the first time period may also be referred to as a "monitoring time window" or a "first time window". For example, the monitoring result may indicate the overall accuracy of multiple prediction results predicted by the AI ​​model within the first time period. For another example, assuming that the AI ​​model can only predict one prediction result within the first time period, the monitoring result may indicate the accuracy of a prediction result predicted by the AI ​​model within the first time period. This embodiment of the present application is not limited to this.

[0131] In some implementations, the monitoring results can be used to determine whether the AI ​​model enters an active state or an inactive state. For example, when the monitoring results indicate that the accuracy of the prediction result is high, the AI ​​model can enter an active state, or the AI ​​model enters an active state from a monitoring state. Conversely, if the monitoring results indicate that the accuracy of the prediction result is low, the AI ​​model can enter an inactive state, or the AI ​​model enters an inactive state from a monitoring state.

[0132] It should be noted that the above-mentioned communication may include communication between a terminal device and a network device. Of course, communication may also include communication between terminal devices, and this embodiment of the application does not limit this. For ease of understanding, this embodiment of the application takes communication between a terminal device and a network device as an example for description.

[0133] As described above, the monitoring results are used to indicate the accuracy of the prediction results. Accordingly, embodiments of the present application also provide a method for determining the accuracy of the prediction results. In some implementations, the accuracy of the prediction results of the AI ​​model in the monitoring state can be determined based on the beam measurement results of the target beam set. In other words, the beam measurement results of the target beam set are used as labels to determine the accuracy of the AI ​​model's prediction results on the target beam set.

[0134] The beam measurement results may be, for example, beam measurement results determined by the beam scanning method described above. Alternatively, the beam measurement results may be beam measurement results determined by traversing the beam combinations in the target beam set described above, where the target beam set may be the aforementioned beam set A. Of course, in the embodiment of the present application, the beam measurement results may also be beam measurement results obtained by measuring some beam combinations in beam set A, which is not limited in the embodiment of the present application.

[0135] The following describes a method for determining the accuracy of a prediction result based on beam measurement results, applicable to embodiments of the present application. In some implementations, the accuracy of the prediction result can be determined based on one or more of the following: the beam indicated by the prediction result; the link quality corresponding to the beam indicated by the prediction result; the beam indicated by the beam measurement result; and the link quality corresponding to the beam indicated by the beam measurement result.

[0136] For ease of understanding, the following describes a method for determining the accuracy of prediction results applicable to an embodiment of the present application in conjunction with determination methods 1 to 2.

[0137] Judgment method 1, taking the accuracy of the prediction result of the AI ​​model based on the beam indicated by the prediction result and the beam indicated by the beam measurement result as an example. The accuracy of the prediction result of the AI ​​model can be determined based on the number of times the beam indicated by the prediction result and the beam indicated by the beam measurement result do not match within the first time period. Therefore, this judgment method can also be called the "prediction failure occurrence number judgment method". For example, when the number of mismatches is greater than the threshold number x (also called the "first threshold x"), the accuracy of the prediction result of the AI ​​model is low, or the prediction result of the AI ​​model is inaccurate, or the AI ​​model is not applicable to the current wireless communication environment. On the contrary, when the number of mismatches is less than or equal to the threshold number x, the accuracy of the prediction result of the AI ​​model is high, or the prediction result of the AI ​​model is accurate, or the AI ​​model is applicable to the current wireless communication environment.

[0138] In some implementations, the threshold number x may be configured by the network device, or determined by the terminal device. Of course, the threshold number may also be predefined by the protocol.

[0139] Taking the following beam selection process as an example, it is assumed that the terminal device triggers the beam scanning process for the target beam set K times in the first time period, and the indexes of the K optimal transmit beams indicated by the beam measurement results are {m1, m2…m K}, the K L1-RSRP values ​​corresponding to the optimal transmit beam are {r1, r2…r K Accordingly, in the first time period, the AI ​​model performs K prediction processes (or called "online reasoning processes"), and the prediction results indicate that the indexes of the K optimal transmit beams are {m1', m2', ...m K '}, and the L1-RSRP values ​​corresponding to the K optimal transmit beams are {r1',r2',...r K '}, where K is a positive integer greater than or equal to.

[0140] Accordingly, based on the above-described determination method 1, if the number of times the indices of the K optimal transmit beams indicated by the prediction result do not match the indices of the K optimal transmit beams indicated by the beam measurement result is greater than the threshold value x, then the AI ​​model is not applicable to the current wireless communication environment. Conversely, if the number of times the indices of the K optimal transmit beams indicated by the prediction result do not match the indices of the K optimal transmit beams indicated by the beam measurement result is less than or equal to the threshold value x, then the AI ​​model is applicable to the current wireless communication environment.

[0141] In some implementations, considering that the larger the number K is, the more likely mismatches will occur, in this embodiment of the present application, a correspondence between K and threshold x may be established, or K and threshold x may be used to form a parameter combination {K, x}. In this way, when the terminal device determines K, it may also determine the threshold x based on the correspondence between K and threshold x.

[0142] In some implementations, for terminal devices or application scenarios where the AI ​​model's output prediction results are relatively loose, a higher threshold number can be configured to make it easier for the AI ​​model to pass monitoring and improve the AI ​​model's working efficiency. For terminal devices or application scenarios where the AI ​​model's output prediction results are relatively strict, a higher threshold number can be configured to obtain more reliable link performance.

[0143] The above-mentioned terminal devices that are relatively loose with respect to the prediction results output by the AI ​​model may include IoT terminal devices; correspondingly, terminal devices that are relatively strict with respect to the prediction results output by the AI ​​model may include vehicle-mounted terminal devices, mobile phones, etc.

[0144] The above application scenarios may include the service type of the service to be transmitted. For example, the application scenario where the prediction results output by the AI ​​model are relatively loose may be the scenario where ordinary services are transmitted. Correspondingly, the application scenario where the prediction results output by the AI ​​model are relatively strict may be the scenario where URLLC services are transmitted.

[0145] In addition, it should be noted that when the accuracy of the prediction result is determined based on the above-mentioned determination method 1, the prediction result and the beam measurement result may not indicate the link quality corresponding to the beam (for example, L1-RSRP). Of course, when the accuracy of the prediction result is determined based on the above-mentioned determination method 1, the prediction result and the beam measurement result may also indicate the link quality corresponding to the beam (for example, L1-RSRP), and this embodiment of the present application is not limited to this.

[0146] Judgment method 2, taking the link quality corresponding to the beam indicated by the prediction result (referred to as "link quality 1") and the link quality corresponding to the beam indicated by the beam measurement result (referred to as "link quality 2") as an example, to determine the accuracy of the prediction result of the AI ​​model. The accuracy of the prediction result of the AI ​​model can be determined based on the difference between link quality 1 and link quality 2 in the first time period. Therefore, this judgment method can also be called "link quality judgment method". For example, when the above difference is greater than the difference threshold (also called the "second threshold"), the accuracy of the prediction result of the AI ​​model is low, or the prediction result of the AI ​​model is inaccurate, or the AI ​​model is not applicable to the current wireless communication environment. On the contrary, when the above difference is less than or equal to the difference threshold, the accuracy of the prediction result of the AI ​​model is high, or the prediction result of the AI ​​model is accurate, or the AI ​​model is applicable to the current wireless communication environment.

[0147] It should be noted that the above difference can be understood as a difference, that is, the difference between link quality 1 and link quality 2. The above difference can also be a ratio, that is, the percentage of the difference between link quality 1 and link quality 2 to link quality 2, which is not limited in this embodiment of the present application.

[0148] In some implementations, the link quality 1 may be the average of multiple link qualities indicated by multiple prediction results obtained by the AI ​​model during the first time period. Taking link quality determination based on L1-RSRP as an example, assuming that K L1-RSRP values ​​are obtained from the predictions during the first time period, the link quality 1 may be the average of the K L1-RSRP values.

[0149] In addition, the link quality 1 may also be the maximum value of multiple link qualities indicated by multiple prediction results obtained by the AI ​​model during the first time period. Taking the determination of link quality based on L1-RSRP as an example, assuming that K L1-RSRP values ​​are obtained by prediction during the first time period, the link quality 1 may be the maximum value of the K L1-RSRP values. Of course, in the embodiments of the present application, the link quality 1 is not specifically limited. For example, the link quality 1 may also be the minimum value of multiple link qualities indicated by multiple prediction results.

[0150] In other implementations, the link quality 2 may be determined based on an average of multiple link qualities indicated by multiple beam measurement results obtained multiple times during the first time period. Taking the determination of link quality based on L1-RSRP as an example, assuming that K L1-RSRP values ​​are obtained from measurements during the first time period, the link quality 2 may be determined based on the average of the K L1-RSRP values.

[0151] In addition, the link quality 2 may also be the maximum value of multiple link qualities indicated by multiple beam measurement results obtained through multiple measurements during the first time period. Taking the determination of link quality based on L1-RSRP as an example, assuming that K L1-RSRP values ​​are obtained from measurements during the first time period, the link quality 2 may correspondingly be the maximum value of the K L1-RSRP values. Of course, in this embodiment of the present application, the link quality 2 is not specifically limited. For example, the link quality 2 may also be the minimum value of multiple link qualities indicated by multiple prediction results.

[0152] In some implementations, the difference threshold may be configured by the network device, or determined by the terminal device. Of course, the difference threshold may also be predefined by the protocol.

[0153] Taking the selection process of the following beam as an example, it is assumed that the terminal device triggers the beam scanning process for the target beam set K times in the first time period, and the indexes of the K optimal transmit beams indicated by the beam measurement results are {m1, m2, m3, m4, m5, m6, m7, m8, m9, m10, m11, m12, m13, m14, m15, m16, m17, m18, m19, m20, m21, m22, m23, m24, m25, m26, m37, m38, m49, m50, m51, m61, m7 2， …m K}, the K L1-RSRP values ​​corresponding to the optimal transmit beam are {r1, r 2， …r K Accordingly, in the first time period, the AI ​​model performs K prediction processes (or called "online reasoning processes"), and the prediction results indicate that the indexes of the K optimal transmit beams are {m1', m2', ...m K '}, and the L1-RSRP values ​​corresponding to the K optimal transmit beams are {r1',r2',...r K '}, where K is a positive integer greater than or equal to . Accordingly, the link quality 1 is based on K L1-RSRP values ​​{r1', r2', ... r K '} determines that the link quality 2 is based on K L1-RSRP values ​​{r1, r 2， …r K}Sure.

[0154] Based on the above determination method 2, if the difference between link quality 1 and link quality 2 is greater than the difference threshold, the AI ​​model is not applicable to the current wireless communication environment. Conversely, if the difference between link quality 1 and link quality 2 is less than or equal to the difference threshold, the AI ​​model is applicable to the current wireless communication environment.

[0155] In some implementations, for terminal devices or application scenarios where the AI ​​model's output prediction results are relatively loose, a higher difference threshold can be configured to make it easier for the AI ​​model to pass monitoring and improve the AI ​​model's working efficiency. For terminal devices or application scenarios where the AI ​​model's output prediction results are relatively strict, a higher difference threshold can be configured to obtain more reliable link performance.

[0156] The above-mentioned terminal devices that are relatively loose with respect to the prediction results output by the AI ​​model may include IoT terminal devices; correspondingly, terminal devices that are relatively strict with respect to the prediction results output by the AI ​​model may include vehicle-mounted terminal devices, mobile phones, etc.

[0157] The above-mentioned application scenarios may include the types of services to be transmitted. For example, an application scenario with relatively loose requirements for the prediction results output by the AI ​​model may be a scenario for transmitting ordinary services. Correspondingly, an application scenario with relatively strict requirements for the prediction results output by the AI ​​model may be a scenario for transmitting ultra-reliable and low latency communications (URLLC) services.

[0158] It should be noted that the above-mentioned determination method 1 and determination method 2 can be used separately or in combination. For example, when the difference between link quality 1 and link quality 2 is greater than the difference threshold, and the number of mismatches is greater than the threshold number, it can be determined that the accuracy of the prediction result output by the AI ​​model is low. Conversely, when the difference between link quality 1 and link quality 2 is less than or equal to the difference threshold, or the number of mismatches is less than or equal to the threshold number, it can be determined that the accuracy of the prediction result output by the AI ​​model is high.

[0159] Of course, other determination methods may also be used in the embodiments of the present application. For example, the accuracy of the prediction result may be determined solely by comparing the beam identifier indicated by the prediction result with the preconfigured beam. For example, the accuracy of the prediction result may also be determined solely by comparing the link quality indicated by the prediction result with a threshold. This embodiment of the present application is not limited to this.

[0160] As described above, in the monitoring state, the accuracy of the prediction result can be determined based on the beam measurement result, and the beam measurement result can be obtained based on beam scanning. Therefore, in some implementations, the network device can configure the time-frequency resources for the beam scanning for the terminal device. That is, the network device sends first configuration information to the terminal device, and the first configuration information is used to configure the time-frequency resources for the beam scanning. The time-frequency resources used for the beam scanning may include the time-frequency resources occupied by the reference signal transmitted by the beam.

[0161] In some cases, the terminal device may need to send beam measurement results to the network device. Therefore, when configuring the time-frequency resources for beam scanning, the network device may also simultaneously configure the time-frequency resources for sending the beam measurement results. Of course, in the embodiment of the present application, the above two time-frequency resources can also be configured separately using different configuration information.

[0162] Typically, both the network device and the terminal device may need to know the monitoring results of the AI ​​model. For example, the network device and the terminal device need to determine whether to perform beam scanning based on the monitoring results. Therefore, in an embodiment of the present application, if the AI ​​model is deployed on the terminal device, the terminal device can send the monitoring results of the AI ​​model to the network device. If the AI ​​model is deployed on the network device, the network device can send the monitoring results of the AI ​​model to the terminal device.

[0163] The previous section describes the AI ​​model status and monitoring methods. The following section describes the applicable scope of AI models based on different AI model deployment methods. These include deploying AI models on network devices or on terminal devices.

[0164] In some implementations, the above-mentioned AI model can be deployed on a network device. In this case, the AI ​​model can be used to perform beam prediction for some or all terminal devices in the cell. At this time, it can be understood that the AI ​​model is a cell-specific model shared by terminal devices in the cell. Of course, in this case, the network device can also deploy multiple AI models, and different AI models can perform beam prediction for different terminal devices. The embodiments of the present application are not limited to this.

[0165] It should be noted that in some scenarios, even if the AI ​​model is a cell-specific model, due to the differences in the wireless communication environments in which different terminal devices are located, the state changes of the AI ​​models for different terminal devices are not synchronized. In other words, in this case, the state of the AI ​​model can be terminal-specific or user-specific (ue-specific). For example, the AI ​​model is cell-specific and its model parameters are synchronized for all terminal devices, but due to different channel changes in different terminal devices, the monitoring results of the AI ​​model may be slightly different. For terminal device 1, if the monitoring results in the first time period indicate that the prediction results of the AI ​​model are accurate, the state of the AI ​​model will change from the monitoring state to the activated state. For terminal device 2, if the monitoring results in the first time period indicate that the prediction results of the AI ​​model are inaccurate, the state of the AI ​​model will change from the monitoring state to the inactivated state.

[0166] Of course, if the differences between the wireless communication environments in which the terminal devices are located are not taken into account, the state of the AI ​​model can also be cell-specific, that is, the state of the AI ​​model corresponding to the terminal devices in the cell can be the same.

[0167] In other implementations, the AI ​​model can be deployed on a terminal device. In this case, the AI ​​model can be used to perform beam prediction for the terminal device. In this case, the AI ​​model can be understood as terminal-specific or user-specific, and accordingly, the state of the AI ​​model is also terminal-specific or user-specific.

[0168] It should be noted that for terminal-specific AI models, the terminal device can obtain the weight of the AI ​​model from the network device. Of course, the terminal device can also train the AI ​​model itself to obtain a terminal-specific AI model.

[0169] As described above, the AI ​​model can be cell-specific or user-specific. Accordingly, the online update process for AI models with different scopes of application may differ slightly. If the AI ​​model is cell-specific, it needs to perform beam prediction for multiple terminal devices in the cell. Therefore, if the AI ​​model is inactive for one or more terminal devices, the AI ​​model may not be updated online. If the AI ​​model is user-specific, if the AI ​​model is inactive for the terminal device, the AI ​​model can be updated online.

[0170] In the embodiment of the present application, for AI models with different scopes of application, the training data used in their training process may also be different. If the AI ​​model is a cell-specific model, the training data of the AI ​​model may include training data collected from all or part of the terminal devices in the cell. If the AI ​​model is a user-specific model, the training data of the AI ​​model may include training data local to the terminal device. Among them, the training data can be referred to the previous introduction, and for the sake of brevity, it will not be repeated here.

[0171] The following describes the online training process for a user-specific AI model, using the example of a user-specific AI model. In some implementations, to simplify the training process, the AI ​​model weights can be configured by the network device. Of course, to improve the accuracy of prediction results, the AI ​​model can also be trained based on the terminal device's own training data and the weights configured by the network device.

[0172] Assuming that the AI ​​model is deployed on the terminal device, the terminal device can obtain multiple beam measurement results for beam set B in the monitoring state and / or inactive state, and use them as labels for the prediction results for beam set A, to build a small-scale training data set for fine-tuning the AI ​​model, and use the training data set to fine-tune the weights of the AI ​​model. This helps to reduce the computing power requirements of the terminal device for online updates of the AI ​​model, and reduces the requirements for the performance of the terminal device for online updates of the AI ​​model. Of course, if the computing power requirements of the terminal device for online updates of the AI ​​model are not considered, the AI ​​model training can be completed by the terminal device alone.

[0173] The previous section introduced the monitoring state of the AI ​​model. The following describes two ways to enter the monitoring state provided by the embodiments of this application. In Method 1, the AI ​​model can enter the monitoring state periodically. In Method 2, the AI ​​model can enter the monitoring state aperiodically.

[0174] Method 1: The AI ​​model can periodically enter the monitoring state.

[0175] Considering that the wireless communication of the terminal device may change, resulting in a decrease in the accuracy of the prediction results of the AI ​​model, the AI ​​model is configured to periodically enter the monitoring state. In other words, the AI ​​model can enter the monitoring state every T time units, where T is a positive integer greater than or equal to 1. In addition, a time unit can be understood as a period of time, which is not limited in the embodiments of the present application.

[0176] For ease of understanding, the following describes a method for the AI ​​model to periodically enter a monitoring state in an embodiment of the present application in conjunction with Figure 13. The method shown in Figure 13 includes steps S1310 to S1360.

[0177] In step S1310 , the AI ​​model is in an active state and lasts for T time units.

[0178] In some implementations, the prediction results output by the activated AI model can be directly used for beam pairing between the network device and the terminal device.

[0179] In step S1320, the AI ​​model enters the monitoring state from the activation state and lasts for W time units, where W is a positive integer greater than or equal to 1.

[0180] In some implementations, the AI ​​model may perform a prediction process once per time unit. Accordingly, within W time units, the AI ​​model may perform a maximum of W prediction processes. In other words, in this case, the first time period mentioned above may be W time units. Accordingly, K of the K prediction processes performed in the first time period is less than or equal to W, i.e., 1≤K≤W.

[0181] At this time, K downlink beam scanning processes for beam set A can be triggered within W time units, and the indexes of the K optimal transmit beams indicated by the beam measurement results are {m1, m2…m K}, the K L1-RSRP values ​​corresponding to the optimal transmit beam are {r1, r2…r K Accordingly, within W time units, the AI ​​model performs K prediction processes (or “online inference processes”), and the prediction results indicate that the indices of the K optimal transmit beams are {m1', m2', ...m K '}, and the L1-RSRP values ​​corresponding to the K optimal transmit beams are {r1',r2'…r K '}, where K is a positive integer greater than or equal to.

[0182] In step S1330, a monitoring result is determined based on the beam measurement result and the prediction result.

[0183] If the monitoring result indicates that the accuracy of the prediction result of the AI ​​model is high, the process proceeds to step S1340. On the contrary, if the monitoring result indicates that the accuracy of the prediction result of the AI ​​model is low, the process proceeds to step S1350.

[0184] It should be noted that the prediction results can be judged based on the beam measurement results based on any of the judgment methods introduced above to determine the monitoring results. The specific methods can be found in the above introduction. For the sake of brevity, they will not be repeated here.

[0185] In step S1340 , the AI ​​model enters the activation state from the monitoring state and lasts for T time units.

[0186] Accordingly, the AI ​​model can continue to perform beam prediction within T time units.

[0187] In step S1350 , the AI ​​model enters the inactive state from the monitoring state and lasts for T time units.

[0188] In some implementations, beam measurement may be performed on the target beam set within T time units to determine the beam used for communication between the terminal device and the network device based on the beam measurement results.

[0189] In addition, in some implementations, the AI ​​model can be trained online within T time units to update the parameters of the AI ​​model. Of course, the AI ​​model can also not be trained online within these T time units.

[0190] It should be noted that after the above step S1340 or step S1350, the AI ​​model can enter the monitoring state of the next cycle, see step S1360, that is, enter the monitoring state and last for W time units.

[0191] In some implementations, one or more parameters of the above-mentioned time unit T, the length W of the monitoring time window, and the number of monitoring times K can be configured in a predefined or preconfigured manner. Of course, the above-mentioned parameters can also be configured by a network device. For example, the network device can be configured through radio resource control (RRC), medium access control control element (MAC CE), or downlink control information (DCI) signaling. Of course, the network device can also be configured through other signaling, and the embodiments of the present application are not limited to this.

[0192] In other implementations, the network device may further adjust the above parameters. For example, in scenarios where the channel changes rapidly (e.g., on a high-speed train), the network device may configure a smaller time unit T. Conversely, in scenarios where the channel changes slowly (e.g., indoors), the network device may configure a longer time unit T.

[0193] As mentioned above, AI models can be deployed on network devices or terminal devices. The following describes the process of periodically entering the monitoring state for these two deployment methods.

[0194] Figure 14 uses the example of an AI model deployed on a network device to illustrate the process of periodically entering the monitoring state. It should be noted that Figure 14 focuses on the interaction between the network device and the terminal device. The terminology and specific determination methods involved can be found in the previous section and will not be further elaborated below for the sake of brevity.

[0195] Assume that the AI ​​model is a cell-specific model. The method shown in Figure 14 includes steps S1410 to S1480.

[0196] In step S1410 , the AI ​​model is currently in an activated state.

[0197] In some implementations, referring to FIG13 , the AI ​​model is in an activated state for both terminal device 1 and terminal device 2 in the cell, and lasts for T time units.

[0198] In step S1420, during the monitoring time window, the AI ​​model enters the monitoring state from the activation state.

[0199] The above monitoring time window is configured periodically. Accordingly, the AI ​​model will periodically enter the monitoring state during the monitoring time window. Continuing with FIG13 , the length of the monitoring time window can be W time units.

[0200] In step S1430, the network device sends first configuration information to terminal device 1 and terminal device 2 respectively, where the first configuration information is used to configure time-frequency resources for beam scanning and time-frequency resources for sending the beam measurement results.

[0201] In some implementations, the time-frequency resources used for beam scanning may include time-frequency resources for terminal device 1 to perform K beam scans, and time-frequency resources for terminal device 2 to perform K beam scans. Accordingly, the time-frequency resources used to send the beam measurement results include time-frequency resources used by terminal device 1 to report K beam measurement results, and time-frequency resources used by terminal device 2 to report K beam measurement results.

[0202] In step S1440, the network device monitors the prediction results of the AI ​​model in the monitoring state.

[0203] In some implementations, the terminal device 1 performs K beam scans on the time-frequency resources of the K beam scans and sends the beam measurement result 1 to the network device. Accordingly, the network device uses the beam measurement result 1 of the beam scan to monitor the K prediction results 1 of the AI ​​model for the terminal device 1 and obtains the monitoring result 1, which indicates that the prediction result of the AI ​​model is accurate for the terminal device 1.

[0204] Accordingly, the terminal device 2 performs K beam scans on the time-frequency resources of the K beam scans, and sends the beam measurement result 2 to the network device. Accordingly, the network device uses the beam measurement result 2 of the beam scan to monitor the K prediction results 2 of the AI ​​model for the terminal device 2, and obtains the monitoring result 2, which indicates that the prediction result of the AI ​​model for the terminal device 2 is inaccurate.

[0205] In step S1450, the network device sends indication information 1 to terminal device 1, instructing the AI ​​model to enter the activation state.

[0206] Accordingly, referring to FIG13 , for terminal device 1 , the AI ​​model will enter the activation state from the monitoring state and last for T time units.

[0207] In step S1460, the network device sends indication information 2 to terminal device 2, instructing the AI ​​model to enter an inactive state.

[0208] Accordingly, referring to FIG13 , for terminal device 2 , the AI ​​model will enter the inactive state from the monitoring state and last for T time units.

[0209] In step S1470, the network device sends indication information 3 to the terminal device 2, where the indication information 3 is used to trigger the terminal device 2 to perform a beam measurement process based on beam scanning (for example, the beam measurement process described above).

[0210] In step S1480, terminal device 2 enters a beam measurement process based on beam scanning.

[0211] It should be noted that, since the AI ​​model is deployed on the network device and is a cell-specific (or cell-shared) model. If the AI ​​model enters an inactive state for one or some terminal devices in the cell, and the model is updated online for these terminal devices, it may cause the weights of the updated AI model to no longer use those terminal devices whose original AI model status is active. That is to say, for terminal device 2, when the AI ​​model is in an inactive state, if the AI ​​model is updated online at this time, it may cause the updated AI model to no longer be applicable to terminal device 1. Therefore, in the embodiment of the present application, under the above circumstances, the AI ​​model may not be updated online.

[0212] Figure 15 uses the example of an AI model deployed on a terminal device to illustrate the process of periodically entering the monitoring state. It should be noted that Figure 15 focuses on the interaction between network devices and terminal devices. The terminology and specific determination methods involved can be found in the previous section and will not be further elaborated below for the sake of brevity.

[0213] Assume that the AI ​​model is a user-specific model, and AI model 1 is deployed on terminal device 1, and AI model 2 is deployed on terminal device 2. The method shown in Figure 15 includes steps S1510 to S1580.

[0214] In step S1510 , AI model 1 and AI model 2 are currently in an active state.

[0215] In some implementations, referring to FIG. 13 , AI Model 1 and AI Model 2 are in an active state for T time units.

[0216] In step S1520, during the monitoring time window, AI model 1 and AI model 2 enter the monitoring state from the activation state.

[0217] The above monitoring time window is configured periodically. Accordingly, AI Model 1 and AI Model 2 will periodically enter the monitoring state during the monitoring time window. Continuing to refer to FIG13 , the length of the monitoring time window can be W time units.

[0218] In step S1530, the network device sends first configuration information to terminal device 1 and terminal device 2 respectively, where the first configuration information is used to configure time-frequency resources for beam scanning and time-frequency resources for sending the beam measurement results.

[0219] In some implementations, the time-frequency resources used for beam scanning may include time-frequency resources for terminal device 1 to perform K beam scans, and time-frequency resources for terminal device 2 to perform K beam scans. Accordingly, the time-frequency resources used to send the beam measurement results include time-frequency resources used by terminal device 1 to report K beam measurement results, and time-frequency resources used by terminal device 2 to report K beam measurement results.

[0220] In step S1540, terminal device 1 and terminal device 2 monitor the prediction results of AI model 1 and AI model 2 respectively in the monitoring state.

[0221] In some implementations, the terminal device 1 performs K beam scans on the time-frequency resources of the K beam scans, uses the beam measurement result 1 of the beam scan to monitor the K prediction results 1 of the AI ​​model 1, and obtains the monitoring result 1, which indicates that the prediction result of the AI ​​model 1 is accurate for the terminal device 1.

[0222] Correspondingly, the terminal device 2 performs K beam scans on the time-frequency resources of the K beam scans, and uses the beam measurement result 2 of the beam scan to monitor the K prediction results 2 of the AI ​​model for the terminal device 2, and obtains the monitoring result 2, which indicates that the prediction result of the AI ​​model 2 is inaccurate.

[0223] In step S1550, terminal device 1 sends indication information 1 to the network device, instructing AI model 1 to enter the activation state.

[0224] Accordingly, referring to FIG13 , for the terminal device 1 , the AI ​​model 1 will enter the activation state from the monitoring state and last for T time units.

[0225] In step S1560, terminal device 2 sends indication information 2 to the network device, instructing AI model 2 to enter an inactive state.

[0226] Accordingly, referring to FIG13 , for the terminal device 2 , the AI ​​model 2 will enter the inactive state from the monitoring state and last for T time units.

[0227] In some implementations, when the AI ​​model 2 enters the inactive state, the terminal device 2 can perform an online update on the AI ​​model, which helps the AI ​​model 2 enter the active state after the next monitoring state. The specific online update method can be found in the above description and will not be repeated here for the sake of brevity.

[0228] In step S1570, the network device sends indication information 3 to the terminal device 2, where the indication information 3 is used to trigger the terminal device 2 to perform a beam measurement process based on beam scanning (for example, the beam measurement process described above).

[0229] In step S1580, the terminal device 2 enters a beam measurement process based on beam scanning.

[0230] It should be noted that in the embodiment of the present application, since the AI ​​model is user-specific, the weight of the AI ​​model may be different for different terminal devices. Of course, in the embodiment of the present application, the weight of the AI ​​model for different terminal devices can also be the same, and the embodiment of the present application does not limit this.

[0231] In an embodiment of the present application, the AI ​​model can periodically enter the monitoring state, which helps to reduce the transmission resources required to transmit the indication information indicating that the AI ​​model has entered the monitoring state.

[0232] The above describes the method for the AI ​​model to periodically enter the monitoring state in the embodiment of the present application in conjunction with Figures 13 to 15. The following describes the method for the AI ​​model to aperiodically enter the monitoring state in conjunction with Figures 16 to 19.

[0233] Method 2: The AI ​​model can enter the monitoring state non-periodically.

[0234] To improve the flexibility of AI model monitoring, network devices or terminal devices can trigger the AI ​​model to enter the monitoring state.

[0235] In some implementations, the network device may instruct the terminal device to trigger the AI ​​model to enter the monitoring state. Of course, the terminal device may instruct the network device to trigger the AI ​​model to enter the monitoring state. Alternatively, the terminal device may autonomously trigger the AI ​​model to enter the monitoring state. Alternatively, the network device may autonomously trigger the AI ​​model to enter the monitoring state. In the embodiments of the present application, the specific method for the AI ​​model to enter the monitoring state is not limited.

[0236] If the network device instructs the terminal device to trigger the AI ​​model to enter the monitoring state, the indication information (also called "first indication information") can be carried in the MAC CE or DCI. Of course, the indication information can also be carried in other information, and this embodiment of the present application is not limited to this.

[0237] In an embodiment of the present application, the duration W of the AI ​​model entering the monitoring state and / or the number of predictions K executed by the AI ​​model in the monitoring state can be predefined by the protocol or configured by the network device. In some implementations, when the network device sends the above-mentioned first configuration information, it can also configure the above-mentioned one or more parameters through the first configuration information. Of course, the network device can also send special configuration information to configure the above-mentioned one or more parameters.

[0238] In other implementations, the network device may not be configured with the duration W for entering the monitoring state. In this case, the indication information (for example, indication information 1 or indication information 2 above) that the network device instructs the terminal device to switch the AI ​​model state can be used to instruct the AI ​​model to end the monitoring state and enter the activation state or non-opportunistic state from the monitoring state.

[0239] In addition, the network device may not be configured with the prediction number K, and this parameter may be implemented by the terminal device.

[0240] In the embodiments of the present application, the conditions for triggering the AI ​​model to enter the monitoring state are not limited. In some implementations, when the terminal device or network device detects a change in the wireless communication environment, the AI ​​model can be triggered to enter the monitoring state. Among them, the change in the wireless communication environment may include that the terminal device has a larger moving distance, or that the terminal device has a faster moving speed, etc. In other implementations, the AI ​​model can also be triggered to enter the monitoring state based on the link quality between the network device and the terminal device. For example, when the link quality between the network device and the terminal device is poor over a period of time, and / or the number of communication failures between the terminal device and the network device is large, the AI ​​model can be triggered to enter the monitoring state. In other implementations, the AI ​​model can also be triggered to enter the monitoring state based on the length of time the AI ​​model is in the active state or the inactive state. For example, when the AI ​​model is in the inactive state for a long period of time, the AI ​​model can be triggered to enter the monitoring state.

[0241] Figure 16 uses the example of an AI model deployed on a network device to illustrate the process of aperiodically entering the monitoring state. It should be noted that Figure 16 focuses on the interaction between the network device and the terminal device. The terminology and specific determination methods involved can be found in the previous section and will not be further elaborated below for the sake of brevity.

[0242] Assume that the AI ​​model is a cell-specific model. The method shown in Figure 16 includes steps S1610 to S1680.

[0243] In step S1610 , the AI ​​model is currently in an activated state.

[0244] In some implementations, the AI ​​model is in an activated state for both terminal device 1 and terminal device 2 in the cell. In the embodiment of the present application, there is no limit on the time that the AI ​​model is in an activated state.

[0245] In step S1620, the network device sends first configuration information to terminal device 1 and terminal device 2 respectively, where the first configuration information is used to configure time-frequency resources for beam scanning, time-frequency resources for sending the beam measurement results, and trigger the AI ​​model to enter a monitoring state.

[0246] In some implementations, the time-frequency resources used for beam scanning may include time-frequency resources for terminal device 1 to perform K beam scans, and time-frequency resources for terminal device 2 to perform K beam scans. Accordingly, the time-frequency resources used to send the beam measurement results include time-frequency resources used by terminal device 1 to report K beam measurement results, and time-frequency resources used by terminal device 2 to report K beam measurement results.

[0247] In step S1630, in response to the first configuration information, the AI ​​model enters the monitoring state from the activation state.

[0248] In step S1640, the network device monitors the prediction results of the AI ​​model in the monitoring state.

[0249] In some implementations, the terminal device 1 performs K beam scans on the time-frequency resources of the K beam scans and sends the beam measurement result 1 to the network device. Accordingly, the network device uses the beam measurement result 1 of the beam scan to monitor the K prediction results 1 of the AI ​​model for the terminal device 1 and obtains the monitoring result 1, which indicates that the prediction result of the AI ​​model is accurate for the terminal device 1.

[0250] Accordingly, the terminal device 2 performs K beam scans on the time-frequency resources of the K beam scans, and sends the beam measurement result 2 to the network device. Accordingly, the network device uses the beam measurement result 2 of the beam scan to monitor the K prediction results 2 of the AI ​​model for the terminal device 2, and obtains the monitoring result 2, which indicates that the prediction result of the AI ​​model for the terminal device 2 is inaccurate.

[0251] In step S1650, the network device sends indication information 1 to terminal device 1, instructing the AI ​​model to enter the activation state.

[0252] In some implementations, the indication information 1 is also used to indicate to the terminal device 1 that the monitoring state of the AI ​​model has ended.

[0253] In step S1660, the network device sends indication information 2 to terminal device 2, instructing the AI ​​model to enter an inactive state.

[0254] In some implementations, the indication information 2 is also used to indicate to the terminal device 2 that the monitoring state of the AI ​​model has ended.

[0255] In step S1670, the network device sends indication information 3 to the terminal device 2, where the indication information 3 is used to trigger the terminal device 2 to perform a beam measurement process based on beam scanning (for example, the beam measurement process described above).

[0256] In step S1680, terminal device 2 enters a beam measurement process based on beam scanning.

[0257] It should be noted that, since the AI ​​model is deployed on the network device and is a cell-specific (or cell-shared) model. If the AI ​​model enters an inactive state for one or some terminal devices in the cell, and the model is updated online for these terminal devices, it may cause the weights of the updated AI model to no longer use those terminal devices whose original AI model status is active. That is to say, for terminal device 2, when the AI ​​model is in an inactive state, if the AI ​​model is updated online at this time, it may cause the updated AI model to no longer be applicable to terminal device 1. Therefore, in the embodiment of the present application, under the above circumstances, the AI ​​model may not be updated online.

[0258] Figure 16 takes the example of a network device triggering an AI model to enter a monitoring state to introduce the method of an embodiment of the present application. The following is introduced in conjunction with Figure 17 using the example of a terminal device triggering an AI model to enter a monitoring state.

[0259] Assume that the AI ​​model is a cell-specific model. The method shown in Figure 17 includes steps S1710 to S1780.

[0260] In step S1710, for terminal device 1, the AI ​​model is currently in an inactive state.

[0261] In step S1720, the terminal device sends request 1 to the network device to request the AI ​​model to enter the monitoring state.

[0262] In step S1730, in response to request 1, the network device sends first configuration information to terminal device 1, where the first configuration information is used to configure time-frequency resources for beam scanning and time-frequency resources for sending the beam measurement results.

[0263] In step S1740, the network device monitors the prediction results of the AI ​​model in the monitoring state.

[0264] In some implementations, the terminal device 1 performs K beam scans on the time-frequency resources of the K beam scans and sends the beam measurement result 1 to the network device. Accordingly, the network device uses the beam measurement result 1 of the beam scan to monitor the K prediction results 1 of the AI ​​model for the terminal device 1 and obtains the monitoring result 1, which is used to indicate the accuracy of the prediction results of the AI ​​model for the terminal device 1.

[0265] In step S1750, if the monitoring result indicates that the prediction result of the AI ​​model is accurate, the network device sends indication information 1 to terminal device 1, instructing the AI ​​model to enter the activation state.

[0266] In some implementations, the indication information 1 is also used to indicate to the terminal device 1 that the monitoring state of the AI ​​model has ended.

[0267] In step S1760, if the monitoring result indicates that the prediction result of the AI ​​model is inaccurate, the network device sends an indication message 2 to the terminal device 1, instructing the AI ​​model to enter an inactive state.

[0268] In some implementations, the indication information 2 is also used to indicate to the terminal device 2 that the monitoring state of the AI ​​model has ended.

[0269] Figure 18 uses the example of an AI model deployed on a network device to illustrate the process of aperiodically entering the monitoring state. It should be noted that Figure 18 focuses on the interaction between the network device and the terminal device. The terminology and specific determination methods involved can be found in the previous section and are omitted for brevity.

[0270] Assume that the AI ​​model is a user-specific model, and AI model 1 is deployed on terminal device 1, and AI model 2 is deployed on terminal device 2. The method shown in Figure 18 includes steps S1810 to S1880.

[0271] In step S1810 , AI model 1 and AI model 2 are currently in an active state.

[0272] In some implementations, AI Model 1 and AI Model 2 are in an active state and last for T time units.

[0273] In step S1820, the network device sends first configuration information to terminal device 1 and terminal device 2 respectively, where the first configuration information is used to configure time-frequency resources for beam scanning, time-frequency resources for sending the beam measurement results, and trigger AI model 1 and AI model 2 to enter the monitoring state respectively.

[0274] In some implementations, the time-frequency resources used for beam scanning may include time-frequency resources for terminal device 1 to perform K beam scans, and time-frequency resources for terminal device 2 to perform K beam scans. Accordingly, the time-frequency resources used to send the beam measurement results include time-frequency resources used by terminal device 1 to report K beam measurement results, and time-frequency resources used by terminal device 2 to report K beam measurement results.

[0275] In step S1830 , in response to the first configuration information, AI model 1 and AI model 2 enter the monitoring state from the activation state.

[0276] In step S1840, terminal device 1 and terminal device 2 monitor the prediction results of AI model 1 and AI model 2 respectively in the monitoring state.

[0277] In some implementations, the terminal device 1 performs K beam scans on the time-frequency resources of the K beam scans, uses the beam measurement result 1 of the beam scan to monitor the K prediction results 1 of the AI ​​model 1, and obtains the monitoring result 1, which indicates that the prediction result of the AI ​​model 1 is accurate for the terminal device 1.

[0278] Correspondingly, the terminal device 2 performs K beam scans on the time-frequency resources of the K beam scans, and uses the beam measurement result 2 of the beam scan to monitor the K prediction results 2 of the AI ​​model for the terminal device 2, and obtains the monitoring result 2, which indicates that the prediction result of the AI ​​model 2 is inaccurate.

[0279] In step S1850, terminal device 1 sends indication information 1 to the network device, instructing AI model 1 to enter the activation state.

[0280] In some implementations, the above-mentioned indication information 1 is also used to indicate the end of the monitoring state of the AI ​​model.

[0281] In step S1860, terminal device 2 sends indication information 2 to the network device, instructing AI model 2 to enter an inactive state.

[0282] In some implementations, the above-mentioned indication information 2 is also used to indicate the end of the monitoring state of the AI ​​model.

[0283] In some implementations, when the AI ​​model 2 enters the inactive state, the terminal device 2 can perform an online update on the AI ​​model, which helps the AI ​​model 2 enter the active state after the next non-periodic monitoring state. The specific online update method can be found in the above description and will not be repeated here for the sake of brevity.

[0284] In step S1870, the network device sends indication information 3 to the terminal device 2, where the indication information 3 is used to trigger the terminal device 2 to perform a beam measurement process based on beam scanning (for example, the beam measurement process described above).

[0285] In step S1880, terminal device 2 enters a beam measurement process based on beam scanning.

[0286] It should be noted that in the embodiment of the present application, since the AI ​​model is user-specific, the weight of the AI ​​model may be different for different terminal devices. Of course, in the embodiment of the present application, the weight of the AI ​​model for different terminal devices can also be the same, and the embodiment of the present application does not limit this.

[0287] In an embodiment of the present application, the AI ​​model can enter the monitoring state non-periodically, which helps to reduce the transmission resources required to transmit the indication information indicating that the AI ​​model has entered the monitoring state.

[0288] Figure 18 takes the example of a network device triggering an AI model to enter a monitoring state to introduce the method of an embodiment of the present application. The following text is introduced in conjunction with Figure 19 using the example of a terminal device triggering an AI model to enter a monitoring state to introduce it.

[0289] Assuming that the AI ​​model is a user-specific model, the method shown in FIG19 includes steps S1910 to S1980.

[0290] In step S1910, for terminal device 1, the AI ​​model is currently in an inactive state.

[0291] In step S1920, the terminal device sends request 1 to the network device to request the AI ​​model to enter the monitoring state.

[0292] In step S1930, in response to request 1, the network device sends first configuration information to terminal device 1, where the first configuration information is used to configure time-frequency resources for beam scanning and time-frequency resources for sending the beam measurement results.

[0293] In step S1940, the terminal device monitors the prediction results of the AI ​​model in the monitoring state.

[0294] In some implementations, the terminal device 1 performs K beam scans on the time-frequency resources of the K beam scans, and uses the beam measurement result 1 of the beam scan to monitor the K prediction results 1 of the AI ​​model for the terminal device 1, and obtains the monitoring result 1, which is used to indicate the accuracy of the prediction results of the AI ​​model for the terminal device 1.

[0295] In step S1950, if the monitoring result indicates that the prediction result of the AI ​​model is accurate, the terminal device 1 sends an indication message 1 to the network device, instructing the AI ​​model to enter an activation state.

[0296] In some implementations, the indication information 1 is also used to indicate to the terminal device 1 that the monitoring state of the AI ​​model has ended.

[0297] In step S1960, if the monitoring result indicates that the prediction result of the AI ​​model is inaccurate, the terminal device 1 sends an indication message 2 to the network device, instructing the AI ​​model to enter an inactive state.

[0298] In some implementations, the indication information 2 is also used to indicate to the terminal device 2 that the monitoring state of the AI ​​model has ended.

[0299] It should be noted that in one or more of the embodiments described above in combination with Figures 14 to 19, indication information 1 can be sent simultaneously with indication information 2. Of course, indication information 1 can also be sent at different times from indication information 2. This is not limited to the embodiments of the present application.

[0300] In addition, in the above embodiment, indication information 3 can be the same as indication information 1. That is, when indication information 1 instructs the AI ​​model to enter the inactive state, it also instructs the triggering of the beam measurement process based on beam scanning. Of course, in the above embodiment, indication information 3 can also be different from indication information 1.

[0301] The method embodiment of the present application is described in detail above in conjunction with Figures 1 to 19. The device embodiment of the present application is described in detail below in conjunction with Figures 20 to 22. It should be understood that the description of the method embodiment corresponds to the description of the device embodiment. Therefore, for parts not described in detail, reference can be made to the above method embodiment.

[0302] FIG20 is a schematic diagram of a terminal device according to an embodiment of the present application. The terminal device 2000 shown in FIG20 includes: a processing unit 2010 .

[0303] Processing unit 2010 is used to determine the state of a first model; wherein the first model is used to perform beam prediction for a target beam set, and the state of the first model includes one or more of an activated state, an inactivated state, and a monitoring state; wherein the monitoring state is used to determine a monitoring result of the first model, and the monitoring result is used to indicate the accuracy of the prediction result of the first model.

[0304] In a possible implementation, in the monitoring state, the accuracy of the prediction result of the first model is determined based on the beam measurement result of the target beam set.

[0305] In one possible implementation, the accuracy of the prediction result of the first model is determined based on one or more of the following: the number of times the beam indicated by the prediction result does not match the beam indicated by the beam measurement result within the first time period; the difference between the link quality corresponding to the beam indicated by the prediction result within the first time period and the link quality corresponding to the beam indicated by the beam measurement result.

[0306] In one possible implementation, if the number of times the beam indicated by the prediction result does not match the beam indicated by the beam measurement result within the first time period is greater than a first threshold, the processing unit determines that the prediction result of the first model is inaccurate; if the number of times the beam indicated by the prediction result does not match the beam indicated by the beam measurement result within the first time period is less than or equal to the first threshold, the processing unit determines that the prediction result of the first model is accurate.

[0307] In one possible implementation, if the difference between the link quality corresponding to the beam indicated by the prediction result in the first time period and the link quality corresponding to the beam indicated by the beam measurement result is greater than a second threshold, the processing unit determines that the prediction result of the first model is inaccurate; if the difference between the link quality corresponding to the beam indicated by the prediction result in the first time period and the link quality corresponding to the beam indicated by the beam measurement result is less than or equal to the second threshold, the processing unit determines that the prediction result of the first model is accurate.

[0308] In a possible implementation, the beam measurement result of the target beam set is determined based on a beam scanning method.

[0309] In one possible implementation, the terminal device also includes: a first receiving unit, used to receive first configuration information sent by a network device, the first configuration information is used to configure the time-frequency resources used for the beam scanning, and / or the first configuration information is used to configure the time-frequency resources used to send the beam measurement results.

[0310] In a possible implementation manner, in the inactive state, beam management is performed on the target beam set based on a beam scanning manner.

[0311] In a possible implementation, the first model periodically enters the monitoring state.

[0312] In a possible implementation, the first model enters the monitoring state non-periodically.

[0313] In one possible implementation, the terminal device also includes: a second receiving unit, used to receive first indication information sent by a network device, wherein the first indication information is used to instruct the terminal device to trigger the first model to enter the monitoring state; or the terminal device autonomously triggers the first model to enter the monitoring state.

[0314] In one possible implementation, the processing unit is further used to: determine that the first model enters the activated state from the monitoring state if the monitoring result indicates that the prediction result of the first model is accurate; and / or determine that the first model enters the inactivated state from the monitoring state if the monitoring result indicates that the prediction result of the first model is inaccurate.

[0315] In one possible implementation, the sending unit is used to send the monitoring results of the first model to the network device if the first model is deployed on the terminal device; or, the third receiving unit is used to receive the monitoring results of the first model sent by the network device if the first model is deployed on the network device.

[0316] In one possible implementation, the fourth receiving unit is used to receive second configuration information sent by the network device, where the second configuration information is used to configure the duration of the first time window, where the first time window indicates the time period during which the first model is in the monitoring state; and / or the second configuration information is used to configure the number of monitoring times for monitoring the accuracy of the prediction results of the first model within the first time window.

[0317] In one possible implementation, the duration of the first time window in which the monitoring state lasts and / or the number of monitoring times are indicated in a preconfigured manner, wherein the first time window indicates the time period in which the first model is in the monitoring state, and the number of monitoring times is the number of monitoring times for monitoring the accuracy of the prediction results of the first model within the first time window.

[0318] In one possible implementation, if the first model is deployed on the terminal device, the first model is used to perform beam prediction for the terminal device; or, if the first model is deployed on a network device, the first model is used to perform beam prediction for some or all terminal devices in a cell.

[0319] Figure 21 is a schematic diagram of a network device according to an embodiment of the present application. The network device shown in Figure 21 includes: a processing unit 2110.

[0320] The processing unit 2110 is used to determine the state of the first model; wherein the first model is used to perform beam prediction for a target beam set, and the state of the first model includes one or more of an activated state, an inactivated state, and a monitoring state; wherein the monitoring state is used to determine the monitoring result of the first model, and the monitoring result is used to indicate the accuracy of the prediction result of the first model.

[0321] In a possible implementation, in the monitoring state, the accuracy of the prediction result of the first model is determined based on the beam measurement result of the target beam set.

[0322] In one possible implementation, the accuracy of the prediction result of the first model is determined based on one or more of the following: the number of times the beam indicated by the prediction result does not match the beam indicated by the beam measurement result within the first time period; the difference between the link quality corresponding to the beam indicated by the prediction result within the first time period and the link quality corresponding to the beam indicated by the beam measurement result.

[0323] In one possible implementation, the method further includes: if the number of times that the beam indicated by the prediction result does not match the beam indicated by the beam measurement result within the first time period is greater than a first threshold, the processing unit determines that the prediction result of the first model is inaccurate; if the number of times that the beam indicated by the prediction result does not match the beam indicated by the beam measurement result within the first time period is less than or equal to the first threshold, the processing unit determines that the prediction result of the first model is accurate.

[0324] In one possible implementation, the method further includes: if the difference between the link quality corresponding to the beam indicated by the prediction result in the first time period and the link quality corresponding to the beam indicated by the beam measurement result is greater than a second threshold, the processing unit determines that the prediction result of the first model is inaccurate; if the difference between the link quality corresponding to the beam indicated by the prediction result in the first time period and the link quality corresponding to the beam indicated by the beam measurement result is less than or equal to the second threshold, the processing unit determines that the prediction result of the first model is accurate.

[0325] In a possible implementation, the beam measurement result of the target beam set is determined based on a beam scanning method.

[0326] In one possible implementation, the network device also includes: a first sending unit, used to send first configuration information to the terminal device, the first configuration information indicates the time-frequency resources used for the beam scanning, and / or the first configuration information indicates the time-frequency resources used to send the beam measurement results.

[0327] In a possible implementation manner, in the inactive state, beam management is performed on the target beam set based on a beam scanning manner.

[0328] In a possible implementation, the first model periodically enters the monitoring state.

[0329] In a possible implementation, the first model enters the monitoring state non-periodically.

[0330] In one possible implementation, the network device also includes: a second sending unit, used to send first indication information to the terminal device, the first indication information is used to instruct the terminal device to trigger the first model to enter the monitoring state; or the terminal device autonomously triggers the first model to enter the monitoring state.

[0331] In one possible implementation, the processing unit is further used to: determine that the first model enters the activated state from the monitoring state if the monitoring result indicates that the prediction result of the first model is accurate; and / or determine that the first model enters the inactivated state from the monitoring state if the monitoring result indicates that the prediction result of the first model is inaccurate.

[0332] In one possible implementation, the first receiving unit is used to receive the monitoring results of the first model sent by the terminal device if the first model is deployed on the terminal device; or, the third sending unit is used to send the monitoring results of the first model to the terminal device if the first model is deployed on the network device.

[0333] In one possible implementation, the network device also includes: a fourth sending unit, used to send second configuration information to the terminal device, the second configuration information is used to configure the duration of the first time window, the first time window indicating the time period during which the first model is in the monitoring state; and / or, the second configuration information is used to configure the number of monitoring times for monitoring the accuracy of the prediction results of the first model within the first time window.

[0334] In one possible implementation, the duration of the first time window in which the monitoring state lasts and / or the number of monitoring times are indicated in a preconfigured manner, wherein the first time window indicates the time period in which the first model is in the monitoring state, and the number of monitoring times is the number of monitoring times for monitoring the accuracy of the prediction results of the first model within the first time window.

[0335] In one possible implementation, if the first model is deployed on a terminal device, the first model is used to perform beam prediction for the terminal device; or, if the first model is deployed on the network device, the first model is used to perform beam prediction for some or all terminal devices in a cell.

[0336] In an optional embodiment, the processing unit 2010 may be a processor 2210. The terminal device 2000 may further include a transceiver 2230 and a memory 2220, as specifically shown in FIG22 .

[0337] In an optional embodiment, the processing unit 2110 may be a processor 2210. The network device 2100 may further include a transceiver 2230 and a memory 2220, as specifically shown in FIG22 .

[0338] Figure 22 is a schematic block diagram of a communication device according to an embodiment of the present application. The dashed lines in Figure 22 indicate that the unit or module is optional. Device 2200 may be used to implement the method described in the above method embodiment. Device 2200 may be a chip, a terminal device, or a network device.

[0339] The device 2200 may include one or more processors 2210. The processor 2210 may support the device 2200 to implement the method described in the above method embodiment. The processor 2210 may be a general-purpose processor or a special-purpose processor. For example, the processor may be a central processing unit (CPU). Alternatively, the processor may be another general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic device, discrete hardware component, etc. The general-purpose processor may be a microprocessor or the processor may be any conventional processor, etc.

[0340] The apparatus 2200 may further include one or more memories 2220. The memories 2220 store programs that can be executed by the processor 2210, causing the processor 2210 to perform the methods described in the above method embodiments. The memories 2220 may be independent of the processor 2210 or integrated into the processor 2210.

[0341] The apparatus 2200 may further include a transceiver 2230. The processor 2210 may communicate with other devices or chips via the transceiver 2230. For example, the processor 2210 may transmit and receive data with other devices or chips via the transceiver 2230.

[0342] The present application also provides a computer-readable storage medium for storing a program. The computer-readable storage medium can be applied to a terminal or network device provided in the present application, and the program enables a computer to execute the method performed by the terminal or network device in each embodiment of the present application.

[0343] The present application also provides a computer program product. The computer program product includes a program. The computer program product can be applied to a terminal or network device provided in the present application, and the program causes a computer to execute the method performed by the terminal or network device in each embodiment of the present application.

[0344] The embodiments of the present application also provide a computer program. The computer program can be applied to the terminal or network device provided in the embodiments of the present application, and the computer program enables a computer to execute the method performed by the terminal or network device in each embodiment of the present application.

[0345] It should be understood that the terms "system" and "network" in this application can be used interchangeably. In addition, the terms used in this application are only used to explain the specific embodiments of this application and are not intended to limit this application. The terms "first", "second", "third", and "fourth" in the specification and claims of this application and the accompanying drawings are used to distinguish different objects rather than to describe a specific order. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions.

[0346] In the embodiments of this application, the term "indication" may refer to a direct indication, an indirect indication, or an indication of an association. For example, "A indicates B" may refer to a direct indication of B, e.g., B can obtain information through A; it may refer to an indirect indication of B, e.g., A indicates C, e.g., B can obtain information through C; or it may refer to an association between A and B.

[0347] In the embodiment of the present application, "B corresponding to A" means that B is associated with A and B can be determined based on A. However, it should be understood that determining B based on A does not mean determining B based solely on A, but B can also be determined based on A and / or other information.

[0348] In the embodiments of the present application, the term "corresponding" may indicate a direct or indirect correspondence between the two, or an association relationship between the two, or a relationship between indication and indication, configuration and configuration, etc.

[0349] In the embodiments of the present application, "pre-definition" or "pre-configuration" may be implemented by pre-storing corresponding codes, tables, or other methods that can be used to indicate relevant information in a device (e.g., a terminal device and a network device). The present application does not limit the specific implementation method. For example, pre-definition may refer to information defined in a protocol.

[0350] In the embodiments of the present application, the “protocol” may refer to a standard protocol in the communications field, for example, it may include an LTE protocol, an NR protocol, and related protocols used in future communication systems, and the present application does not limit this.

[0351] In the embodiments of this application, the term "and / or" is simply a description of the association relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this document generally indicates that the related objects are in an "or" relationship.

[0352] In various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0353] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the units is merely 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 devices or units, which can be electrical, mechanical or other forms.

[0354] 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 network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0355] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0356] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center via a wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) method. The computer-readable storage medium can be any available medium that can be read by a computer or a data storage device such as a server or data center that includes one or more available media integrated therein. The available medium may be a magnetic medium (eg, a floppy disk, a hard disk, a magnetic tape), an optical medium (eg, a digital versatile disc (DVD)), or a semiconductor medium (eg, a solid state disk (SSD)).

[0357] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. A method for beam management, characterized in that include: The terminal device determines the state of the first model; The first model is used to perform beam prediction for a target beam set, and a state of the first model includes one or more of an activated state, an inactivated state, and a monitoring state; The monitoring state is used to determine the monitoring result of the first model, and the monitoring result is used to indicate the accuracy of the prediction result of the first model.

2. The method according to claim 1, characterized in that In the monitoring state, the accuracy of the prediction result of the first model is determined based on the beam measurement result of the target beam set.

3. The method according to claim 2, characterized in that The accuracy of the prediction results of the first model is determined based on one or more of the following: the number of times the beam indicated by the prediction result does not match the beam indicated by the beam measurement result within the first time period; The difference between the link quality corresponding to the beam indicated by the prediction result and the link quality corresponding to the beam indicated by the beam measurement result in the first time period.

4. The method according to claim 3, characterized in that The method further comprises: If the number of times that the beam indicated by the prediction result and the beam indicated by the beam measurement result do not match within the first time period is greater than a first threshold, the terminal device determines that the prediction result of the first model is inaccurate; If the number of times that the beam indicated by the prediction result does not match the beam indicated by the beam measurement result within the first time period is less than or equal to the first threshold, the terminal device determines that the prediction result of the first model is accurate.

5. The method according to claim 3, characterized in that The method further comprises: If a difference between the link quality corresponding to the beam indicated by the prediction result and the link quality corresponding to the beam indicated by the beam measurement result in the first time period is greater than a second threshold, the terminal device determines that the prediction result of the first model is inaccurate; If the difference between the link quality corresponding to the beam indicated by the prediction result in the first time period and the link quality corresponding to the beam indicated by the beam measurement result is less than or equal to a second threshold, the terminal device determines that the prediction result of the first model is accurate.

6. The method according to any one of claims 2 to 5, characterized in that The beam measurement result of the target beam set is determined based on a beam scanning manner.

7. The method according to claim 6, characterized in that The method further comprises: The terminal device receives first configuration information sent by the network device, where the first configuration information is used to configure time-frequency resources for the beam scanning, and / or the first configuration information is used to configure time-frequency resources for sending the beam measurement results.

8. The method according to any one of claims 1 to 7, characterized in that In the inactive state, beam management is performed on the target beam set based on a beam scanning manner.

9. The method according to any one of claims 1 to 8, characterized in that The first model periodically enters the monitoring state.

10. The method according to any one of claims 1 to 8, characterized in that The first model enters the monitoring state non-periodically.

11. The method according to claim 10, characterized in that The method further comprises: The terminal device receives first indication information sent by a network device, where the first indication information is used to instruct the terminal device to trigger the first model to enter the monitoring state; or The terminal device autonomously triggers the first model to enter the monitoring state.

12. The method according to any one of claims 1 to 11, characterized in that The method further comprises: If the monitoring result indicates that the prediction result of the first model is accurate, the terminal device determines that the first model enters the activation state from the monitoring state; and / or If the monitoring result indicates that the prediction result of the first model is inaccurate, the terminal device determines that the first model enters the inactive state from the monitoring state.

13. The method according to any one of claims 1 to 12, characterized in that The method further comprises: If the first model is deployed on the terminal device, the terminal device sends the monitoring result of the first model to the network device; or, If the first model is deployed on a network device, the terminal device receives the monitoring result of the first model sent by the network device.

14. The method according to any one of claims 1 to 13, characterized in that The method further comprises: The terminal device receives second configuration information sent by the network device, where the second configuration information is used to configure the duration of a first time window, where the first time window indicates a time period during which the first model is in the monitoring state; and / or The second configuration information is used to configure the number of monitoring times for monitoring the accuracy of the prediction result of the first model within the first time window.

15. The method according to any one of claims 1 to 14, characterized in that The duration of the first time window of the monitoring state and / or the number of monitoring times are indicated in a preconfigured manner, The first time window indicates the time period during which the first model is in the monitoring state, and the number of monitoring times is the number of monitoring times for monitoring the accuracy of the prediction result of the first model within the first time window.

16. The method according to any one of claims 1 to 15, characterized in that If the first model is deployed on the terminal device, the first model is used to perform beam prediction for the terminal device; or, If the first model is deployed on a network device, the first model is used to perform beam prediction for some or all terminal devices in the cell.

17. A method for beam management, characterized in that include: The network device determines a state of the first model; The first model is used to perform beam prediction for a target beam set, and a state of the first model includes one or more of an activated state, an inactivated state, and a monitoring state; The monitoring state is used to determine the monitoring result of the first model, and the monitoring result is used to indicate the accuracy of the prediction result of the first model.

18. The method according to claim 17, characterized in that In the monitoring state, the accuracy of the prediction result of the first model is determined based on the beam measurement result of the target beam set.

19. The method according to claim 18, characterized in that The accuracy of the prediction results of the first model is determined based on one or more of the following: the number of times the beam indicated by the prediction result does not match the beam indicated by the beam measurement result within the first time period; The difference between the link quality corresponding to the beam indicated by the prediction result and the link quality corresponding to the beam indicated by the beam measurement result in the first time period.

20. The method according to claim 19, characterized in that The method further comprises: If the number of times that the beam indicated by the prediction result and the beam indicated by the beam measurement result do not match within the first time period is greater than a first threshold, the network device determines that the prediction result of the first model is inaccurate; If the number of times that the beam indicated by the prediction result does not match the beam indicated by the beam measurement result within the first time period is less than or equal to the first threshold, the terminal device determines that the prediction result of the first model is accurate.

21. The method according to claim 19, wherein The method further comprises: If a difference between the link quality corresponding to the beam indicated by the prediction result and the link quality corresponding to the beam indicated by the beam measurement result in the first time period is greater than a second threshold, the network device determines that the prediction result of the first model is inaccurate; If the difference between the link quality corresponding to the beam indicated by the prediction result in the first time period and the link quality corresponding to the beam indicated by the beam measurement result is less than or equal to a second threshold, the terminal device determines that the prediction result of the first model is accurate.

22. The method according to any one of claims 17 to 21, characterized in that The beam measurement result of the target beam set is determined based on a beam scanning manner.

23. The method according to claim 22, characterized in that The method further comprises: The network device sends first configuration information to the terminal device, where the first configuration information indicates the time-frequency resources used for the beam scanning, and / or the first configuration information indicates the time-frequency resources used for sending the beam measurement result.

24. The method according to any one of claims 17 to 23, wherein: In the inactive state, beam management is performed on the target beam set based on a beam scanning manner.

25. The method according to any one of claims 17 to 24, characterized in that The first model periodically enters the monitoring state.

26. The method according to any one of claims 17 to 24, characterized in that The first model enters the monitoring state non-periodically.

27. The method according to claim 26, characterized in that The method further comprises: The network device sends first indication information to the terminal device, where the first indication information is used to instruct the terminal device to trigger the first model to enter the monitoring state; or The terminal device autonomously triggers the first model to enter the monitoring state.

28. The method according to any one of claims 17 to 27, characterized in that The method further comprises: If the monitoring result indicates that the prediction result of the first model is accurate, the network device determines that the first model enters the activation state from the monitoring state; and / or If the monitoring result indicates that the prediction result of the first model is inaccurate, the network device determines that the first model enters the inactive state from the monitoring state.

29. The method according to any one of claims 17 to 28, characterized in that The method further comprises: If the first model is deployed on the terminal device, the network device receives the monitoring result of the first model sent by the terminal device; or, If the first model is deployed on a network device, the network device sends the monitoring result of the first model to the terminal device.

30. The method according to any one of claims 17 to 29, wherein: The method further comprises: The network device sends second configuration information to the terminal device, where the second configuration information is used to configure the duration of a first time window, where the first time window indicates a time period during which the first model is in the monitoring state; and / or The second configuration information is used to configure the number of monitoring times for monitoring the accuracy of the prediction result of the first model within the first time window.

31. The method according to any one of claims 17 to 29, wherein: The duration of the first time window of the monitoring state and / or the number of monitoring times are indicated in a preconfigured manner, The first time window indicates the time period during which the first model is in the monitoring state, and the number of monitoring times is the number of monitoring times for monitoring the accuracy of the prediction result of the first model within the first time window.

32. The method according to any one of claims 17 to 31, characterized in that If the first model is deployed in a terminal device, the first model is used to perform beam prediction for the terminal device; or, If the first model is deployed on the network device, the first model is used to perform beam prediction for some or all terminal devices in the cell.

33. A terminal device, characterized in that: include: a processing unit for determining a state of the first model; The first model is used to perform beam prediction for a target beam set, and a state of the first model includes one or more of an activated state, an inactivated state, and a monitoring state; The monitoring state is used to determine the monitoring result of the first model, and the monitoring result is used to indicate the accuracy of the prediction result of the first model.

34. The terminal device according to claim 33, characterized in that In the monitoring state, the accuracy of the prediction result of the first model is determined based on the beam measurement result of the target beam set.

35. The terminal device according to claim 34, characterized in that The accuracy of the prediction results of the first model is determined based on one or more of the following: the number of times the beam indicated by the prediction result does not match the beam indicated by the beam measurement result within the first time period; The difference between the link quality corresponding to the beam indicated by the prediction result and the link quality corresponding to the beam indicated by the beam measurement result in the first time period.

36. The terminal device according to claim 35, characterized in that If the number of times that the beam indicated by the prediction result and the beam indicated by the beam measurement result do not match within the first time period is greater than a first threshold, the processing unit determines that the prediction result of the first model is inaccurate; If the number of times that the beam indicated by the prediction result does not match the beam indicated by the beam measurement result in the first time period is less than or equal to the first threshold, the processing unit determines that the prediction result of the first model is accurate.

37. The terminal device according to claim 35, characterized in that If a difference between the link quality corresponding to the beam indicated by the prediction result and the link quality corresponding to the beam indicated by the beam measurement result in the first time period is greater than a second threshold, the processing unit determines that the prediction result of the first model is inaccurate; If the difference between the link quality corresponding to the beam indicated by the prediction result and the link quality corresponding to the beam indicated by the beam measurement result in the first time period is less than or equal to a second threshold, the processing unit determines that the prediction result of the first model is accurate.

38. The terminal device according to any one of claims 33 to 37, characterized in that: The beam measurement result of the target beam set is determined based on a beam scanning manner.

39. The terminal device according to claim 38, characterized in that The terminal device further includes: A first receiving unit is used to receive first configuration information sent by a network device, where the first configuration information is used to configure time-frequency resources for the beam scanning, and / or the first configuration information is used to configure time-frequency resources for sending the beam measurement results.

40. The terminal device according to any one of claims 33 to 39, characterized in that: In the inactive state, beam management is performed on the target beam set based on a beam scanning manner.

41. The terminal device according to any one of claims 33 to 40, characterized in that: The first model periodically enters the monitoring state.

42. The terminal device according to any one of claims 33 to 40, characterized in that: The first model enters the monitoring state non-periodically.

43. The terminal device according to claim 42, characterized in that The terminal device further includes: A second receiving unit is configured to receive first indication information sent by a network device, where the first indication information is used to instruct the terminal device to trigger the first model to enter the monitoring state; or The terminal device autonomously triggers the first model to enter the monitoring state.

44. The terminal device according to any one of claims 33 to 43, characterized in that: The processing unit is further configured to: If the monitoring result indicates that the prediction result of the first model is accurate, determining that the first model enters the activation state from the monitoring state; and / or If the monitoring result indicates that the prediction result of the first model is inaccurate, it is determined that the first model enters the inactive state from the monitoring state.

45. The terminal device according to any one of claims 33 to 44, characterized in that: a sending unit, configured to send a monitoring result of the first model to a network device if the first model is deployed on the terminal device; or The third receiving unit is configured to receive a monitoring result of the first model sent by the network device if the first model is deployed on the network device.

46. ​​The terminal device according to any one of claims 33 to 45, characterized in that: The terminal device further includes: a fourth receiving unit, configured to receive second configuration information sent by a network device, where the second configuration information is used to configure a duration of a first time window, where the first time window indicates a time period during which the first model is in the monitoring state; and / or The second configuration information is used to configure the number of monitoring times for monitoring the accuracy of the prediction result of the first model within the first time window.

47. The terminal device according to any one of claims 33 to 46, characterized in that: The duration of the first time window of the monitoring state and / or the number of monitoring times are indicated in a preconfigured manner, The first time window indicates the time period during which the first model is in the monitoring state, and the number of monitoring times is the number of monitoring times for monitoring the accuracy of the prediction result of the first model within the first time window.

48. The terminal device according to any one of claims 33 to 47, characterized in that: If the first model is deployed on the terminal device, the first model is used to perform beam prediction for the terminal device; or, If the first model is deployed on a network device, the first model is used to perform beam prediction for some or all terminal devices in the cell.

49. A network device, characterized in that include: a processing unit for determining a state of the first model; The first model is used to perform beam prediction for a target beam set, and a state of the first model includes one or more of an activated state, an inactivated state, and a monitoring state; The monitoring state is used to determine the monitoring result of the first model, and the monitoring result is used to indicate the accuracy of the prediction result of the first model.

50. The network device according to claim 49, wherein: In the monitoring state, the accuracy of the prediction result of the first model is determined based on the beam measurement result of the target beam set.

51. The network device according to claim 50, characterized in that The accuracy of the prediction results of the first model is determined based on one or more of the following: the number of times the beam indicated by the prediction result does not match the beam indicated by the beam measurement result within the first time period; The difference between the link quality corresponding to the beam indicated by the prediction result and the link quality corresponding to the beam indicated by the beam measurement result in the first time period.

52. The network device according to claim 51, wherein: If the number of times that the beam indicated by the prediction result and the beam indicated by the beam measurement result do not match within the first time period is greater than a first threshold, the processing unit determines that the prediction result of the first model is inaccurate; If the number of times that the beam indicated by the prediction result does not match the beam indicated by the beam measurement result in the first time period is less than or equal to the first threshold, the processing unit determines that the prediction result of the first model is accurate.

53. The network device according to claim 51, wherein: If a difference between the link quality corresponding to the beam indicated by the prediction result and the link quality corresponding to the beam indicated by the beam measurement result in the first time period is greater than a second threshold, the processing unit determines that the prediction result of the first model is inaccurate; If the difference between the link quality corresponding to the beam indicated by the prediction result and the link quality corresponding to the beam indicated by the beam measurement result in the first time period is less than or equal to a second threshold, the processing unit determines that the prediction result of the first model is accurate.

54. The network device according to any one of claims 49 to 53, characterized in that: The beam measurement result of the target beam set is determined based on a beam scanning manner.

55. The network device according to claim 54, characterized in that The network device further includes: A first sending unit is used to send first configuration information to a terminal device, where the first configuration information indicates the time-frequency resources used for the beam scanning, and / or the first configuration information indicates the time-frequency resources used for sending the beam measurement result.

56. The network device according to any one of claims 49 to 55, characterized in that: In the inactive state, beam management is performed on the target beam set based on a beam scanning manner.

57. The network device according to any one of claims 49 to 56, characterized in that: The first model periodically enters the monitoring state.

58. The network device according to any one of claims 49 to 57, characterized in that: The first model enters the monitoring state non-periodically.

59. The network device according to claim 58, wherein: The network device further includes: A second sending unit is configured to send first indication information to a terminal device, where the first indication information is used to instruct the terminal device to trigger the first model to enter the monitoring state; or The terminal device autonomously triggers the first model to enter the monitoring state.

60. The network device according to any one of claims 49 to 59, characterized in that: The processing unit is further configured to: If the monitoring result indicates that the prediction result of the first model is accurate, determining that the first model enters the activation state from the monitoring state; and / or If the monitoring result indicates that the prediction result of the first model is inaccurate, it is determined that the first model enters the inactive state from the monitoring state.

61. The network device according to any one of claims 49-60, characterized in that A first receiving unit is configured to receive a monitoring result of the first model sent by the terminal device if the first model is deployed on the terminal device; or The third sending unit is used to send the monitoring result of the first model to the terminal device if the first model is deployed on the network device.

62. The network device according to any one of claims 49 to 61, characterized in that: The network device further includes: a fourth sending unit, configured to send second configuration information to the terminal device, where the second configuration information is used to configure the duration of the first time window, where the first time window indicates a time period during which the first model is in the monitoring state; and / or The second configuration information is used to configure the number of monitoring times for monitoring the accuracy of the prediction result of the first model within the first time window.

63. The network device according to any one of claims 49 to 62, characterized in that: The duration of the first time window of the monitoring state and / or the number of monitoring times are indicated in a preconfigured manner, The first time window indicates the time period during which the first model is in the monitoring state, and the number of monitoring times is the number of monitoring times for monitoring the accuracy of the prediction result of the first model within the first time window.

64. The network device according to any one of claims 49 to 63, characterized in that: If the first model is deployed on a terminal device, the first model is used to perform beam prediction for the terminal device; or, If the first model is deployed on the network device, the first model is used to perform beam prediction for some or all terminal devices in the cell.

65. A terminal device, characterized in that: The terminal comprises a transceiver, a memory and a processor, wherein the memory is used to store a program, and the processor is used to call the program in the memory and control the transceiver to receive or send a signal, so that the terminal executes the method according to any one of claims 1 to 16.

66. A network device, characterized in that The network device comprises a transceiver, a memory and a processor, wherein the memory is used to store a program, and the processor is used to call the program in the memory and control the transceiver to receive or send a signal so that the network device executes the method according to any one of claims 16 to 32.

67. A device, characterized in that The device comprises a processor configured to call a program from a memory so as to enable the device to execute the method according to any one of claims 1 to 32.

68. A chip, characterized in that: The device comprises a processor configured to call a program from a memory so that a device equipped with the chip executes the method according to any one of claims 1 to 32.

69. A computer-readable storage medium, characterized in that A program is stored thereon, and the program causes a computer to execute the method according to any one of claims 1 to 32.

70. A computer program product, characterized in that The method comprises a program for causing a computer to execute the method according to any one of claims 1 to 32.

71. A computer program, characterized in that The computer program causes a computer to execute the method according to any one of claims 1 to 32.

Images