An information prediction method and device

Abstract

The application provides an information prediction method and device. A terminal device obtains first indication information based on a first historical downlink reference signal, and the first indication information indicates signal quality or a channel state of the first historical downlink reference signal. The terminal device inputs the first indication information into a first network model, obtains a first feature vector output by the first network model, and sends the first feature vector to a network device. The network device obtains a second feature vector based on the first feature vector and a second network model, and sends the second feature vector to the terminal device. The terminal device inputs the second feature vector into a third network model, and obtains second indication information corresponding to a downlink reference signal at a next moment predicted by the third network model. In the prediction, the terminal device and the network device cooperatively perform an inference task, fully utilize the computing resources of the network device, and reduce the occupation of the computing resources of the terminal device. Moreover, the terminal device only needs to transmit a low-dimensional feature vector to the network device, and reduces communication overhead.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to an information prediction method and device. Background Art

[0002] With the continuous development of artificial intelligence (AI) and machine learning (ML) technologies, AI / ML has been introduced into wireless communication networks, significantly improving the performance of wireless communication systems. Starting with 3GPP Release 18, the use of AI for wireless network optimization has been proposed, with a focus on key wireless AI use cases such as channel feedback and beam management.

[0003] When using AI / ML technology to predict some information in wireless communication systems, it is necessary to pre-train the relevant network model and then use the trained network model to complete the reasoning task.

[0004] Currently, training or inference tasks are typically performed collaboratively by terminal devices and network devices. However, during this collaborative process, a large amount of data needs to be transmitted between the network devices and the terminal devices, resulting in significant communication overhead. Summary of the Invention

[0005] In order to solve the above problems, the present application provides an information prediction method and device to reduce communication overhead.

[0006] In a first aspect, the present application provides an information prediction method. This method can be performed, for example, by a terminal device, or by a component configured in the terminal device (such as a circuit, chip, or chip system), or by a logic module or software that implements all or part of the terminal device's functions. In this method, the terminal device obtains first indication information based on a first historical downlink reference signal. The first indication information indicates the signal quality of the first historical downlink reference signal or the channel state of a channel corresponding to the first historical downlink reference signal. The terminal device inputs the first indication information into a first network model corresponding to the first indication information, obtains a first eigenvector output by the first network model, and sends the first eigenvector to the network device. The terminal device receives a second eigenvector sent by the network device. The second eigenvector is obtained by the network device based on the first eigenvector and the second network model. The terminal device inputs the second eigenvector into a third network model and obtains second indication information corresponding to a downlink reference signal at a next moment, as predicted by the third network model. The second indication information indicates the signal quality of the downlink reference signal at the next moment or the channel state of the channel corresponding to the downlink reference signal at the next moment. As can be seen, during the information prediction process, the terminal device and the network device collaborate to perform the inference task, fully utilizing the computing resources on the network device side and reducing the computing resources occupied by the terminal device side. Moreover, during the inference process, the terminal device only needs to transmit low-dimensional feature vectors to the network device, reducing communication overhead.

[0007] The parameter scales corresponding to the first network model and the third network model are both smaller than the parameter scale corresponding to the second network model. That is, the two network models with smaller parameter scales are deployed on the terminal device side, and the network model with larger parameter scale is deployed on the network device side, reducing the load on the terminal device side.

[0008] In one possible implementation, if the first historical downlink reference signal is a first historical channel state reference signal, the second indication information includes a channel state and / or a precoding matrix identifier corresponding to the channel state reference signal at the next moment. If the first historical downlink parameter signal is a first historical downlink beam, the second indication information includes a beam quality and / or a beam identifier corresponding to the downlink beam at the next moment, where the beam identifier indicates a beam whose beam quality meets a first preset condition.

[0009] In this embodiment, if the first historical downlink reference signal is a first historical channel state reference signal, the prediction model composed of the first network model, the second network model, and the third network model is used to predict the channel state at the next moment or a precoding matrix identifier that matches the channel state at the next moment. If the first historical downlink reference signal is a first historical downlink beam, the prediction model composed of the first network model, the second network model, and the third network model is used to predict the beam quality of the downlink beam at the next moment and / or a beam identifier whose beam quality meets a first preset condition.

[0010] In one possible implementation, if the second indication information includes a channel state corresponding to a channel state reference signal at a next moment, the method further includes: the terminal device determining a precoding matrix identifier based on the channel state; and the terminal device transmitting the precoding matrix identifier to the network device. In this way, the network device can use the precoding matrix indicated by the precoding matrix identifier to preprocess the transmitted signal, thereby reducing interference during transmission and improving signal transmission quality.

[0011] In one possible implementation, if the second indication information includes the beam quality corresponding to the downlink beam at the next moment, the method further includes: the terminal device determining a beam whose beam quality meets the first preset condition, obtaining a beam identifier for the beam; and the terminal device transmitting the beam identifier to the network device. In this implementation, the terminal device may transmit the beam identifier of the optimal receive beam to the network device, so that the network device can transmit signals using the optimal receive beam, thereby improving signal transmission stability.

[0012] The first historical downlink beam and the downlink beam at the next moment are time-domain downlink beams, that is, the time-domain downlink beam can be predicted.

[0013] In one possible implementation, the training process of the above network model includes:

[0014] (1) The parameters of the network model on the terminal device side and the network model on the network device side are updated

[0015] The terminal device obtains a first training sample set, which includes third indication information corresponding to the second historical downlink reference signal and a first sample label, the third indication information indicates the signal quality of the second historical downlink reference signal or the channel state of the channel corresponding to the second historical downlink reference signal, and the first sample label refers to the signal quality of the next historical downlink reference signal corresponding to the second historical downlink reference signal or the channel state of the channel corresponding to the next historical downlink reference signal; the terminal device inputs the third indication information into the first network model to obtain a third eigenvector output by the first network model; the terminal device sends the third eigenvector to the network device; the terminal device receives a fourth eigenvector sent by the network device, which is a fourth eigenvector generated by the network device based on the third eigenvector. and the second network model; the terminal device inputs the fourth eigenvector into the third network model to obtain a first prediction result output by the third network model, where the first prediction result refers to the signal quality of the next historical downlink reference signal or the channel state of the channel corresponding to the next historical downlink reference signal; the terminal device determines the first gradient corresponding to the third network model according to the first sample label and the first prediction result, and updates the parameters of the third network model using the first gradient; the terminal device sends the first gradient to the network device, so that the network device updates the parameters of the second network model using the first gradient; the terminal device obtains the second gradient corresponding to the first network model, and updates the parameters of the first network model using the second gradient; and repeats the above training process until the second preset condition is met.

[0016] Through this training method, only low-dimensional feature vectors need to be transmitted between the terminal device and the network device, avoiding the transmission of large numbers of training samples over the air interface and saving network overhead. Furthermore, the first and third network models deployed on the terminal device side have simple structures, reducing training overhead on the terminal device side. Furthermore, when training the second network model on the network side, feature vectors sent by multiple terminal devices can be used for joint training, improving the generalization of the second network model and accelerating model convergence.

[0017] The terminal device obtains the second gradient corresponding to the first network model, which may include the following implementation methods:

[0018] One approach is to determine the second gradient corresponding to the first network model based on the first sample label, the first prediction result, and the parameters of the first network model. Specifically, the terminal device calculates the loss function of the first network model using the first sample label and the first prediction result, and then derives the loss function with respect to the parameters of the first network model to obtain the second gradient corresponding to the first network model.

[0019] Another is that the terminal device receives a second gradient corresponding to the first network model sent by the network device, where the second gradient is determined by the network device based on the first sample label, the first prediction result, and the parameters of the first network model.

[0020] In order to use the resources of the network device to calculate the second gradient, the terminal device needs to send the parameters of the first network model, the first sample label, the first prediction result, and the loss function of the first network model to the network device.

[0021] (2) The second network model on the network device side has been trained, and only the two network models on the terminal device side need to be trained

[0022] The terminal device obtains a second training sample set, which includes fourth indication information corresponding to the third historical downlink reference signal and a second sample identifier, the fourth indication information indicates the quality of the third historical downlink reference signal or the channel state of the channel corresponding to the third historical downlink reference signal, and the second sample label indicates the signal quality of the next historical downlink reference signal corresponding to the third historical downlink reference signal or the channel state of the channel corresponding to the next historical downlink reference signal; the terminal device inputs the fourth indication information into the first network model to obtain the fifth eigenvector output by the first network model; the terminal device sends the fifth eigenvector to the network device; the terminal device receives the sixth eigenvector sent by the network device, and the sixth eigenvector The eigenvector is obtained by the network device based on the fifth eigenvector and the second network model; the terminal device inputs the sixth eigenvector into the third network model to obtain a second prediction result output by the third network model, where the second prediction result refers to the signal quality of the next historical downlink reference signal or the channel state of the channel corresponding to the next historical downlink reference signal; the terminal device determines the third gradient corresponding to the third network model based on the second sample label and the second prediction result, and uses the third gradient to update the parameters of the third network model; the terminal device obtains the fourth gradient corresponding to the first network model, and uses the fourth gradient to update the parameters of the first network model; and repeats the above training process until the second preset condition is met.

[0023] It should be noted that the second implementation method can be applied to a variety of scenarios. One is that the two network models on the terminal device side have been trained, but the prediction accuracy of the third network model is less than the preset threshold, then the second training method is used to perform incremental training on the two network models on the terminal device side; the other is when a new terminal device is connected to the above-mentioned network device, the trained second network model can be used to train the first network model and the third network model on the new terminal device side.

[0024] Among them, the specific implementation of the terminal device obtaining the fourth gradient can refer to the relevant description of the terminal device obtaining the second gradient.

[0025] In a second aspect, an information prediction method is provided. The method can be performed by a network device, or by a component configured in the network device (such as a circuit, chip, or chip system), or by a logic module or software capable of implementing all or part of the network device's functions. In this method, the network device receives a first feature vector sent by a terminal device, the first feature vector being obtained by the terminal device based on first indication information corresponding to a first historical downlink reference signal and a first network model corresponding to the first indication information; the network device inputs a second feature vector into a second network model to obtain a second feature vector output by the second network model; and the network device sends the second feature vector to the terminal device, so that the terminal device predicts second indication information corresponding to a downlink reference signal at a next moment based on the second feature vector and a third network model. The second indication information indicates the signal quality of the downlink reference signal at the next moment or the channel state of the channel corresponding to the downlink reference signal at the next moment.

[0026] The second aspect is the implementation on the network device side corresponding to the first aspect. The explanation, supplement and description of the beneficial effects of the first aspect are also applicable to the second aspect and will not be repeated here.

[0027] In a possible implementation, the parameter scale corresponding to the first network model and the parameter scale corresponding to the third network model are both smaller than the parameter scale corresponding to the second network model.

[0028] In one possible implementation, the first historical downlink reference signal is a first historical channel state reference signal, and the second indication information includes the channel state and / or precoding matrix identifier corresponding to the downlink channel state reference signal at the next moment; or, the first historical downlink reference signal is a first historical downlink beam, and the second indication information includes the beam quality and / or beam identifier corresponding to the downlink beam at the next moment, and the beam identifier indicates a beam whose beam quality meets the first preset condition.

[0029] In a possible implementation, if the second indication information includes the channel state corresponding to the channel state reference signal at the next moment, the method further includes: the network device receives a precoding matrix identifier sent by the terminal device, where the precoding matrix is ​​determined by the terminal device according to the channel state.

[0030] In one possible implementation, if the second indication information includes the beam quality corresponding to the downlink beam at the next moment, the method also includes: the network device receives the beam identifier sent by the terminal device, and the beam identifier indicates the beam that meets the first preset condition selected by the terminal device based on the beam quality.

[0031] In a possible implementation, the first historical downlink beam and the downlink beam at the next moment are time-domain downlink beams.

[0032] In one possible implementation, the method further includes: the network device receives a third eigenvector sent by the terminal device, the third eigenvector is obtained by the terminal device based on third indication information and the first network model, the third indication information indicating the quality of the second historical downlink reference signal or the channel state of the channel corresponding to the second historical downlink reference signal; the network device inputs the third eigenvector into the second network model to obtain a fourth eigenvector output by the second network model; the network device sends the fourth eigenvector to the terminal device, so that the terminal device obtains a first prediction result based on the fourth eigenvector and the third network model, the first prediction result refers to the signal quality of the next historical downlink reference signal corresponding to the second historical downlink reference signal or the channel state of the channel corresponding to the next historical downlink reference signal; the network device receives a first gradient sent by the terminal device, the first gradient is a gradient for the third network model determined by the terminal device based on the first prediction result and the first sample label, the first sample label refers to the signal quality of the next historical downlink reference signal or the channel state of the channel corresponding to the next historical downlink reference signal; the network device updates the parameters of the second network model based on the first gradient; and repeats the above training process to indicate that the second preset condition is met.

[0033] In one possible implementation, the method further includes: the network device determines a second gradient based on the first sample label, the first prediction result, and the parameters of the first network model; the network device sends the second gradient to the terminal device, so that the terminal device updates the parameters of the first network model according to the second gradient.

[0034] In a possible implementation, the method further includes: the network device receiving the first sample label, the first prediction result, and the parameters of the first network model sent by the terminal device.

[0035] In one possible implementation, the method also includes: the network device receives a fifth eigenvector sent by the terminal device, the fifth eigenvector is obtained by the terminal device based on the fourth indication information and the first network model, and the fourth indication information indicates the quality of the third historical downlink reference signal or the channel state of the channel corresponding to the third historical downlink reference signal; the network device inputs the fifth eigenvector into the second network model to obtain the sixth eigenvector output by the second network model; the network device sends the sixth eigenvector to the terminal device, so that the terminal device obtains a second prediction result based on the sixth eigenvector and the third network model, and the second prediction result refers to the signal quality of the next historical downlink reference signal corresponding to the third historical downlink reference signal or the channel state of the channel corresponding to the next historical downlink reference signal; repeats the above training process to indicate that the second preset condition is met.

[0036] According to a third aspect, a communication device is provided, comprising a processing module and a transceiver module. The processing module is configured to obtain first indication information based on a first historical downlink reference signal, the first indication information indicating the signal quality of the first historical downlink reference signal or the channel state of a channel corresponding to the first historical downlink reference signal; input the first indication information into a first network model corresponding to the first indication information to obtain a first eigenvector output by the first network model; the transceiver module is configured to send the first eigenvector to a network device; the transceiver module is further configured to receive a second eigenvector sent by the network device, the second eigenvector being obtained by the network device based on the first eigenvector and the second network model; and the processing module is further configured to input the second eigenvector into a third network model to obtain second indication information corresponding to a downlink reference signal at a next moment predicted by the third network model, the second indication information indicating the signal quality of the downlink reference signal at the next moment or the channel state of the channel corresponding to the downlink reference signal at the next moment.

[0037] In a fourth aspect, a communication device is provided, comprising a transceiver module and a processing module. The transceiver module is configured to receive a first eigenvector sent by a terminal device, the first eigenvector being obtained by the terminal device based on first indication information corresponding to a first historical downlink reference signal and a first network model corresponding to the first indication information; the processing module is configured to input the first eigenvector into a second network model to obtain a second eigenvector output by the second network model; and the transceiver module is further configured to send a second eigenvector to the terminal device, so that the terminal device predicts second indication information corresponding to a downlink reference signal at a next moment based on the second eigenvector and a third network model, the second indication information indicating the signal quality of the downlink reference signal at the next moment or the channel state of the channel corresponding to the downlink reference signal at the next moment.

[0038] The third and fourth aspects are the device-side implementations corresponding to the first and second aspects. The explanations, supplements and descriptions of the beneficial effects of the first and second aspects are also applicable to the third and fourth aspects and will not be repeated here.

[0039] In a fifth aspect, a communication device is provided, comprising a processor. The processor is coupled to a memory and can be configured to execute instructions or data in the memory to implement the method of the first aspect. Optionally, the device further comprises a memory. Optionally, the device further comprises a communication interface, the processor being coupled to the communication interface.

[0040] In one implementation, the communication interface may be a transceiver, or an input / output interface.

[0041] In another implementation, the device is a chip configured in a terminal device. When the device is a chip configured in a terminal device, the communication interface may be an input / output interface.

[0042] In a sixth aspect, a communication device is provided, comprising a processor. The processor is coupled to a memory and can be configured to execute instructions or data in the memory to implement the method of the second aspect. Optionally, the device further comprises a memory. Optionally, the device further comprises a communication interface, the processor being coupled to the communication interface.

[0043] In one implementation, the communication interface may be a transceiver, or an input / output interface.

[0044] In another implementation, the device is a chip configured in a network device. When the device is a chip configured in a network device, the communication interface may be an input / output interface.

[0045] In a seventh aspect, a processor is provided, comprising: an input circuit, an output circuit, and a processing circuit, wherein the processing circuit is configured to receive a signal through the input circuit and transmit a signal through the output circuit, so that the processor executes the method in any aspect.

[0046] In a specific implementation, the processor may be one or more chips, the input circuit may be an input pin, the output circuit may be an output pin, and the processing circuit may be a transistor, a gate circuit, a trigger, or various logic circuits. The input signal received by the input circuit may be, for example, but not limited to, received and input by a receiver, and the signal output by the output circuit may be, for example, but not limited to, output to and transmitted by a transmitter. The input circuit and the output circuit may be the same circuit, which functions as an input circuit and an output circuit at different times. The embodiments of the present application do not limit the specific implementation of the processor and various circuits.

[0047] In an eighth aspect, a communication device is provided, comprising a processor and a memory. The processor is configured to read instructions stored in the memory and receive signals via a receiver and transmit signals via a transmitter to execute the method of any of the above aspects.

[0048] Optionally, there are one or more processors and one or more memories.

[0049] In a ninth aspect, a computer program product is provided, comprising: a computer program (also referred to as code, or instruction), which enables a computer to execute the method in any one of the above aspects when the computer program is executed.

[0050] In a tenth aspect, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program (also referred to as code, or instructions) which, when executed on a computer, enables the computer to execute the method in any one of the above aspects.

[0051] In an eleventh aspect, embodiments of the present application provide a chip system comprising one or more processors configured to retrieve and execute instructions stored in a memory, thereby executing the methods described in each of the above aspects. The chip system may be comprised of a chip or may include a chip and other discrete components.

[0052] Among them, the chip system may include an input circuit or interface for sending information or data, and an output circuit or interface for receiving information or data.

[0053] In a twelfth aspect, a communication system is provided, comprising the aforementioned terminal device and network device. Optionally, the system may further comprise other devices that communicate with the terminal device and / or the network device. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 A communication system structure diagram provided in an embodiment of the present application;

[0055] Figure 2 An interactive diagram of a training prediction mode provided in an embodiment of the present application;

[0056] Figure 3 A schematic diagram of a model training scenario provided in an embodiment of the present application;

[0057] Figure 4 An interaction diagram of a model adaptation process provided in an embodiment of the present application;

[0058] Figure 5 A schematic diagram of a model adaptation scenario provided in an embodiment of the present application;

[0059] Figure 6 An interactive diagram of an information prediction process provided by an embodiment of the present application;

[0060] Figure 7 A schematic diagram of an information prediction scenario provided in an embodiment of the present application;

[0061] Figure 8 A structural diagram of a communication device provided in an embodiment of the present application;

[0062] Figure 9 A structural diagram of another communication device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0063] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application. The terms used in the following embodiments are only for the purpose of describing specific embodiments and are not intended to be limiting of the present application. As used in the specification and appended claims of the present application, the singular expressions "one", "a kind of", "said", "above", "the" and "this" are intended to also include expressions such as "one or more", unless there is a clear contrary indication in the context. It should also be understood that in the embodiments of the present application, "one or more" refers to one, two or more; "and / or" describes the association relationship of associated objects, indicating that three relationships may exist; for example, A and / or B can represent: the existence of A alone, the existence of A and B at the same time, and the existence of B alone, where A and B can be singular or plural. The character " / " generally indicates that the related objects before and after are in an "or" relationship.

[0064] References to "one embodiment" or "some embodiments" in this specification mean that a particular feature, structure, or characteristic described in conjunction with that embodiment is included in one or more embodiments of the present application. Thus, phrases such as "in one embodiment," "in some embodiments," "in other embodiments," and "in yet other embodiments" appearing in various places in this specification do not necessarily refer to the same embodiment, but rather mean "one or more but not all embodiments," unless otherwise specifically emphasized. The terms "including," "comprising," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0065] The "multiple" involved in the embodiments of the present application means greater than or equal to two. It should be noted that in the description of the embodiments of the present application, the words "first" and "second" are only used for the purpose of distinguishing the description and cannot be understood as indicating or implying relative importance or order.

[0066] AI / ML has been introduced into wireless communication networks and has been widely used in numerous air interface technology scenarios. Common applications include channel feedback and beam management, such as predicting channel state information (CSI) based on AI models or adjusting beam width and direction based on AI models. When using models to perform AI tasks, such as training or inference, significant communication overhead is incurred between terminal devices and network equipment.

[0067] Based on this, this application proposes a model structure that divides the prediction model into three parts: an input network, a core network, and an output network. The input and output networks are deployed on the terminal device side, while the core network is deployed on the network device side. The core network is the main component of this model, and the input and output networks can be lightweight linear networks. This allows only low-dimensional features to be transmitted between the terminal device and the network device when performing the aforementioned tasks, eliminating the need to transmit raw data (such as training samples or input data), reducing communication overhead.

[0068] Among them, when conducting model training, the core network can be trained using training samples from multiple terminal devices to improve the generalization of the core network.

[0069] Additionally, after model training is complete, model adaptation can be performed for newly added devices or devices with predicted performance degradation. Unlike model training, model adaptation does not require training the core network on the network device side; it only requires training the input and output networks on the device side, enabling rapid model adaptation.

[0070] The terminal devices in the embodiments of the present application include various devices with wireless communication functions, which can be used to connect people, objects, machines, etc. The terminal devices can be widely used in various scenarios, such as: cellular communication, D2D, V2X, peer to peer (P2P), M2M, MTC, IoT, virtual reality (VR), augmented reality (AR), industrial control, autonomous driving, telemedicine, smart grid, smart furniture, smart office, smart wearable, smart transportation, smart city drones, robots, remote sensing, passive sensing, positioning, navigation and tracking, autonomous delivery, etc.

[0071] The terminal device can be a terminal in any of the above scenarios, such as an MTC terminal, an IoT terminal, etc. The terminal device may be a user equipment (UE) of the third generation partnership project (3GPP) standard, a terminal, a fixed device, a mobile station device or a mobile device, a subscriber unit, a handheld device, a vehicle-mounted device, a wearable device, a cellular phone, a smart phone, a session initialization protocol (SIP) phone, a wireless data card, a personal digital assistant (PDA), a computer, a tablet computer, a notebook computer, a wireless modem, a handheld device (handset), a laptop computer, a computer with wireless transceiver function, a smart book, a vehicle, a satellite, a global positioning system (GPS) device, a target tracking device, an aircraft (such as a drone, a helicopter, a multi-copter, a quadcopter, or an airplane), a ship, a remote control device, a smart home device, an industrial device, or a device built into the above-mentioned device (such as a communication module, a modem or a chip in the above-mentioned device), or other processing devices connected to a wireless modem. For the sake of convenience of description, the terminal device will be described below by taking the terminal or UE as an example.

[0072] It should be understood that in some scenarios, a UE can also be used to act as a base station. For example, a UE can act as a scheduling entity that provides sidelink signals between UEs in scenarios such as V2X, D2D, or P2P.

[0073] In the embodiments of the present application, the device for implementing the function of the terminal device can be the terminal device, or it can be a device that can support the terminal device to implement the function, such as a chip system or chip, which can be installed in the terminal device. In the embodiments of the present application, the chip system can be composed of a chip, or it can include a chip and other discrete devices.

[0074] 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. The 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 reception point (TRP), transmitting point (TP), master station, auxiliary station, multi-standard wireless (motor slide retainer, 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.

[0075] A base station may 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 may also refer to a communication module, a modem, or a chip used to be provided in the aforementioned device or apparatus. A base station may also be a mobile switching center and a device that performs base station functions in D2D, V2X, and M2M communications, a network-side device in a 6G network, or a device that performs base station functions in future communication systems. A base station may support networks with the same or different access technologies. The embodiments of this application do not limit the specific technology and specific device form adopted by the network equipment.

[0076] 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.

[0077] In the embodiments of the present application, the device for implementing the function of the network device can be the network device, or it can be a device that can support the network device to implement the function, such as a chip system or chip, which can be installed in the network device. In the embodiments of the present application, the chip system can be composed of a chip, or it can include a chip and other discrete devices.

[0078] Network devices and terminal devices can be deployed on land, including indoors or outdoors, handheld or vehicle-mounted; they can also be deployed on the water surface; they can also be deployed on aircraft, balloons and satellites in the air. The embodiments of this application do not limit the scenarios in which network devices and terminal devices are located. In addition, terminal devices and network devices can be hardware devices, or they can be software functions running on dedicated hardware, software functions running on general-purpose hardware, such as virtualization functions instantiated on a platform (e.g., a cloud platform), or entities including dedicated or general-purpose hardware devices and software functions. This application does not limit the specific forms of terminal devices and network devices.

[0079] See also Figure 1 , which is a schematic diagram of the communication system provided by this application. The wireless communication system includes an access network 100 and a core network 200. Optionally, the communication system may also include the Internet 300. The wireless access network 100 may be a next-generation (e.g., 6G or higher) wireless access network, or a traditional (e.g., 5G, 4G, 3G, or 2G) wireless access network.

[0080] One or more terminal devices 120 (only one of which is shown in the figure) can be connected to each other or to one or more network devices in the access network 100. Figure 1 In the example, the access network 100 includes two network devices 110 .

[0081] Figure 1 The above is for illustrative purposes only. The wireless communication system may also include other devices, such as core network devices, wireless relay devices, and / or wireless backhaul devices. In actual applications, the wireless communication system may simultaneously include multiple network devices (also called access network devices) and multiple terminal devices, without limitation. A network device may simultaneously serve one or more terminal devices. A terminal device may also simultaneously access one or more network devices.

[0082] When using a model to perform reasoning tasks, the model needs to be trained in advance. The following will first introduce how to train the generated model in conjunction with a specific embodiment. From the above, it can be seen that this application splits the model for performing information prediction into three parts, namely the input network, the core network, and the output network. Such a model structure can make full use of the computing power resources on the network device side to complete training, while transmitting low-dimensional feature vectors through the air interface to replace the training data set, reducing training overhead.

[0083] See also Figure 2 , which is an interactive diagram of a model training method provided in an embodiment of the present application. Specifically, the method includes:

[0084] S201: The terminal device obtains a first training sample set, where the first training sample set includes third indication information corresponding to a second historical downlink reference signal and a first sample label.

[0085] The third indication information indicates the signal quality of the second historical downlink reference signal, or indicates the channel state of the channel corresponding to the second historical downlink reference signal, and the first sample label refers to the signal quality of the next historical downlink reference signal corresponding to the second historical downlink reference signal, or the first sample label refers to the channel state of the channel corresponding to the next historical downlink reference signal. It should be noted that the third indication information matches the first sample label. If the third indication information indicates signal quality, the first sample label also indicates signal quality; if the third indication information indicates channel state, the first sample label also indicates channel state.

[0086] For example, define the first training sample set as: ,in t is the sample number, C represents a complex set, is the training set size, and UE i In the sample t The sampled input value and the corresponding true value. Among them, the dimension F x =N sample *K*N r *N t ，F y =N pre *K* N t ，N sample is the sampling window size (the number of channel tensors sampled at a time), N pre is the prediction window size (for example, the number of channel precoding matrices predicted at one time), K is the number of carrier frequencies, N r is the number of receiving antennas on the UE side, N t is the number of transmitting antennas on the network side. In addition, the sample t Corresponding UE i The input network and output network parameter sets are and , the core network parameter set is , and there are

[0087] The corresponding model functions of the three are recorded as .

[0088] Typically, the second historical downlink reference signal includes downlink reference signals sent by the network device at multiple historical moments, and the transmission time of the next historical downlink reference signal is later than the latest transmission time of the second historical downlink reference signal. For example, the second historical downlink reference signal includes the first k historical downlink reference signals, and the next historical downlink reference signal is the k+1th historical downlink reference signal.

[0089] In this embodiment, if the model is used to predict the channel state at a future time, the second historical downlink reference signal is a channel state reference signal (CSI-RS); if the model is used to predict the beam quality at a future time, the second historical downlink reference signal is a downlink beam, for example, a time domain downlink beam.

[0090] It should be noted that the model in this embodiment usually only predicts one type of information. For different prediction targets, multiple models need to be trained and generated, but the structure of each model is the same, that is, it includes a core network deployed on the network device side and an input network and output network deployed on the terminal device side.

[0091] S202: The terminal device inputs the third indication information into the first network model to obtain a third feature vector output by the first network model.

[0092] The first network model is the input network deployed on the terminal device side, and the third feature vector is extracted from the third indication information through the first network model.

[0093] For example, UE i The feedforward output characteristics of the side input network are:

[0094] (1-1)

[0095] Among them, the function value of the above formula is the third eigenvector, F in Identifies the dimension of the feature vector output by the input network, F in <F x , F in <F y .

[0096] S203: The terminal device sends the third eigenvector to the network device, and correspondingly, the network device receives the third eigenvector.

[0097] S204: The network device inputs the third eigenvector into the second network model to obtain a fourth eigenvector output by the second network model.

[0098] The second network model is a core network deployed on the network device side, and the fourth eigenvector is extracted from the third eigenvector through the second network model. The dimension of the fourth eigenvector is lower than the dimension of the third eigenvector.

[0099] For example, the feedforward output characteristics of the core network on the network side are:

[0100] (1-2)

[0101] Among them, the function value of the above formula is the fourth eigenvector, F out Represents the dimension of the feature vector output by the core network, F out <F x ，F out <F y .

[0102] S205: The network device sends the fourth eigenvector to the terminal device, and correspondingly, the terminal device receives the fourth eigenvector.

[0103] S206: The terminal device inputs the fourth eigenvector into the third network model to obtain a first prediction result output by the third network model.

[0104] In this embodiment, the terminal device uses the fourth eigenvector as the input of the third network model, and outputs the first prediction result through the third network model.

[0105] The first prediction result refers to the signal quality of the next historical downlink reference signal corresponding to the second historical downlink reference signal or the channel state of the channel corresponding to the next historical downlink reference signal. The representation of the first prediction result matches the representation of the first sample label. If the first sample label is signal quality, the first prediction result is signal quality; if the first sample label is channel state, the first prediction result is channel state.

[0106] Among them, the third network model corresponds to the output network deployed on the terminal device side, and the prediction result is obtained through the third network model.

[0107] For example, UE i The feedforward output of the output network is:

[0108] (1-3)

[0109] Among them, the function value of the above formula is the first prediction result.

[0110] S207: The terminal device determines a first gradient corresponding to the third network model according to the first sample label and the first prediction result, and updates parameters of the third network model using the first gradient.

[0111] In this embodiment, after obtaining the first prediction result corresponding to the third indication information, the terminal device calculates the loss value between the first sample label and the first prediction result based on the loss function corresponding to the third network model, and derives the parameters of the third network model to obtain a first gradient. Simultaneously, the parameters of the third network model are updated using the first gradient.

[0112] The terminal device calculates the first gradient corresponding to the third network model, which can be referred to the following formula:

[0113] (1-4)

[0114] in, l out,i ( x ) is the loss function of the third network model, and the function value of the above formula is the first gradient.

[0115] The terminal device updates the parameters of the third network model using the first gradient, which can be seen in the following formula:

[0116] (1-5)

[0117] Among them, the function value of the above formula is the updated parameter, is the learning rate of the third network model at training sample t.

[0118] S208: The terminal device sends the first gradient to the network device, and correspondingly, the network device receives the first gradient.

[0119] S209: The network device updates the parameters of the second network model according to the first gradient.

[0120] In this embodiment, if multiple terminal devices are jointly training a model with the network device, the network device receives the first gradients sent by the multiple terminal devices and fuses the multiple first gradients using a gradient fusion function to obtain the gradient corresponding to the second network model, and uses this gradient to update the parameters of the second network model. If a single terminal device is training a model with the network model, the network device directly uses the first gradient to update the parameters of the second network model.

[0121] It should be noted that if multiple terminal devices are jointly training a model with the network device, each terminal device will be deployed with the first and third network models, and all terminal devices will share the second network model on the network device. This training method improves the generalizability of the second network model and accelerates model convergence.

[0122] For example, the network device uses the gradient fusion function A(x) to calculate the gradient corresponding to the second network model:

[0123] (1-6)

[0124] The network model uses the fused gradient to update the parameters of the second network model:

[0125] (1-7)

[0126] The function value of the above formula is the parameter of the updated second network model. is the learning rate of the second network model at training sample t.

[0127] S210: The terminal device obtains a second gradient corresponding to the first network model, and uses the second gradient to update parameters of the first network model.

[0128] In this embodiment, the terminal device also needs to obtain the second gradient corresponding to the first network model, so as to use the second gradient to update the parameters of the first network model and then continue the next round of training process.

[0129] The terminal device can obtain the second gradient in the following ways:

[0130] In one implementation, the terminal device determines the second gradient corresponding to the first network model based on the first sample label, the first prediction result, and the parameters of the first network model. In this implementation, the terminal device calculates the loss value between the first sample label and the first prediction result based on the loss function corresponding to the first network model, and derives the parameters of the first network model to obtain the second gradient.

[0131] Alternatively, the terminal device sends the first sample label, the first prediction result, and the parameters of the first network model to the network device, and the network device determines the second gradient based on the first sample label, the first prediction result, and the parameters of the first network model. In this implementation, the network device calculates the loss between the first sample label and the first prediction result based on the loss function corresponding to the first network model, and derives the parameters of the first network model to obtain the second gradient.

[0132] For example, this is achieved by the following formula:

[0133] (1-8)

[0134] Among them, the function value of the above formula is the second gradient, l in is the loss function of the first network model.

[0135] The terminal device updates the parameters of the first network model using the following formula:

[0136] (1-9)

[0137] The function value of the above formula is the parameter of the updated first network model. is the learning rate of the first network model at training sample t.

[0138] After updating the parameters of each network model, the terminal device restarts the next round of training until the second preset condition is met. The second preset condition can be set according to the actual application, for example, the second preset condition is N iterations, or the second preset condition is that the loss function value of the third network model is less than a preset threshold.

[0139] To understand the model training process, see Figure 3 The model training diagram shown in Figure 3 As shown, the data transmitted between the terminal device and the network device consists of low-dimensional feature vectors and gradients, which reduces communication overhead compared to transmitting training samples. Furthermore, multiple terminal devices can participate in training, improving the convergence speed and generalization of the second network model. Furthermore, the first and third network models deployed by the terminal device have simple structures, which can reduce training overhead on the terminal device side.

[0140] After model training is complete, adaptation training is performed for newly added UEs or UEs predicted to have degraded performance. Unlike model training, this process does not modify the core network on the network equipment side; only the input and output networks on the UE to be adapted are trained. Because the input and output networks are lightweight, the adaptation process requires a much smaller dataset than the training phase, facilitating rapid model adaptation and deployment.

[0141] See also Figure 4 The interactive diagram of the model training method specifically includes:

[0142] S401: The terminal device obtains a second training sample set, where the second training sample set includes fourth indication information corresponding to a third historical downlink reference signal and a second sample label.

[0143] The fourth indication information indicates the signal quality of the third historical downlink reference signal, or indicates the channel state of the channel corresponding to the third historical downlink reference signal, and the second sample label refers to the signal quality of the next historical downlink reference signal corresponding to the third historical downlink reference signal, or the second sample label refers to the channel state of the channel corresponding to the next historical downlink reference signal. It should be noted that the fourth indication information matches the second sample label. If the fourth indication information indicates signal quality, the second sample label also indicates signal quality; if the fourth indication information indicates channel state, the second sample label also indicates channel state.

[0144] In this embodiment, the second training sample set is defined as , where the meaning of each parameter can refer to the above definition of the first training sample set, but the sample size of the second training sample is much smaller than that of the first training sample, that is, .

[0145] Typically, the third historical downlink reference signal includes downlink reference signals sent by the network device at multiple historical moments, and the transmission time of the next historical downlink reference signal is later than the latest transmission time of the third historical downlink reference signal. For example, the third historical downlink reference signal includes the first n historical downlink reference signals, and the next historical downlink reference signal is the n+1th historical downlink reference signal.

[0146] In the present application, the signal type of the third historical downlink reference signal is consistent with the signal type of the second historical downlink reference signal. If the model is used to predict the channel state at a future time, the second historical downlink reference signal and the third historical downlink reference signal are both CSI-RS; if the model is used to predict the beam quality at a future time, the second historical downlink reference signal and the third historical downlink reference signal are both downlink beams, for example, both are time domain downlink beams.

[0147] S402: The terminal device inputs the fourth indication information into the first network model to obtain a fifth eigenvector output by the first network model.

[0148] In this embodiment, the terminal device uses the fourth indication information as input of the first network model, and extracts the fifth feature vector from the fourth indication information through the first network model.

[0149] For example, UE i The feedforward output characteristics of the side input network are:

[0150] (2-1)

[0151] The function value of the above formula is the fifth eigenvector.

[0152] S403: The terminal device sends the fifth eigenvector to the network device, and correspondingly, the network device receives the fifth eigenvector.

[0153] S404: The network device inputs the fifth eigenvector into the second network model to obtain a sixth eigenvector output by the second network model.

[0154] In this embodiment, the network device uses the fifth eigenvector as input to the second network model and extracts a sixth eigenvector from the fifth eigenvector using the second network model, wherein the dimension of the sixth eigenvector is lower than that of the fifth eigenvector.

[0155] For example, the feedforward output characteristics of the core network on the network side are:

[0156] (2-2)

[0157] The function value of the above formula is the sixth eigenvector.

[0158] S405: The network device sends the sixth eigenvector to the terminal device, and correspondingly, the terminal device receives the sixth eigenvector.

[0159] S406: The terminal device inputs the sixth eigenvector into the third network model to obtain a second prediction result output by the third network model.

[0160] In this embodiment, the terminal device uses the sixth eigenvector as input of the third network model, and outputs the second prediction result through the third network model.

[0161] The second prediction result refers to the signal quality of the next historical downlink reference signal corresponding to the third historical downlink reference signal or the channel state of the channel corresponding to the next historical downlink reference signal. The representation of the second prediction result matches the representation of the second sample label. If the second sample label is signal quality, the second prediction result is signal quality; if the second sample label is channel state, the second prediction result is channel state.

[0162] For example, the feedforward output of the UEi output network is:

[0163] (2-3)

[0164] Among them, the function value of the above formula is the second prediction result.

[0165] S407: The terminal device determines a third gradient corresponding to the third network model according to the second sample label and the second prediction result, and updates the parameters of the third network model using the third gradient.

[0166] In this embodiment, after obtaining the second prediction result corresponding to the fourth indication information, the terminal device calculates the loss value between the second sample label and the second prediction result based on the loss function corresponding to the third network model, and derives the parameters of the third network model to obtain a third gradient. Simultaneously, the parameters of the third network model are updated using the third gradient.

[0167] The terminal device calculates the third gradient corresponding to the third network model, which can be seen from the following formula:

[0168] (2-4)

[0169] The terminal device updates the parameters of the third network model using the third gradient, which can be seen in the following formula:

[0170] (2-5)

[0171] in, is the learning rate of the third network model at training sample t.

[0172] S408: The terminal device obtains a fourth gradient corresponding to the first network model, and uses the fourth gradient to update parameters of the first network model.

[0173] In this embodiment, the terminal device also needs to obtain the fourth gradient corresponding to the first network model, so as to update the parameters of the first network model using the fourth gradient, and then continue the next round of training process.

[0174] The terminal device can obtain the fourth gradient in the following ways:

[0175] In one implementation, the terminal device determines the fourth gradient corresponding to the first network model based on the second sample label, the second prediction result, and the parameters of the first network model. In this implementation, the terminal device calculates the loss value between the second sample label and the second prediction result based on the loss function corresponding to the first network model, and derives the parameters of the first network model to obtain the fourth gradient.

[0176] Alternatively, the terminal device sends the second sample label, the second prediction result, and the parameters of the first network model to the network device, and the network device determines the fourth gradient based on the second sample label, the second prediction result, and the parameters of the first network model. In this implementation, the network device calculates the loss value between the second sample label and the second prediction result based on the loss function corresponding to the first network model, and derives the parameters of the first network model to obtain the fourth gradient.

[0177] For example, this is achieved by the following formula:

[0178] (2-6)

[0179] in, l in ( x ) is the loss function of the first network model.

[0180] The terminal device updates the parameters of the first network model using the following formula:

[0181] (2-7)

[0182] After updating the parameters of the first network model and the second network model, the terminal device restarts the next round of training until the second preset condition is met. The second preset condition can be set according to the actual application, for example, the second preset condition is N iterations, or the second preset condition is that the loss function value of the third network model is less than a preset threshold.

[0183] To understand the above training, see Figure 5 , which is a model adaptation framework diagram provided by an embodiment of the present application, such as Figure 5 As shown in the figure, the trained core network is used to perform model adaptation for newly added terminal devices or those with predicted performance degradation. This adaptation algorithm not only eliminates the need to update the parameters of the core network model, but only updates the parameters of the input and output networks of the adapted terminal device, significantly reducing computational complexity. Furthermore, because the local input and output networks of the terminal device are lightweight models, only a small amount of adaptation data is required to complete the adaptation.

[0184] After completing model training or adaptation, the above model can be used to perform reasoning tasks, which will be described in detail below with reference to specific embodiments.

[0185] See also Figure 6 , which is an information prediction method provided by an embodiment of the present application, such as Figure 6 As shown, the method includes:

[0186] S601: The terminal device obtains first indication information based on a first historical downlink reference signal.

[0187] In this embodiment, after receiving the first historical downlink reference signal sent by the network device, the terminal device obtains the first indication information corresponding to the first historical downlink reference signal. The first indication information indicates the signal quality of the first historical downlink reference signal or the channel state of the channel corresponding to the first historical downlink reference signal. The signal quality can be described by signal-to-noise ratio, bit error rate, received signal strength indication, reference signal received power, etc., so the first indication information can include but is not limited to one or more of signal-to-noise ratio, bit error rate, received signal strength indication, and reference signal received power; the channel state can be described by information such as channel quality indicator, precoding matrix indicator PMI, precoding type indicator PTI, etc. Therefore, the first indication information can include but is not limited to one or more of channel quality indicator, precoding matrix indicator PMI, precoding type indicator PTI, etc.

[0188] S602: The terminal device inputs the first indication information into the first network model corresponding to the first indication information, and obtains a first feature vector output by the first network model.

[0189] After obtaining the first indication information, the terminal device uses the first indication information as input to the first network model, and extracts the features of the first indication information, that is, the first feature vector, through the first network model.

[0190] For example, the output of the UE input network is:

[0191] (3-1)

[0192] The function value corresponding to the above formula is the first eigenvector, and the function value is transmitted to the network device.

[0193] S603: The terminal device sends a first feature vector to the network device, and correspondingly, the network device receives the first feature vector.

[0194] S604: The network device inputs the first feature vector into the second network model to obtain a second feature vector output by the second network model.

[0195] In this embodiment, the network device uses the first feature vector as input to the second network model, and extracts the first feature vector again through the second network model to obtain a second feature vector, wherein the dimension of the second feature vector is lower than that of the first feature vector.

[0196] For example, the output of the core network device on the network device side is:

[0197] (3-2)

[0198] The function value corresponding to the above formula is the second eigenvector, and the function value is transmitted back to the terminal device.

[0199] S605: The network device sends the second feature vector to the terminal device, and correspondingly, the terminal device receives the second feature vector.

[0200] S606: The terminal device inputs the second eigenvector into the third network model to obtain second indication information corresponding to the downlink reference signal at the next moment predicted by the third network model.

[0201] In this embodiment, the network device uses the second eigenvector as input to a third network model, and further extracts the second eigenvector using the third network model to obtain second indication information. The second indication information indicates the signal quality of a downlink reference signal at the next moment or the channel status of a channel corresponding to the downlink reference signal at the next moment. Typically, the second indication information is of the same type as the first indication information. If the first indication information is signal quality, the second indication information is signal quality; if the first indication information is channel status, the second indication information is channel status.

[0202] For example, the output of the UE-side output network is:

[0203] (3-3)

[0204] The function value of the above formula is the second indication information.

[0205] In a preferred implementation, the parameter scale corresponding to the first network model and the parameter scale corresponding to the third network model are both smaller than the parameter scale corresponding to the second network model.

[0206] In a preferred implementation, the first historical downlink reference signal is a first historical channel state reference signal, and the second indication information includes the channel state and / or precoding matrix identifier corresponding to the channel state reference signal at a next moment. Alternatively, the first historical downlink reference signal is a first historical downlink beam, and the second indication information includes the beam quality and / or beam identifier corresponding to the downlink beam at the next moment, where the beam identifier indicates a beam whose beam quality meets a first preset condition.

[0207] The content included in the second indication information is related to the training process. For example, if during the training process, the first sample label or the second sample label only includes the channel state or the precoding matrix identifier, the second indication information also only includes the channel state or the precoding matrix identifier; if the first sample label or the second sample label both includes the channel state and the precoding matrix identifier, the second indication information also includes the channel state or the precoding matrix identifier. Similarly, if the first sample label or the second sample label only includes the beam quality or the beam identifier, the second indication information also only includes the beam instruction or the beam identifier; if the first sample label or the second sample label both includes the beam quality and the beam identifier, the second indication information also includes the beam instruction and the beam identifier.

[0208] When the second indication information only includes a channel state corresponding to a channel state reference signal at a next moment, the method further includes: the terminal device determining a precoding matrix identifier based on the channel state; and the terminal device transmitting the precoding matrix identifier to the network device. Thus, the network device encodes the downlink signal using the precoding matrix indicated by the precoding matrix identifier, so that the terminal device can receive the downlink signal.

[0209] When the second indication information includes only the beam quality corresponding to the downlink beam at the next moment, the method further includes: the terminal device determining a beam whose beam quality meets the first preset condition and obtaining a beam identifier for the beam; and the terminal device transmitting the beam identifier to the network device. In this way, the network device uses the beam identifier to indicate that the indicated beam is transmitted in a specific direction. The first historical downlink beam and the downlink beam at the next moment are time-domain downlink beams.

[0210] To understand the reasoning process, see Figure 7 The inference framework diagram shown in the figure takes the second indication information as the channel state, matches the current best PMI after obtaining the channel state, and sends it to the network device as the basis for the network device to adjust the precoding parameters. Figure 7 As can be seen, during the inference process, the terminal device and the network device collaborate to perform the inference task, fully utilizing the computing resources on the network device side and reducing the computing resources occupied by the terminal device side. Moreover, during the inference process, the terminal device only needs to transmit low-dimensional feature vectors to the network device, reducing communication overhead.

[0211] It should be understood that Figures 2 to 7 The flowcharts or scenario diagrams shown are only for ease of understanding and are not intended to limit the embodiments of the present application to the examples shown in the diagrams. In fact, those skilled in the art will Figures 2 to 7 The examples in can be equivalently transformed to obtain more implementation methods.

[0212] Combined with the above Figures 2 to 7 , describes in detail the communication method provided by the embodiment of the present application. Figure 8 and Figure 9 The device embodiments of the present application are described in detail. It should be understood that the device embodiments of the present application can execute the various methods of the aforementioned embodiments of the present application, that is, the specific working processes of the following various products can refer to the corresponding processes in the aforementioned method embodiments.

[0213] In each of the above embodiments, the terminal device may perform some or all of the steps in each embodiment; the network device may perform some or all of the steps in each embodiment. These steps or operations are merely examples, and the embodiments of the present application may also perform other operations or variations of various operations. In addition, the various steps may be performed in a different order as presented in the embodiments, and it is possible that not all of the operations in the embodiments of the present application need to be performed. Moreover, the size of the sequence number of each step 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.

[0214] Figure 8 : is a schematic block diagram of a communication device provided in an embodiment of the present application. Figure 8 As shown, the communication device 800 may include a communication module 820. The communication module 820 can implement corresponding communication functions, which may be internal communication functions of the communication device 800 or communication functions between the communication device 800 and other devices. Optionally, the communication module 820 may also be referred to as a communication interface or a transceiver module. Optionally, the communication device 800 also includes a processing module 810. The processing module 810 can implement corresponding processing functions.

[0215] Optionally, the communication device 800 further includes a storage module, which can be used to store instructions and / or data; the processing module 810 can read the instructions and / or data in the storage module, so that the communication device 800 implements the aforementioned method embodiment.

[0216] In one possible design, the communication device 800 may correspond to the terminal device in the above method embodiments, or a component configured in the terminal device (such as a circuit, chip, or chip system). The communication device 800 can be used to execute the steps or processes executed by the terminal device in any of the above method embodiments.

[0217] Illustratively, the processing module 810 is configured to obtain first indication information based on a first historical downlink reference signal, where the first indication information indicates a signal quality of the first historical downlink reference signal or a channel state of a channel corresponding to the first historical downlink reference signal; input the first indication information into a first network model corresponding to the first indication information, and obtain a first eigenvector output by the first network model;

[0218] The communication module 820 is configured to send a first feature vector to a network device; and receive a second feature vector sent by the network device, where the second feature vector is obtained by the network device based on the first feature vector and a second network model.

[0219] The processing module 810 is also used to input the second eigenvector into the third network model to obtain the second indication information corresponding to the downlink reference signal at the next moment predicted by the third network model, where the second indication information indicates the signal quality of the downlink reference signal at the next moment or the channel state of the channel corresponding to the downlink reference signal at the next moment.

[0220] In some implementations, the parameter scale corresponding to the first network model and the parameter scale corresponding to the third network model are both smaller than the parameter scale corresponding to the second network model.

[0221] In some embodiments, the first historical downlink reference signal is a first historical channel state reference signal, and the second indication information includes the channel state and / or precoding matrix identifier corresponding to the channel state reference signal at the next moment; or, the first historical downlink reference signal is a first historical downlink beam, and the second indication information includes the beam quality and / or beam identifier corresponding to the downlink beam at the next moment, and the beam identifier indicates a beam whose beam quality meets a first preset condition.

[0222] In some implementations, if the second indication information includes a channel state corresponding to a channel state reference signal at a next moment, the processing module 810 is further configured to determine a precoding matrix identifier according to the channel state;

[0223] The communication module 820 is further configured to send the precoding matrix identifier to the network device.

[0224] In some embodiments, if the second indication information includes the beam quality corresponding to the downlink beam at the next moment, the processing module 810 is also used to determine the beam whose beam quality meets the first preset condition and obtain the beam identifier of the beam; the communication module 820 is also used to send the beam identifier to the network device.

[0225] In some implementations, the first historical downlink beam and the next moment downlink beam are time domain downlink beams.

[0226] In some embodiments, the processing module 810 is further configured to obtain a first training sample set, where the first training sample set includes third indication information corresponding to a second historical downlink reference signal and a first sample label, the third indication information indicating a signal quality of the second historical downlink reference signal or a channel state of a channel corresponding to the second historical downlink reference signal, and the first sample label indicating a signal quality of a next historical downlink reference signal corresponding to the second historical downlink reference signal or a channel state of a channel corresponding to the next historical downlink reference signal; input the third indication information into the first network model to obtain a third eigenvector output by the first network model;

[0227] The communication module 820 is further configured to send the third eigenvector to the network device; receive a fourth eigenvector sent by the network device, where the fourth eigenvector is obtained by the network device based on the third eigenvector and the second network model;

[0228] The processing module 810 is further configured to input the fourth eigenvector into the third network model to obtain a first prediction result output by the third network model, where the first prediction result refers to the signal quality of the next historical downlink reference signal or the channel state of the channel corresponding to the next historical downlink reference signal; determine a first gradient corresponding to the third network model based on the first sample label and the first prediction result, and update parameters of the third network model using the first gradient;

[0229] The communication module 820 is further configured to send the first gradient to the network device, so that the network device updates the parameters of the second network model using the first gradient;

[0230] The processing module 810 is further configured to obtain a second gradient corresponding to the first network model, and update the parameters of the first network model using the second gradient;

[0231] Repeat the above training process until the second preset condition is met.

[0232] In some embodiments, the processing module 810 is specifically configured to determine a second gradient corresponding to the first network model according to the first sample label, the first prediction result, and the parameters of the first network model.

[0233] In some embodiments, the communication module 820 is further used to receive a second gradient corresponding to the first network model sent by the network device, where the second gradient is determined by the network device based on the first sample label, the first prediction result, and the parameters of the first network model.

[0234] In some implementations, the communication module 820 is further configured to send the parameters of the first network model, the first sample label, and the first prediction result to the network device.

[0235] In some embodiments, the processing module 810 is further configured to obtain a second training sample set, where the second training sample set includes fourth indication information corresponding to a third historical downlink reference signal and a second sample label, where the fourth indication information indicates the quality of the third historical downlink reference signal or the channel state of a channel corresponding to the third historical downlink reference signal, and the second sample label indicates the signal quality of a next historical downlink reference signal corresponding to the third historical downlink reference signal or the channel state of a channel corresponding to the next historical downlink reference signal; input the fourth indication information into the first network model to obtain a fifth eigenvector output by the first network model;

[0236] The communication module 820 is further configured to send the fifth eigenvector to the network device; receive a sixth eigenvector sent by the network device, where the sixth eigenvector is obtained by the network device based on the fifth eigenvector and the second network model;

[0237] The processing module 810 is further configured to input the sixth eigenvector into the third network model to obtain a second prediction result output by the third network model, where the second prediction result refers to the signal quality of the next historical downlink reference signal or the channel state of the channel corresponding to the next historical downlink reference signal; determine a third gradient corresponding to the third network model based on the second sample label and the second prediction result, and update the parameters of the third network model using the third gradient;

[0238] The processing module 810 is further configured to obtain a fourth gradient corresponding to the first network model, and update parameters of the first network model using the fourth gradient;

[0239] Repeat the above training process until the second preset condition is met.

[0240] The above is only an example, and for detailed steps or processes, please refer to the description of the aforementioned embodiments.

[0241] In one possible design, the communication device 800 may correspond to the network device in the above method embodiments, or a component configured in the network device (such as a circuit, chip, or chip system). The communication device 800 can be used to execute the steps or processes executed by the network device in any of the above method embodiments.

[0242] Illustratively, the communication module 820 is configured to receive a first feature vector sent by a terminal device, where the first feature vector is obtained by the terminal device based on first indication information corresponding to a first historical downlink reference signal and a first network model corresponding to the first indication information;

[0243] The processing module 810 is configured to input the first feature vector into a second network model to obtain a second feature vector output by the second network model;

[0244] The communication module 820 is also used to send the second eigenvector to the terminal device, so that the terminal device predicts the second indication information corresponding to the downlink reference signal at the next moment based on the second eigenvector and the third network model, and the second indication information indicates the signal quality of the downlink reference signal at the next moment or the channel state of the channel corresponding to the downlink reference signal at the next moment.

[0245] In a possible implementation, the parameter scale corresponding to the first network model and the parameter scale corresponding to the third network model are both smaller than the parameter scale corresponding to the second network model.

[0246] In one possible implementation, the first historical downlink reference signal is a first historical channel state reference signal, and the second indication information includes the channel state and / or precoding matrix identifier corresponding to the downlink channel state reference signal at the next moment; or, the first historical downlink reference signal is a first historical downlink beam, and the second indication information includes the beam quality and / or beam identifier corresponding to the downlink beam at the next moment, and the beam identifier indicates a beam whose beam quality meets a first preset condition.

[0247] In some embodiments, if the second indication information includes the channel state corresponding to the channel state reference signal at the next moment, the communication module 820 is further used to receive a precoding matrix identifier sent by the terminal device, and the precoding matrix is ​​determined by the terminal device according to the channel state.

[0248] In some embodiments, if the second indication information includes the beam quality corresponding to the downlink beam at the next moment, the communication module 820 is also used to receive the beam identifier sent by the terminal device, and the beam identifier indicates the beam that meets the first preset condition and is selected by the terminal device based on the beam quality.

[0249] In some implementations, the first historical downlink beam and the next moment downlink beam are time domain downlink beams.

[0250] In some embodiments, the communication module 820 is further configured to receive a third eigenvector sent by the terminal device, where the third eigenvector is obtained by the terminal device based on third indication information and the first network model, and the third indication information indicates a quality of a second historical downlink reference signal or a channel state of a channel corresponding to the second historical downlink reference signal;

[0251] The processing module 810 is further configured to input the third eigenvector into the second network model to obtain a fourth eigenvector output by the second network model;

[0252] The communication module 820 is further configured to send the fourth eigenvector to the terminal device, so that the terminal device obtains a first prediction result based on the fourth eigenvector and the third network model, where the first prediction result refers to the signal quality of the next historical downlink reference signal corresponding to the second historical downlink reference signal or the channel state of the channel corresponding to the next historical downlink reference signal; receive a first gradient sent by the terminal device, where the first gradient is a gradient for the third network model determined by the terminal device based on the first prediction result and a first sample label, where the first sample label refers to the signal quality of the next historical downlink reference signal or the channel state of the channel corresponding to the next historical downlink reference signal;

[0253] The processing module 810 is further configured to update the parameters of the second network model according to the first gradient;

[0254] The above training process is repeatedly executed, indicating that the second preset condition is met.

[0255] In some embodiments, the processing module 810 is further configured to determine a second gradient based on the first sample label, the first prediction result, and the parameters of the first network model;

[0256] The communication module 820 is further configured to send the second gradient to the terminal device, so that the terminal device updates the parameters of the first network model according to the second gradient.

[0257] In some embodiments, the communication module 820 is further configured to receive the first sample label, the first prediction result, and the parameters of the first network model sent by the terminal device.

[0258] In some embodiments, the communication module 820 is further configured to receive a fifth eigenvector sent by the terminal device, where the fifth eigenvector is obtained by the terminal device according to fourth indication information and the first network model, where the fourth indication information indicates a quality of a third historical downlink reference signal or a channel state of a channel corresponding to the third historical downlink reference signal;

[0259] The processing module 810 is further configured to input the fifth eigenvector into the second network model to obtain a sixth eigenvector output by the second network model;

[0260] The communication module 820 is further configured to send the sixth eigenvector to the terminal device, so that the terminal device obtains a second prediction result based on the sixth eigenvector and the third network model, where the second prediction result refers to the signal quality of the next historical downlink reference signal corresponding to the third historical downlink reference signal or the channel state of the channel corresponding to the next historical downlink reference signal;

[0261] The above training process is repeatedly executed, indicating that the second preset condition is met.

[0262] The above is only an example, and for detailed steps or processes, please refer to the description of the aforementioned embodiments.

[0263] Figure 9 is another schematic block diagram of a communication device 900 provided in an embodiment of the present application. The communication device 900 may be a chip, chip system, or processor, etc., that implements the above-described method in a terminal device or network device. The communication device 900 may be used to implement the method described in the above-described method embodiment. For details, please refer to the description of the above-described method embodiment.

[0264] like Figure 9 As shown, the communication device 900 may include one or more processors 910, which may also be referred to as processing units or processing modules, and may implement certain control functions. The processor 910 may be a general-purpose processor or a dedicated processor, for example, a baseband processor or a central processing unit. The baseband processor may be used to process communication protocols and communication data, and the central processing unit may be used to control the communication device 900 (e.g., base station, baseband chip, user, user chip), execute software programs, and process software program data.

[0265] In an optional design, the processor 910 may also store instructions and / or data, which can be executed by the processor 910 to enable the communication device 900 to perform the method described in the above method embodiment.

[0266] In another optional design, the communication device 900 may include a communication interface 920 for implementing receiving and transmitting functions. For example, the communication interface 920 may be a transceiver circuit, an interface, an interface circuit, or a transceiver. The transceiver circuit, interface, interface circuit, or transceiver for implementing the receiving and transmitting functions may be separate or integrated. The transceiver circuit, interface, interface circuit, or transceiver may be used for reading and writing code / data, or the transceiver circuit, interface, interface circuit, or transceiver may be used for transmitting or delivering signals.

[0267] Optionally, the communication device 900 may include one or more memories 930, which may store instructions. The instructions may be executed on the processor 910, causing the communication device 900 to perform the method described in the above method embodiment. Optionally, the memory 930 may also store data. Optionally, the processor 910 may also store instructions and / or data. The processor 910 and memory 930 may be provided separately or integrated together.

[0268] It should be understood that, in one possible design, each step in the method embodiment provided in the present application can be completed by an integrated logic circuit of the hardware in the processor or by instructions in the form of software. The steps of the method disclosed in conjunction with the embodiments of the present application can be directly embodied as being executed by a hardware processor, or can be executed by a combination of hardware and software modules in the processor. The software module can be located in a storage medium mature in the art, such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory, and the processor reads the information in the memory and completes the steps of the above method in combination with its hardware. To avoid repetition, it will not be described in detail here.

[0269] In one implementation, the communication device 900 may correspond to the terminal device in the above-mentioned method embodiment and may be used to execute the various steps and / or processes performed by the terminal device in the above-mentioned method embodiment. The processor 910 may be used to execute instructions stored in the memory 930, and when the processor 910 executes the instructions stored in the memory, the processor 910 is used to execute the various steps and / or processes of the above-mentioned method embodiment corresponding to the terminal device.

[0270] In another implementation, the communication device 900 may correspond to the network device in the above-mentioned method embodiment, and may be used to execute the various steps and / or processes performed by the network device in the above-mentioned method embodiment. The processor 910 may be used to execute instructions stored in the memory 930, and when the processor 910 executes the instructions stored in the memory, the processor 910 is used to execute the various steps and / or processes of the above-mentioned method embodiment corresponding to the network device.

[0271] It should be understood that the processing device may be one or more chips. For example, the processing device may be a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), a system on chip (SoC), a central processor unit (CPU), a network processor (NP), a digital signal processor (DSP), a microcontroller unit (MCU), a programmable logic device (PLD), or other integrated chips.

[0272] It is understood that the memory in the embodiments of the present application may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. The non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of RAM are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), and direct RAM bus RAM (DR RAM). It should be noted that the memory of the systems and methods described herein is intended to include, but is not limited to, these and any other suitable types of memory.

[0273] Based on the methods provided in the embodiments of the present application, the present application also provides a chip system, which includes one or more processors configured to retrieve and execute instructions stored in a memory, thereby executing the methods of the embodiments of the present application. The chip system can be composed of a chip or can include a chip and other discrete devices.

[0274] Among them, the chip system may include an input circuit or interface for sending information or data, and an output circuit or interface for receiving information or data.

[0275] According to the method provided in the embodiment of the present application, the present application also provides a communication system, which includes the aforementioned terminal device and network device.

[0276] According to the method provided in the embodiments of the present application, the present application also provides a computer program product, which includes: computer program code, when the computer program code is run on a computer, enables the computer to execute the various steps or processes executed by the terminal device and network device in any of the aforementioned method embodiments.

[0277] According to the method provided in the embodiments of the present application, the present application also provides a computer-readable storage medium, which stores program code. When the program code runs on a computer, the computer executes the various steps or processes performed by the terminal device and network device in any of the aforementioned method embodiments.

[0278] The computer-readable storage medium may be the aforementioned volatile memory or non-volatile memory, or may include both volatile memory and non-volatile memory.

[0279] In the embodiments of this application, each term and English abbreviation is provided for convenience of description and shall not constitute any limitation to this application. This application does not exclude the possibility of defining other terms that can achieve the same or similar functions in existing or future agreements.

[0280] In the above embodiments, all or part of the embodiments may be implemented using software, hardware, firmware, or any combination thereof. When implemented using software, all or part of the embodiments may be implemented in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer instructions are loaded and executed on a computer, the processes or functions described in the embodiments of the present application are generated in whole or in part.

[0281] 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.

[0282] It should be understood that in the various embodiments of the present application, the size of the serial number of each process 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.

[0283] In short, the above description is only a preferred embodiment of the technical solution of this application and is not intended to limit the scope of protection of this application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of this application shall be included in the scope of protection of this application.

Claims

1. An information prediction method, characterized in that: The method comprises: The terminal device obtains first indication information based on the first historical downlink reference signal, where the first indication information indicates a signal quality of the first historical downlink reference signal or a channel state of a channel corresponding to the first historical downlink reference signal; The terminal device inputs the first indication information into a first network model corresponding to the first indication information, and obtains a first feature vector output by the first network model; The terminal device sends the first feature vector to the network device; The terminal device receives a second feature vector sent by the network device, where the second feature vector is obtained by the network device based on the first feature vector and a second network model; The terminal device inputs the second eigenvector into a third network model to obtain second indication information corresponding to a downlink reference signal at a next moment predicted by the third network model, where the second indication information indicates a signal quality of the downlink reference signal at the next moment or a channel state of a channel corresponding to the downlink reference signal at the next moment; The first network model, the second network model and the third network model constitute a prediction model, and the prediction model is used to predict the channel state or signal quality at a future moment.

2. The method according to claim 1, characterized in that The parameter scale corresponding to the first network model and the parameter scale corresponding to the third network model are both smaller than the parameter scale corresponding to the second network model.

3. The method according to claim 1, characterized in that The first historical downlink reference signal is a first historical channel state reference signal, and the second indication information includes a channel state and / or a precoding matrix identifier corresponding to a channel state reference signal at a next moment; Alternatively, the first historical downlink reference signal is a first historical downlink beam, and the second indication information includes the beam quality and / or beam identifier corresponding to the downlink beam at the next moment, and the beam identifier indicates a beam whose beam quality meets a first preset condition.

4. The method according to claim 3, characterized in that If the second indication information includes a channel state corresponding to a channel state reference signal at a next moment, the method further includes: The terminal device determines a precoding matrix identifier according to the channel state; The terminal device sends the precoding matrix identifier to the network device.

5. The method according to claim 3, characterized in that If the second indication information includes a beam quality corresponding to a downlink beam at a next moment, the method further includes: The terminal device determines a beam whose beam quality satisfies the first preset condition, and obtains a beam identifier of the beam; The terminal device sends the beam identifier to the network device.

6. The method according to claim 3 or 5, characterized in that The first historical downlink beam and the next moment downlink beam are time domain downlink beams.

7. The method according to claim 1, characterized in that The method further comprises: The terminal device obtains a first training sample set, where the first training sample set includes third indication information corresponding to a second historical downlink reference signal and a first sample label, where the third indication information indicates a signal quality of the second historical downlink reference signal or a channel state of a channel corresponding to the second historical downlink reference signal, and the first sample label refers to a signal quality of a next historical downlink reference signal corresponding to the second historical downlink reference signal or a channel state of a channel corresponding to the next historical downlink reference signal; The terminal device inputs the third indication information into the first network model to obtain a third feature vector output by the first network model; The terminal device sends the third feature vector to the network device; The terminal device receives a fourth eigenvector sent by the network device, where the fourth eigenvector is obtained by the network device based on the third eigenvector and the second network model; The terminal device inputs the fourth eigenvector into the third network model to obtain a first prediction result output by the third network model, where the first prediction result refers to the signal quality of the next historical downlink reference signal or the channel state of the channel corresponding to the next historical downlink reference signal; The terminal device determines a first gradient corresponding to the third network model according to the first sample label and the first prediction result, and updates parameters of the third network model using the first gradient; The terminal device sends the first gradient to the network device, so that the network device updates the parameters of the second network model by using the first gradient; The terminal device obtains a second gradient corresponding to the first network model, and updates parameters of the first network model using the second gradient; Repeat the steps of the terminal device acquiring the first training sample set and subsequent steps until a second preset condition is met.

8. The method according to claim 7, characterized in that The terminal device obtains a second gradient corresponding to the first network model, including: The terminal device determines a second gradient corresponding to the first network model according to the first sample label, the first prediction result, and the parameters of the first network model.

9. The method according to claim 7, characterized in that The terminal device obtains a second gradient corresponding to the first network model, including: The terminal device receives a second gradient corresponding to the first network model sent by the network device, where the second gradient is determined by the network device according to the first sample label, the first prediction result, and the parameters of the first network model.

10. The method according to claim 9, characterized in that The method further comprises: The terminal device sends the parameters of the first network model, the first sample label, and the first prediction result to the network device.

11. The method according to any one of claims 1-5, 7-10, characterized in that: The method further comprises: The terminal device obtains a second training sample set, where the second training sample set includes fourth indication information corresponding to a third historical downlink reference signal and a second sample label, where the fourth indication information indicates a quality of the third historical downlink reference signal or a channel state of a channel corresponding to the third historical downlink reference signal, and the second sample label indicates a signal quality of a next historical downlink reference signal corresponding to the third historical downlink reference signal or a channel state of a channel corresponding to the next historical downlink reference signal; The terminal device inputs the fourth indication information into the first network model to obtain a fifth eigenvector output by the first network model; The terminal device sends the fifth eigenvector to the network device; The terminal device receives a sixth eigenvector sent by the network device, where the sixth eigenvector is obtained by the network device based on the fifth eigenvector and the second network model; The terminal device inputs the sixth eigenvector into the third network model to obtain a second prediction result output by the third network model, where the second prediction result refers to the signal quality of the next historical downlink reference signal or the channel state of the channel corresponding to the next historical downlink reference signal; The terminal device determines a third gradient corresponding to the third network model according to the second sample label and the second prediction result, and updates parameters of the third network model using the third gradient; The terminal device obtains a fourth gradient corresponding to the first network model, and updates parameters of the first network model using the fourth gradient; Repeat the steps of the terminal device acquiring the second training sample set and subsequent steps until the second preset condition is met.

12. An information prediction method, characterized in that: The method comprises: The network device receives a first feature vector sent by the terminal device, where the first feature vector is obtained by the terminal device based on first indication information corresponding to a first historical downlink reference signal and a first network model corresponding to the first indication information; The network device inputs the first feature vector into a second network model to obtain a second feature vector output by the second network model; The network device sends the second eigenvector to the terminal device, so that the terminal device predicts second indication information corresponding to a downlink reference signal at a next moment based on the second eigenvector and a third network model, where the second indication information indicates a signal quality of the downlink reference signal at the next moment or a channel state of a channel corresponding to the downlink reference signal at the next moment; The first network model, the second network model and the third network model constitute a prediction model, and the prediction model is used to predict the channel state or signal quality at a future moment.

13. The method according to claim 12, characterized in that The parameter scale corresponding to the first network model and the parameter scale corresponding to the third network model are both smaller than the parameter scale corresponding to the second network model.

14. The method according to claim 12, characterized in that The first historical downlink reference signal is a first historical channel state reference signal, and the second indication information includes a channel state and / or a precoding matrix identifier corresponding to a downlink channel state reference signal at a next moment; Alternatively, the first historical downlink reference signal is a first historical downlink beam, and the second indication information includes the beam quality and / or beam identifier corresponding to the downlink beam at the next moment, and the beam identifier indicates a beam whose beam quality meets a first preset condition.

15. The method according to claim 14, characterized in that If the second indication information includes a channel state corresponding to a channel state reference signal at a next moment, the method further includes: The network device receives a precoding matrix identifier sent by the terminal device, where the precoding matrix is ​​determined by the terminal device according to the channel state.

16. The method according to claim 14, characterized in that If the second indication information includes a beam quality corresponding to a downlink beam at a next moment, the method further includes: The network device receives the beam identifier sent by the terminal device, where the beam identifier indicates the beam that meets the first preset condition and is selected by the terminal device based on the beam quality.

17. The method according to claim 14 or 16, characterized in that The first historical downlink beam and the next moment downlink beam are time domain downlink beams.

18. The method according to claim 12, characterized in that The method further comprises: The network device receives a third eigenvector sent by the terminal device, where the third eigenvector is obtained by the terminal device according to third indication information and the first network model, and the third indication information indicates a quality of a second historical downlink reference signal or a channel state of a channel corresponding to the second historical downlink reference signal; The network device inputs the third eigenvector into the second network model to obtain a fourth eigenvector output by the second network model; The network device sends the fourth eigenvector to the terminal device, so that the terminal device obtains a first prediction result based on the fourth eigenvector and the third network model, where the first prediction result refers to a signal quality of a next historical downlink reference signal corresponding to the second historical downlink reference signal or a channel state of a channel corresponding to the next historical downlink reference signal; The network device receives, by the terminal device, a first gradient sent by the terminal device, where the first gradient is a gradient for the third network model determined by the terminal device according to the first prediction result and a first sample label, where the first sample label refers to a signal quality of a next historical downlink reference signal or a channel state of a channel corresponding to the next historical downlink reference signal; The network device updates the parameters of the second network model according to the first gradient; Repeat the steps of the network device receiving the third feature vector sent by the terminal device and subsequent steps until a second preset condition is met.

19. The method according to claim 18, characterized in that The method further comprises: The network device determines a second gradient according to the first sample label, the first prediction result, and the parameters of the first network model; The network device sends the second gradient to the terminal device, so that the terminal device updates the parameters of the first network model according to the second gradient.

20. The method according to claim 19, characterized in that The method further comprises: The network device receives the first sample label, the first prediction result, and the parameters of the first network model sent by the terminal device.

21. The method according to any one of claims 12-16, 18-20, characterized in that: The method further comprises: The network device receives a fifth eigenvector sent by the terminal device, where the fifth eigenvector is obtained by the terminal device according to fourth indication information and the first network model, where the fourth indication information indicates a quality of a third historical downlink reference signal or a channel state of a channel corresponding to the third historical downlink reference signal; The network device inputs the fifth eigenvector into the second network model to obtain a sixth eigenvector output by the second network model; The network device sends the sixth eigenvector to the terminal device, so that the terminal device obtains a second prediction result based on the sixth eigenvector and the third network model, where the second prediction result refers to a signal quality of a next historical downlink reference signal corresponding to the third historical downlink reference signal or a channel state of a channel corresponding to the next historical downlink reference signal; Repeat the steps of receiving, by the network device, the fifth eigenvector sent by the terminal device and subsequent steps until a second preset condition is met.

22. A communication device, characterized in that: The device comprises at least one processor coupled to a memory, wherein the memory stores a program or instruction, and the processor executes the program or instruction so that the device is used to perform the information prediction method according to any one of claims 1 to 21.

23. A computer-readable storage medium, characterized in that A computer program or instruction is stored thereon, and when the computer program or instruction is executed, the computer is caused to perform the information prediction method according to any one of claims 1 to 21.

24. A communication system, characterized in that: Comprising the communication device as claimed in claim 22.

25. A chip system, characterized in that: The chip system includes one or more processors, which are used to call and execute instructions stored in the memory from the memory, so that the method as claimed in any one of claims 1 to 21 is executed.

Images