Method and apparatus for wireless communication

Abstract

A method and apparatus for wireless communication are provided. The method comprises: a first terminal device determines a first association ID, the first association ID being used for CSI prediction; the first terminal device sends a reasoning result of the CSI prediction to a network device; wherein the input parameter of the CSI prediction is determined according to the first association ID, and / or the reasoning result is determined according to the first association ID, and / or the update of a first model corresponding to the CSI prediction is related to the first association ID.

Description

Technical Field

[0001] The present application relates to the field of communication technology, and more particularly, to a method and apparatus for wireless communication. Background Art

[0002] In communication systems, channel state information (CSI) feedback helps network devices optimize transmission parameters to improve system performance. Introducing artificial intelligence (AI) technology into the CSI feedback process enables more efficient CSI prediction and compression. Improving AI-based CSI prediction accuracy is a technical challenge that needs to be addressed. Summary of the Invention

[0003] The present application provides a method and apparatus for wireless communication. The following describes various aspects of the embodiments of the present application.

[0004] In a first aspect, a method for wireless communication is provided, including: a first terminal device determines a first association identifier (identity, ID), where the first association ID is used for CSI prediction; the first terminal device sends an inference result of the CSI prediction to a network device; wherein an input parameter of the CSI prediction is determined according to the first association ID, and / or the inference result is determined according to the first association ID, and / or an update of a first model corresponding to the CSI prediction is related to the first association ID.

[0005] In a second aspect, a method for wireless communication is provided, including: a network device receives an inference result of a CSI prediction from a first terminal device; wherein a first association ID determined by the first terminal device is used for the CSI prediction; an input parameter of the CSI prediction is determined according to the first association ID, and / or the inference result is determined according to the first association ID, and / or the update of a first model corresponding to the CSI prediction is related to the first association ID.

[0006] According to a third aspect, a device for wireless communication is provided, which is a first terminal device, and includes: a determination unit for determining a first association ID, wherein the first association ID is used for CSI prediction; a sending unit for sending an inference result of the CSI prediction to a network device; wherein the input parameters of the CSI prediction are determined according to the first association ID, and / or the inference result is determined according to the first association ID, and / or the update of the first model corresponding to the CSI prediction is related to the first association ID.

[0007] In a fourth aspect, a device for wireless communication is provided, which is a network device, and includes: a receiving unit for receiving an inference result of a CSI prediction from a first terminal device; wherein a first association ID determined by the first terminal device is used for the CSI prediction; the input parameters of the CSI prediction are determined according to the first association ID, and / or the inference result is determined according to the first association ID, and / or the update of the first model corresponding to the CSI prediction is related to the first association ID.

[0008] In a fifth aspect, a communication device is provided, comprising a memory and a processor, wherein the memory is used to store a program, and the processor is used to call the program in the memory to execute the method described in the first aspect or the second aspect.

[0009] In a sixth aspect, a device is provided, comprising a processor for calling a program from a memory to execute the method as described in the first aspect or the second aspect.

[0010] In a seventh aspect, a chip is provided, comprising a processor for calling a program from a memory so that a device equipped with the chip executes the method described in the first aspect or the second aspect.

[0011] In an eighth aspect, a computer-readable storage medium is provided, on which a program is stored, wherein the program enables a computer to execute the method as described in the first aspect or the second aspect.

[0012] In a ninth aspect, a computer program product is provided, comprising a program, wherein the program enables a computer to execute the method as described in the first aspect or the second aspect.

[0013] In a tenth aspect, a computer program is provided, which enables a computer to execute the method as described in the first aspect or the second aspect.

[0014] In an embodiment of the present application, a first terminal device can perform CSI prediction based on a first association ID. The first association ID can be used to determine input parameters or inference results for the CSI prediction, and can also be used to update a first model for CSI prediction. Thus, the inference results of the CSI prediction obtained by the first terminal device are not determined solely by the first model, but are also corrected or updated based on the first association ID, thereby improving the accuracy of the CSI prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 This is an example diagram of the system architecture of a wireless communication system to which the embodiments of the present application are applicable.

[0016] Figure 2 It is a schematic diagram of a network architecture applicable to the embodiments of the present application.

[0017] Figure 3A and Figure 3B It is a structural diagram of a wireless protocol stack applicable to the embodiments of the present application.

[0018] Figure 4 Schematic diagram of neurons in a neural network to which embodiments of the present application are applicable.

[0019] Figure 5 Schematic diagram of a neural network applicable to the embodiments of the present application.

[0020] Figure 6 Schematic diagram of a convolutional neural network applicable to an embodiment of the present application.

[0021] Figure 7 This is a flow chart of a method for wireless communication proposed in an embodiment of the present application.

[0022] Figure 8 yes Figure 7 A flowchart of a possible implementation of the method is shown.

[0023] Figure 9 This is a flowchart of another method for wireless communication proposed in an embodiment of the present application.

[0024] Figure 10 This is a schematic diagram of the structure of a device for wireless communication provided in an embodiment of the present application.

[0025] Figure 11 It is a structural diagram of another device for wireless communication provided in an embodiment of the present application.

[0026] Figure 12 It is a structural diagram of a communication device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0027] The technical solutions in the embodiments of the present application will be described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments.

[0028] Communication system architecture

[0029] Figure 1 1 is a diagram illustrating an example system architecture of a wireless communication system 100 to which embodiments of the present application may be applied. The wireless communication system 100 may include a network device 110 and a terminal device 120. The network device 110 may be a device that communicates with the terminal device 120. The network device 110 may provide communication coverage for a specific geographic area and may communicate with the terminal device 120 located within the coverage area.

[0030] Figure 1 A network device and multiple terminal devices are exemplarily shown, for example, Figure 1 Optionally, the wireless communication system 100 may include multiple network devices and each network device may include other number of terminal devices within its coverage area, which is not limited in this embodiment of the present application.

[0031] Optionally, the wireless communication system 100 may further include other network entities such as a network controller and a mobility management entity, which is not limited in the embodiments of the present application.

[0032] It should be understood that the technical solutions of the embodiments of the present application can be applied to various communication systems, such as: fifth-generation (5G) system or new radio (NR) system, long term evolution (LTE) system, LTE frequency division duplex (FDD) system, LTE time division duplex (TDD) system, enhanced long term evolution (LTE-A) system, enhanced 5G (5G advanced) system, etc. The technical solutions provided in the present application can also be applied to future communication systems, such as sixth-generation (6G) mobile communication systems, satellite communication systems, and the like.

[0033] The communication system in the embodiment of the present application can be applied to a carrier aggregation (CA) scenario, a dual connectivity (DC) scenario, and a standalone (SA) networking scenario.

[0034] The terminal device in the embodiments of the present application may also be referred to as user equipment (UE), access terminal, user unit, user station, mobile station, mobile station (MS), mobile terminal (MT), remote station, remote terminal, mobile device, user terminal, terminal, wireless communication device, user agent or user device. The terminal device in the embodiments of the present application may refer to a device that provides voice and / or data connectivity to a user and can be used to connect people, objects and machines, such as a handheld device with wireless connection function, a vehicle-mounted device, etc. The terminal device in the embodiment of the present application can be a mobile phone, a tablet computer, a laptop computer, a handheld computer, a camera device, a mobile internet device (MID), a wearable device, a virtual reality (VR) device, an augmented reality (AR) device, a wireless terminal in industrial control, a wireless terminal in self-driving, a wireless terminal in remote medical surgery, a wireless terminal in a smart grid, a wireless terminal in transportation safety, a wireless terminal in a smart city, a wireless terminal in a smart home, etc. Optionally, the terminal device can be used to act as a base station. For example, the terminal device can act as a scheduling entity that provides sidelink signals between UEs in vehicle-to-everything (V2X) or device-to-device (D2D). For example, a cellular phone and a car communicate with each other using sidelink signals. Cellular phones and smart home devices communicate with each other without relaying the communication signal through a base station.

[0035] The network device in the embodiment of the present application may be a device for communicating with a terminal device, and the network device may also be referred to as an access network device or a wireless access network device, such as a base station (BS). The network device in the embodiment of the present application may refer to a radio access network (RAN) node or a next generation RAN (NG-RAN) node (or device) that connects a user device to a wireless network. A base station can broadly cover various names as follows, or be replaced with the following names, such as: NodeB, evolved NodeB (eNB), next generation NodeB (gNB), relay station, transmitting and receiving point (TRP), transmitting point (TP), master station (MeNB), secondary station (SeNB), multi-standard radio (MSR) node, home base station, network controller, access node, wireless node, access point (AP), transmission node, transceiver node, baseband unit (BBU), remote radio unit (RRU), active antenna unit (AAU), remote radio head (RRH), central unit (CU), distributed unit (DU), positioning node, etc. A base station can be a macro base station, a micro base station, a relay node, a donor node or the like, or a combination thereof. A base station can also refer to a communication module, a modem or a chip used to be set in the aforementioned device or apparatus. The base station can also be a mobile switching center and a device that performs base station functions in D2D, V2X, and machine-to-machine (M2M) communications, a network-side device in a 6G network, or a device that performs base station functions in future communication systems. The base station can support networks with the same or different access technologies. The embodiments of this application do not limit the specific technology and specific device form used by the network equipment.

[0036] 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.

[0037] In some deployments, the network device in the embodiments of the present application may refer to a CU or a DU, or the network device may include a CU and a DU. The gNB may also include an AAU.

[0038] The network equipment and terminal devices can be deployed on land, including indoors or outdoors, handheld or vehicle-mounted; they can also be deployed on water; they can also be deployed in the air on aircraft, balloons, and satellites. The embodiments of this application do not limit the scenarios in which the network equipment and terminal devices are located.

[0039] It should be understood that all or part of the functions of the communication device in this application can also be implemented through software functions running on hardware, or through virtualization functions instantiated on a platform (such as a cloud platform).

[0040] Figure 2 A schematic diagram of a network architecture 200 of an embodiment of the present application is exemplarily shown. The network architecture 200 illustrates the network architecture of a 5G NR / LTE / LTE-A system, which may also be referred to as a 5G system (5G system, 5GS) / evolved packet system (Evolved Packet System, EPS) network architecture. The network architecture 200 includes at least one of a network device 110, a terminal device 120, a 5G core network (5G core network, 5GC) / evolved packet core (EPC) 210, a home subscriber server (HSS) / unified data management (UDM) 220, and an Internet service 230. Figure 2 The network device and terminal device in the figure are illustrated by taking RAN and UE as examples respectively.

[0041] like Figure 2As shown, the network device 110 provides user plane protocol and control plane protocol termination towards the terminal device 120. The network device 110 is connected to the 5GC / EPC 210 via the S1 / NG interface. The 5GC / EPC 210 includes a mobility management entity (MME) / authentication management field (AMF) / session management function (SMF) 211, other MME / AMF / SMF 214, a service gateway (S-GW) / user plane function (UPF) 212, and a packet data network gateway (P-GW) / UPF 213. The MME / AMF / SMF 211 is a control node that handles signaling between the terminal device 120 and the 5GC / EPC 210. Generally, the MME / AMF / SMF 211 provides bearer and connection management. All user Internet Protocol (IP) packets are routed through S-GW / UPF 212, which is itself connected to P-GW / UPF 213. The P-GW provides UE IP address allocation and other functions. P-GW / UPF 213 is connected to Internet Services 230. Internet Services 230 includes operator-specific Internet Protocol services, which may include the Internet, intranet, IP Multimedia Subsystem (IMS), and packet-switched streaming services. While network architecture 200 provides packet-switched services, those skilled in the art will readily appreciate that the various concepts presented herein can be extended to networks providing circuit-switched services or other cellular networks.

[0042] Figure 3A and Figure 3B A schematic diagram of the wireless protocol stack structure of an embodiment of the present application is shown respectively. Figure 3A and Figure 3B Let's take the 5G wireless protocol stack as an example. The 5G wireless protocol stack is divided into two planes: the user plane (UP) protocol stack and the control plane (CP) protocol stack. The user plane protocol stack is the protocol cluster used for user data transmission, and the control plane protocol stack is the protocol cluster used for control signaling transmission in the 5G system. The names of the various protocol stack layers are as follows:

[0043] like Figure 3AAs shown in the figure, the user plane protocol stack includes, from top to bottom, the service data adaptation protocol (SDAP) layer, the packet data convergence protocol (PDCP) layer, the radio link control (RLC) layer, the medium access control (MAC) layer, and the physical (PHY) layer.

[0044] like Figure 3B As shown in FIG, the control plane protocol stack includes, from top to bottom, a non-access stratum (NAS); a radio resource control (RRC) layer, a PDCP layer, an RLC layer, a MAC layer, and a PHY layer.

[0045] It should be understood that the different layers in the aforementioned protocol stack have different functions, and that communication between terminal devices and network devices is achieved through inter-layer interaction. With the development of artificial intelligence (AI) technology, AI-assisted computing has penetrated into the processing and implementation methods of the aforementioned protocol stack. For example, AI algorithms can be applied to the scheduling algorithms of the MAC layer and the encoding and decoding algorithms of the PHY layer to improve the performance of communication algorithms.

[0046] As an example, Figure 3A and Figure 3B The wireless protocol architecture in is applicable to the terminal device in this application, such as UE.

[0047] As an example, Figure 3A and Figure 3B The wireless protocol architecture in is applicable to the network devices in this application, such as gNB.

[0048] It should be understood that the interpretation of the terminology in the embodiments of the present application can refer to the specification protocols TS36 series, TS37 series and TS38 series of the 3rd Generation Partnership Project (3GPP), but can also refer to the specification protocols of the Institute of Electrical and Electronics Engineers (IEEE).

[0049] For ease of understanding, some relevant technical knowledge involved in the embodiments of this application is first introduced. The following related technologies can be combined with the technical solutions of the embodiments of this application as optional solutions, and they all fall within the scope of protection of the embodiments of this application. The embodiments of this application include at least part of the following contents.

[0050] Neural Networks

[0051] AI research, exemplified by neural networks, has achieved remarkable results in many fields and will continue to impact people's lives and production for a long time to come. A neural network can be understood as a computational model consisting of multiple interconnected neuron nodes. In a neural network, the strength of the connections between nodes can be expressed as weighted values ​​corresponding to the input signals, also known as parameters. Each neuron node performs a weighted summation of different input signals and generates an output using a specific activation function. Neurons can implement nonlinear mapping based on activation functions.

[0052] by Figure 4 The neuron shown in the figure is used as an example. The input of the neuron can be recorded as A, and each dimension of the input is recorded as a j , the corresponding weighted value is recorded as w j , where j is 1, 2, ..., n. The input of the neuron can also be set with a bias term to adjust the output, such as Figure 4 The constant 1 in is denoted by b (the corresponding weight value is b). The weight value and the summation unit (SU) together enhance or weaken the input. The output of SU can be input into the activation function f to obtain the output t.

[0053] Common neural networks include convolutional neural network (CNN), recurrent neural network (RNN), deep neural network (DNN), etc.

[0054] Combined with the following Figure 5 A neural network applicable to the embodiments of the present application is introduced. Figure 5 The neural network shown can be divided into three categories according to the positions of different layers: input layer 510, hidden layer 520, and output layer 530. Generally speaking, the first layer is the input layer 510, the last layer is the output layer 530, and the intermediate layers between the first and last layers are all hidden layers 520.

[0055] The input layer 510 is used to input data, where the input data can be, for example, a received signal received by a receiver. The hidden layer 520 is used to process the input data, for example, decompress the received signal. Hidden layers can also be called intermediate layers. The output layer 530 is used to output processed output data, for example, a decompressed signal.

[0056] See also Figure 5 A neural network consists of multiple layers, each of which contains multiple neurons. The neurons between layers can be fully connected or partially connected. For connected neurons, the output of the neurons in the previous layer can serve as the input of the neurons in the next layer.

[0057] For ease of understanding, let's take CNN as an example and combine it with Figure 6 The following is an example of multiple layers in a neural network. CNN is a deep neural network with a convolutional structure. Figure 6 As shown, the structure of CNN may include an input layer 610, a convolutional layer 620, a pooling layer 630, a fully connected layer 640, and an output layer 650. The convolutional layer 620, the pooling layer 630, and the fully connected layer 640 are intermediate layers of the CNN.

[0058] It should be noted that if Figure 6 The CNN shown is only an example of a convolutional neural network. In specific applications, the convolutional neural network can also exist in the form of other network models, which is not limited in the embodiments of the present application.

[0059] CSI Feedback

[0060] CSI feedback is a crucial component of wireless communication systems. CSI feedback from terminal devices helps network equipment accurately understand the state of the wireless channel, thereby optimizing transmission parameters and improving system performance. The following describes CSI feedback using 5G NR as an example.

[0061] In 5G NR, two types of codebooks are defined: Type 1 and Type 2. A set of precoding matrices is used to describe channel state information (CSI). These precoding matrices are measured based on the channel state information reference signal (CSI-RS). The Type 1 codebook is a standard precision codebook, primarily used to support single-user multiple input multiple output (MIMO) transmission; the Type 2 codebook is a high-precision codebook, primarily used to support multi-user MIMO transmission to improve system spectrum efficiency. Terminal devices can transmit this information to network equipment through a feedback mechanism for scheduling and precoding.

[0062] As an example, all CSI codebook designs focus on feedback (FB) based on current CSI-RS measurements. As an implementation, a network device (e.g., gNB) periodically sends CSI-RS to a terminal device, and the terminal device periodically provides CSI feedback (CSI-FB). The network device applies the CSI feedback to generate a precoding matrix indication (PMI) until the next CSI feedback is available. In addition, assuming that the channel does not change substantially over time (e.g., within a few milliseconds), the network device can use older CSI feedback. As another implementation, the network device can send a cluster of CSI-RSs, and the terminal device can send CSI feedback to the network device based on at least one CSI-RS in the cluster of CSI-RSs. The network device applies the CSI feedback to generate a PMI (related to time) until the next CSI feedback is available.

[0063] In both implementations in the above examples, the terminal device requires frequent CSI feedback, and is more suitable for low-mobility terminal devices.

[0064] As an example, the UE may measure the channel quality by analyzing a reference signal (CSI-RS) transmitted by the base station.

[0065] As an example, CSI parameters may include channel quality indicator (CQI), rank indicator (RI) and PMI. CQI is a quantitative value reported by the terminal device to the network device to indicate the channel quality of the downlink. CQI reflects the maximum modulation and coding scheme that the terminal device can receive under the current channel conditions to ensure a certain bit error rate. RI indicates the number of parallel transmitted data streams in the MIMO system. RI reflects the multipath propagation characteristics of the channel and the rank of the channel matrix, which is usually related to the spatial degrees of freedom of the channel. PMI is a feedback from the user equipment on the downlink channel status, indicating which precoding matrix the network device should select for signal transmission. The precoding matrix is ​​a linear transformation matrix used to process signals in a multi-antenna system.

[0066] It should be noted that in the NR system, when the downlink (DL) / uplink (UL) channel reciprocity is good, CSI feedback based on CSI-RS is not required, because the sounding reference signal (SRS) can be used for DL ​​CSI acquisition. However, when the DL / UL channel reciprocity is not good enough, CSI feedback based on CSI-RS measurement is required. In this case, since the DL channel may be very different from the UL channel, DL CSI cannot be obtained through SRS alone.

[0067] Due to the complexity of wireless channels and the massive MIMO and high-frequency, large-bandwidth communications supported by 5G, CSI information is typically very high-dimensional and rich in details, resulting in a huge amount of bandwidth resources required for feedback.

[0068] In 5G systems, physical layer and CSI feedback information plays a crucial role in wireless communications. Regulations mandate the compression of physical layer and CSI feedback to reduce radio resource consumption and improve system efficiency, known as physical layer and CSI feedback compression. Given the importance of feedback information, effective compression is a key means of improving system performance.

[0069] To improve the efficiency and accuracy of CSI feedback, AI technology has been introduced to promote the development and application of wireless communication technologies. As an example, CSI compression technology based on AI / machine learning (ML) offers greater flexibility, lower distortion, and greater adaptability. Therefore, leveraging the power of AI technology can achieve more efficient and lower-distortion CSI compression.

[0070] During the CSI feedback process, AI-based CSI feedback enhancement focuses on the following two areas:

[0071] 1. Space-frequency domain CSI compression based on bilateral AI model;

[0072] 2. Time-domain CSI prediction based on a unilateral AI model.

[0073] A space-frequency domain CSI compression solution based on a bilateral AI model deploys a CSI generation model on the terminal device and a corresponding CSI reconstruction model on the network equipment. These two models work together to complete CSI compression, feedback, and reconstruction tasks. This significantly compresses CSI on the terminal device while ensuring that network equipment can obtain more accurate CSI data, thereby optimizing resource scheduling and other operations.

[0074] Unlike space-frequency domain CSI compression solutions based on bilateral AI models, time-domain CSI prediction based on unilateral AI models uses a unilateral model on the terminal device side, meaning the CSI prediction model is deployed on the terminal device. This solution inputs historical CSI measurement data into the CSI prediction model and then outputs CSI predictions for future moments, effectively addressing the timeliness of CSI feedback. In certain scenarios, CSI prediction based on AI / ML models can also improve CSI accuracy.

[0075] As an example, the overall process of CSI prediction based on AI / ML models can be data preparation and preprocessing, input feature construction, AI model reasoning (prediction), post-processing, and performance evaluation and optimization.

[0076] It should be understood that the above introduction to CSI feedback based on 5G is only an example. With the development of 5G mobile communication systems as a basis, support can be provided for subsequent new communication technologies. For example, a new waveform for providing coverage in the terahertz band of 6G mobile communication technology. New technologies also include, for example, multi-antenna transmission technology that applies full-dimensional MIMO (FD-MIMO), array antennas, and large antennas; lenses based on metamaterials and antennas for improving the coverage of terahertz band signals; high-dimensional spatial multiplexing technology using orbital angular momentum (OAM); reconfigurable intelligent surface (RIS); full-duplex technology for increasing the frequency efficiency of 6G mobile communication technology and improving system networks; AI-based communication technology for achieving system optimization by utilizing satellites and AI from the design stage and internalizing end-to-end AI support functions; and next-generation distributed computing technology for implementing services at a complexity level that exceeds the operating capacity limit of UE by utilizing ultra-high-performance communication and computing resources. Therefore, in 6G and subsequent communication system technologies, air interface CSI prediction and compression are important research directions.

[0077] The above article introduces SI-based CSI feedback enhancement. To achieve space-frequency domain CSI compression based on bilateral AI models and time-domain CSI prediction based on unilateral AI models, how to collect model-related data has become a key research issue.

[0078] Based on this, an embodiment of the present application proposes a method for wireless communication. In this method, multiple terminal devices including a first terminal device (e.g., a terminal device) can send multiple CSIs to a network device (e.g., a network device) so that the network device determines a first sample set related to CSI prediction. The first sample set can be used to determine a plurality of sequence pairs including a first sequence and a second sequence, respectively, and the sampling time of the samples in the first sequence is earlier than the sampling time of the samples in the second sequence. It can be seen that when the network device collects data related to CSI prediction, it determines multiple sequence pairs based on the sampling time, which helps to improve the rationality of data collection and improve the timeliness of CSI prediction.

[0079] For ease of understanding, the following Figure 7 The method proposed in the embodiment of the present application is described in detail. Figure 7 It is introduced from the perspective of the interaction between the first terminal device and the network device.

[0080] The first terminal device is any terminal device or terminal side device described above, such as a UE.

[0081] In some embodiments, the first terminal device may communicate with the network device. In one embodiment, the first terminal device may receive a reference signal sent by the network device to perform channel estimation. In one embodiment, the first terminal device may send an SRS to the network device.

[0082] In some embodiments, the first terminal device may be a communication device that deploys the first model. The first model deployed by the first terminal device may be used for inference or prediction. When the first model is deployed on a terminal-side device corresponding to the first terminal device, the terminal-side device may be an auxiliary device that communicates with the first terminal device, or may be another communication device on the network side. For example, the terminal-side device may be a network relay.

[0083] As an example, the terminal-side device may be a server that trains the first model, i.e., a first device. The first device may perform inference based on the deployed first model and send the inference results or prediction results to the first terminal device. For example, the first device may be a non-3GPP entity belonging to the terminal device vendor, such as an OTT server.

[0084] As an embodiment, the first model may be a model related to CSI prediction, and may also be referred to as a CSI predictor. For example, the first model may be an AI / ML model related to CSI prediction. In another example, the first model may be a baseline model related to CSI prediction.

[0085] As an embodiment, the first model can implement CSI prediction based on a unilateral AI model, such as time-domain CSI prediction.

[0086] As an embodiment, the first model can work in conjunction with the second model deployed on the network device side to implement CSI prediction based on the bilateral AI model, such as space-frequency domain CSI prediction.

[0087] In some embodiments, the first terminal device may only support channel estimation, or support both data collection and channel estimation. As an embodiment, the first terminal device may report whether it supports both data collection and channel estimation, or only supports channel estimation.

[0088] As an embodiment, the first terminal device may be deployed with at least one transmitting antenna and at least one receiving antenna.

[0089] In some embodiments, the cell where the first terminal device is located is a first cell. The first terminal device communicates with a network device of the first cell.

[0090] The first terminal device is one of a plurality of terminal devices. The first terminal device can be any of the plurality of terminal devices. The plurality of terminal devices can be a variety of terminal devices and / or terminal-side devices. In some embodiments, the plurality of terminal devices can be a plurality of terminal devices of the same type, so that the network device can perform device group-based communication for the plurality of terminal devices. In some embodiments, at least two of the plurality of terminal devices are of different types. For example, the plurality of terminal devices can be a plurality of terminal devices of different types.

[0091] The network device is any of the network devices or network (NW) side devices described above, such as a base station. The network device may be a network device corresponding to the first cell. The network device may provide services for all terminal devices in the first cell.

[0092] In some embodiments, a network device may be deployed with a second model. The second model may be an AI / ML model. As an embodiment, the second model may work in conjunction with the first model to perform joint reasoning. When the second model is used for CSI prediction, it is also referred to as a CSI predictor.

[0093] As an example, the network device may collect data to construct a data set related to the AI / ML model. As an example, the network device may sample the collected data at a certain time interval to obtain multiple samples arranged in a certain time sequence.

[0094] As an example, network devices can pre-process the collected data to obtain better samples. Through a reasonable data processing process, the training effect of the AI / ML model and the accuracy of CSI prediction can be improved.

[0095] As an embodiment, the network device may train the AI / ML model deployed on the terminal device and then send the trained model to the terminal device. For example, the network device may train a first model and send the trained first model to the first terminal device.

[0096] In some embodiments, a network device may be deployed with multiple receive antennas and multiple transmit antennas. The multiple antennas may form beams.

[0097] See also Figure 7 , Figure 7 The process shown includes step S710 and step S720, which are described below. Figure 7 The dashed line in the box indicates that the step is optional.

[0098] In step S710, a first terminal device receives a first reference signal sent by a network device. The first terminal device may receive the first reference signal from the network device and measure the first reference signal to determine the CSI of a downlink channel.

[0099] The first reference signal can be any downlink signal used to determine the downlink channel state. As an embodiment, the first reference signal can be a CSI-RS. For data collection on the NW side, the network device can use a data collection method based on CSI-RS measurements. For example, when DL / UL channel reciprocity is poor, the network device determines the downlink channel state based on the CSI.

[0100] In some embodiments, the first reference signal may be one of the following: a CSI reference signal (i.e., CSI-RS) used only for data collection; or a CSI-RS used for both data collection and channel estimation. The CSI-RS used only for data collection may be of type 1, and the CSI-RS used for both data collection and channel estimation may be of type 2.

[0101] It should be understood that the types of CSI-RS may also include non-zero power, zero power consumption, interference measurement resource, etc. Type 1 and Type 2 in this application may be configured in parallel with these CSI-RS types, or may include these CSI-RS types.

[0102] As an embodiment, when the first reference signal is only used for data collection, the CSI-RS-based data collection can be independent of the traditional channel estimation purpose. For example, the network device sends the first reference signal for data collection and the second reference signal for channel estimation separately.

[0103] As an embodiment, when the first reference signal is a CSI-RS used only for data collection, the transmission resources of the first reference signal and the transmission resources of the CSI reference signal (CSI-RS) used for channel estimation are independent of each other. The independence of transmission resources can be understood as the mutual independence of the configuration methods of the transmission resources, the mutual independence of the configuration information of the transmission resources, the mutual spacing of the transmission resources in the resource pool, and the mutual independence of the processing units or processing modules of the transmission resources. In other words, the transmission resources of the first reference signal and the transmission resources of the CSI reference signal used for channel estimation are not related to each other.

[0104] As an implementation of the above embodiment, the time-frequency resources of the first reference signal are different from the time-frequency resources of the CSI-RS used for channel estimation. For example, in an NR system, a network device may design a new CSI-RS configuration for data collection, in which the CSI-RS frequency and time resources in the CSI-RS configuration are different from the frequency and time resources of the CSI-RS used for traditional channel estimation.

[0105] As another implementation of the above embodiment, the network can optimize the CSI-RS configuration for data collection to distinguish between the CSI-RS used for data collection and the CSI-RS used for channel estimation. For example, the network device can increase the sampling density or adjust the beam distribution.

[0106] For example, a higher sampling density configuration can be achieved by increasing the periodic resource interval of the CSI-RS. In the traditional CSI-RS periodic configuration, the CSI-RS is transmitted every 5 ms. The network can adjust to transmit CSI-RS every 1 ms to enable high-frequency sampling. To distinguish between data collection and channel estimation, 4 out of every 5 CSI-RS transmissions are used for data collection, and the remaining 1 can be used for channel estimation.

[0107] For another example, the transmission beam of the CSI-RS can be adjusted to a finer beam, and the CSI-RS in certain beams can be indicated to be used only for data collection.

[0108] As an embodiment, when the first reference signal is a CSI-RS used only for data collection, the transmission resources of the first reference signal and the transmission resources of the CSI reference signal used for channel estimation can be shared. Sharing of transmission resources means that the network device can simultaneously indicate the transmission resources of the first reference signal and the transmission resources of the CSI-RS used for channel estimation through a single indication information.

[0109] As an embodiment, when the first reference signal is used for data collection and channel estimation, it means that the first reference signal is compatible with the CSI-RS for traditional channel estimation. CSI-RS supports both data collection and channel estimation and can be applied to scenarios with limited resources. In such scenarios, the configuration needs to take into account the performance of both data collection and traditional channel estimation. For example, maintaining the configuration of the standard NR CSI-RS but adding sampling mode selection or data extension can make this CSI-RS configuration compatible with CSI data collection and normal channel estimation.

[0110] In some embodiments, for CSI prediction using a UE-side model (e.g., a first model), data collection and resource configuration for model training may be based on channel measurement resources (CMR). The configuration may be related to the CSI-RS resource type. The configuration may support at least one of periodic, semi-persistent, and non-periodic modes. As described above, for CSI-RS resources, the network device may separately configure the first reference signal resources for model input measurement and the CSI-RS resources for real CSI measurement. The network device may also configure that the resources of the two can be shared.

[0111] In some embodiments, the first reference signal may be a CSI-RS specific to the first cell and / or a CSI-RS specific to the first terminal device. Regardless of being specific to the first cell or the first terminal device, the first reference signal is a CSI-RS that the first terminal device can receive and measure.

[0112] As an embodiment, when the first reference signal is a CSI-RS specific to the first cell, the first reference signal corresponds to the first cell. For a cell-specific CSI-RS, all terminal devices located in the cell can measure CSI data, which is then used for data collection on the terminal device side and the network device side. For cell-specific configurations, CSI-RS resources can be sent through broadcast or system information. For example, when the first reference signal corresponds to the first cell, the transmission resources of the first reference signal are sent through broadcast or system information. System information is also called a system message and can be sent through a system information block (SIB).

[0113] As an embodiment, when the first reference signal is a CSI-RS specific to the first terminal device, the first reference signal may correspond to a first type of terminal device including the first terminal device. The first type of terminal device may be a personalized UE or a specific UE with a CSI-RS related to data collection. When the CSI-RS used for data collection is configured as a CSI-RS specific to the first type of terminal device, only the first type of terminal device can use these CSI-RSs for data collection. At this time, the CSI-RS is only for the configured specific UE, so that the specific UE for personalized data collection collects CSI data, thereby improving the adaptability of the personalized model.

[0114] In the above embodiment, when the first reference signal corresponds to the first type of terminal device, the transmission resource of the first reference signal can be sent through dedicated signaling. For example, for UE-specific configuration, the CSI-RS resource can be indicated through dedicated signaling, and the CSI-RS configuration signaling can be conveyed to the UE using radio resource control (RRC) signaling.

[0115] Exemplarily, the content of the CSI-RS configuration signaling may include one or more of the following: CSI-RS type (such as type 1, type 2); frequency domain and time domain resource location; antenna port information; beam direction information (for beamforming).

[0116] In the above embodiment, in order to implement UE-specific CSI-RS transmission, the network can configure a specific resource group and a unique CSI-RS identifier, that is, an identification (ID). In other words, the network can configure a dedicated CSI-RS resource group and identifier for the first type of terminal device. For example, the network allocates specific frequency and time resources to the first terminal device for CSI-RS transmission. The resource group should avoid conflicts with the CSI-RS configuration of other terminal devices in the cell.

[0117] Exemplarily, when the first reference signal corresponds to the first terminal device, the network device may select a set of physical resource blocks (PRBs) and subframe positions for the first terminal device, such as frequency domain resources: f1, f2, ..., fn; time domain resources: t1, t2, ..., tm; and use a specific CSI-RS configuration type (e.g., type 1 or type 2).

[0118] Exemplarily, when the first reference signal corresponds to the first terminal device, the first reference signal includes a first identifier indicating the first terminal device. The first identifier may be a unique identifier that distinguishes the first terminal device from other devices. In order to assign a unique identifier to a specific CSI-RS, the first identifier may be associated with a radio network temporary identifier (RNTI) of the first terminal device.

[0119] Exemplarily, when the first reference signal corresponds to the first terminal device, the network device may send the first reference signal to the first terminal device via broadcast, system information, or dedicated signaling.

[0120] In some embodiments, after determining the transmission resource of the first reference signal, the first terminal device may receive and measure the first reference signal. The measurement result of the first reference signal is used by the first terminal device to determine at least one CSI to be sent to the network device.

[0121] As an embodiment, for the resources receiving the CSI-RS, the first terminal device may measure the channel quality thereof and generate data.

[0122] In some embodiments, before sending the CSI to the network device, the first terminal device may process the CSI-RS measurement results. For example, the first terminal device may input the measurement results as raw CSI data into a local processing model to obtain processed CSI data. The local processing model may be, for example, a CSI data generation model deployed on the first terminal device.

[0123] As an embodiment, data processing performed on the original CSI data may include channel estimation, data preprocessing, etc. Channel estimation may be estimating channel characteristics based on the received CSI-RS. Data preprocessing may be performing denoising and normalization on the estimation results.

[0124] Continue to see Figure 7 In step S720, the first terminal device sends at least one CSI to the network device. The network device may receive multiple CSIs from multiple terminal devices. The multiple CSIs include the at least one CSI sent by the first terminal device.

[0125] The first terminal device may report at least one CSI to the network device via a CSI report. In some embodiments, the at least one CSI sent by the first terminal device is used to collect data related to CSI prediction. In other words, the first terminal device sends at least one CSI related to CSI prediction to the network device via a CSI report.

[0126] As an embodiment, when the first terminal device performs CSI prediction based on the first model, the data collection is related to the first model.

[0127] As an embodiment, at least one CSI is used for data collection, and these CSIs are measurement results, not prediction results.

[0128] The CSI report sent by the first terminal device can be a traditional CSI report or an AI / ML-based CSI report. In some embodiments, the central processing unit (CPU) of the first terminal device can share or calculate separately between the traditional CSI report and the AI / ML-based CSI report. For example, the processing unit can share or calculate separately between AI / ML-related features / functions. For example, the traditional framework of CSI-RS resources and port counts can also be reused or separated.

[0129] As an embodiment, the AI / ML-based CSI report may be a CSI report obtained after prediction by the first model.

[0130] As an embodiment, the CSI reported through the traditional CSI report may be the first CSI, and the CSI reported through the AI / ML-based CSI report may be the second CSI. That is, the first CSI is not related to the CSI prediction, and the second CSI is related to the CSI prediction. The at least one CSI sent by the first terminal device includes at least one first CSI and / or at least one second CSI. It can be seen that the at least one CSI used for data collection may be CSI that has not been processed based on the first model, or may be CSI that has been processed based on the first model.

[0131] As an embodiment, the first terminal device may include a first processing unit and a second processing unit. For example, the CPU in the first terminal device may include two independent processing units. The first processing unit is configured to determine the first CSI, and the second processing unit is configured to determine the second CSI. Since the second CSI is related to CSI prediction, the second processing unit may be configured to execute the first model.

[0132] As an embodiment, the first terminal device may determine the first CSI and the second CSI using the same processing unit. The same processing unit is used to execute the first model. In other words, the processing unit may share the calculation of the first CSI and the second CSI based on AI / ML functions.

[0133] As an embodiment, the first processing unit and the second processing unit can respectively perform operations related to different types of CSI-RS. For example, the first processing unit can be used to process the resources and port counts of a CSI-RS that only participates in channel estimation (legacy CSI-RS); the second processing unit can be used to process the resources and port counts of a CSI-RS that only participates in data collection (Type 1 CSI-RS), or the second processing unit can be used to process the resources and port counts of a CSI-RS that participates in both channel estimation and data collection (Type 2 CSI-RS).

[0134] As an embodiment, the first terminal device can perform related operations for different types of CSI-RS using the same processing unit. For example, the same processing unit can be used to process resources and port counts for traditional CSI-RS and Type 1 / Type 2 CSI-RS. The same processing unit can reuse the traditional CSI-RS processing framework or set a new processing framework based on AI / ML.

[0135] As an embodiment, the multiple CSIs from the multiple terminal devices may include at least one first CSI and / or at least one second CSI.

[0136] In some embodiments, multiple CSIs sent by multiple terminal devices to a network device are concentrated within a certain time period to facilitate sampling by the network device. For example, a first terminal device sends at least one CSI within a first time period, and multiple terminal devices send multiple CSIs within the first time period.

[0137] The first terminal device may send at least one CSI to the network device in a variety of ways. As an embodiment, for data collection on the NW side, the CSI may be reported via layer 1 (L1) signaling or RRC signaling.

[0138] As an embodiment, the first terminal device may send at least one CSI via RRC signaling.

[0139] As an embodiment, the first terminal device may directly send the original CSI data, or may feed back the processed CSI data to the network device via a physical uplink shared channel (PUSCH). The content reported by the first terminal device may include: measured CSI data (e.g., processed CSI data) and / or a measurement report associated with a CSI-RS identifier.

[0140] The plurality of CSIs is used by the network device to determine a first sample set related to CSI prediction. The first sample set may be referred to as a first data set or a CSI prediction data set. For the first terminal device, the first sample set is related to a first model for performing CSI prediction.

[0141] As an embodiment, the samples in the first sample set are measurement results of different CSI-RSs, and may also be referred to as data samples.

[0142] As an embodiment, the samples in the first sample set are used to indicate status information of different downlink channels, and may also be referred to as channel samples.

[0143] As an embodiment, the first sample set may be associated with at least one of the following information: the total number of samples in the first sample set; the dimensions of the first sample set; the total duration of a sampling window of the network device; and the sampling interval of the network device. The total number of samples and the dimensions of the first sample set are related to the sample types in the first sample set. The total duration of the sampling window and the sampling interval are related to the method for determining samples.

[0144] When a network device determines a first sample set related to CSI prediction based on multiple CSIs, the first sample set may be determined based on one or more methods. In one embodiment, the network device may directly collect multiple CSIs to form the first sample set. In another embodiment, the network device may sample all CSIs sent by multiple terminal devices to determine the first sample set. In another embodiment, the network device may preprocess the multiple CSIs to improve the performance of the CSI prediction model.

[0145] It should be understood that the first terminal device may also construct a first sample set based on at least one CSI obtained by measurement to train the first model.

[0146] In some embodiments, when receiving multiple CSIs from multiple terminal devices, the network device may perform sampling based on a fixed time interval S to generate multiple samples to construct a data set related to CSI prediction. The time interval S is, for example, 1 ms or 10 ms.

[0147] As an embodiment, in the construction of the first sample set, N can be generated by sampling. samples samples.

[0148] As an embodiment, the size of the first sample set may represent the total number of samples in the first sample set. The total number of samples in the first sample set is determined according to at least one of the following: the number of antenna ports used by the network device to receive CSI; the number of antenna ports used by the plurality of terminal devices to transmit CSI; the number of subbands used to transmit the plurality of CSIs; or the number of the plurality of terminal devices.

[0149] For example, the total number of samples in the first sample set can be (N Rx ×N Tx ×N sb ×N UEs ), where N Rx Indicates the number of antenna ports on the network device that receive CSI, N Tx Indicates the number of transmitting antenna ports of the first terminal device, N sb Indicates the number of subbands, N UEs Indicates the number of multiple terminal devices.

[0150] For another example, when the first terminal device collects data, the total number of samples in the first sample set may be (N Rx ×N Tx ×N sb ).

[0151] In some embodiments, the dimensions of the first sample set are determined by at least one of: the number of antenna ports through which the network device receives CSI; the number of antenna ports through which multiple terminal devices transmit CSI; the number of subbands used to transmit multiple CSIs; and the number of multiple terminal devices.

[0152] As an embodiment, the dimension of the first sample set can be expressed as (N Rx ,N Tx ,N sb ,N UEs ).

[0153] In some embodiments, the total duration of the sampling window for sampling by the second device can be determined according to the CSI transmission resource used for data collection. As an embodiment, the total duration of the sampling window can include a time window for historical CSI and a time window for predicted CSI.

[0154] As an embodiment, the total duration of the sampling window may be set to T. The number of samples of each channel may be expressed as T / S.

[0155] In some embodiments, the sampling interval of the second device may be fixed or dynamically changing. That is, the network device may dynamically adjust the sampling rate. As an embodiment, the network device may adjust the sampling interval S of the CSI according to the rate of channel change. For example, the sampling interval in a low-speed environment (such as an indoor scene) is longer (such as 10ms, 100ms) to reduce redundant data. As another example, the sampling interval in a high-speed environment (such as a vehicle / UAV scene) is shorter (such as 1ms) to capture fast fading characteristics. As another example, the network device may adopt an adaptive sampling interval to adjust the sampling interval as the channel Doppler shift is adjusted.

[0156] In some embodiments, to accurately predict CSI, the sampled CSI data needs to be properly processed to ensure its quality and consistency. This processing may include pre-processing for data alignment or format alignment, or a first processing to increase data diversity.

[0157] Optionally, the preprocessing of multiple CSIs may include normalization, dynamic time warping, format construction, etc.

[0158] As an example, due to additional characteristics such as Doppler shift and path loss, the amplitude and phase range of CSI is relatively large, so the sample data can be normalized. Normalizing the CSI amplitude and phase separately can ensure that the data input range is consistent.

[0159] For example, the normalized sample H(t) at time t can be expressed as: Where H(t)′ represents the normalized sample, μ and σ are the mean and standard deviation of the CSI data, respectively.

[0160] As an embodiment, in order to correct the time misalignment problem caused by different sampling rates or delays, the time sequence of the CSI data can be adjusted to perform dynamic time warping. In order to cope with the time variability of the channel, the first sample set can be variable to improve flexibility.

[0161] As an embodiment, the sampled CSI data needs to be constructed into a format suitable for AI model input. This construction method can be for input based on mathematical modeling or input based on frequency domain features. For example, the processing of CSI data can be to convert CSI data to the frequency domain through fast Fourier transform (FFT) and extract features. For another example, for scenarios such as MIMO, the processing of CSI data can be based on spatial joint feature extraction.

[0162] In the above embodiment, the network device may use an autoencoder to reduce the dimension of the CSI data to extract important features. For example, a sliding window technique may be used to convert a time series into a feature vector of fixed length.

[0163] As an embodiment, since there may be packet loss or measurement errors during CSI-RS measurement and CSI transmission, it is necessary to interpolate or fill in missing data. As an embodiment, the data is first processed by adding noise, time shifting, frequency offsetting, etc. to increase data diversity and improve the generalization ability of the model.

[0164] Optionally, the first sample set may include CSI data after a first processing is performed on multiple CSIs. The first processing may be used to enhance time-varying channel data. This data enhancement technique can generate a variety of channel scenarios to improve the adaptability of the CSI prediction model to time-varying channels. This data enhancement technique may include time shift transformation, noise perturbation, frequency domain transformation, etc.

[0165] As an embodiment, the first processing may include at least one of the following: adjusting the time tag of at least one CSI among multiple CSIs; adjusting at least one CSI among multiple CSIs based on frequency offset; adjusting at least one CSI among multiple CSIs through random noise.

[0166] As an embodiment, the time shift transformation may include adjusting a time tag of at least one CSI among the multiple CSIs. By adjusting the time tag of the CSI sequence, a delay effect may be simulated.

[0167] As an embodiment, the noise perturbation may include adjusting at least one of the multiple CSIs by using random noise. By adding random noise, the anti-interference capability of the model may be enhanced.

[0168] As an embodiment, the frequency domain change may include adjusting at least one of the multiple CSIs based on a frequency offset. By adding a frequency offset to the CSI data, the model can be adapted to different carrier configurations.

[0169] In some embodiments, the CSI prediction model can be jointly modeled based on spatial features. In massive MIMO or distributed antenna systems, spatial correlation is crucial for CSI prediction. The network device can add CSI of neighboring antennas or neighboring users as additional features in the CSI data samples. As an embodiment, the first sample set may also include at least one of the following: CSI received by all antenna ports of the network device; CSI received by beams adjacent to the directions of multiple CSI receive beams; auxiliary information of multiple terminal devices; antenna configurations of multiple terminal devices and / or network devices.

[0170] As an example, the network device can include CSI data from adjacent antenna ports in the data samples, so that the first sample set includes CSI received by all antenna ports. In a MIMO system, CSI from adjacent antenna ports is typically highly correlated. By adding CSI information from adjacent antenna ports, CSI data from adjacent antennas can be spliced ​​into the input features to enhance the model's predictive capabilities.

[0171] For example, if the CSI data of the nth antenna port at time t is H(n,t), the CSI data of its adjacent ports can be added at the same time [H(n-1,t),H(n,t),H(n+1,t)]. The CSI data of adjacent antennas can be combined to form a new feature vector. Assume that the original number of samples of CSI data is (N Rx ×N Tx ×N sb ×N UEs ), then the number of samples of enhanced CSI input data is (3N Rx ×N Tx ×N sb ×N UEs ).

[0172] As an embodiment, the prediction capability of the model may also be enhanced by introducing CSI in adjacent beam directions into the first sample set.

[0173] As an example, auxiliary information from multiple terminal devices can include combining multi-user CSI and utilizing spatial correlation for collaborative prediction. In distributed antenna or multi-user MIMO (MU-MIMO) systems, CSIs of different users are correlated, especially in multipath environments. By inputting CSI data of neighboring users into the model together, spatial collaborative prediction can be achieved.

[0174] As an embodiment, the auxiliary information of multiple terminal devices may also include information such as the terminal device's location, speed, and path loss. Because factors such as the terminal device's location, speed, and path loss can affect CSI, this auxiliary information can be used as additional input features. For example, the first sample set may include data such as the first terminal device's location information (x, y, z), speed v, and path loss L.

[0175] As an embodiment, the antenna configuration of multiple terminal devices and / or network devices can optimize the sample features. The parameters of the antenna configuration may include beamforming parameters (or beamforming vectors), transmit receive unit (TxRU) mapping, power allocation, and other parameters that have an important impact on CSI prediction. Adding antenna configuration parameters to the feature set (first sample set) can help the model model CSI changes more accurately. For example, adding the antenna beamforming weights and TxRU mapping method to the CSI data sample results in a dimension of (N Tx ,N sb ) beamforming matrix, the number of samples of the enhanced input feature data (first sample set) is (N Rx ×N Tx ×N sb ×N UEs+N Tx ×N sb ).

[0176] It should be noted that the TxRU is a radio frequency unit in a wireless device that manages the transmission and reception of multiple antennas. Different TxRU mapping methods will affect the channel characteristics, CSI feedback, and precoding strategy. TxRU virtualization refers to the use of logical mapping to merge multiple physical antennas into virtual antenna units to optimize beamforming or MIMO performance. When TxRU virtualization is not used, each TxRU is directly mapped to a physical antenna element. In the context of CSI prediction, a baseline refers to a standard configuration or default reference scheme used to evaluate the performance of different configurations or optimization schemes. The antenna element is a larger-scale antenna mapping scheme, that is, a physical array size that is different from the baseline.

[0177] For example, for a 32-port CSI-RS, the baseline configuration can be [N1, N2, P] = [2, 8, 2], where: N1 = 2, indicating the number of antennas in the first dimension (usually polarization direction); N2 = 8, indicating the number of antennas in the second dimension (usually horizontal direction); P = 2, indicating the number of different polarization layers.

[0178] As an embodiment, the first sample set may include part or all of the above-mentioned information. For example, the historical CSI measurement data H(t) may be combined with relevant environmental information such as antenna configuration and user location.

[0179] In some embodiments, the data in the first sample set is variable. As an embodiment, the first sample set can be updated based on at least one of a sampling interval adjustment, a change in data distribution, and a change in channel distribution. As an embodiment, the first sample set can be regularly updated based on new data collected through online learning. For example, new data is collected through online learning to regularly update the training data set. Environmental labels such as antenna angle, transmit (TX) / receive (RX) mapping, etc. can be introduced into the training set.

[0180] As an embodiment, the first sample set may be updated based on adjustment of the dynamic sampling rate.

[0181] As an embodiment, the first sample set may be updated based on dynamic changes in data distribution.

[0182] As an example, in a wireless environment, the channel distribution varies over time and space, and the first sample set can be updated based on the dynamic changes in the channel distribution. The dynamic update of the first sample set can reflect the transitions between various communication scenarios, such as different geographic locations (urban, suburban, rural), different antenna configurations (MIMO, massive MIMO), and different environmental conditions (static, multipath, obstruction).

[0183] It should be noted that when the processed data is used as input data for AI / ML models, the CSI predictions output by the AI / ML models may deviate from the actual situation and require certain post-processing to ensure the rationality of the data. This is because the AI ​​model captures the temporal correlation of CSI (modeling long-range dependencies) through recursive neural networks and uses convolutional layers to extract the spatial patterns of CSI for prediction output.

[0184] For example, when normalized data is used as input to an AI model, denormalization can be used to restore the data to its original scale. For the sample H(t) mentioned above, the denormalization process can be: H(t) = H(t)′σ + μ.

[0185] For example, smoothing filters such as Kalman filtering or moving average can be used to reduce prediction errors, or the physical consistency of CSI can be checked to ensure that the predicted CSI value does not exceed predefined physical boundaries.

[0186] The first sample set is used to determine multiple sequence pairs. The multiple sequence pairs can be at least two sequence groups. The sequence pairs include a first sequence and a second sequence. In other words, each sequence pair includes a first sequence and a second sequence that correspond to each other.

[0187] As an embodiment, the first sequence includes a plurality of samples, and the second sequence includes a plurality of samples different from the first sequence.

[0188] As an embodiment, the first sequence is a feature sequence corresponding to historical CSI, and the second sequence is a tag sequence corresponding to predicted CSI. For example, in the context of CSI prediction, let K represent the amount of historical CSI to be fed into the CSI prediction model, and let N represent the number of time slot indices of future CSI to be predicted. Then, the first sequence includes K samples, and the second sequence includes N samples.

[0189] As an embodiment, samples of the first sequence and the second sequence both correspond to the first channel, that is, future CSI data samples and historical CSI data samples correspond to the same channel element, so as to improve the accuracy of the model.

[0190] As an embodiment, the samples of the first sequence and the second sequence correspond to different channels. The different channels may have a certain correlation.

[0191] In some embodiments, the network device can perform time series modeling on multiple CSI data to support CSI prediction. The first sequence and the second sequence each include multiple samples arranged in chronological order, so multiple sequence pairs can also be referred to as multiple time series pairs. The time sequence can be determined based on the sampling time of the samples, or based on other times related to the sampling time. For example, the time when multiple terminal devices respectively send multiple CSIs or the time when the network device receives multiple CSIs.

[0192] The sampling time of the samples in the first sequence is earlier than the sampling time of the samples in the second sequence. As an embodiment, the sampling time of the last sample in the first sequence is earlier than the sampling time of the first sample in the second sequence. For example, for the time series pair construction of CSI samples, X can be sampled from the past K time steps. i As model input, sample Y from the next N time steps i as a prediction target.

[0193] As an embodiment, the sampling time of the last sample in the first sequence is separated from the sampling time of the first sample in the second sequence by one or more sampling intervals. When the sampling time is separated by one sampling interval, the samples in the first sequence and the second sequence are continuous based on the sampling interval. When the sampling time is separated by multiple sampling intervals, the samples in the first sequence and the second sequence are discontinuous.

[0194] In some embodiments, multiple sequence pairs are determined based on a first sliding window. Multiple sequence pairs can be determined based on a sliding window mechanism to ensure temporal and spatial correlation between the first and second sequences. As an example, the data samples for CSI prediction include a feature sequence (historical CSI) and a tag sequence (future CSI), and can be constructed using a sliding window technique for time series. As previously mentioned, the sample dimensionality still depends on the number of receive antennas, transmit antennas, subbands, and users.

[0195] In some embodiments, the duration of the first sliding window is determined according to the number of samples in the first sequence and / or the number of samples in the second sequence, so as to ensure that the first sequence and the second sequence with a determined number of samples can be obtained based on the first sliding window.

[0196] As an embodiment, the duration of the first sliding window is determined according to the number of samples in the first sequence to ensure the number of historical CSI samples. In this scenario, the samples in the second sequence can be any number of samples after the first sequence.

[0197] As an embodiment, the duration of the first sliding window is determined according to the number of samples in the second sequence to ensure the number of future CSI samples. In this scenario, the samples in the first sequence can be any number of samples before the second sequence.

[0198] As an embodiment, the duration of the first sliding window is determined based on the number of samples in the first sequence and the number of samples in the second sequence. For example, the duration of the first sliding window is greater than or equal to the sum of the duration of the window corresponding to the first sequence (also referred to as the historical CSI window) and the duration of the window corresponding to the second sequence (also referred to as the future CSI window).

[0199] In the above embodiment, the duration of the window may be expressed by the number of time slots, or the number of symbols or subframes.

[0200] For example, the network device can construct N through the first sliding window seq For CSI samples, each pair of CSI samples contains a historical CSI sequence (feature) and a future CSI sequence (target). The number of CSI time series pairs that can be constructed based on the first sliding window is N. seq It can be expressed as: N seq =N samples -(K+N)+1; As mentioned above, K represents the number of historical CSIs fed to the CSI prediction model (the number of samples in the historical time window); N represents the number of time slot indices of the predicted future CSI; N samples Indicates the total sample size.

[0201] For N seq For CSI samples (first sequence and second sequence), when i∈[1,N seq ], the two sequences of the i-th pair of CSI samples (historical CSI feature sequence and future CSI target sequence) can be represented separately.

[0202] The first sequence (historical CSI feature sequence) can correspond to the model input (X i ), is fed as input features to the CSI prediction model. The sample sequence from (i-1)+1 to (i-1)+K can be expressed as X i =[H i ,H i+1 ,…H i+K―1 ] The sequence dimension is (K,N Rx ,N Tx ,N sb ,N UEs ).

[0203] The second sequence (future CSI target sequence) can correspond to the output label (Y i), represents the label sequence (future CSI sequence) to be predicted by the CSI prediction model. The samples in the second sequence are the time samples corresponding to the first sequence. The sample sequence from (i-1)+K+1 to (i-1)+K+N+1 can be expressed as Y i =[H i+K ,H i+K+1 ,…H i+K+N―1 ]. The sequence dimension is (N,N Rx ,N Tx ,N sb ,N UEs ), which is the predicted future CSI label.

[0204] In some embodiments, for the first sample set after determining multiple sequence pairs, the dimension of the first sample set may be composed of the dimension of the first sequence (a feature sequence representing historical CSI) and the dimension of the second sequence (a target sequence representing future CSI). In the above example, the dimension of the multiple first sequences in the first sample set is (N seq ,K,N Rx ,N Tx ,N sb ,N UEs ), the dimension of the multiple second sequences is N seq ,N,N Rx ,N Tx ,N sb ,N UEs ), so the total number of samples (total number of samples) of the first sample set can also be expressed as (N Rx ×N Tx ×N sb ×N UEs )×N seq .

[0205] In some embodiments, multiple sequence pairs can be determined based on multiple sliding windows, and the duration of the multiple sliding windows is different. Multiple sequence pairs can also be determined based on a multi-scale time window method to collect more reasonable data. As an embodiment, short-term and long-term CSI change trends are extracted synchronously based on multiple sliding windows with different time spans. That is, multiple sliding windows can be sampled for different scenarios respectively. For example, a short-term window (such as 5ms) is used to capture fast channel changes. A long-term window (such as 100ms) is used to capture slow channel fading trends. Sampling based on multiple sliding windows of different durations can enhance the model's adaptability to CSI changes at different time scales. Combining short-term and long-term information helps to improve the accuracy of model predictions.

[0206] As an embodiment, the multiple sliding windows may include a short-term time window, a medium-term time window, and a long-term time window. The short-term time window is primarily used to capture rapidly changing short-term CSI trends. The medium-term time window is primarily used to focus on channel changes on a medium time scale. The long-term time window is primarily used to identify channel fading trends over a longer period of time.

[0207] In the above embodiment, the duration of the short-term time window can be less than 10ms or less than 20ms; the duration of the medium-term time window can be 10ms~50ms or 20ms~100ms; the duration of the long-term time window can be greater than 50ms or greater than 100ms.

[0208] As an embodiment, any two adjacent sliding windows in the plurality of sliding windows are continuous. Within a given time length, multiple sliding windows of different lengths are continuous. Multiple sliding windows can be spliced ​​or cross-combined to perform periodic data sampling.

[0209] As an embodiment, at least two sliding windows in the multiple sliding windows include the same time domain resources. When the multiple sliding windows partially overlap, samples may be repeated.

[0210] As an embodiment, at least two sliding windows among the multiple sliding windows are discontinuous, which can reduce the number of samples within a given time.

[0211] As an embodiment, the sampling intervals of multiple sliding windows are the same, which helps to reflect the sampling requirements of different scenarios.

[0212] As an embodiment, the sampling intervals of multiple sliding windows are different, which helps to unify the number of samples in different scenarios.

[0213] As an embodiment, each of the multiple sliding windows can determine multiple sequence pairs. That is, each sliding window includes a historical CSI window and a future CSI window. Therefore, each sliding window can determine a sequence pair with reference to the first sliding window.

[0214] As an embodiment, multiple sliding windows are used to determine the first sequence in the sequence pair, and a time window after the multiple sliding windows is used to determine the second sequence in the sequence pair. For example, multiple sliding windows are used to determine K historical CSI samples.

[0215] As an implementation method, let H(t) be the CSI sample matrix at time t, with dimension (N Rx ,N Tx ,N sb ,N UEs ); T represents the total time series length; K shortis the length of the short time window (such as 10ms), K medium is the length of the time window (such as 50ms), K long is the length of the long time window (such as 100ms); N is the number of future time steps that need to be predicted; S is the sampling interval, such as 1ms.

[0216] On the time series {H(1), H(2), …, H(T)}, three time windows of different scales can be constructed, as follows.

[0217] The sequence in a short time window can be expressed as X short (i)=[H(iS),H(i-2S),...,H(iK short )], where X short The dimension of (i) can be expressed as (K short ,N Rx ,N Tx ,N sb ,N UEs ).

[0218] The series in the medium-term time window can be expressed as X medium (i)=[H(iS),H(i-2S),...,H(iK medium )], where X medium The dimension of (i) can be expressed as (K medium ,N Rx ,N Tx ,N sb ,N UEs ).

[0219] The series in the long-term time window can be expressed as X long (i)=[H(iS),H(i-2S),...,H(iK long )], where X long The dimension of (i) can be expressed as (K long ,N Rx ,N Tx ,N sb ,N UEs ).

[0220] For a given time length T, the final multi-scale time window sample X(i) is composed of short, medium and long-term windows, which can be expressed as X(i) = [X short (i),X medium (i),X long (i)]. The final dimension of the input data is (K short +K medium +K long ,N Rx ,NTx ,N sb ,N UEs ).

[0221] The target prediction label sequence can be expressed as Y(i) = [H(i+1), H(i+2), ..., H(i+N)].

[0222] Based on this target, we predict the label sequence. The goal can be to train a function f(·) to minimize the prediction error:

[0223]

[0224] Among them, θ is the model parameter, and the loss function to be minimized can usually be the mean-square error (MSE):

[0225]

[0226] In some embodiments, the first sample set can also be determined through distributed collection. For example, CSI data can be collected in a distributed manner across multiple network devices or multiple terminal devices. In this scenario, federated learning can be used to train the model without the need for centralized data storage. Multiple network devices or multiple terminal devices can each update data locally and periodically send model updates rather than the data itself.

[0227] In some embodiments, due to the time-varying nature of the channel, the AI ​​model needs to be adaptive. As an example, the first model can be adapted as the data changes. For example, through online learning, the first model can adapt to the time-varying CSI.

[0228] As an example, CSI parameters can be continuously updated based on online learning or during model training, allowing the model to adapt to the latest channel conditions at all times. For example, a sliding window mechanism can be used to update model weights using only the most recent CSI data. Another example is controlling the frequency of model updates through an adaptive learning rate to prevent overfitting.

[0229] Taking adaptive learning rate as an example, the model update can be expressed as: Among them, θ t represents the current model parameters, ε is the learning rate, x t ,y t is the newly arrived CSI data.

[0230] Combined with the above Figure 7This paper introduces various methods for collecting data for CSI prediction. This data collection generates a large sample set, which can lead to excessive power consumption and air interface overhead for terminal devices when transmitted over the air interface. Therefore, air interface delivery of the sample set becomes a problem. Furthermore, after the network determines the first sample set, the entire sample set may not be sent from a single network device to a single terminal device; it may be sent from multiple network devices to a single terminal device.

[0231] To address the above problem, the original data set can be split into subsets, each of which has a limited number of data samples. In some embodiments, the first sample set can be split into multiple sample subsets, each of which includes the same or different numbers of samples.

[0232] As an embodiment, the NW side can split the first sample set into M*D subsets, each of which is transmitted from the network device to the terminal device. In other words, the first sample set can be split into multiple subsets and transmitted by M network devices, and each network device can send D subsets to D terminal devices.

[0233] As an example, in a multi-beam scenario, each beam can be associated with a sample subset. After the network device trains the model, it can upload the model to the first device. All users associated with a beam can share this sample subset.

[0234] In some embodiments, each terminal device does not need to receive the entire sample set, but only needs to receive one or more sample subsets. As an embodiment, model training on the terminal device side is usually performed on the first device described above. Each terminal device may only need to receive a subset of the original data set. In the case where multiple terminal devices correspond to the same first device, the first device can recombine all subsets received and uploaded by multiple terminal devices to restore the original data set, which is then used for model training.

[0235] In some embodiments, all sample subsets of the first sample set are associated with the ID of the first sample set. In other words, each sample subset is associated with a common dataset ID to facilitate reassembly of the datasets.

[0236] As an embodiment, the first terminal device receives one or more sample subsets in the first sample set according to the identifier of the first sample set. The one or more sample subsets can be used to train and / or update the first model.

[0237] It should be noted that the process of sample set delivery does not need to be carried out in real time. Since the sample set delivery and retraining on the terminal device side are non-real-time and overhead-insensitive, there is no need to specify a complex process to specifically avoid redundant transmission of subsets, nor is there a need to handle retransmission of subsets. For the first sample set with a unique data set ID, the terminal device side can combine its component subsets on the first device (such as an OTT server). If the first device still lacks any component subsets, the terminal device side can (through its terminal device) request the NW to send the missing subsets until the first device has completely received all component subsets.

[0238] In some embodiments, the CSI prediction model is typically trained in one scenario. To reduce the cost of training in different scenarios, the model can be migrated to another scenario, for example, migrating a first model trained in a static environment to a dynamic environment. This shows that models for different scenarios do not need to be trained separately, but rather adjusted based on the training results for a specific scenario.

[0239] As an embodiment, the first model may include a general model layer and a dedicated model layer. When the first model is migrated from a network device to a first terminal device, the general model layer is determined by the network device, and the dedicated model layer is determined by the first terminal device. Alternatively, when the first terminal device uses a model trained by the first device, the first device trains the general model layer, and the dedicated model layer is determined by the first terminal device.

[0240] As an example, a model is pre-trained on a general CSI dataset and fine-tuned using less new environment data. For example, a CSI prediction model trained by a network device is pre-trained using a general first sample set, and the first terminal device can fine-tune its corresponding first model using dedicated data.

[0241] As an example, in CSI prediction tasks, some model layers (general model layers) can remain unchanged, and only some high-level (dedicated model layer) parameters can be fine-tuned to reduce computational overhead. For example, multiple terminal devices using the same third-party OTT service provider will have a common CSI prediction AI / ML model. Each terminal device can use this common AI / ML model and only use a few parameters specific to its environment to fine-tune the model.

[0242] As an example, when using federated learning, CSI data from different terminal devices is used for training separately without centralized storage. In this scenario, the construction of the first sample set needs to consider the data distribution of different users. Storing historical CSI locally on the terminal device and performing local training helps protect user privacy and reduce data transmission costs.

[0243] As described above, the first reference signal can be identified as corresponding to the first terminal device through the first identifier. The first terminal device can also report the first identifier when sending at least one CSI, so that the network device can perform data matching based on the first terminal device corresponding to the first identifier.

[0244] In some embodiments, the data matched based on the first identifier may include CSI data measured by the first terminal device and related environmental parameters (e.g., antenna configuration, frequency range). The network device may use the collected data to train a personalized model and optimize resource allocation to enhance the channel prediction capability of the first terminal device. For example, the network device may dynamically adjust the configuration parameters of the CSI-RS (e.g., frequency, time resources, or beam direction) based on real-time feedback.

[0245] For ease of understanding, the following Figure 8 The method of collecting data from network devices and optimizing resource allocation is exemplified. Figure 8 The UE here represents the first terminal device, and the NW represents the network device.

[0246] See also Figure 8 , step S810 can refer to Figure 7 The process is different from step S710 in the process and will not be described in detail.

[0247] In step S820, the UE generates original CSI data, where the original CSI data is determined according to a measurement result of the first reference signal.

[0248] In step S830, the UE processes data and generates a measurement report.

[0249] In step S840 , the UE uploads CSI (including processed CSI data) via the PUSCH.

[0250] In step S850, the NW receives and stores the data reported by the UE.

[0251] In step S860, the NW trains a model, which can be a personalized model for the UE or a universal model.

[0252] In step S870, the NW optimizes the CSI-RS resource configuration. The NW can dynamically adjust the CSI-RS parameters. When the first reference signal carries the first identifier, personalized data collection for the UE can be achieved through the specific identifier and resource group configuration.

[0253] Depend on Figure 8 It can be seen that the network equipment can dynamically update the CSI-RS configuration according to the CSI reported by the terminal equipment to adapt to different environmental requirements.

[0254] Combined with the above Figures 7 and 8This paper describes various methods for collecting data for CSI prediction to determine a first sample set. This first sample set can be used to train a CSI prediction model, enabling the first terminal device to perform CSI prediction through model inference. However, in actual communications, some scenarios may affect prediction accuracy.

[0255] As an example, the training data of the CSI prediction model cannot cover all communication scenarios. When the current scenario exceeds the corresponding scenario of the training data, the accuracy of the model prediction decreases. Taking the CSI prediction model determined based on CSI-RS measurement as an example, CSI-RS usually uses digital beamforming, which has a larger beam width and is less sensitive to data distribution. The data samples for the training model can be directly derived from a single CSI-RS measurement without considering the beam sorting problem. Even if, as mentioned above, the model is trained using a mixed training data set (first sample set) generated by multiple network-side conditions (such as antenna tilt, TxRU mapping, and environmental scenarios) during the training phase, it is impossible to cover all communication scenarios.

[0256] For example, the CSI prediction model used by multiple terminal devices is typically a generalized model that is applicable to most scenarios, but its accuracy may decrease in certain scenarios. No two network devices are 100% identical. For CSI prediction, training different models for two base stations with downtilt angles of 3° and 6°, respectively, would incur unnecessary costs. To reduce training time and costs, multiple terminal devices primarily perform CSI prediction based on a unified generalized model. This generalized model can be trained on a variety of scenarios to improve its adaptability under different conditions. However, these scenarios will differ from actual communication scenarios and cannot cover all scenarios.

[0257] In summary, in some scenarios, when multiple terminal devices including the first terminal device perform actual predictions based on the CSI prediction model, the accuracy of the model may be low. How to improve the accuracy of the model in these scenarios becomes a technical problem that needs to be solved.

[0258] To address this issue, embodiments of the present application also propose a method for wireless communication. In this method, a first terminal device can use a first association ID to determine CSI prediction input parameters, inference results, and model updates, thereby resolving the issue of potential inconsistencies between the training scenario and inference conditions of the CSI prediction model. By supplementing the association ID mechanism, the accuracy of the CSI prediction model can be effectively improved.

[0259] For ease of understanding, the following Figure 9 The method proposed in the embodiment of the present application is described in detail. Figure 9This is also introduced from the perspective of the interaction between the first terminal device and the network device. For the sake of simplicity, Figure 7 Terms that have been explained in will not be repeated.

[0260] See also Figure 9 In step S910, the first terminal device receives first indication information from the network device.

[0261] The first indication information can be used by the first terminal device to determine the first association ID, and can also be used by the first terminal device to determine whether to trigger the use of the first association ID. The first terminal device is not aware of the scenario changes on the network side. Therefore, in some scenarios, the network device needs to send the first indication information to the first terminal device so that the first terminal device can determine the first association ID or trigger the first association ID.

[0262] In some embodiments, the association ID can be associated with one or more types of information to address various factors that may reduce model accuracy. As an example, the association ID can be associated with network devices to address issues such as reduced prediction accuracy due to significant changes in network parameters. As an example, the association ID can be associated with inference conditions to address issues such as reduced prediction accuracy due to significant differences between current inference conditions and training conditions.

[0263] It should be noted that inference conditions may refer to the relevant conditions for CSI prediction inference. These relevant conditions may be one or more pieces of information that affect the accuracy of CSI model prediction. In other words, inference conditions may include one or more factors that cause inconsistencies between training and inference in CSI prediction. In actual communications, multiple factors may affect data distribution.

[0264] As an embodiment, the inference conditions may include the different geographical locations, different antenna configurations, and different environmental conditions mentioned above.

[0265] As an embodiment, the inference condition may include relevant information of the network device, such as the antenna layout of the network device, TxRU mapping, antenna downward tilt angle (antenna tilt angle), site direction, deployment scenario, relative height, etc.

[0266] In the above embodiment, the antenna tilt angle is also referred to as the antenna tilt angle. Different antenna tilt angles have little impact on the performance of the CSI prediction model (usually <3%), but may need to be identified in some cases.

[0267] In the above embodiment, TxRU mapping may refer to different mapping modes. Different subarray sizes or virtualization configurations may result in significant performance variations (up to 77% loss in some cases), and therefore need to be identified.

[0268] In the above embodiments, different cells or sites may have unique conditions, such as user distribution, rate, or antenna layout. Therefore, a generalized evaluation is required to identify potential additional NW-side conditions to ensure consistency between training and inference by using associated IDs. For example, the evaluation can be based on one or more of the following aspects: various antenna tilt angles, various TxRU mappings, and the use or absence of TxRU virtualization.

[0269] Among them, the scenario with TxRU virtualization can be, for example, for 32-port CSI-RS, [N1, N2, P] = [2, 8, 2] as the baseline, [4, 4, 2] as the optional, with (8 × 8 × 2) and (12 × 8 × 2) antenna elements. The scenario without TxRU virtualization can be, for example, for 32-port CSI-RS, [N1, N2, P] = [2, 8, 2] as the baseline, [4, 4, 2] as the optional, with (2 × 8 × 2) and (4 × 4 × 2) antenna elements respectively.

[0270] As an embodiment, the inference condition may further include a transmission scenario between the network device and the terminal device. The transmission scenario may refer to an indoor / outdoor environment, a line of sight (LOS) / non line of sight (NLOS) scenario.

[0271] In some embodiments, the first indication information may be carried in one or more of the following signaling or information: RRC signaling, medium access control (MAC) control element (CE), downlink control information (DCI), and system information.

[0272] In some embodiments, the first indication information may be used to indicate at least one of the following: a first correlation ID, whether a trigger condition of the first correlation ID is satisfied, and a first mapping rule.

[0273] As an embodiment, when the first indication information indicates the first association ID, the first association ID can be used by the first terminal device to determine or update the current association ID. In other words, the first indication information can be used by the first terminal device to determine that the current association ID is the first association ID, or to update the current association ID to the first association ID.

[0274] As an implementation method, to reduce management complexity, the first indication information can prioritize reusing existing association IDs to avoid generating new IDs for each new network device. The NW side can also dynamically allocate association IDs through the first indication information, triggering an ID change (update) only when performance degrades. The triggering condition for an association ID change can be at least one of the following: a channel prediction error exceeding a threshold, a block error rate (BLER) exceeding a threshold, or a path loss deviation exceeding a corresponding threshold.

[0275] In the above embodiment, if the change condition of the association ID is triggered, the first indication information can be used to allocate a new ID to ensure that the first terminal device can adapt to the new association ID. The allocation of the new ID can be applied to different scenarios according to the transmission method of the first indication information.

[0276] As an embodiment, the transmission method of the first indication information is used to determine the update method of the current association ID. For example, when the first indication information is transmitted via RRC signaling, it can be used for long-term configuration of the association ID, which is a cell-level update method. For another example, when the first indication information is transmitted via MAC CE, it can be used for short-term adjustment of the association ID, which is a dynamic adaptation-level update method. For another example, when the first indication information is transmitted via DCI or SIB, it can be a fast update or periodic broadcast update method, respectively, as shown in Table 1.

[0277] Table 1

[0278]

[0279]

[0280] As an embodiment, the first indication information may be used to indicate whether the current inference condition satisfies the triggering condition of the first correlation ID. The triggering condition of the first correlation ID belongs to the first condition, which will be described in detail below in conjunction with step S920.

[0281] As an embodiment, the first indication information may be used to indicate a first mapping rule. The first mapping rule may refer to a mapping rule between an inference condition / environmental measurement result and an association ID. The first terminal device may determine the first mapping rule based on the first indication information and determine the first association ID based on the current inference condition and / or environmental measurement result.

[0282] As an implementation manner, the network device may predefine a first mapping rule and share the first mapping rule with multiple terminal devices. Based on the first mapping rule, the network device does not need to frequently transmit specific association IDs.

[0283] As an implementation, the first mapping rule may indicate a correspondence between a combination of multiple inference conditions and an associated ID. The combination of multiple inference conditions may be a combination of specific condition parameters or a combination of parameter ranges, which is not limited here.

[0284] As an example, for ID = f (antenna tilt angle, TxRU mapping); the fixed mapping ID of [tilt angle = 102°, TxRU = [2×8×2]] can be set to ID_01, or the fixed mapping ID of [tilt angle within the range of [100°~110°], TxRU = [2×2]] can be set to ID_11.

[0285] As an implementation manner, the first terminal device may derive the current association ID through environmental measurements.

[0286] In step S920, the first terminal device determines a first association identifier ID.

[0287] In some embodiments, the first terminal device may determine the first association ID based on the first indication information. In other words, the first terminal device may implement an association ID mechanism based on network allocation. This has been described in step S910 and will not be repeated here.

[0288] In some embodiments, the first terminal device can independently determine the first association ID. In some scenarios, the first terminal device can assume that the downlink Tx beam or beam set / list using the same association ID has similar properties. However, the network side may not know the exact quantitative value that two data samples should be classified using the same association ID. For example, the network device may not be sure whether 3° and 7° should be classified using the same correlation ID or different correlation IDs. Therefore, the terminal device needs to report.

[0289] As an embodiment, the first terminal device may execute the association ID mechanism based on a reporting method. As an embodiment, the first terminal device may send its capability report or related auxiliary information to the network device. The capability report or auxiliary information may indicate the association ID mechanism related information supported by the first terminal device. The network device may send the first indication information based on the capability report or auxiliary information.

[0290] The first association ID is used for CSI prediction. Alternatively, the first terminal device may perform CSI prediction based on the first association ID, or the inference result of the CSI prediction may be determined based on the first association ID. As previously described, in the association ID mechanism, the association ID may correspond to one or more factors that affect the CSI prediction result, thereby improving prediction accuracy in specific scenarios.

[0291] In some embodiments, the first association ID may correspond to a network device. The first association ID may be one of multiple association IDs corresponding to the network device. In other words, the network side may assign multiple association IDs to each network device to cope with different communication scenarios.

[0292] In some embodiments, the first association ID may correspond to the current inference condition, so that the first terminal device can perform CSI prediction based on the first association ID according to the current inference condition. When the current inference condition is not applicable to the generalized model or the current first model, the first association ID can effectively avoid the problem of reduced prediction accuracy due to the special inference condition.

[0293] As an embodiment, the current first model may be an AI / ML model for CSI prediction. As an AI / ML model, the first model may be updated based on certain training data.

[0294] In one embodiment, the current inference condition may include at least one of the following: an antenna tilt angle of the network device, a TxRU mapping method, and a transmission scenario between the first terminal device and the network device. The antenna tilt angle of the network device and the TxRU mapping method may refer to network-side conditions of the association ID. The transmission scenario may represent the relative positional relationship between the first terminal device and the network device.

[0295] As an embodiment, the first association ID may be one of multiple association IDs.

[0296] As an implementation manner, multiple association IDs correspond one-to-one to multiple inference conditions, that is, a unique association ID is assigned to each inference condition (such as tilt angle, TxRU mapping).

[0297] As another implementation, multiple association IDs correspond to multiple reasoning condition groups, respectively. Each reasoning condition group may include a combination of two or three of the three conditions of the current reasoning condition, and form a corresponding association ID.

[0298] It should be understood that each reasoning condition group may also be a combination of any multiple of the reasoning conditions described above.

[0299] As another implementation, multiple association IDs correspond one-to-one to parameter ranges of multiple inference conditions. One association ID can correspond to one parameter range of one inference condition. One parameter range of an inference condition can represent network conditions with similar impacts. That is, network conditions with similar impacts are grouped into one group and represented by a single association ID. For example, ID_Group_A can represent all conditions with inclination angles in the range of [100°~110°]. ID_Group_B can represent the condition that all TxRUs are mapped to [2×8×2, 4×4×2]. ID_Group_C can represent the condition that all inclination angles are in the range of [100°~110°] and TxRUs are mapped to [2×8×2, 4×4×2].

[0300] In some embodiments, using the first association ID for CSI prediction may include at least one of the following: input parameters for CSI prediction are determined according to the first association ID, inference results are determined according to the first association ID, and updates of the first model corresponding to the CSI prediction are related to the first association ID.

[0301] As an example, if the input parameters for CSI prediction are determined solely based on the first association ID, this means that the input parameters are adjusted to adapt to current reasoning conditions or environmental measurement results while the first model remains unchanged. The first sample set may include one or more sample subsets for specific scenarios. The first sequence of input parameters in the sample subsets can be used to correct the prediction input for the specific scenario.

[0302] As an example, if the CSI prediction inference result is determined based solely on the first correlation ID, this means that the inference result is adjusted to adapt to current inference conditions or environmental measurement results while the first model remains unchanged. The first sample set may include one or more sample subsets for specific scenarios. The second sequence of output labels in these sample subsets can be used to correct the CSI prediction result for specific scenarios.

[0303] As an embodiment, the first association ID can be used by the first terminal device to update the current first model. The first model can be the generalized model described above, or another model that is not applicable to the current inference conditions. The first sample set can include one or more sample subsets for specific scenarios. Sequence pairs in the sample subsets can be used to update the first model.

[0304] In some embodiments, the use of the first correlation ID is triggered based on certain conditions. The network device or first terminal device can introduce the correlation ID only when a specific condition significantly impacts model performance; or the applicable scope of the correlation ID can be flexibly adjusted based on real-time performance requirements. Triggering the correlation ID mechanism based on certain conditions can reduce the level of detail of the correlation ID, thereby reducing complexity.

[0305] In one embodiment, in response to a first condition being met, the first terminal device determines a first association ID. That is, use of the first association ID is triggered based on whether the first condition is met. If the first condition is met, the first terminal device performs CSI prediction based on the first association ID. If the first condition is not met, the first terminal device directly performs CSI prediction based on the deployed first model.

[0306] In one embodiment, the first association ID corresponds to a current inference condition. When the first condition is met, the input parameters or inference results of the CSI prediction may be modified based on the current inference condition, or the first model may be updated based on the first association ID. Because the first association ID corresponds to the current inference condition, the first terminal device may determine the current inference condition based on the first association ID and search the first sample set for one or more sequence pairs corresponding to the current inference condition.

[0307] In some embodiments, the first condition may include at least one of the following: the network device sends a first indication information for triggering the association ID; the difference between the training condition of the first model and the current reasoning condition predicted by the CSI is greater than a first threshold; the performance degradation value of the first model is greater than a second threshold; the current reasoning condition predicted by the CSI is a specific condition.

[0308] As an embodiment, the first indication information may be used to instruct the first terminal device to trigger the association ID mechanism and determine the first association ID. When a significant performance impact occurs within a specific cell, the network device may notify the first terminal device of one or more relevant association IDs via downlink signaling. The core of the association ID is to distinguish additional conditions on the network side to ensure a match between training data and inference conditions.

[0309] In the above embodiment, the first association ID may include one or more association IDs indicated by the first indication information. As mentioned above, the network device may send the first indication information via RRC, MAC CE, DCI or system information.

[0310] As an example, since the Association ID is only introduced when CSI prediction performance is affected under certain specific conditions, a threshold value can be used to determine whether the Association ID should be triggered or whether a new Association ID should be allocated. A first threshold value can be set for the difference between the training condition and the current inference condition. The relationship between this difference and the first threshold value can be used to determine whether the first condition is met. For example, if the difference is greater than the first threshold value, the first condition is met; if the difference is less than or equal to the first threshold value, the first condition is not met.

[0311] In the above embodiment, for CSI prediction, it is assumed that the prediction error of the model under different network device configurations is measured using mean squared error. The first threshold can be determined based on the mean squared error. For example, if two network devices with different configurations (e.g., different antenna tilt angles) result in a prediction error exceeding the threshold τ (the first threshold), it is necessary to trigger the association ID or assign a new association ID.

[0312] As an example, the impact on CSI prediction performance can be directly reflected as abnormal performance during the inference phase, and the threshold value can be directly set to correspond to the performance degradation. A second threshold value can be set for performance degradation of the first model. The relationship between the performance degradation value and the second threshold value can be used to determine whether the first condition is met. For example, when the performance degradation value is greater than the second threshold value, the first condition is met; when it is less than or equal to the first threshold value, the first condition is not met.

[0313] In the above embodiment, when the first terminal device detects that a significant difference between the training condition and the inference condition causes the performance to drop by more than a second threshold (for example, the performance error exceeds 3%), the association ID is dynamically enabled.

[0314] As an example, the current inference condition for CSI prediction is a specific condition, which may mean that the training conditions of the first model do not include the current inference condition, or that the current inference condition is an extreme network condition. Extreme network conditions may include special antenna tilt angles (such as very steep tilt angles) or rare environmental scenarios (such as complete NLOS scenarios). Model performance under these conditions may not be able to meet CSI prediction requirements through generalization of the model, so the specific condition is prompted by triggering the association ID.

[0315] In step S930, the first terminal device sends the inference result of the CSI prediction to the network device. The inference result of the CSI prediction can be post-processed as described above, which will not be repeated here.

[0316] As an embodiment, when the first model is not updated, the inference result of the CSI prediction is the inference result of the first model.

[0317] As an embodiment, when the first model is updated to the second model, the inference result of the CSI prediction is the inference result of the second model.

[0318] Combined with the above Figure 9This paper introduces a method for CSI prediction based on Association ID. If the network device or the first terminal device finds that certain conditions have a significant impact on model performance (for example, a specific TxRU mapping), the Association ID can be introduced only for these conditions, that is, conditional Association ID. Alternatively, in non-ideal situations (for example, certain configurations significantly affect performance), the network device or the first terminal device can also provide supplementary support through the Association ID mechanism, that is, the Association ID mechanism can be used in certain special environments.

[0319] Combined with the above Figures 1 to 9 , describes the method embodiment of the present application in detail. Figures 10 to 12 , the device embodiment of the present application is described in detail. It should be understood that the description of the device embodiment corresponds to the description of the method embodiment, so for parts not described in detail, reference can be made to the previous method embodiment.

[0320] Figure 10 1000 is a schematic block diagram of an apparatus for wireless communication according to an embodiment of the present application. The apparatus 1000 may be any of the first terminal devices described above. Figure 10 The illustrated apparatus 1000 includes a determining unit 1010 and a sending unit 1020 .

[0321] The sending unit 1010 may be configured to determine a first association ID, where the first association ID is used for CSI prediction.

[0322] The sending unit 1020 can be used to send the inference result of the CSI prediction to the network device; wherein, the input parameters of the CSI prediction are determined according to the first association ID, and / or, the inference result is determined according to the first association ID, and / or, the update of the first model corresponding to the CSI prediction is related to the first association ID.

[0323] Optionally, in response to satisfying the first condition, the first association ID is used for CSI prediction.

[0324] Optionally, the first condition includes at least one of the following: the network device sends a first indication information for triggering the association ID; the difference between the training condition of the first model and the current reasoning condition predicted by the CSI is greater than a first threshold; the performance degradation value of the first model is greater than a second threshold; the current reasoning condition predicted by the CSI is a specific condition.

[0325] Optionally, the current inference condition includes at least one of the following: the antenna tilt angle of the network device, the transceiver unit TxRU mapping method, and the transmission scenario between the first terminal device and the network device.

[0326] Optionally, the first association ID corresponds to a current reasoning condition. When the first condition is met, the input parameter or the reasoning result is modified according to the current reasoning condition, or the first model is updated according to the first association ID.

[0327] Optionally, the device 1000 also includes a receiving unit, which can be used to receive first indication information; wherein the first indication information is used to indicate at least one of the following: a first association ID, whether the current reasoning condition meets the trigger condition of the first association ID, and a first mapping rule.

[0328] Optionally, when the first indication information indicates a first association ID, the first association ID is used by the first terminal device to update a current association ID, and the first association ID is one of multiple association IDs corresponding to the network device.

[0329] Optionally, the transmission method of the first indication information is used to determine the update method of the current association ID.

[0330] Optionally, the first mapping rule is used by the first terminal device to determine the first association ID according to the current reasoning condition and / or the environmental measurement result.

[0331] Optionally, the first indication information is carried in one or more of the following: RRC signaling, MAC CE, DCI, and system information.

[0332] Optionally, the first association ID is one of multiple association IDs, and the multiple association IDs correspond one-to-one to multiple reasoning conditions, or the multiple association IDs correspond one-to-one to multiple reasoning condition groups, or the multiple association IDs correspond one-to-one to parameter ranges of multiple reasoning conditions.

[0333] Optionally, the first model is an artificial intelligence (AI) model or a machine learning (ML) model for CSI prediction.

[0334] Figure 11 FIG1 is a schematic block diagram of another apparatus for wireless communication according to an embodiment of the present application. The apparatus 1100 may be any of the network devices described above. Figure 11 The illustrated apparatus 1100 includes a receiving unit 1110 .

[0335] Receiving unit 1110 can be used to receive the inference result of CSI prediction from the first terminal device; wherein, the first association ID determined by the first terminal device is used for CSI prediction; the input parameters of the CSI prediction are determined according to the first association ID, and / or, the inference result is determined according to the first association ID, and / or, the update of the first model corresponding to the CSI prediction is related to the first association ID.

[0336] Optionally, in response to satisfying the first condition, the first association ID is used for CSI prediction.

[0337] Optionally, the first condition includes at least one of the following: the network device sends a first indication information for triggering the association ID; the difference between the training condition of the first model and the current reasoning condition predicted by the CSI is greater than a first threshold; the performance degradation value of the first model is greater than a second threshold; the current reasoning condition predicted by the CSI is a specific condition.

[0338] Optionally, the current inference condition includes at least one of the following: an antenna tilt angle of the network device, a TxRU mapping method, and a transmission scenario between the first terminal device and the network device.

[0339] Optionally, the first association ID corresponds to a current reasoning condition. When the first condition is met, the input parameter or the reasoning result is modified according to the current reasoning condition, or the first model is updated according to the first association ID.

[0340] Optionally, the apparatus 1100 further includes a sending unit, which can be used to send a first indication message to the first terminal device; wherein the first indication message is used to indicate at least one of the following: a first association ID, whether the current inference condition meets the trigger condition of the first association ID, and a first mapping rule.

[0341] Optionally, when the first indication information indicates a first association ID, the first association ID is used by the first terminal device to update a current association ID, and the first association ID is one of multiple association IDs corresponding to the network device.

[0342] Optionally, the transmission method of the first indication information is used to determine the update method of the current association ID.

[0343] Optionally, the first mapping rule is used by the first terminal device to determine the first association ID according to the current reasoning condition and / or the environmental measurement result.

[0344] Optionally, the first indication information is carried in one or more of the following: RRC signaling, MAC CE, DCI, and system information.

[0345] Optionally, the first association ID is one of multiple association IDs, and the multiple association IDs correspond one-to-one to multiple reasoning conditions, or the multiple association IDs correspond one-to-one to multiple reasoning condition groups, or the multiple association IDs correspond one-to-one to parameter ranges of multiple reasoning conditions.

[0346] Optionally, the first model is an AI model or an ML model for CSI prediction.

[0347] Figure 12 Shown is a schematic structural diagram of a communication device according to an embodiment of the present application. Figure 12The dotted line in the figure indicates that the unit or module is optional. The apparatus 1200 can be used to implement the method described in the above method embodiment. The apparatus 1200 can be a chip, a terminal device, or a network device.

[0348] The device 1200 may include one or more processors 1210. The processor 1210 may support the device 1200 to implement the method described in the above method embodiment. The processor 1210 may be a general-purpose processor or a special-purpose processor. For example, the processor may be a central processing unit (CPU). Alternatively, the processor may be another general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic device, discrete hardware component, etc. The general-purpose processor may be a microprocessor or the processor may be any conventional processor, etc.

[0349] The apparatus 1200 may further include one or more memories 1220. The memories 1220 store programs that can be executed by the processor 1210, causing the processor 1210 to perform the methods described in the above method embodiments. The memories 1220 may be independent of the processor 1210 or integrated into the processor 1210.

[0350] The apparatus 1200 may further include a transceiver 1230. The processor 1210 may communicate with other devices or chips via the transceiver 1230. For example, the processor 1210 may transmit and receive data with other devices or chips via the transceiver 1230.

[0351] The present application also provides a computer-readable storage medium for storing a program. The computer-readable storage medium can be applied to a terminal device or network device provided in the present application, and the program enables a computer to execute the method performed by the terminal device or network device in each embodiment of the present application.

[0352] The computer-readable storage medium may be any available medium that can be read by a computer or a data storage device such as a server or data center that includes one or more available media. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, or a magnetic tape), an optical medium (e.g., a digital versatile disc (DVD)), or a semiconductor medium (e.g., a solid state drive (SSD)).

[0353] The present application also provides a computer program product. The computer program product includes a program. The computer program product can be applied to a terminal device or network device provided in the present application, and the program causes a computer to execute the method performed by the terminal device or network device in each embodiment of the present application.

[0354] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center via wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means.

[0355] The present application also provides a computer program that can be applied to a terminal device or network device provided in the present application, and enables a computer to execute the method performed by the terminal or network device in each embodiment of the present application.

[0356] The terms "system" and "network" in this application may be used interchangeably. In addition, the terms used in this application are only used to explain the specific embodiments of this application and are not intended to limit this application. The terms "first," "second," "third," and "fourth" in the specification and claims of this application and the accompanying drawings are used to distinguish different objects rather than to describe a specific order. In addition, the terms "including" and "having," as well as any variations thereof, are intended to cover non-exclusive inclusions.

[0357] In the embodiments of this application, the term "indication" may refer to a direct indication, an indirect indication, or an indication of an association. For example, "A indicates B" may refer to a direct indication of B, e.g., B can obtain information through A; it may refer to an indirect indication of B, e.g., A indicates C, e.g., B can obtain information through C; or it may refer to an association between A and B.

[0358] In the embodiments of the present application, the term "corresponding" may indicate a direct or indirect correspondence between the two, or an association relationship between the two, or a relationship between indication and indication, configuration and configuration, etc.

[0359] In the embodiments of the present application, "pre-definition" or "pre-configuration" may be implemented by pre-storing corresponding codes, tables, or other methods that can be used to indicate relevant information in a device (e.g., including a terminal device and a network device). The present application does not limit the specific implementation method. For example, pre-definition may refer to a definition in a protocol.

[0360] In the embodiments of the present application, the "protocol" may refer to a standard protocol in the communication field, for example, it may include an LTE protocol, a NR protocol, and related protocols used in future communication systems, and this application does not limit this.

[0361] In the embodiments of the present application, determining B based on A does not mean determining B only based on A. B can also be determined based on A and / or other information.

[0362] In the embodiments of this application, the term "and / or" is simply a description of the association relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this document generally indicates that the related objects are in an "or" relationship.

[0363] In the embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0364] 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.

[0365] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0366] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0367] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. A method for wireless communication, characterized in that include: The first terminal device determines a first association identifier ID, where the first association ID is used for channel state information CSI prediction; The first terminal device sends the inference result of the CSI prediction to the network device; The input parameters of the CSI prediction are determined according to the first association ID, and / or the inference result is determined according to the first association ID, and / or the update of the first model corresponding to the CSI prediction is related to the first association ID.

2. The method according to claim 1, characterized in that In response to a first condition being met, the first association ID is used for the CSI prediction.

3. The method according to claim 2, characterized in that The first condition includes at least one of the following: The network device sends first indication information for triggering the association ID; a difference between a training condition of the first model and a current inference condition of the CSI prediction is greater than a first threshold; The performance degradation value of the first model is greater than a second threshold; The current reasoning condition for the CSI prediction is a specific condition.

4. The method according to claim 3, characterized in that The current inference condition includes at least one of the following: the antenna tilt angle of the network device, the transceiver unit TxRU mapping method, and the transmission scenario between the first terminal device and the network device.

5. The method according to claim 3 or 4, characterized in that The first association ID corresponds to the current reasoning condition. When the first condition is met, the input parameter or the reasoning result is modified according to the current reasoning condition, or the first model is updated according to the first association ID.

6. The method according to any one of claims 1 to 5, characterized in that The method further comprises: The first terminal device receives first indication information; The first indication information is used to indicate at least one of the following: the first association ID, whether the current inference condition meets the triggering condition of the first association ID, and the first mapping rule.

7. The method according to claim 6, characterized in that When the first indication information indicates the first association ID, the first association ID is used by the first terminal device to update the current association ID, and the first association ID is one of multiple association IDs corresponding to the network device.

8. The method according to claim 7, characterized in that The transmission method of the first indication information is used to determine the update method of the current association ID.

9. The method according to claim 6, characterized in that The first mapping rule is used by the first terminal device to determine the first association ID according to the current reasoning condition and / or the environmental measurement result.

10. The method according to any one of claims 6 to 9, characterized in that The first indication information is carried in one or more of the following: radio resource control RRC signaling, media access control MAC control element CE, downlink control information DCI, and system information.

11. The method according to any one of claims 1 to 10, characterized in that The first association ID is one of multiple association IDs, and the multiple association IDs correspond one-to-one to multiple reasoning conditions, or the multiple association IDs correspond one-to-one to multiple reasoning condition groups, or the multiple association IDs correspond one-to-one to parameter ranges of multiple reasoning conditions.

12. The method according to any one of claims 1 to 11, characterized in that The first model is an artificial intelligence (AI) model or a machine learning (ML) model used for the CSI prediction.

13. A method for wireless communication, characterized in that: include: The network device receives an inference result of a channel state information (CSI) prediction from the first terminal device; In which, the first association identification ID determined by the first terminal device is used for the CSI prediction; the input parameters of the CSI prediction are determined according to the first association ID, and / or the inference result is determined according to the first association ID, and / or the update of the first model corresponding to the CSI prediction is related to the first association ID.

14. The method according to claim 13, characterized in that In response to a first condition being met, the first association ID is used for the CSI prediction.

15. The method according to claim 14, characterized in that The first condition includes at least one of the following: The network device sends first indication information for triggering the association ID; a difference between a training condition of the first model and a current inference condition of the CSI prediction is greater than a first threshold; The performance degradation value of the first model is greater than a second threshold; The current reasoning condition for the CSI prediction is a specific condition.

16. The method according to claim 15, characterized in that The current inference condition includes at least one of the following: the antenna tilt angle of the network device, the transceiver unit TxRU mapping method, and the transmission scenario between the first terminal device and the network device.

17. The method according to claim 15 or 16, characterized in that The first association ID corresponds to the current reasoning condition. When the first condition is met, the input parameter or the reasoning result is modified according to the current reasoning condition, or the first model is updated according to the first association ID.

18. The method according to any one of claims 13 to 17, characterized in that The method further comprises: The network device sends first indication information to the first terminal device; The first indication information is used to indicate at least one of the following: the first association ID, whether the current inference condition meets the triggering condition of the first association ID, and the first mapping rule.

19. The method according to claim 18, characterized in that When the first indication information indicates the first association ID, the first association ID is used by the first terminal device to update the current association ID, and the first association ID is one of multiple association IDs corresponding to the network device.

20. The method according to claim 19, characterized in that The transmission method of the first indication information is used to determine the update method of the current association ID.

21. The method according to claim 18, wherein The first mapping rule is used by the first terminal device to determine the first association ID according to the current reasoning condition and / or the environmental measurement result.

22. The method according to any one of claims 18 to 21, characterized in that The first indication information is carried in one or more of the following: radio resource control RRC signaling, media access control MAC control element CE, downlink control information DCI, and system information.

23. The method according to any one of claims 13 to 22, characterized in that The first association ID is one of multiple association IDs, and the multiple association IDs correspond one-to-one to multiple reasoning conditions, or the multiple association IDs correspond one-to-one to multiple reasoning condition groups, or the multiple association IDs correspond one-to-one to parameter ranges of multiple reasoning conditions.

24. The method according to any one of claims 13 to 23, characterized in that The first model is an artificial intelligence (AI) model or a machine learning (ML) model used for the CSI prediction.

25. A device for wireless communication, characterized in that: The apparatus is a first terminal device, and the apparatus includes: a determining unit, configured to determine a first association identifier ID, where the first association ID is used for channel state information CSI prediction; A sending unit, configured to send the inference result of the CSI prediction to a network device; The input parameters of the CSI prediction are determined according to the first association ID, and / or the inference result is determined according to the first association ID, and / or the update of the first model corresponding to the CSI prediction is related to the first association ID.

26. The device according to claim 25, characterized in that In response to a first condition being met, the first association ID is used for the CSI prediction.

27. The device according to claim 26, characterized in that The first condition includes at least one of the following: The network device sends first indication information for triggering the association ID; a difference between a training condition of the first model and a current inference condition of the CSI prediction is greater than a first threshold; The performance degradation value of the first model is greater than a second threshold; The current reasoning condition for the CSI prediction is a specific condition.

28. The device according to claim 27, characterized in that The current inference condition includes at least one of the following: the antenna tilt angle of the network device, the transceiver unit TxRU mapping method, and the transmission scenario between the first terminal device and the network device.

29. The device according to claim 27 or 28, characterized in that The first association ID corresponds to the current reasoning condition. When the first condition is met, the input parameter or the reasoning result is modified according to the current reasoning condition, or the first model is updated according to the first association ID.

30. The device according to any one of claims 25 to 29, characterized in that The device further comprises: A receiving unit, configured to receive first indication information; The first indication information is used to indicate at least one of the following: the first association ID, whether the current inference condition meets the triggering condition of the first association ID, and the first mapping rule.

31. The device according to claim 30, characterized in that When the first indication information indicates the first association ID, the first association ID is used by the first terminal device to update the current association ID, and the first association ID is one of multiple association IDs corresponding to the network device.

32. The device according to claim 31, characterized in that The transmission method of the first indication information is used to determine the update method of the current association ID.

33. The device according to claim 32, characterized in that The first mapping rule is used by the first terminal device to determine the first association ID according to the current reasoning condition and / or the environmental measurement result.

34. The device according to any one of claims 30 to 33, characterized in that The first indication information is carried in one or more of the following: radio resource control RRC signaling, media access control MAC control element CE, downlink control information DCI, and system information.

35. The device according to any one of claims 25 to 34, characterized in that The first association ID is one of multiple association IDs, and the multiple association IDs correspond one-to-one to multiple reasoning conditions, or the multiple association IDs correspond one-to-one to multiple reasoning condition groups, or the multiple association IDs correspond one-to-one to parameter ranges of multiple reasoning conditions.

36. The device according to any one of claims 25 to 35, characterized in that The first model is an artificial intelligence (AI) model or a machine learning (ML) model used for the CSI prediction.

37. A device for wireless communication, characterized in that: The device is a network device, and the device includes: A receiving unit, configured to receive an inference result of a channel state information (CSI) prediction from a first terminal device; In which, the first association identification ID determined by the first terminal device is used for the CSI prediction; the input parameters of the CSI prediction are determined according to the first association ID, and / or the inference result is determined according to the first association ID, and / or the update of the first model corresponding to the CSI prediction is related to the first association ID.

38. The device according to claim 37, characterized in that In response to a first condition being met, the first association ID is used for the CSI prediction.

39. The device according to claim 38, characterized in that The first condition includes at least one of the following: The network device sends first indication information for triggering the association ID; a difference between a training condition of the first model and a current inference condition of the CSI prediction is greater than a first threshold; The performance degradation value of the first model is greater than a second threshold; The current reasoning condition for the CSI prediction is a specific condition.

40. The device according to claim 39, characterized in that The current inference condition includes at least one of the following: the antenna tilt angle of the network device, the transceiver unit TxRU mapping method, and the transmission scenario between the first terminal device and the network device.

41. The device according to claim 39 or 40, characterized in that The first association ID corresponds to the current reasoning condition. When the first condition is met, the input parameter or the reasoning result is modified according to the current reasoning condition, or the first model is updated according to the first association ID.

42. The device according to any one of claims 37 to 41, characterized in that The device further comprises: A sending unit, configured to send first indication information to the first terminal device; The first indication information is used to indicate at least one of the following: the first association ID, whether the current inference condition meets the triggering condition of the first association ID, and the first mapping rule.

43. The device according to claim 42, characterized in that When the first indication information indicates the first association ID, the first association ID is used by the first terminal device to update the current association ID, and the first association ID is one of multiple association IDs corresponding to the network device.

44. The device according to claim 43, characterized in that The transmission method of the first indication information is used to determine the update method of the current association ID.

45. The device according to claim 42, characterized in that The first mapping rule is used by the first terminal device to determine the first association ID according to the current reasoning condition and / or the environmental measurement result.

46. ​​The device according to any one of claims 42 to 45, characterized in that The first indication information is carried in one or more of the following: radio resource control RRC signaling, media access control MAC control element CE, downlink control information DCI, and system information.

47. The device according to any one of claims 37 to 46, characterized in that The first association ID is one of multiple association IDs, and the multiple association IDs correspond one-to-one to multiple reasoning conditions, or the multiple association IDs correspond one-to-one to multiple reasoning condition groups, or the multiple association IDs correspond one-to-one to parameter ranges of multiple reasoning conditions.

48. The device according to any one of claims 37 to 47, characterized in that The first model is an artificial intelligence (AI) model or a machine learning (ML) model used for the CSI prediction.

49. A communication device, characterized in that The system comprises a memory and a processor, wherein the memory is used to store a program, and the processor is used to call the program in the memory to execute the method according to any one of claims 1 to 24.

50. A device, characterized in that The device comprises a processor configured to call a program from a memory to execute the method according to any one of claims 1 to 24.

51. A chip, characterized in that: The device comprises a processor configured to call a program from a memory so that a device equipped with the chip executes the method according to any one of claims 1 to 24.

52. A computer-readable storage medium, characterized in that A program is stored thereon, the program causing a computer to execute the method according to any one of claims 1 to 24.

53. A computer program product, characterized in that The method comprises a program for causing a computer to execute the method according to any one of claims 1 to 24.

54. A computer program, characterized in that The computer program causes a computer to execute the method according to any one of claims 1 to 24.

Images