Terminal, wireless communication method, base station and system

Abstract

A terminal according to one embodiment of the present disclosure has a reception unit for receiving student model information pertaining to a student model in knowledge distillation, and a transmission unit for transmitting information pertaining to an output that is based on the student model information. The embodiment of the present disclosure enables suitable overhead reduction, channel estimation, and resource utilization.

Description

[Technical Field]

[0001] The present disclosure relates to a terminal and a wireless communication method in a next-generation mobile communication system. 、 base station and systems Regarding. [Background technology]

[0002] In Universal Mobile Telecommunications System (UMTS) networks, Long Term Evolution (LTE) was specified with the aim of achieving higher data rates and lower latency (Non-Patent Document 1). Also, LTE-Advanced (3GPP Rel. 10-14) was specified with the aim of achieving higher capacity and more advanced features than LTE (Third Generation Partnership Project (3GPP) Release (Rel.) 8, 9).

[0003] Successor systems to LTE (e.g., 5th generation mobile communication system (5G), 5G+ (plus), 6th generation mobile communication system (6G), New Radio (NR), 3GPP Rel. 15 or later) are also being considered. [Prior art documents] [Non-patent literature]

[0004] [Non-Patent Document 1] 3GPP TS 36.300 V8.12.0 “Evolved Universal Terrestrial Radio Access (E-UTRA) and Evolved Universal Terrestrial Radio Access Network (E-UTRAN); Overall description; Stage 2 (Release 8)”, April 2010 Summary of the Invention [Problem to be solved by the invention]

[0005] Regarding future wireless communication technologies, the use of artificial intelligence (AI) technologies such as machine learning (ML) for network / device control and management is being considered.

[0006] It is desirable to be able to appropriately control AI model training / model inference in terminals (user terminals, User Equipment (UE)). However, the specific content of such control has not yet been studied. If this control is not properly specified, it may be impossible to achieve appropriate overhead reduction / high-precision channel estimation / high-efficiency resource utilization, which may hinder improvements in communication throughput / communication quality.

[0007] Therefore, the present disclosure provides a terminal and a wireless communication method that can achieve suitable overhead reduction / channel estimation / resource utilization. 、 base station and systems One of the aims is to provide [Means for solving the problem]

[0008] A terminal according to one aspect of the present disclosure includes a receiving unit that receives student model information related to a student model in knowledge distillation, and a processing unit that receives a student model based on the student model information. Execute an inference process using the student model, Information about the output Wireless communication to the base station a transmitting unit for transmitting the The student model information includes an index of the student model, and the information about the output of the inference process includes an index of a student model that is a basis for a student model to be retrained at the base station. . [Effects of the Invention]

[0009] According to one aspect of the present disclosure, it is possible to achieve favorable overhead reduction / channel estimation / resource utilization. [Brief explanation of the drawings]

[0010] [Figure 1]FIG. 1 is a diagram illustrating an example of a framework for managing AI models. [Figure 2] FIG. 2 is a diagram illustrating an example of an AI model. [Figure 3] FIG. 3 is a diagram illustrating an example of designation of an AI model according to the first embodiment. [Figure 4] FIG. 4 is a diagram illustrating an example of updating an AI model according to the first embodiment. [Figure 5] FIG. 5 is a diagram illustrating an example of fine tuning. [Figure 6] FIG. 6 is a diagram illustrating an example of transfer learning. [Figure 7] FIG. 7 is a diagram illustrating an example of updating an AI model according to the first embodiment. [Figure 8] FIG. 8 is a diagram illustrating an example of designation of an AI model according to the second embodiment. [Figure 9] FIG. 9 is a diagram illustrating an example of updating an AI model according to the second embodiment. [Figure 10] FIG. 10 is a diagram illustrating an example of updating an AI model according to the second embodiment. [Figure 11] FIG. 11 is a diagram illustrating an example of designation of an AI model according to the third embodiment. [Figure 12] FIG. 12 is a diagram illustrating an example of designation of an AI model according to the third embodiment. [Figure 13] 13A and 13B are diagrams illustrating an example of UE training AI model information according to the third embodiment. [Figure 14] 14A and 14B are diagrams illustrating an example of UE training AI model information according to the third embodiment. [Figure 15] FIG. 15 is a diagram illustrating an example of exchange of AI model information regarding federated learning. [Figure 16] 16A to 16C are diagrams showing an example of updated AI model information. [Figure 17] 17A and 17B are diagrams illustrating an example of learning using knowledge distillation. [Figure 18]FIG. 18 is a diagram illustrating an example of a schematic configuration of a wireless communication system according to an embodiment. [Figure 19] FIG. 19 is a diagram illustrating an example of the configuration of a base station according to an embodiment. [Figure 20] FIG. 20 is a diagram illustrating an example of the configuration of a user terminal according to an embodiment. [Figure 21] FIG. 21 is a diagram illustrating an example of the hardware configuration of a base station and a user terminal according to an embodiment. [Figure 22] FIG. 22 is a diagram illustrating an example of a vehicle according to an embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0011] (Application of Artificial Intelligence (AI) technology to wireless communications) Regarding future wireless communication technologies, the use of AI technologies such as machine learning (ML) for network / device control and management is being considered.

[0012] For example, for future wireless communication technologies, the use of AI techniques is being considered to improve channel state information (CSI) feedback (e.g., reducing overhead, improving accuracy, prediction), improve beam management (e.g., improving accuracy, prediction in the time / space domain), and improve positioning (e.g., improving position estimation / prediction).

[0013] Figure 1 shows an example of a framework for managing AI models. In this example, each stage related to an AI model is shown as a block. This example is also referred to as lifecycle management of an AI model.

[0014] The data collection stage corresponds to a stage of collecting data for generating / updating an AI model. The data collection stage may include data reduction (e.g., determining which data to transfer for model training / model inference), data transfer (e.g., transferring data to an entity (e.g., UE, gNB) that performs model training / model inference), etc.

[0015] In the model training stage, model training is performed based on the data (training data) transferred from the collection stage. This stage may include data preparation (e.g., performing data preprocessing, cleaning, formatting, transformation, etc.), model training / validation, model testing (e.g., verifying that the trained model meets a performance threshold), model exchange (e.g., transferring the model for distributed learning), and model deployment / update (deploying / updating the model to the entity that will perform model inference).

[0016] In the model inference stage, model inference is performed based on the data (inference data) transferred from the collection stage. This stage may include data preparation (e.g., performing data preprocessing, cleaning, formatting, transformation, etc.), model inference, model performance feedback (feeding back model performance to the entity training the model), and output (providing model output to the actor).

[0017] An actor stage may include action triggers (e.g., deciding whether to trigger an action on another entity), feedback (e.g., feeding back information needed for training data / inference data / performance feedback), etc.

[0018] For example, training of a model for mobility optimization may be performed in, for example, the Operation, Administration and Maintenance (Management) / gNodeB (gNB) in the network (NW). The former has advantages in interoperability, large-capacity storage, operator manageability, and model flexibility (feature engineering, etc.). The latter has advantages in that it does not require latency for model updates or data exchange for model deployment. Inference of the above model may be performed in, for example, the gNB.

[0019] Also, depending on the use case, the entity that performs the training / inference may be different.

[0020] For example, for AI-assisted beam management based on measurement reports, the OAM / gNB may perform model training and the gNB may perform model inference.

[0021] For AI-assisted UE-assisted positioning, a Location Management Function (LMF) may perform model training and the LMF may perform model inference.

[0022] For CSI feedback / channel estimation using an autoencoder, the OAM / gNB / UE may perform model training, and the gNB / UE may (jointly) perform model inference.

[0023] For AI-assisted beam management or AI-assisted UE-based positioning based on beam measurements, the OAM / gNB / UE may perform model training and the UE may perform model inference.

[0024] As mentioned above, it is preferable to be able to train AI models in the UE, as the UE may have direct access to data that is not directly available to the gNB / OAM. For example, model training of an autoencoder for CSI feedback requires complete downlink CSI information (or channel information), which is only directly accessible by the UE.

[0025] If the autoencoder model is trained in the gNB / OAM, the UE needs to transmit complete downlink CSI information (or channel information), which results in large communication overhead. On the other hand, if the autoencoder model is trained in the UE, the UE needs to store complete downlink CSI information (or channel information), which results in poor storage and computation resources in the UE compared to the gNB / OAM.

[0026] For these reasons, it is desirable to be able to appropriately control model training / model inference in UE. However, the specific content of this control has not yet been studied. If this control is not properly specified, it may be impossible to achieve appropriate overhead reduction / high-accuracy channel estimation / high-efficiency resource utilization, which may hinder improvements in communication throughput / communication quality.

[0027] Therefore, the present inventors have conceived a suitable method for controlling model training / model inference. Note that each embodiment of the present disclosure may be applied when AI / prediction is not used.

[0028] In one embodiment of the present disclosure, a terminal (user equipment (UE)) / base station (BS) trains an ML model in a training mode and executes the ML model in an inference mode (also called an inference mode). In the inference mode, the accuracy of the ML model trained in the training mode may be validated.

[0029] In the present disclosure, the UE / BS may input channel state information, reference signal measurements, etc. to the ML model and output highly accurate channel state information / measurements / beam selection / position, future channel state information / radio link quality, etc.

[0030] In this disclosure, AI may be interpreted as an object (also referred to as a subject, object, data, function, program, etc.) that has (performs) at least one of the following characteristics: · inferences based on observed or collected information; · making choices based on information observed or collected; · Predictions based on observed or collected information.

[0031] In the present disclosure, an object may be, for example, an apparatus, device, etc., such as a terminal or a base station. Also, in the present disclosure, an object may correspond to a program / model / entity that operates in the apparatus.

[0032] In addition, in the present disclosure, an ML model may be interpreted as an object having (implementing) at least one of the following characteristics: - Producing estimates by feeding information, · Predicting estimates by giving information, · Discover features by providing information, · Selecting behavior by providing information.

[0033] In addition, in this disclosure, the terms ML model, model, AI model, predictive analytics, predictive analysis model, etc. may be interchangeable. The ML model may be derived using at least one of regression analysis (e.g., linear regression analysis, multiple regression analysis, logistic regression analysis), support vector machine, random forest, neural network, deep learning, etc. In this disclosure, the term model may be interchangeable with at least one of encoder, decoder, tool, etc.

[0034] Based on input information, the ML model outputs at least one piece of information, such as an estimate, a prediction, a selected action, or a classification.

[0035] ML models may include supervised learning, unsupervised learning, and reinforcement learning. Supervised learning may be used to learn general rules that map inputs to outputs. Unsupervised learning may be used to learn features of data. Reinforcement learning may be used to learn behaviors to maximize a goal.

[0036] In this disclosure, terms such as generate, calculate, derive, etc. may be interchangeable. In this disclosure, terms such as implement, operate, operate, execute, etc. may be interchangeable. In this disclosure, terms such as train, learn, update, retrain, etc. may be interchangeable. In this disclosure, terms such as infer, after-training, live use, actual use, etc. may be interchangeable. Signal may be interchangeable with signal / channel.

[0037] In this disclosure, the training mode may refer to a mode in which the UE / BS transmits / receives signals for an ML model (i.e., a mode of operation during training). In this disclosure, the inference mode may refer to a mode in which the UE / BS implements an ML model (e.g., implements a trained ML model to predict outputs) (i.e., a mode of operation during inference).

[0038] In the present disclosure, training mode may refer to a mode in which a particular signal transmitted in inference mode is transmitted with a large overhead (e.g., a large amount of resources).

[0039] In the present disclosure, the training mode may refer to a mode that refers to a first configuration (e.g., a first DMRS configuration, a first CSI-RS configuration, or a first CSI reporting configuration). In the present disclosure, the inference mode may refer to a mode that refers to a second configuration (e.g., a second DMRS configuration, a second CSI-RS configuration, or a second CSI reporting configuration) that is different from the first configuration. The first configuration may be configured with at least one more time resource, frequency resource, code resource, or port (antenna port) related to measurement than the second configuration. Note that, for example, the CSI reporting configuration may include a configuration related to an autoencoder.

[0040] Hereinafter, embodiments of the present disclosure will be described in detail with reference to the drawings. Wireless communication methods according to the embodiments may be applied independently or in combination.

[0041] In the following embodiments, to explain an ML model for communication between a UE and a BS, the relevant entities are a UE and a BS, but application of each embodiment of the present disclosure is not limited to this. For example, for communication between other entities (e.g., communication between UEs), the UE and BS in the following embodiments may be read as a first UE and a second UE. In other words, the UE, BS, etc. in the present disclosure may all be read as any UE / BS.

[0042] In the present disclosure, "A / B" and "at least one of A and B" may be interpreted interchangeably. Also, in the present disclosure, "A / B / C" may mean "at least one of A, B, and C."

[0043] In the present disclosure, terms such as activate, deactivate, indicate (or indicate), select, configure, update, and determine may be read interchangeably. In the present disclosure, terms such as support, control, controllable, operate, and operate may be read interchangeably.

[0044] In the present disclosure, Radio Resource Control (RRC), RRC parameters, RRC messages, higher layer parameters, fields, information elements (IEs), settings, etc. may be interchangeable. In the present disclosure, Medium Access Control (MAC) control elements (CEs), update commands, activation / deactivation commands, etc. may be interchangeable.

[0045] In the present disclosure, higher layer signaling may be, for example, any one of Radio Resource Control (RRC) signaling, Medium Access Control (MAC) signaling, broadcast information, and the like, or a combination thereof.

[0046] In the present disclosure, MAC signaling may use, for example, a MAC Control Element (MAC CE), a MAC Protocol Data Unit (PDU), etc. Broadcast information may be, for example, a Master Information Block (MIB), a System Information Block (SIB), Remaining Minimum System Information (RMSI), Other System Information (OSI), etc.

[0047] In the present disclosure, physical layer signaling may be, for example, Downlink Control Information (DCI), Uplink Control Information (UCI), and the like.

[0048] In this disclosure, the terms index, identifier (ID), indicator, resource ID, etc. may be interchangeable. In this disclosure, the terms sequence, list, set, group, cluster, subset, etc. may be interchangeable.

[0049] In this disclosure, the terms panel, UE panel, panel group, beam, beam group, precoder, Uplink (UL) transmitting entity, Transmission / Reception Point (TRP), base station, Spatial Relation Information (SRI), spatial relation, SRS Resource Indicator (SRI), Control Resource Set (CORESET), Physical Downlink Shared Channel (PDSCH), Codeword (CW), Transport Block (TB), Reference Signal (RS), antenna port (e.g., Demodulation Reference Signal (DMRS) port), antenna port group (e.g., DMRS port group), group (e.g., spatial relation group, Code Division Multiplexing (CDM) group, reference signal group, CORESET group, Physical Uplink Control Channel (PUCCH) group, PUCCH resource group), resource (e.g., reference signal resource, SRS resource), resource set (e.g., reference signal resource set), CORESET pool, downlink Transmission Configuration Indication state (TCI state) (DL TCI state), uplink TCI state (UL TCI state), unified TCI state, common TCI state, Quasi-Co-Location (QCL), QCL assumption, etc. may be read as interchangeable.

[0050] In the present disclosure, CSI-RS, non-zero power (NZP) CSI-RS, zero power (ZP) CSI-RS, and CSI interference measurement (CSI-IM) may be interchangeable. Furthermore, CSI-RS may include other reference signals.

[0051] In this disclosure, the measured / reported RS may refer to the RS measured / reported for a CSI report.

[0052] In the present disclosure, timing, time, duration, slot, subslot, symbol, subframe, etc. may be read interchangeably.

[0053] In the present disclosure, the terms direction, axis, dimension, domain, polarization, polarization component, etc. may be read interchangeably.

[0054] In the present disclosure, the RS may be, for example, a CSI-RS, an SS / PBCH block (SS block (SSB)), etc. Also, the RS index may be a CSI-RS Resource Indicator (CRI), an SS / PBCH Block Indicator (SSBRI), etc.

[0055] In this disclosure, estimation, prediction, and inference may be used interchangeably. Also, in this disclosure, estimate, predict, and infer may be used interchangeably.

[0056] In the present disclosure, the terms autoencoder, encoder, decoder, etc. may be replaced with at least one of a model, an ML model, a neural network model, an AI model, an AI algorithm, etc. Furthermore, the term autoencoder may be replaced with any autoencoder, such as a stacked autoencoder or a convolutional autoencoder. The encoder / decoder of the present disclosure may employ a model such as a Residual Network (ResNet), a DenseNet, or a RefineNet.

[0057] In addition, in this disclosure, the terms encoder, encoding, encoding, modification / change / control by an encoder, etc. may be interchangeable. In addition, in this disclosure, the terms decoder, decoding, decoding, modification / change / control by a decoder, etc. may be interchangeable.

[0058] In the present disclosure, channel measurement / estimation may be performed using at least one of, for example, a Channel State Information Reference Signal (CSI-RS), a Synchronization Signal (SS), a Synchronization Signal / Physical Broadcast Channel (SS / PBCH) block, a Demodulation Reference Signal (DMRS), a Sounding Reference Signal (SRS), etc.

[0059] In the present disclosure, the CSI may include at least one of a Channel Quality Indicator (CQI), a Precoding Matrix Indicator (PMI), a CSI-RS Resource Indicator (CRI), a SS / PBCH Block Resource Indicator (SSBRI), a Layer Indicator (LI), a Rank Indicator (RI), Layer 1 Reference Signal Received Power (L1-RSRP), Reference Signal Received Quality (L1-RSRQ), Signal to Interference plus Noise Ratio (L1-SINR), Signal to Noise Ratio (L1-SNR), etc.

[0060] In the present disclosure, UCI, CSI report, CSI feedback, feedback information, feedback bit, etc. may be interchangeable. Also, in the present disclosure, bit, bit string, bit sequence, sequence, value, information, value obtained from a bit, information obtained from a bit, etc. may be interchangeable.

[0061] In the present disclosure, a layer (for an encoder) may be interchangeably read as a layer (such as an input layer or an intermediate layer) used in an AI model. The layer in the present disclosure may correspond to at least one of an input layer, an intermediate layer, an output layer, a batch normalization layer, a convolutional layer, an activation layer, a dense layer, a normalization layer, a pooling layer, an attention layer, a dropout layer, a fully connected layer, etc.

[0062] (Wireless communication method) In this disclosure, AI model information may refer to information including at least one of the following: - Information on the input / output of the AI ​​model, Pre-processing / post-processing information for input / output of AI models, - Information on AI model parameters, Training information for the AI ​​model (training information), Inference information for AI models, Performance information about AI models.

[0063] Here, the input / output information of the AI ​​model may include information about at least one of the following: Input / output data content (e.g., RSRP, SINR, amplitude / phase information in the channel matrix (or precoding matrix), information on the angle of arrival (AoA), information on the angle of departure (AoD), location information), Input / output data types (e.g., immutable values, floating-point numbers), Quantization interval (quantization step size) of input / output data (e.g., 1 dBm for L1-RSRP), The range that the input / output data can take (e.g., [0, 1]).

[0064] In the present disclosure, the information on AoA may include information on at least one of an azimuth angle of arrival and a zenith angle of arrival (ZoA). Also, the information on AoD may include information on at least one of an azimuth angle of departure and a zenith angle of departure (ZoD).

[0065] In the present disclosure, location information may be location information related to a UE / NW. The location information may include at least one of information (e.g., latitude, longitude, altitude) obtained using a positioning system (e.g., a satellite positioning system (e.g., Global Navigation Satellite System (GNSS), Global Positioning System (GPS)), etc.), information on a base station adjacent to (or serving) the UE (e.g., a base station / cell identifier (ID), a BS-UE distance, a direction / angle of the BS (UE) as seen from the UE (BS), coordinates of the BS (UE) as seen from the UE (BS) (e.g., coordinates on the X, Y, and Z axes), etc.), a specific address of the UE (e.g., an Internet Protocol (IP) address), etc. The location information of the UE is not limited to information based on the position of the BS, but may also be information based on a specific point.

[0066] The location information may include information about its implementation (e.g., location / position / orientation of antennas, location / orientation of antenna panels, number of antennas, number of antenna panels, etc.).

[0067] The location information may include mobility information, which may include information indicating at least one of information indicating a mobility type, a moving speed of the UE, an acceleration of the UE, and a moving direction of the UE.

[0068] Here, the mobility type may correspond to at least one of a fixed location UE, a movable / moving UE, a no mobility UE, a low mobility UE, a middle mobility UE, a high mobility UE, a cell-edge UE, a not-cell-edge UE, etc.

[0069] The pre-processing / post-processing information for the input / output of the AI ​​model may include information about at least one of the following: Whether to apply normalization (e.g., Z-score normalization, min-max normalization), Parameters for normalization (e.g., mean / variance for Z-score normalization, min / max for min-max normalization), Whether to apply a specific numeric conversion method (e.g., one hot encoding, label encoding, etc.), - Selection rules for whether or not to use as training data.

[0070] For example, for input information x, Z-score normalization (x new = (x-μ) / σ, where μ is the mean of x and σ is the standard deviation) of the normalized input information x new may be input to the AI ​​model, and the output from the AI ​​model is y out may be subjected to post-processing to obtain the final output y.

[0071] The information about the parameters of the AI ​​model may include information about at least one of the following: Weight information in AI models (e.g., neuron coefficients (connection coefficients)), · AI model structure, -Types of AI models as model components (e.g., Residual Network (ResNet), DenseNet, RefineNet, Transformer model, CRBlock, Recurrent Neural Network (RNN), Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU)), · Functions of an AI model as model components (e.g., decoder, encoder).

[0072] The weight information in the AI ​​model may include information on at least one of the following: · Bit width (size) of weight information, Quantization interval of weight information, - The range of weight information possible Weight parameters in AI models, - Information on the differences from the AI ​​model before the update (if updating), Weight initialization method (e.g., zero initialization, random initialization (based on normal / uniform / truncated normal distribution), Xavier initialization (for sigmoid function), He initialization (for Rectified Linear Units (ReLU))).

[0073] The structure of the AI ​​model may also include information about at least one of the following: Number of layers, Layer type (e.g., convolutional layer, activation layer, dense layer, normalization layer, pooling layer, attention layer), Layer information, time series specific parameters (e.g., bidirectionality, time step), Parameters for training (e.g., type of feature (L2 regularization, dropout feature, etc.), where (e.g., after which layer) to put this feature).

[0074] The layer information may include information regarding at least one of the following: The number of neurons in each layer kernel size, stride for pooling / convolutional layers, Pooling method (MaxPooling, AveragePooling, etc.), Residual block information, Number of heads, Normalization methods (batch normalization, instance normalization, layer normalization, etc.), Activation functions (Sigmoid, tanh function, ReLU, leaky ReLU information, Maxout, Softmax).

[0075] FIG. 2 is a diagram showing an example of an AI model. This example shows an AI model including a ResNet as model component #1, a Transformer model as model component #2, a dense layer, and a normalization layer. In this way, one AI model may be included as a component of another AI model. Note that FIG. 2 may also show an AI model in which processing proceeds from left to right.

[0076] The training information for the AI ​​model may include information about at least one of the following: Information for the optimization algorithm (e.g., type of optimization (Stochastic Gradient Descent (SGD)), AdaGrad, Adam, etc.), optimization parameters (learning rate, momentum information, etc.), Loss function information (e.g., information about loss function metrics (Mean Absolute Error (MAE)), Mean Square Error (MSE), Cross Entropy Loss, NLL Loss, KL Divergence, etc.)), Parameters to be frozen for training (e.g., layers, weights), Parameters to be updated (e.g. layers, weights), Parameters (e.g., layers, weights) that should be (be used as) initial parameters for training, How to train / update the AI ​​model (e.g., recommended number of epochs, batch size, number of data used for training).

[0077] The inference information for the AI ​​model may include information regarding decision tree branch pruning, parameter quantization, etc.

[0078] The performance information about the AI ​​model may include information about the expected value of a loss function defined for the AI ​​model.

[0079] AI model information regarding a specific AI model may be predetermined in a standard or may be notified to a UE from a network (NW). An AI model defined in a standard may be called a reference AI model. AI model information regarding a reference AI model may be called reference AI model information.

[0080] Note that the AI ​​model information in the present disclosure may include an index for identifying the AI ​​model (for example, may be referred to as an AI model index). The AI ​​model information in the present disclosure may include an AI model index in addition to or instead of the input / output information of the AI ​​model described above. The association between the AI ​​model index and the AI ​​model information (for example, input / output information of the AI ​​model) may be predetermined in a standard or may be notified to the UE from the NW.

[0081] First Embodiment The first embodiment relates to controlling the use of a reference AI model.

[0082] The UE may receive reference AI model information from the NW, and may deploy (make available or deploy) the reference AI model based on the received reference AI model information.

[0083] 3 is a diagram illustrating an example of specifying an AI model according to the first embodiment. In this example, AI models #1 to #3 are defined in the standard. A UE that receives AI model information indicating AI model #3 from the NW may deploy AI model #3.

[0084] The UE may directly use the reference AI model based on (eg, specified by) the received reference AI model information for inference.

[0085] The UE may also train a reference AI model based on the received reference AI model information.

[0086] In the present disclosure, the training / updating of the AI ​​model in the UE may be performed based on a data set held by the UE (for example, data obtained from measurement results).

[0087] FIG. 4 is a diagram illustrating an example of updating an AI model according to the first embodiment. In this example, AI models #1-#2 are defined in the standard. A UE that receives AI model information indicating AI model #2 from the NW may deploy AI model #2. The UE may also update AI model #2.

[0088] Here, fine tuning and transfer learning will be described regarding updating of an AI model. Fig. 5 is a diagram showing an example of fine tuning. Fig. 6 is a diagram showing an example of transfer learning.

[0089] As shown in Fig. 5, in fine-tuning, an entity performing training performs training using parameters of a trained AI model in the source domain as initial parameters of an AI model in the target domain (pre-trained AI model), and derives a trained AI model in the target domain. That is, all layers of the trained AI model in the source domain may be transferred to all layers of the AI ​​model in the target domain, or some may be used as is without being updated (may be fixed (frozen)).

[0090] As shown in FIG. 6, in transfer learning (which may also be referred to as network-based deep transfer learning), the entity performing the training reuses some of the layers trained in the source domain as layers in the target domain. These layers (front (left) layers in the figure) may be used as feature extractors. That is, only some of the layers of the AI ​​model trained in the source domain may be transferred to corresponding layers in the AI ​​model in the target domain, and these layers may be fixed (frozen) or fine-tuned. Furthermore, the remaining layers in the target domain may not have corresponding layers in the source domain and may be newly trained (or fine-tuned) in the target domain.

[0091] The UE may not update parameters (e.g., layers, weights) that are obtained from the reference AI model information and that should be frozen for training during training. The UE may update parameters that are obtained from the reference AI model information and that should be updated (fine-tuned) during training.

[0092] The UE may perform training using parameters that should be (are used as) initial parameters for training obtained from the reference AI model information.

[0093] 7 is a diagram illustrating an example of updating an AI model according to the first embodiment. In this example, the UE receives an instruction from the NW to update the reference AI model #2. However, the instruction includes information indicating that the first three layers should be frozen and the last two layers should be fine-tuned. Based on the instruction, the UE may derive an updated AI model #2 in which the first three layers of the reference AI model #2 are frozen and the last two layers are fine-tuned.

[0094] According to the first embodiment described above, the reference AI model can be appropriately used in the UE.

[0095] <Second embodiment> The second embodiment relates to control of the use of an AI model trained by a NW (hereinafter also referred to as a NW-trained AI model).

[0096] The NW may train a reference AI model. The NW may transmit AI model information about the trained reference AI model (in other words, the NW-trained AI model) to the UE. The AI ​​model information may be referred to as NW-trained AI model information.

[0097] The UE may receive, from the NW, reference AI model information on which the NW training AI model information is based. The UE may also receive NW training AI model information from the NW. The NW training AI model information may include any one or a combination of the following information instead of or in addition to the AI ​​model information already described: The reference AI model (index) on which the NW-trained AI model was based, -Information on the difference of the NW-trained AI model from the reference AI model.

[0098] The UE may deploy (make available or deploy) the NW training AI model based on the received NW training AI model information.

[0099] 8 is a diagram illustrating an example of specifying an AI model according to the second embodiment. In this example, reference AI models #1 and #2 are defined in the standard. The network trains the reference AI model #2 and derives the network training AI model #2.

[0100] The NW notifies the UE of AI model information indicating the reference AI model #2, and also notifies the UE of information about the NW training AI model #2 (e.g., information about updated differences). The UE that receives this information can determine the NW training AI model #2 taking into account the information about the updated differences for the reference AI model #2.

[0101] The UE may use the NW training AI model based on the received NW training AI model information as is for inference.

[0102] The UE may also train the NW training AI model based on the received NW training AI model information.

[0103] FIG. 9 is a diagram illustrating an example of updating an AI model according to the second embodiment. In this example, similar to FIG. 8, it is assumed that the UE has determined the NW training AI model #2. The UE may further update the NW training AI model #2. For example, the UE may update the same layers of the NW training AI model #2 as the original reference AI model #2 (the first three layers shown in the figure) based on the data set that the UE has.

[0104] The UE may not update parameters (e.g., layers, weights) that are obtained from the NW training AI model information and that should be frozen for training during training. The UE may update parameters that are obtained from the NW training AI model information and that should be updated during training.

[0105] The UE may perform training using parameters that should be (are used as) initial parameters for training obtained from the NW training AI model information.

[0106] 10 is a diagram illustrating an example of updating an AI model according to the second embodiment. In this example, the UE receives an instruction from the NW to update the NW-trained AI model #2. However, the instruction includes information indicating that the first three layers should be frozen and the last two layers should be fine-tuned. Based on the instruction, the UE may derive an updated AI model #2 in which the first three layers of the NW-trained AI model #2 are frozen and the last two layers are fine-tuned.

[0107] According to the second embodiment described above, the NW training AI model can be appropriately used in the UE.

[0108] <Third embodiment> The third embodiment relates to controlling the use of an AI model trained by a UE (hereinafter also referred to as a UE-trained AI model).

[0109] The UE may train the reference AI model / NW training AI model to derive the UE training AI model.

[0110] The UE may transmit AI model information regarding the UE training AI model (hereinafter also referred to as UE training AI model information) to the NW using physical layer signaling (e.g., UCI), higher layer signaling (e.g., RRC signaling, MAC CE), a specific signal / channel, or a combination thereof. The UE training AI model information may include any one or a combination of the following information instead of or in addition to the AI ​​model information already described: -Reference AI model / NW training AI model (index) on which the UE training AI model is based, -Information on the difference between the UE training AI model and the reference AI model / NW training AI model.

[0111] FIG. 11 is a diagram illustrating an example of specifying an AI model according to the third embodiment. In this example, the UE derives a UE training AI model. The UE transmits UE training AI model information (which may include performance information of the AI ​​model) to the NW. The NW may apply the UE training AI model indicated by the information.

[0112] FIG. 12 is a diagram showing an example of specifying an AI model according to the third embodiment. In this example, the UE derives a UE training AI model from a certain reference AI model. The differences between these models are some of the layers colored in the diagram. The UE transmits UE training AI model information to the NW. This information may include information on the reference AI model, information on the differences, and (performance information of the AI ​​model). The NW can determine the UE training AI model by taking into account the updated difference information for the reference AI model indicated by this information.

[0113] The UE may report information regarding which weights / layers are updated (eg, in the UE training AI model information).

[0114] 13A-13B and 14A-14B are diagrams illustrating an example of UE training AI model information according to the third embodiment. In this example, the UE trains an autoencoder (encoder / decoder) and reports the UE training AI model information for the decoder to the NW as needed.

[0115] FIG. 13A corresponds to the case where the UE does not train an AI model, and since both the encoder and decoder are not updated, UE training AI model information is not transmitted to the NW.

[0116] Figure 13B corresponds to the case where only the encoder is updated. The UE may not include decoder information in the UE training AI model information, but may include information on the estimation performance of the updated model, for example. The case of Figure 13B can be used when encoder modification alone is sufficient to address changes in the statistical properties of the input.

[0117] Figure 14A corresponds to a case where the encoder and part of the decoder are updated. The UE may include information about the decoder (information about the updated part) in the UE training AI model information, for example, information about the estimation performance of the updated model, and transmit it. The case of Figure 14A can be used when relatively small modifications to the decoder are sufficient to respond to changes in the statistical properties of the input.

[0118] Figure 14B corresponds to the case where both the encoder and decoder are updated. The UE may include decoder information in the UE training AI model information, such as information on the estimated performance of the updated model, a reference model report, etc. The case of Figure 14B may be used when a relatively large modification of the decoder is required due to changes in the statistical properties of the input.

[0119] From the viewpoint of reducing communication overhead, it is preferable that the UE applies the report of Fig. 13B when the performance is sufficient in Fig. 13B, the report of Fig. 14A when the performance is sufficient in Fig. 14A, and the report of Fig. 14B when the performance is insufficient in Fig. 14A. In this way, the UE may dynamically control the communication overhead involved in notifying the AI ​​model.

[0120] [Distributed learning] The transmission of the UE training AI model information in the third embodiment may be used for distributed learning performed using a server and multiple clients in a network. In this case, the UE training AI model information may be referred to as client training AI model information.

[0121] Distributed learning may be federated learning, which includes at least one of the following steps: Each client reports its computational resources to the server. The server provides the global AI model to selected clients. Each selected client trains a local AI model based on the global AI model and its own data; Each selected client reports the parameters of its local AI model to the server. The server combines the reported local AI models to update a global AI model and provides the updated global AI model to one or more clients (e.g., all clients).

[0122] At least some of the above steps may be repeated. Figure 15 is a diagram showing an example of AI model information exchange for federated learning.

[0123] For example, in the iteration of the i-th iteration (i is an integer), the server (in this example, the NW) creates a global AI model G i The client then sends the AI ​​model information to the client (UE in this example). The client then creates a global AI model G based on the data set it has. i Update the local AI model L i The client obtains the local AI model L i The updated information about i is fed back (reported) to the server.

[0124] The server generates a global AI model G based on updates i from one or more clients. i Update the global AI model G i+1 In this way, associative learning can proceed.

[0125] The NW may send AI model information related to the global AI model to the UE. The AI ​​model information may be referred to as global AI model information. The global AI model information may include reference AI model information.

[0126] Global AI model information may include only any or a combination of the following information: - Reference AI model (index) on which the global AI model is based, Global AI model (index), -Information on the difference of the global AI model from the reference AI model.

[0127] In addition, the Global AI Model G i+1 For AI model information, please visit Global AI Model G i It may also include differential information from the AI ​​model (information).

[0128] The update information i may include a model ID for the global AI model / local AI model, or the global AI model G i It may include information on the difference from the AI ​​model (information) or may correspond to updated AI model information, which will be described later in the fourth embodiment.

[0129] The UE may update the local AI model L in update i or instead of update i. i Intermediate information (e.g., gradients, features, etc.) may be reported.

[0130] The server may be a gNB / OAM / LMF / UE, and the client may be a UE / gNB. For example, when the UE is the client, there is an advantage that it is not necessary to transmit a large amount of data for training to the gNB. For example, when the gNB is the client, there is an advantage that it is not necessary to transmit a large amount of data for training to the OAM / LMF.

[0131] In the present disclosure, the server, NW, gNB, OAM, LMF, and UE may be interchangeable. In the present disclosure, the client, NW, UE, and gNB may be interchangeable. In the present disclosure, the gNB / UE may be interchangeable with other devices in the network / core network.

[0132] A client (e.g., a UE) may report information about resources for AI model training in a current iteration (hereinafter also referred to as resource information) to a server (e.g., a gNB). The resource information may include information about at least one of the following: capabilities related to training AI models (e.g., computational resources); Information about the datasets available (e.g., structure / type / format / amount of datasets), Information about the client (e.g., UE status).

[0133] The above-mentioned computational resources may indicate, for example, at least one of the number of available floating-point operations per second (FLOPS), Central Processing Unit (CPU) / Graphics Processing Unit (GPU) performance, available memory amount, etc.

[0134] The information about the client may include information about at least one of the following: Information about the type of client (device) (e.g., device model name / model number, communication chip name / model number, etc.), Estimated outcome / quality (e.g. Block Error Rate (BLER) / CQI / RSRP / RSRQ / SINR), The willingness of clients to handle training (e.g., whether one or more clients are willing to handle training of the AI ​​model); The remaining amount of data that the client can exchange with the server (e.g., the amount of data that the client can exchange), Battery power remaining.

[0135] According to the third embodiment described above, the UE training AI model can be appropriately used.

[0136] <Fourth embodiment> The fourth embodiment relates to information on the difference between an AI model and another AI model (information on the difference between another AI model based on an AI model). The difference information may also be called updated AI model information.

[0137] The updated AI model information may be, for example, at least one of the following information already mentioned: - Information on the differences between the AI ​​model before the update and the AI ​​model after the update, - Information on the difference between the network-trained AI model and the reference AI model, -Information on the difference between the UE training AI model and the reference AI model / NW training AI model.

[0138] The updated AI model information may include weight information expressed in absolute values, or may include weight information expressed in differential values ​​between the AI ​​model before the update and the AI ​​model after the update.

[0139] 16A-16C are diagrams showing an example of updated AI model information. In this example, updated AI model information is shown when a NW training AI model is trained from the reference model shown in FIG. 16A.

[0140] 16B shows weight information expressed as absolute values ​​included in updated AI model information. In this example, the last layer has not been updated and therefore may not be reported. The UE may not expect to report / receive non-updated AI model information.

[0141] Figure 16C shows weight information expressed as differential values ​​included in the updated AI model information. Since differential values ​​are generally expected to be smaller than absolute values, a reduction in the amount of information can be expected. Furthermore, if the amount of information in Figure 16C is the same as in Figure 16B, the granularity of each weight information can be reduced, allowing for a more detailed model representation.

[0142] If the reference model information also includes information about the structure of the AI ​​model, the updated AI model information does not need to include information about the structure of the AI ​​model, because the UE and base station can determine the AI ​​model based on the same AI model structure by exchanging only weight information.

[0143] According to the fourth embodiment described above, the signaling overhead for AI model deployment can be appropriately reduced.

[0144] <Fifth embodiment> A fifth embodiment relates to knowledge distillation.

[0145] In knowledge distillation, a single large model or multiple ensemble models with good predictive accuracy is prepared as a teacher model, and the knowledge contained in the teacher model is used to train a lightweight, easily deployable student model. The student model obtained through distillation is expected to be as accurate as the teacher model but more lightweight.

[0146] 17A and 17B are diagrams showing an example of learning using knowledge distillation. In normal learning, training is performed on an AI model by inputting a certain data set (learning data) so that the output (output from layer m) obtained is close to the correct answer.

[0147] Note that for the output from layer m, the variables / values ​​before one or more functions (e.g., sigmoid function / softmax function) are applied may be called logits, and the variables / values ​​after the one or more functions are applied may be called probs. The sum of probs for all possible classes of the correct answer is 1.

[0148] In the normal learning shown in FIG. 17A, the correct label of the learning data is used as the hard target, and learning is performed so that the loss between the probabilities and the hard target (also called hard target loss, inference reference loss, etc.) is small.

[0149] On the other hand, in learning using knowledge distillation shown in FIG. 17B, the teacher's output (prob) is used as a soft target, and learning is performed to minimize the loss between the student's prob and the soft target (also called soft target loss or distillation loss). In knowledge distillation, the one or more functions may be, for example, a temperature softmax function. Note that the soft target may correspond to information regarding the probability that the input belongs to each class.

[0150] Let v be the logit of the teacher model, p be the prob, z be the logit of the student model, and q be the prob. Then, the prob for class i of the teacher model when the temperature parameter is T is i is p i =exp(v i / T) / Σ j exp(v j / T), and the probabilities q for class i of the teacher model when the temperature parameter is T i , q i =exp(z i / T) / Σ j exp(z j / T).

[0151] The soft target loss is the p i and student model p i Based on this, the cross entropy may be calculated, for example, soft target loss = -Σ i p i log(q i ) may be required.

[0152] The hard target loss is q for T=1. i It may be determined based on the following.

[0153] In the fifth embodiment, the UE may receive AI model information (hereinafter also referred to as student model information) related to a student model. The UE may obtain an output (train the student model) using a student model specified based on the received student model information, and report information related to the output to the NW. This step may be referred to as re-training the student model.

[0154] The student model information may include any or a combination of the following information instead of or in addition to the AI ​​model information already described: · Student model (index), Teacher model (index), Information about the datasets used to calculate the teacher model / student model; -Soft target values ​​for teacher model / student model, - The output value from a layer of the teacher model / student model, Loss function values ​​of teacher model / student model, Soft target values ​​for teacher / student models, - The output value from a layer for the teacher model / student model, Loss function values ​​for teacher / student models.

[0155] The information about the dataset may correspond to at least one of, for example, a batch of data, an index indicating the dataset (or data), the dataset (or data) itself, etc. The information about the value of the loss function may include information about one or both of a soft target loss and a hard target loss.

[0156] The UE may transmit information about the student model to be retrained (hereinafter also referred to as retrained student model information) to the NW using physical layer signaling (e.g., UCI), higher layer signaling (e.g., RRC signaling, MAC CE), a specific signal / channel, or a combination thereof. The retrained student model information may include any one or a combination of the following information instead of or in addition to the AI ​​model information already described: (Index of) the student model on which the retrained student model was based, Retraining information about the student model's knowledge (e.g., soft targets, outputs from certain layers), Information about the dataset used for retraining, Retraining the intermediate information of the student model (e.g., gradients, features, etc.), Information about the hard target loss for retraining student models.

[0157] The information about the dataset may include information about the content of the data contained in the dataset.

[0158] The network may further update the student model based on the soft targets identified by the information about the student model to be retrained and the soft targets in the training data. This updating of the network may also be referred to as retraining.

[0159] When the UE receives information about the teacher model (hereinafter also referred to as teacher model information) from the NW, the UE may use the teacher model information to retrain the student model.

[0160] According to the fifth embodiment, even if the UE uses a lightweight student model taking into account its computational power, by reporting retrained student model information to the NW, the NW can re-distill knowledge to generate a more accurate / lightweight student model.

[0161] <Supplementary information> Any notification from the NW to the UE in the above-described embodiments may be performed using physical layer signaling (e.g., DCI), higher layer signaling (e.g., RRC signaling, MAC CE), specific signals / channels (e.g., PDCCH, PDSCH, reference signals), or a combination thereof.

[0162] If the notification is made by a MAC CE, the MAC CE may be identified by including a new Logical Channel ID (LCID) in the MAC subheader.

[0163] When the notification is made by DCI, the notification may be made by a specific field of the DCI, a Radio Network Temporary Identifier (RNTI) used to scramble the Cyclic Redundancy Check (CRC) bits assigned to the DCI, the format of the DCI, etc.

[0164] Furthermore, any notification from the NW to the UE in the above-described embodiments may be performed periodically, semi-persistently, or aperiodically.

[0165] In the above-described embodiments, the encoder / decoder may be interchangeably interpreted as an AI model deployed in the UE / base station. That is, the present disclosure is not limited to the case where an autoencoder is used, but may be applied to the case where inference is performed using any model. Furthermore, the target that the UE / base station compresses using the encoder in the present disclosure is not limited to CSI (or channel / precoding matrix), but may be any information.

[0166] At least one of the above embodiments may be applied only to UEs that have reported or support a particular UE capability.

[0167] The specific UE capabilities may indicate at least one of the following: Supporting specific processing / operations / control / information for at least one of the above embodiments; Support acquisition / reporting of channel characteristic information (e.g., LOS, NLOS, location information); The maximum number of FLOPs of an AI model that the UE can deploy. The maximum number of parameters for an AI model that can be deployed by the UE; The reference model supported by the UE; -Layers / algorithms / functions supported by the UE, ·Numeracy ability, ·Data collection capabilities.

[0168] Furthermore, the above-mentioned specific UE capabilities may be capabilities that are applied across all frequencies (commonly regardless of frequency), capabilities for each frequency (e.g., cell, band, BWP), capabilities for each frequency range (e.g., Frequency Range 1 (FR1), FR2, FR3, FR4, FR5, FR2-1, FR2-2), or capabilities for each subcarrier spacing (SubCarrier Spacing (SCS)).

[0169] Furthermore, the specific UE capability may be a capability that is applied across all duplexing methods (commonly regardless of the duplexing method), or may be a capability for each duplexing method (for example, Time Division Duplex (TDD) or Frequency Division Duplex (FDD)).

[0170] Furthermore, at least one of the above-described embodiments may be applied when specific information related to the above-described embodiments is configured in the UE by higher layer signaling. For example, the specific information may be information indicating that the use of an AI model is enabled, any RRC parameters for a specific release (e.g., Rel. 18), etc.

[0171] If the UE does not support at least one of the specific UE capabilities or is not configured with the specific information, the UE may apply, for example, Rel. 15 / 16 behavior.

[0172] At least one of the above-described embodiments may be used for (compressing) transmission of information between a UE and a base station other than CSI feedback. For example, a UE may report information on its location (or positioning) / information on its location estimation in a Location Management Function (LMF) to a network according to at least one of the above-described embodiments (e.g., by generating it using an encoder). The information may be information on a Channel Impulse Response (CIR) for each subband / antenna port. By reporting this information, the base station can estimate the location of the UE without reporting the angle / time difference of the received signal, etc.

[0173] (Addendum) The following inventions are added regarding one embodiment of the present disclosure. [Appendix 1] A receiving unit that receives NW training model information regarding a model (NW training model) trained by a network (Network (NW)) and reference model information regarding a reference model that is a basis for the NW training model; A terminal having a control unit that determines the NW training model based on the NW training model information and the reference model information. [Appendix 2] The terminal according to Supplementary Note 1, wherein the NW training model information indicates a difference between the NW training model and the reference model, and the reference model information indicates an index related to the reference model. [Appendix 3] The control unit further trains the NW training model. A device as described in Appendix 1 or Appendix 2. [Appendix 4] The NW training model information includes information of parameters to be frozen for training; The terminal according to any one of Supplementary Note 1 to Supplementary Note 3, wherein the control unit further trains the NW training model, and the parameters are not updated.

[0174] (Addendum) The following inventions are added regarding one embodiment of the present disclosure. [Appendix 1] A receiving unit that receives NW training model information regarding a model (NW training model) trained by a network (Network (NW)) or reference model information regarding a reference model that is a basis for the NW training model; A control unit that further trains a model determined based on the NW training model information or the reference model information; a transmitter that transmits information about the trained model. [Appendix 2] 10. The terminal of claim 1, wherein the information about the trained model includes information about which weights or layers are updated. [Appendix 3] 3. The terminal of claim 1, wherein the information about the trained model includes updates to a global model in federated learning. [Appendix 4] 4. The terminal of claim 1, wherein the sending unit sends information about resources for model training in a current iteration in federated learning.

[0175] (Addendum) The following inventions are added regarding one embodiment of the present disclosure. [Appendix 1] a receiving unit for receiving student model information relating to a student model in knowledge distillation; a transmitting unit that transmits information regarding an output based on the student model information. [Appendix 2] 2. The terminal of claim 1, wherein the information about the output includes information about a soft target of the student model. [Appendix 3] 3. The terminal of claim 1, wherein the student model information includes a value of a loss function. [Appendix 4] 4. The terminal of claim 1, wherein the student model information includes information on one or both of a soft target loss and a hard target loss.

[0176] (wireless communication system) The configuration of a wireless communication system according to an embodiment of the present disclosure will be described below. In this wireless communication system, communication is performed using any one of the wireless communication methods according to the above embodiments of the present disclosure or a combination thereof.

[0177] 18 is a diagram illustrating an example of a schematic configuration of a wireless communication system according to an embodiment. The wireless communication system 1 may be a system that realizes communication using Long Term Evolution (LTE) or 5th generation mobile communication system New Radio (5G NR) specified by the Third Generation Partnership Project (3GPP).

[0178] The wireless communication system 1 may also support dual connectivity between multiple Radio Access Technologies (RATs) (Multi-RAT Dual Connectivity (MR-DC)). MR-DC may include dual connectivity between LTE (Evolved Universal Terrestrial Radio Access (E-UTRA)) and NR (E-UTRA-NR Dual Connectivity (EN-DC)), dual connectivity between NR and LTE (NR-E-UTRA Dual Connectivity (NE-DC)), etc.

[0179] In EN-DC, the LTE (E-UTRA) base station (eNB) is the master node (MN), and the NR base station (gNB) is the secondary node (SN). In NE-DC, the NR base station (gNB) is the MN, and the LTE (E-UTRA) base station (eNB) is the SN.

[0180] The wireless communication system 1 may support dual connectivity between multiple base stations within the same RAT (for example, dual connectivity in which both the MN and the SN are NR base stations (gNBs) (NR-NR Dual Connectivity (NN-DC))).

[0181] The wireless communication system 1 may include a base station 11 that forms a macrocell C1 with a relatively wide coverage, and base stations 12 (12a-12c) that are located within the macrocell C1 and form small cells C2 that are smaller than the macrocell C1. A user terminal 20 may be located within at least one of the cells. The locations and numbers of the cells and user terminals 20 are not limited to the embodiment shown in the figure. Hereinafter, when there is no need to distinguish between the base stations 11 and 12, they will be collectively referred to as base station 10.

[0182] The user terminal 20 may be connected to at least one of the multiple base stations 10. The user terminal 20 may utilize at least one of carrier aggregation (CA) using multiple component carriers (CC) and dual connectivity (DC).

[0183] Each CC may be included in at least one of a first frequency band (Frequency Range 1 (FR1)) and a second frequency band (Frequency Range 2 (FR2)). The macro cell C1 may be included in FR1, and the small cell C2 may be included in FR2. For example, FR1 may be a frequency band of 6 GHz or less (sub-6 GHz), and FR2 may be a frequency band above 24 GHz (above-24 GHz). Note that the frequency bands and definitions of FR1 and FR2 are not limited to these, and for example, FR1 may be a frequency band higher than FR2.

[0184] Furthermore, the user terminal 20 may perform communication using at least one of Time Division Duplex (TDD) and Frequency Division Duplex (FDD) in each CC.

[0185] The multiple base stations 10 may be connected by wire (for example, optical fiber conforming to the Common Public Radio Interface (CPRI), an X2 interface, or the like) or wirelessly (for example, NR communication). For example, when NR communication is used as a backhaul between the base stations 11 and 12, the base station 11 corresponding to the upper station may be called an Integrated Access Backhaul (IAB) donor, and the base station 12 corresponding to the relay station (relay) may be called an IAB node.

[0186] The base station 10 may be connected to the core network 30 directly or via another base station 10. The core network 30 may include, for example, at least one of an Evolved Packet Core (EPC), a 5G Core Network (5GCN), a Next Generation Core (NGC), and the like.

[0187] The user terminal 20 may be a terminal compatible with at least one of communication methods such as LTE, LTE-A, and 5G.

[0188] An Orthogonal Frequency Division Multiplexing (OFDM)-based radio access scheme may be used in the radio communication system 1. For example, Cyclic Prefix OFDM (CP-OFDM), Discrete Fourier Transform Spread OFDM (DFT-s-OFDM), Orthogonal Frequency Division Multiple Access (OFDMA), Single Carrier Frequency Division Multiple Access (SC-FDMA), or the like may be used in at least one of the downlink (DL) and uplink (UL).

[0189] The radio access scheme may also be called a waveform. Note that in the wireless communication system 1, other radio access schemes (for example, other single-carrier transmission schemes, other multi-carrier transmission schemes) may be used as the UL and DL radio access schemes.

[0190] In the wireless communication system 1, a downlink shared channel (Physical Downlink Shared Channel (PDSCH)) shared by each user terminal 20, a broadcast channel (Physical Broadcast Channel (PBCH)), a downlink control channel (Physical Downlink Control Channel (PDCCH)), etc. may be used as the downlink channel.

[0191] Furthermore, in the wireless communication system 1, an uplink shared channel (Physical Uplink Shared Channel (PUSCH)) shared by each user terminal 20, an uplink control channel (Physical Uplink Control Channel (PUCCH)), a random access channel (Physical Random Access Channel (PRACH)), etc. may be used as an uplink channel.

[0192] The PDSCH transmits user data, higher layer control information, System Information Block (SIB), etc. The PUSCH may transmit user data, higher layer control information, etc. Furthermore, the PBCH may transmit Master Information Block (MIB).

[0193] Lower layer control information may be transmitted by the PDCCH. The lower layer control information may include, for example, Downlink Control Information (DCI) including scheduling information for at least one of the PDSCH and the PUSCH.

[0194] Note that the DCI for scheduling the PDSCH may be referred to as a DL assignment, a DL DCI, etc., and the DCI for scheduling the PUSCH may be referred to as an UL grant, a UL DCI, etc. Note that the PDSCH may be interpreted as DL data, and the PUSCH may be interpreted as UL data.

[0195] A control resource set (CORESET) and a search space may be used to detect the PDCCH. The CORESET corresponds to resources to search for DCI. The search space corresponds to a search region and a search method for PDCCH candidates. One CORESET may be associated with one or more search spaces. The UE may monitor a CORESET associated with a certain search space based on the search space configuration.

[0196] One search space may correspond to PDCCH candidates corresponding to one or more aggregation levels. One or more search spaces may be called a search space set. Note that the terms "search space," "search space set," "search space setting," "search space set setting," "CORESET," "CORESET setting," etc. in the present disclosure may be read interchangeably.

[0197] The PUCCH may transmit uplink control information (UCI) including at least one of channel state information (CSI), acknowledgement information (which may be called, for example, Hybrid Automatic Repeat reQuest ACKnowledgement (HARQ-ACK), ACK / NACK, etc.), and scheduling request (SR). The PRACH may transmit a random access preamble for establishing a connection with a cell.

[0198] In the present disclosure, downlink, uplink, etc. may be expressed without adding "link." Also, various channels may be expressed without adding "Physical" to the beginning.

[0199] In the wireless communication system 1, a synchronization signal (SS), a downlink reference signal (DL-RS), etc. may be transmitted. In the wireless communication system 1, a cell-specific reference signal (CRS), a channel state information reference signal (CSI-RS), a demodulation reference signal (DMRS), a positioning reference signal (PRS), a phase tracking reference signal (PTRS), etc. may be transmitted as DL-RS.

[0200] The synchronization signal may be, for example, at least one of a Primary Synchronization Signal (PSS) and a Secondary Synchronization Signal (SSS). A signal block including the SS (PSS, SSS) and the PBCH (and DMRS for the PBCH) may be referred to as an SS / PBCH block, an SS Block (SSB), or the like. Note that the SS, SSB, and the like may also be referred to as reference signals.

[0201] Furthermore, in the wireless communication system 1, a sounding reference signal (SRS), a demodulation reference signal (DMRS), etc. may be transmitted as an uplink reference signal (UL-RS). Note that the DMRS may also be called a user equipment-specific reference signal (UE-specific reference signal).

[0202] (base station) 19 is a diagram showing an example of the configuration of a base station according to an embodiment. The base station 10 includes a control unit 110, a transceiver unit 120, a transceiver antenna 130, and a transmission line interface 140. Note that the base station may include one or more of each of the control unit 110, the transceiver unit 120, the transceiver antenna 130, and the transmission line interface 140.

[0203] In this example, the functional blocks of the characteristic parts of the present embodiment are mainly shown, and it may be assumed that the base station 10 also has other functional blocks necessary for wireless communication. Some of the processing of each unit described below may be omitted.

[0204] The control unit 110 performs overall control of the base station 10. The control unit 110 can be configured from a controller, a control circuit, and the like that are explained based on common understanding in the technical field to which the present disclosure relates.

[0205] The control unit 110 may control signal generation, scheduling (e.g., resource allocation, mapping), etc. The control unit 110 may control transmission and reception using the transceiver unit 120, the transceiver antenna 130, and the transmission path interface 140, measurement, etc. The control unit 110 may generate data to be transmitted as signals, control information, sequences, etc., and transfer them to the transceiver unit 120. The control unit 110 may perform call processing (setting up, releasing, etc.) of communication channels, status management of the base station 10, management of radio resources, etc.

[0206] The transceiver unit 120 may include a baseband unit 121, a radio frequency (RF) unit 122, and a measurement unit 123. The baseband unit 121 may include a transmission processing unit 1211 and a reception processing unit 1212. The transceiver unit 120 may be configured with a transmitter / receiver, an RF circuit, a baseband circuit, a filter, a phase shifter, a measurement circuit, a transceiver circuit, etc., which are described based on common understanding in the technical field related to the present disclosure.

[0207] The transmitting / receiving unit 120 may be configured as an integrated transmitting / receiving unit, or may be composed of a transmitting unit and a receiving unit. The transmitting unit may be composed of a transmission processing unit 1211 and an RF unit 122. The receiving unit may be composed of a reception processing unit 1212, an RF unit 122, and a measurement unit 123.

[0208] The transmitting and receiving antenna 130 can be configured from an antenna described based on common understanding in the technical field to which the present disclosure pertains, such as an array antenna.

[0209] The transceiver 120 may transmit the above-mentioned downlink channel, synchronization signal, downlink reference signal, etc. The transceiver 120 may receive the above-mentioned uplink channel, uplink reference signal, etc.

[0210] The transceiver 120 may form at least one of a transmission beam and a reception beam using digital beamforming (for example, precoding), analog beamforming (for example, phase rotation), or the like.

[0211] The transceiver 120 (transmission processing unit 1211) may perform Packet Data Convergence Protocol (PDCP) layer processing, Radio Link Control (RLC) layer processing (e.g., RLC retransmission control), Medium Access Control (MAC) layer processing (e.g., HARQ retransmission control), etc. on data, control information, etc. obtained from the control unit 110, and generate a bit string to be transmitted.

[0212] The transceiver 120 (transmission processor 1211) may perform transmission processing such as channel coding (which may include error correction coding), modulation, mapping, filtering, Discrete Fourier Transform (DFT) processing (if necessary), Inverse Fast Fourier Transform (IFFT) processing, precoding, and digital-to-analog conversion on the bit string to be transmitted, and output a baseband signal.

[0213] The transmitting / receiving unit 120 (RF unit 122) may perform modulation to a radio frequency band, filtering, amplification, etc. on the baseband signal, and transmit the radio frequency band signal via the transmitting / receiving antenna .

[0214] On the other hand, the transmitting / receiving unit 120 (RF unit 122) may perform amplification, filtering, demodulation to a baseband signal, etc. on the radio frequency band signal received by the transmitting / receiving antenna .

[0215] The transceiver 120 (reception processing unit 1212) may apply reception processing such as analog-to-digital conversion, Fast Fourier Transform (FFT) processing, Inverse Discrete Fourier Transform (IDFT) processing (if necessary), filtering, demapping, demodulation, decoding (which may include error correction decoding), MAC layer processing, RLC layer processing, and PDCP layer processing to the acquired baseband signal, thereby acquiring user data, etc.

[0216] The transceiver 120 (measurement unit 123) may perform measurements on the received signal. For example, the measurement unit 123 may perform Radio Resource Management (RRM) measurements, Channel State Information (CSI) measurements, etc. based on the received signal. The measurement unit 123 may measure received power (e.g., Reference Signal Received Power (RSRP)), received quality (e.g., Reference Signal Received Quality (RSRQ), Signal to Interference plus Noise Ratio (SINR), Signal to Noise Ratio (SNR)), signal strength (e.g., Received Signal Strength Indicator (RSSI)), propagation path information (e.g., CSI), etc. The measurement results may be output to the control unit 110.

[0217] The transmission path interface 140 may transmit and receive signals (backhaul signaling) between devices included in the core network 30, other base stations 10, etc., and may acquire and transmit user data (user plane data), control plane data, etc. for the user terminal 20.

[0218] The transmitting section and receiving section of the base station 10 in the present disclosure may be configured by at least one of the transmitting / receiving section 120, the transmitting / receiving antenna 130, and the transmission path interface 140.

[0219] The control unit 110 may train a reference model. The transmission / reception unit 120 may transmit, to the user terminal 20, information on the trained model and reference model information on the reference model.

[0220] The transceiver 120 may also transmit reference model information about a reference model or information about a model trained based on the reference model to the user terminal 20. The transceiver 120 may also receive information about the reference model or a model further trained based on the reference model.

[0221] Furthermore, the transmitting / receiving unit 120 may transmit student model information relating to a student model in knowledge distillation to the user terminal 20. The transmitting / receiving unit 120 may receive information relating to an output based on the student model information.

[0222] (user terminal) 20 is a diagram showing an example of the configuration of a user terminal according to one embodiment. The user terminal 20 includes a control unit 210, a transceiver unit 220, and a transceiver antenna 230. Note that the user terminal 20 may include one or more of each of the control unit 210, the transceiver unit 220, and the transceiver antenna 230.

[0223] Note that this example mainly shows functional blocks of the characteristic parts of the present embodiment, and it may be assumed that the user terminal 20 also has other functional blocks necessary for wireless communication. Some of the processing of each unit described below may be omitted.

[0224] The control unit 210 performs overall control of the user terminal 20. The control unit 210 can be configured from a controller, a control circuit, and the like that are explained based on common understanding in the technical field to which the present disclosure relates.

[0225] The control unit 210 may control signal generation, mapping, etc. The control unit 210 may also control transmission and reception, measurement, etc. using the transceiver unit 220 and the transceiver antenna 230. The control unit 210 may generate data, control information, sequences, etc. to be transmitted as signals, and transfer them to the transceiver unit 220.

[0226] The transceiver unit 220 may include a baseband unit 221, an RF unit 222, and a measurement unit 223. The baseband unit 221 may include a transmission processing unit 2211 and a reception processing unit 2212. The transceiver unit 220 may be configured from a transmitter / receiver, an RF circuit, a baseband circuit, a filter, a phase shifter, a measurement circuit, a transceiver circuit, etc., which are described based on common understanding in the technical field related to the present disclosure.

[0227] The transmitting / receiving unit 220 may be configured as an integrated transmitting / receiving unit, or may be composed of a transmitting unit and a receiving unit. The transmitting unit may be composed of a transmission processing unit 2211 and an RF unit 222. The receiving unit may be composed of a reception processing unit 2212, an RF unit 222, and a measurement unit 223.

[0228] The transmitting / receiving antenna 230 can be configured as an antenna described based on common understanding in the technical field to which the present disclosure pertains, such as an array antenna.

[0229] The transceiver 220 may receive the above-mentioned downlink channel, synchronization signal, downlink reference signal, etc. The transceiver 220 may transmit the above-mentioned uplink channel, uplink reference signal, etc.

[0230] The transceiver 220 may form at least one of a transmission beam and a reception beam using digital beamforming (for example, precoding), analog beamforming (for example, phase rotation), or the like.

[0231] The transceiver 220 (transmission processing unit 2211) may perform PDCP layer processing, RLC layer processing (e.g., RLC retransmission control), MAC layer processing (e.g., HARQ retransmission control), etc. on data, control information, etc. obtained from the control unit 210, and generate a bit string to be transmitted.

[0232] The transceiver 220 (transmission processor 2211) may perform transmission processing such as channel coding (which may include error correction coding), modulation, mapping, filtering, DFT processing (if necessary), IFFT processing, precoding, and digital-to-analog conversion on the bit string to be transmitted, and output a baseband signal.

[0233] Whether or not to apply DFT processing may be based on the setting of transform precoding. When transform precoding is enabled for a certain channel (e.g., PUSCH), the transceiver unit 220 (transmission processing unit 2211) may perform DFT processing as the transmission processing to transmit the channel using a DFT-s-OFDM waveform, and when transform precoding is not enabled, the transceiver unit 220 may not perform DFT processing as the transmission processing.

[0234] The transmitting / receiving unit 220 (RF unit 222) may perform modulation to a radio frequency band, filtering, amplification, etc. on the baseband signal, and transmit the radio frequency band signal via the transmitting / receiving antenna 230.

[0235] On the other hand, the transmitting / receiving unit 220 (RF unit 222) may perform amplification, filtering, demodulation to a baseband signal, etc. on the radio frequency band signal received by the transmitting / receiving antenna 230.

[0236] The transceiver 220 (reception processing unit 2212) may apply reception processing such as analog-to-digital conversion, FFT processing, IDFT processing (if necessary), filtering, demapping, demodulation, decoding (which may include error correction decoding), MAC layer processing, RLC layer processing, and PDCP layer processing to the acquired baseband signal to acquire user data, etc.

[0237] The transceiver 220 (measurement unit 223) may perform measurements on the received signal. For example, the measurement unit 223 may perform RRM measurement, CSI measurement, etc. based on the received signal. The measurement unit 223 may measure received power (e.g., RSRP), received quality (e.g., RSRQ, SINR, SNR), signal strength (e.g., RSSI), propagation path information (e.g., CSI), etc. The measurement results may be output to the control unit 210.

[0238] The transmitting section and receiving section of the user terminal 20 in the present disclosure may be configured by at least one of the transmitting / receiving section 220 and the transmitting / receiving antenna 230.

[0239] The transmitter / receiver 220 may receive NW training model information about a model (NW training model) trained by a network (Network (NW)) and reference model information about a reference model on which the NW training model is based. The control unit 210 may determine the NW training model based on the NW training model information and the reference model information.

[0240] The NW training model information may indicate a difference between the NW training model and the reference model, and the reference model information may indicate an index related to the reference model.

[0241] The control unit 210 may further train the NW training model.

[0242] The NW training model information includes information of parameters to be frozen for training, and the control unit 210 further trains the NW training model, where the parameters do not need to be updated.

[0243] Furthermore, the transceiver 220 may receive NW training model information regarding a model (NW training model) trained by a network (Network (NW)) or reference model information regarding a reference model on which the NW training model is based. The control unit 210 may further train a model determined based on the NW training model information or the reference model information. The transceiver 220 may transmit information regarding the trained model.

[0244] The information about the trained model may include information about which weights or layers are updated.

[0245] The information about the trained model may include updates to a global model in federated learning.

[0246] The sender 220 may send information about resources for model training in the current iteration of federated learning.

[0247] The transceiver 220 may also receive student model information related to a student model in knowledge distillation. The transceiver 220 may also transmit information related to an output based on the student model information.

[0248] The information about the output may include information about a soft target of the student model.

[0249] The student model information may include a value of a loss function.

[0250] The student model information may include information on one or both of a soft target loss and a hard target loss.

[0251] (Hardware configuration) The block diagrams used to explain the above embodiments show functional blocks. These functional blocks (components) are realized by any combination of at least one of hardware and software. Furthermore, the method for realizing each functional block is not particularly limited. That is, each functional block may be realized using a single device that is physically or logically coupled, or may be realized using two or more physically or logically separated devices that are directly or indirectly connected (for example, using wires, wirelessly, etc.) and these multiple devices. The functional block may also be realized by combining the single device or multiple devices with software.

[0252] Here, the functions include, but are not limited to, judgment, determination, judgment, calculation, computation, processing, derivation, investigation, search, confirmation, reception, transmission, output, access, resolution, selection, election, establishment, comparison, assumption, expectation, deeming, broadcasting, notifying, communicating, forwarding, configuring, reconfiguring, allocating, mapping, and assignment. For example, a functional block (component) that performs transmission may be called a transmitting unit, transmitter, etc. As mentioned above, the implementation method of each is not particularly limited.

[0253] For example, a base station, a user terminal, etc. according to an embodiment of the present disclosure may function as a computer that performs processing of the wireless communication method of the present disclosure. Fig. 21 is a diagram illustrating an example of the hardware configuration of a base station and a user terminal according to an embodiment. The above-described base station 10 and user terminal 20 may be physically configured as a computer device including a processor 1001, a memory 1002, a storage 1003, a communication device 1004, an input device 1005, an output device 1006, a bus 1007, etc.

[0254] In this disclosure, terms such as apparatus, circuit, device, section, unit, etc. may be read interchangeably. The hardware configurations of the base station 10 and the user terminal 20 may be configured to include one or more of the devices shown in the drawings, or may be configured to exclude some of the devices.

[0255] For example, although only one processor 1001 is shown, there may be multiple processors. Furthermore, processing may be performed by one processor, or processing may be performed by two or more processors simultaneously, serially, or in other ways. Furthermore, processor 1001 may be implemented by one or more chips.

[0256] Each function in the base station 10 and the user terminal 20 is realized, for example, by loading predetermined software (programs) onto hardware such as a processor 1001 and a memory 1002, causing the processor 1001 to perform calculations, control communication via the communication device 1004, and control at least one of reading and writing data in the memory 1002 and the storage 1003.

[0257] The processor 1001, for example, runs an operating system to control the entire computer. The processor 1001 may be configured as a central processing unit (CPU) including an interface with peripheral devices, a control device, an arithmetic unit, a register, etc. For example, at least a part of the above-mentioned control unit 110 (210), transmission / reception unit 120 (220), etc. may be realized by the processor 1001.

[0258] The processor 1001 also reads programs (program codes), software modules, data, etc. from at least one of the storage 1003 and the communication device 1004 into the memory 1002, and executes various processes in accordance with these. The programs used are those that cause a computer to execute at least some of the operations described in the above-described embodiments. For example, the control unit 110 (210) may be realized by a control program stored in the memory 1002 and running on the processor 1001, and the other functional blocks may be realized in a similar manner.

[0259] The memory 1002 is a computer-readable recording medium and may be configured by at least one of, for example, a read-only memory (ROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), a random access memory (RAM), or other suitable storage medium. The memory 1002 may also be referred to as a register, a cache, a main memory, or the like. The memory 1002 may store executable programs (program codes), software modules, and the like for implementing a wireless communication method according to an embodiment of the present disclosure.

[0260] Storage 1003 is a computer-readable recording medium and may be constituted by at least one of, for example, a flexible disk, a floppy disk, a magneto-optical disk (e.g., a compact disc (e.g., a Compact Disc ROM (CD-ROM)), a digital versatile disc, a Blu-ray disc), a removable disk, a hard disk drive, a smart card, a flash memory device (e.g., a card, stick, key drive), a magnetic stripe, a database, a server, or other suitable storage medium. Storage 1003 may also be referred to as a secondary storage device.

[0261] The communication device 1004 is hardware (transmission / reception device) for communicating between computers via at least one of a wired network and a wireless network, and is also referred to as, for example, a network device, a network controller, a network card, or a communication module. The communication device 1004 may be configured to include a high-frequency switch, a duplexer, a filter, a frequency synthesizer, etc. to realize at least one of frequency division duplex (FDD) and time division duplex (TDD). For example, the above-mentioned transmission / reception unit 120 (220), transmission / reception antenna 130 (230), etc. may be realized by the communication device 1004. The transmission / reception unit 120 (220) may be implemented as a transmission unit 120a (220a) and a reception unit 120b (220b) that are physically or logically separated.

[0262] The input device 1005 is an input device (for example, a keyboard, a mouse, a microphone, a switch, a button, a sensor, etc.) that receives input from the outside. The output device 1006 is an output device (for example, a display, a speaker, a light emitting diode (LED) lamp, etc.) that outputs to the outside. The input device 1005 and the output device 1006 may be integrated into one device (for example, a touch panel).

[0263] Furthermore, each device, such as the processor 1001 and the memory 1002, is connected by a bus 1007 for communicating information. The bus 1007 may be configured using a single bus, or may be configured using different buses between each device.

[0264] Furthermore, the base station 10 and the user terminal 20 may be configured to include hardware such as a microprocessor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a programmable logic device (PLD), or a field programmable gate array (FPGA), and some or all of the functional blocks may be realized using such hardware. For example, the processor 1001 may be implemented using at least one of these pieces of hardware.

[0265] (Variation) Note that terms explained in the present disclosure and terms necessary for understanding the present disclosure may be replaced with terms having the same or similar meanings. For example, a channel, a symbol, and a signal (signal or signaling) may be interchangeable. A signal may also be a message. A reference signal may be abbreviated as RS, and may also be called a pilot, pilot signal, etc. depending on the applicable standard. A component carrier (CC) may also be called a cell, frequency carrier, carrier frequency, etc.

[0266] A radio frame may be composed of one or more periods (frames) in the time domain. Each of the one or more periods (frames) constituting a radio frame may be called a subframe. Furthermore, a subframe may be composed of one or more slots in the time domain. A subframe may have a fixed time length (e.g., 1 ms) that is independent of numerology.

[0267] Here, the numerology may be a communication parameter applied to at least one of transmission and reception of a signal or channel, and may indicate at least one of, for example, a subcarrier spacing (SCS), a bandwidth, a symbol length, a cyclic prefix length, a transmission time interval (TTI), the number of symbols per TTI, a radio frame structure, a specific filtering process performed by a transceiver in the frequency domain, and a specific windowing process performed by a transceiver in the time domain.

[0268] A slot may be composed of one or more symbols in the time domain (such as an Orthogonal Frequency Division Multiplexing (OFDM) symbol or a Single Carrier Frequency Division Multiple Access (SC-FDMA) symbol), and may be a time unit based on numerology.

[0269] A slot may include multiple minislots. Each minislot may consist of one or multiple symbols in the time domain. A minislot may also be called a subslot. A minislot may consist of fewer symbols than a slot. A PDSCH (or PUSCH) transmitted in a time unit larger than a minislot may be called PDSCH (PUSCH) mapping type A. A PDSCH (or PUSCH) transmitted using a minislot may be called PDSCH (PUSCH) mapping type B.

[0270] A radio frame, a subframe, a slot, a minislot, and a symbol all represent time units for transmitting signals. The radio frame, the subframe, the slot, the minislot, and the symbol may be referred to by other names corresponding to the radio frame, the subframe, the slot, the minislot, and the symbol. Note that the time units such as a frame, a subframe, a slot, a minislot, and a symbol in the present disclosure may be interchangeable.

[0271] For example, one subframe may be referred to as a TTI, or multiple consecutive subframes may be referred to as a TTI, or one slot or one minislot may be referred to as a TTI. That is, at least one of a subframe and a TTI may be a subframe (1 ms) in existing LTE, a period shorter than 1 ms (for example, 1-13 symbols), or a period longer than 1 ms. Note that the unit representing the TTI may be called a slot, minislot, etc., instead of a subframe.

[0272] Here, TTI refers to, for example, the smallest time unit for scheduling in wireless communication. For example, in an LTE system, a base station performs scheduling to allocate radio resources (such as frequency bandwidth and transmission power that can be used by each user terminal) to each user terminal in TTI units. However, the definition of TTI is not limited to this.

[0273] The TTI may be a transmission time unit for a channel-encoded data packet (transport block), a code block, a code word, etc., or may be a processing unit for scheduling, link adaptation, etc. When a TTI is given, the time interval (e.g., the number of symbols) to which a transport block, a code block, a code word, etc. is actually mapped may be shorter than the TTI.

[0274] When one slot or one minislot is called a TTI, one or more TTIs (i.e., one or more slots or one or more minislots) may be the minimum time unit for scheduling. Also, the number of slots (minislots) constituting the minimum time unit for scheduling may be controlled.

[0275] A TTI having a time length of 1 ms may be called a regular TTI (TTI in 3GPP Rel. 8-12), normal TTI, long TTI, regular subframe, normal subframe, long subframe, slot, etc. A TTI shorter than a regular TTI may be called a shortened TTI, short TTI, partial or fractional TTI, shortened subframe, short subframe, minislot, subslot, slot, etc.

[0276] In addition, a long TTI (e.g., a normal TTI, a subframe, etc.) may be interpreted as a TTI having a time length of more than 1 ms, and a short TTI (e.g., a shortened TTI, etc.) may be interpreted as a TTI having a TTI length shorter than the TTI length of a long TTI and equal to or greater than 1 ms.

[0277] A resource block (RB) is a resource allocation unit in the time domain and the frequency domain, and may include one or more consecutive subcarriers in the frequency domain. The number of subcarriers included in an RB may be the same regardless of numerology, for example, 12. The number of subcarriers included in an RB may also be determined based on numerology.

[0278] In addition, an RB may include one or more symbols in the time domain and may have a length of one slot, one minislot, one subframe, or one TTI. One TTI, one subframe, etc. may each be composed of one or more resource blocks.

[0279] Note that one or more RBs may also be called a physical resource block (PRB), a sub-carrier group (SCG), a resource element group (REG), a PRB pair, an RB pair, etc.

[0280] Furthermore, a resource block may be composed of one or more resource elements (REs). For example, one RE may be a radio resource region of one subcarrier and one symbol.

[0281] A Bandwidth Part (BWP), which may also be referred to as a fractional bandwidth, may represent a subset of contiguous common resource blocks (RBs) for a given numerology on a given carrier, where the common RBs may be identified by their index relative to a common reference point of the carrier. PRBs may be defined in a given BWP and numbered within that BWP.

[0282] The BWP may include an UL BWP (a BWP for UL) and a DL BWP (a BWP for DL). One or more BWPs may be configured for a UE within one carrier.

[0283] At least one of the configured BWPs may be active, and the UE may not expect to transmit or receive a given signal / channel outside the active BWP. Note that the terms "cell," "carrier," etc. in this disclosure may be read as "BWP."

[0284] The above-described structures of radio frames, subframes, slots, minislots, symbols, etc. are merely examples. For example, the number of subframes included in a radio frame, the number of slots per subframe or radio frame, the number of minislots included in a slot, the number of symbols and RBs included in a slot or minislot, the number of subcarriers included in an RB, the number of symbols in a TTI, the symbol length, the cyclic prefix (CP) length, etc. may be changed in various ways.

[0285] Furthermore, the information, parameters, etc. described in the present disclosure may be expressed using absolute values, may be expressed using relative values ​​from a predetermined value, or may be expressed using other corresponding information. For example, a radio resource may be indicated by a predetermined index.

[0286] The names used for parameters and the like in this disclosure are not intended to be limiting in any way. Furthermore, the mathematical expressions and the like using these parameters may differ from those explicitly disclosed in this disclosure. The various channels (PUCCH, PDCCH, etc.) and information elements may be identified by any suitable names, and therefore the various names assigned to these various channels and information elements are not intended to be limiting in any way.

[0287] The information, signals, etc. described in this disclosure may be represented using any of a variety of different technologies. For example, data, instructions, commands, information, signals, bits, symbols, chips, etc. that may be referred to throughout the above description may be represented by voltages, currents, electromagnetic waves, magnetic fields or magnetic particles, optical fields or photons, or any combination thereof.

[0288] Furthermore, information, signals, etc. may be output from a higher layer to a lower layer and / or from a lower layer to a higher layer. Information, signals, etc. may be input / output via multiple network nodes.

[0289] Input and output information, signals, etc. may be stored in a specific location (for example, memory) or may be managed using a management table. Input and output information, signals, etc. may be overwritten, updated, or added. Output information, signals, etc. may be deleted. Input information, signals, etc. may be transmitted to another device.

[0290] Notification of information is not limited to the aspects / embodiments described in the present disclosure, and may be performed using other methods. For example, notification of information in the present disclosure may be performed by physical layer signaling (e.g., Downlink Control Information (DCI) and Uplink Control Information (UCI)), higher layer signaling (e.g., Radio Resource Control (RRC) signaling, broadcast information (Master Information Block (MIB) and System Information Block (SIB)), Medium Access Control (MAC) signaling), other signals, or a combination thereof.

[0291] Note that the physical layer signaling may be called Layer 1 / Layer 2 (L1 / L2) control information (L1 / L2 control signal), L1 control information (L1 control signal), etc. Furthermore, the RRC signaling may be called an RRC message, such as an RRC connection setup message or an RRC connection reconfiguration message. Furthermore, the MAC signaling may be notified using, for example, a MAC Control Element (CE).

[0292] Furthermore, notification of specified information (e.g., notification that "it is X") is not limited to explicit notification, but may be made implicitly (e.g., by not notifying the specified information or by notifying other information).

[0293] The determination may be made by a value represented by one bit (0 or 1), by a Boolean value represented by true or false, or by a comparison of numerical values ​​(e.g., comparison with a predetermined value).

[0294] Software shall be construed broadly to mean instructions, instruction sets, code, code segments, program code, programs, subprograms, software modules, applications, software applications, software packages, routines, subroutines, objects, executable files, threads of execution, procedures, functions, etc., whether referred to as software, firmware, middleware, microcode, hardware description language, or otherwise.

[0295] Software, instructions, information, etc. may also be transmitted or received over a transmission medium. For example, if software is transmitted from a website, server, or other remote source using wired technologies (such as coaxial cable, fiber optic cable, twisted pair, Digital Subscriber Line (DSL)), and / or wireless technologies (such as infrared, microwave), then these wired and / or wireless technologies are included within the definition of transmission media.

[0296] As used in this disclosure, the terms "system" and "network" may be used interchangeably. A "network" may refer to devices included in the network (e.g., base stations).

[0297] In the present disclosure, terms such as "precoding," "precoder," "weight (precoding weight)," "Quasi-Co-Location (QCL)," "Transmission Configuration Indication state (TCI state)," "spatial relation," "spatial domain filter," "transmit power," "phase rotation," "antenna port," "antenna port group," "layer," "number of layers," "rank," "resource," "resource set," "resource group," "beam," "beam width," "beam angle," "antenna," "antenna element," "panel," etc. may be used interchangeably.

[0298] In this disclosure, terms such as "base station (BS)," "radio base station," "fixed station," "NodeB," "eNB (eNodeB)," "gNB (gNodeB)," "access point," "transmission point (TP)," "reception point (RP)," "transmission / reception point (TRP)," "panel," "cell," "sector," "cell group," "carrier," "component carrier," etc. may be used interchangeably. Base stations may also be referred to by terms such as macrocell, small cell, femtocell, picocell, etc.

[0299] A base station can accommodate one or more (e.g., three) cells. When a base station accommodates multiple cells, the overall coverage area of ​​the base station can be divided into multiple smaller areas, and each smaller area can be provided with communication service by a base station subsystem (e.g., a small indoor base station (Remote Radio Head (RRH))). The term "cell" or "sector" refers to a part or the entire coverage area of ​​a base station and / or base station subsystem that provides communication service within this coverage.

[0300] In the present disclosure, a base station transmitting information to a terminal may be interpreted as the base station instructing the terminal to control / operate based on the information.

[0301] In this disclosure, the terms "Mobile Station (MS)," "user terminal," "User Equipment (UE)," "terminal," etc. may be used interchangeably.

[0302] A mobile station may also be referred to as a subscriber station, mobile unit, subscriber unit, wireless unit, remote unit, mobile device, wireless device, wireless communication device, remote device, mobile subscriber station, access terminal, mobile terminal, wireless terminal, remote terminal, handset, user agent, mobile client, client, or some other suitable terminology.

[0303] At least one of the base station and the mobile station may be called a transmitting device, a receiving device, a wireless communication device, etc. Note that at least one of the base station and the mobile station may be a device mounted on a moving object, the moving object itself, etc.

[0304] The mobile body is a movable object that can move at any speed and naturally includes cases where the mobile body is stationary. Examples of the mobile body include, but are not limited to, vehicles, transport vehicles, automobiles, motorcycles, bicycles, connected cars, excavators, bulldozers, wheel loaders, dump trucks, forklifts, trains, buses, handcars, rickshaws, ships and other watercraft, airplanes, rockets, satellites, drones, multicopters, quadcopters, balloons, and objects mounted thereon. The mobile body may also be a mobile body that moves autonomously based on an operation command.

[0305] The mobile object may be a vehicle (e.g., a car, an airplane, etc.), an unmanned mobile object (e.g., a drone, an autonomous vehicle, etc.), or a robot (manned or unmanned). Note that at least one of the base station and the mobile station may also include devices that do not necessarily move during communication operations. For example, at least one of the base station and the mobile station may be an Internet of Things (IoT) device such as a sensor.

[0306] 22 is a diagram showing an example of a vehicle according to an embodiment. A vehicle 40 includes a drive unit 41, a steering unit 42, an accelerator pedal 43, a brake pedal 44, a shift lever 45, left and right front wheels 46, left and right rear wheels 47, axles 48, an electronic control unit 49, various sensors (including a current sensor 50, an RPM sensor 51, an air pressure sensor 52, a vehicle speed sensor 53, an acceleration sensor 54, an accelerator pedal sensor 55, a brake pedal sensor 56, a shift lever sensor 57, and an object detection sensor 58), an information service unit 59, and a communication module 60.

[0307] The drive unit 41 is configured with at least one of an engine, a motor, and a hybrid of an engine and a motor, for example. The steering unit 42 includes at least a steering wheel (also called a handle) and is configured to steer at least one of the front wheels 46 and the rear wheels 47 based on the operation of the steering wheel operated by the user.

[0308] The electronic control unit 49 is composed of a microprocessor 61, memory (ROM, RAM) 62, and a communication port (e.g., an input / output (IO) port) 63. Signals are input to the electronic control unit 49 from various sensors 50-58 provided in the vehicle. The electronic control unit 49 may also be called an Electronic Control Unit (ECU).

[0309] The signals from the various sensors 50-58 include a current signal from a current sensor 50 that senses the current of the motor, a rotation speed signal of the front wheels 46 / rear wheels 47 obtained by a rotation speed sensor 51, an air pressure signal of the front wheels 46 / rear wheels 47 obtained by an air pressure sensor 52, a vehicle speed signal obtained by a vehicle speed sensor 53, an acceleration signal obtained by an acceleration sensor 54, a depression amount signal of the accelerator pedal 43 obtained by an accelerator pedal sensor 55, a depression amount signal of the brake pedal 44 obtained by a brake pedal sensor 56, an operation signal of the shift lever 45 obtained by a shift lever sensor 57, and a detection signal for detecting obstacles, vehicles, pedestrians, etc. obtained by an object detection sensor 58.

[0310] The information service unit 59 is composed of various devices, such as a car navigation system, an audio system, speakers, a display, a television, and a radio, for providing (outputting) various information such as driving information, traffic information, and entertainment information, and one or more ECUs for controlling these devices. The information service unit 59 uses information acquired from external devices via the communication module 60 or the like to provide various information / services (for example, multimedia information / multimedia services) to the occupants of the vehicle 40.

[0311] The information service unit 59 may include input devices (e.g., keyboards, mice, microphones, switches, buttons, sensors, touch panels, etc.) that accept input from the outside, and may also include output devices (e.g., displays, speakers, LED lamps, touch panels, etc.) that output to the outside.

[0312] The driving assistance system unit 64 is configured with various devices that provide functions for preventing accidents and reducing the driver's driving burden, such as millimeter-wave radar, Light Detection and Ranging (LiDAR), cameras, positioning locators (e.g., Global Navigation Satellite System (GNSS)), map information (e.g., High Definition (HD) maps, Autonomous Vehicle (AV) maps), gyro systems (e.g., Inertial Measurement Units (IMUs), Inertial Navigation Systems (INSs)), Artificial Intelligence (AI) chips, and AI processors, as well as one or more ECUs that control these devices. The driving assistance system unit 64 also transmits and receives various information via the communication module 60 to realize driving assistance functions or autonomous driving functions.

[0313] The communication module 60 can communicate with the microprocessor 61 and components of the vehicle 40 via the communication port 63. For example, the communication module 60 transmits and receives data (information) via the communication port 63 to and from the drive unit 41, steering unit 42, accelerator pedal 43, brake pedal 44, shift lever 45, left and right front wheels 46, left and right rear wheels 47, axles 48, the microprocessor 61 and memory (ROM, RAM) 62 in the electronic control unit 49, and the various sensors 50-58, which are provided in the vehicle 40.

[0314] The communication module 60 is a communication device that can be controlled by the microprocessor 61 of the electronic control unit 49 and can communicate with an external device. For example, it transmits and receives various information to and from the external device via wireless communication. The communication module 60 may be located either inside or outside the electronic control unit 49. The external device may be, for example, the above-mentioned base station 10 or user terminal 20. Furthermore, the communication module 60 may be, for example, at least one of the above-mentioned base station 10 and user terminal 20 (or may function as at least one of the base station 10 and user terminal 20).

[0315] The communication module 60 may transmit at least one of signals from the above-mentioned various sensors 50-58 input to the electronic control unit 49, information obtained based on the signals, and information based on input from the outside (user) obtained via the information service unit 59 to an external device via wireless communication. The electronic control unit 49, the various sensors 50-58, the information service unit 59, etc. may be referred to as input units that accept input. For example, the PUSCH transmitted by the communication module 60 may include information based on the above-mentioned input.

[0316] The communication module 60 receives various information (traffic information, traffic signal information, vehicle distance information, etc.) transmitted from external devices and displays it on an information service unit 59 provided in the vehicle. The information service unit 59 may also be called an output unit that outputs information (for example, outputs information to a device such as a display or speaker based on the PDSCH received by the communication module 60 (or data / information decoded from the PDSCH)).

[0317] Furthermore, the communication module 60 stores various information received from external devices in a memory 62 that can be used by the microprocessor 61. Based on the information stored in the memory 62, the microprocessor 61 may control the drive unit 41, steering unit 42, accelerator pedal 43, brake pedal 44, shift lever 45, left and right front wheels 46, left and right rear wheels 47, axles 48, various sensors 50-58, and the like provided in the vehicle 40.

[0318] Furthermore, a base station in the present disclosure may be read as a user terminal. For example, the aspects / embodiments of the present disclosure may be applied to a configuration in which communication between a base station and a user terminal is replaced with communication between multiple user terminals (which may be called, for example, Device-to-Device (D2D) or Vehicle-to-Everything (V2X)). In this case, the user terminal 20 may be configured to have the functions of the base station 10 described above. Furthermore, terms such as "uplink" and "downlink" may be read as terms corresponding to terminal-to-terminal communication (for example, "sidelink"). For example, terms such as uplink channel and downlink channel may be read as sidelink channel.

[0319] Similarly, the user terminal in the present disclosure may be read as a base station, in which case the base station 10 may be configured to have the functions of the user terminal 20 described above.

[0320] In the present disclosure, an operation described as being performed by a base station may be performed by its upper node in some cases. It is apparent that in a network including one or more network nodes having a base station, various operations performed for communication with a terminal may be performed by the base station, one or more network nodes other than the base station (such as, but not limited to, a Mobility Management Entity (MME) and a Serving-Gateway (S-GW)), or a combination thereof.

[0321] Each aspect / embodiment described in this disclosure may be used alone, in combination, or switched depending on the implementation. Furthermore, the order of the processing procedures, sequences, flowcharts, etc. of each aspect / embodiment described in this disclosure may be changed unless inconsistent. For example, the methods described in this disclosure present elements of various steps using an example order, and are not limited to the specific order presented.

[0322] Each aspect / embodiment described in the present disclosure may be a technology other than Long Term Evolution (LTE), LTE-Advanced (LTE-A), LTE-Beyond (LTE-B), SUPER 3G, IMT-Advanced, 4th generation mobile communication system (4G), 5th generation mobile communication system (5G), 6th generation mobile communication system (6G), xth generation mobile communication system (xG (x is, for example, an integer or decimal number)), Future Radio Access (FRA), New-Radio Access Technology (RAT), New Radio (NR), New radio access (NX), Future generation radio access (FX), Global System for Mobile communications (GSM (registered trademark)), CDMA2000, Ultra Mobile Broadband (UMB), IEEE 802.11 (Wi-Fi (registered trademark)), IEEE 802.16 (WiMAX (registered trademark)), IEEE The present invention may be applied to systems that use 802.20, Ultra-WideBand (UWB), Bluetooth (registered trademark), or other appropriate wireless communication methods, or to next-generation systems that are expanded, modified, created, or defined based on these. It may also be applied to a combination of multiple systems (e.g., a combination of LTE or LTE-A and 5G).

[0323] As used in this disclosure, the phrase "based on" does not mean "based only on," unless expressly stated otherwise. In other words, the phrase "based on" means both "based only on" and "based at least on."

[0324] As used in this disclosure, any reference to an element using a designation such as "first," "second," etc. does not generally limit the quantity or order of those elements. These designations may be used in this disclosure as a convenient method of distinguishing between two or more elements. Thus, a reference to a first and a second element does not imply that only two elements may be employed or that the first element must in some way precede the second element.

[0325] The term "determining," as used in this disclosure, may encompass a wide variety of actions. For example, "determining" may be considered to be judging, calculating, computing, processing, deriving, investigating, looking up, search, inquiry (e.g., looking up in a table, database, or another data structure), ascertaining, etc.

[0326] Additionally, "determining" may be considered to be "determining" receiving (e.g., receiving information), transmitting (e.g., sending information), input, output, accessing (e.g., accessing data in memory), etc.

[0327] Also, "decision" may be considered to be "deciding" resolving, selecting, choosing, establishing, comparing, etc. In other words, "decision" may be considered to be "deciding" on some action.

[0328] Furthermore, "judgment (decision)" may be read as "assuming," "expecting," "considering," or the like.

[0329] As used in this disclosure, the terms "connected," "coupled," or any variation thereof, refer to any direct or indirect connection or coupling between two or more elements, and may include the presence of one or more intermediate elements between two elements that are "connected" or "coupled" to each other. The coupling or connection between elements may be physical, logical, or a combination thereof. For example, "connected" may be read as "access."

[0330] In this disclosure, when two elements are connected, they may be considered to be "connected" or "coupled" to one another using one or more wires, cables, printed electrical connections, etc., as well as using electromagnetic energy having wavelengths in the radio frequency range, microwave range, light (both visible and invisible) range, etc., as some non-limiting and non-exhaustive examples.

[0331] In the present disclosure, the term "A and B are different" may mean "A and B are different from each other." The term may also mean "A and B are each different from C." Terms such as "separate" and "coupled" may also be interpreted in the same way as "different."

[0332] When used in this disclosure, the terms "include," "including," and variations thereof are intended to be inclusive, similar to the term "comprising." Furthermore, when used in this disclosure, the term "or" is not intended to be an exclusive or.

[0333] In this disclosure, where articles are added by translation, such as a, an, and the in English, the disclosure may include that the nouns following these articles are in the plural form.

[0334] In the present disclosure, terms such as "less than or equal to," "less than," "greater than," "more than," "equal to," etc. may be interchangeable. Furthermore, in the present disclosure, terms meaning "good," "bad," "big," "small," "high," "low," "fast," "slow," etc. may be interchangeable (without being limited to the positive, comparative, or superlative). Furthermore, in the present disclosure, terms meaning "good," "bad," "big," "small," "high," "low," "fast," "slow," etc. may be interchangeable (without being limited to the positive, comparative, or superlative) with "i-th" added (for example, "highest" may be interchangeable with "i-th highest").

[0335] In this disclosure, the terms "of," "for," "regarding," "related to," "associated with," etc. may be read interchangeably.

[0336] Although the invention according to the present disclosure has been described in detail above, it is clear to those skilled in the art that the invention according to the present disclosure is not limited to the embodiments described in the present disclosure. The invention according to the present disclosure can be implemented in modified and altered forms without departing from the spirit and scope of the invention as defined by the description of the claims. Therefore, the description of the present disclosure is intended to be illustrative and explanatory and does not impose any limiting meaning on the invention according to the present disclosure.

Claims

1. a receiving unit for receiving student model information relating to a student model in knowledge distillation; a transmitting unit that executes an inference process using the student model based on the student model information, and transmits information relating to an output of the inference process to a base station by wireless communication; The terminal, wherein the student model information includes an index of the student model, and the information regarding the output of the inference process includes an index of the student model on which the student model to be retrained at the base station is based.

2. The terminal of claim 1 , wherein the information about the output includes information about a soft target of the student model.

3. The terminal of claim 1 , wherein the student model information includes a value of a loss function.

4. The terminal of claim 1 , wherein the student model information includes information on one or both of a soft target loss and a hard target loss.

5. receiving student model information regarding a student model in knowledge distillation; executing an inference process using the student model based on the student model information, and transmitting information about the output of the inference process to a base station by wireless communication; A wireless communication method for a terminal, wherein the student model information includes an index of the student model, and the information regarding the output of the inference process includes an index of the student model that was the basis for the student model to be retrained in the base station.

6. a sending unit for sending student model information related to the student model in the knowledge distillation; a receiving unit that receives information regarding an output of an inference process by the student model based on the student model information from a terminal via wireless communication; A base station, wherein the student model information includes an index of the student model, and the information regarding the output of the inference process includes an index of the student model on which the student model to be retrained is based.

7. A system including a terminal and a base station, The terminal a receiving unit for receiving student model information relating to a student model in knowledge distillation; a transmitting unit that executes an inference process using the student model based on the student model information, and transmits information relating to an output of the inference process to the base station by wireless communication; the student model information includes an index of the student model, and the information about the output of the inference process includes an index of a student model that is a basis for a student model to be retrained at the base station; The base station a transmitting unit that transmits the student model information to the terminal.

Images