Method and apparatus for model transmission in wireless communication systems

Abstract

A method performed by a first device in a wireless communication system according to one embodiment of this specification includes the steps of receiving information about at least one model from a second device and determining a Model Application Time (MAT) associated with the at least one model. The MAT is determined based on at least one of i) model transfer type, ii) number of model transfers, iii) number of the at least one model, and / or iv) model transfer unit.

Description

Technical Field

[0001] This specification relates to a method and apparatus for model transfer in a wireless communication system.

Background Art

[0002] Mobile communication systems were developed to provide voice services while ensuring user mobility. However, mobile communication systems have expanded their scope to include not only voice but also data services. Currently, the explosive increase in traffic has led to a shortage of resources, and users are demanding faster services, so more advanced mobile communication systems are required.

[0003] The requirements for next-generation mobile communication systems are significant and must support the accommodation of explosive data traffic, an epochal increase in transmission rate per user, the accommodation of a significantly increased number of connected devices, a very low end-to-end latency, and high energy efficiency. Therefore, various technologies such as dual connectivity, massive multiple input multiple output (Massive MIMO), in-band full duplex, non-orthogonal multiple access (NOMA), super wideband support, and device networking are being studied.

[0004] Communication functions can be realized based on a model(s). In this regard, model transfer (transfer / delivery) can be performed. For example, a model (or information related to the corresponding model) can be transferred from a first device (e.g., a terminal, a base station, or a first terminal) to a second device (e.g., a base station, a terminal, or a second terminal).

Summary of the Invention

[0005] When determining / defining a consistent point in time at which actions based on a transferred / delivered model can be performed, it can be inefficient in terms of communication reliability / latency. For example, the time required to apply / use a model after a particular model transfer (e.g., the time required to compile the transferred model) can vary depending on the model structure, the number of models, and the capacity of the entity receiving the model. In other words, if a point in time defined to be consistent for all models (and / or all model transfers / deliveries) is used, the following problems may arise:

[0006] As an example, it may be assumed that a time point defined to be consistent across all models (and / or all model transfers / deliveries) is earlier than the time required to apply / use a particular model that has been transmitted. In this case, an action based on the particular model may be triggered at a time when the particular model is unavailable (when it is not ready to be applied / used). Consequently, the action based on the particular model may not execute correctly, or additional signaling may be triggered after the time when the particular model is applicable.

[0007] As an example, it may be assumed that the time defined to be consistent across all models (and / or all model transfers / deliveries) is later than the time required to apply / use a particular model that has been transmitted. In this case, the time at which the operation based on the particular model is performed may be later than the time at which the preparation for use / application of the particular model is complete. In other words, unnecessary delays may occur before the operation based on the particular model begins.

[0008] The purpose of this specification is to propose a method for solving the aforementioned problems.

[0009] The technical problems to be addressed in this specification are not limited to those mentioned above, and other technical problems not mentioned can be clearly understood by a person skilled in the art to which the present invention pertains from the following description. [Means for solving the problem]

[0010] A method performed by a first device in a wireless communication system according to one embodiment of this specification includes the steps of receiving information about at least one model from a second device and determining a Model Application Time (MAT) associated with the at least one model.

[0011] The MAT is determined based on at least one of the following: i) model transfer type, ii) number of model transfers, iii) number of at least one model, and / or iv) model transfer unit.

[0012] The MAT may be a time period related to the time from a reference point until the first device completes the validation, testing, application, deployment, or compilation of the at least one model.

[0013] The reference time may be i) the time when the model transfer procedure is triggered or started by the second device, ii) the time when the first device transmits a response to the trigger of the model transfer procedure, iii) the time when the transmission of the information is started, iv) the time when the transmission of the information is completed, or v) the time when the first device transmits a response to the reception of the information.

[0014] The MAT may be the time it takes or is required for the first device to activate, apply, or prepare a model or a function related to the model.

[0015] The method may further include the steps of reporting to a second device information regarding the minimum MAT supported by the first device, and receiving information related to the MAT from the second device.

[0016] The aforementioned MAT may be greater than or equal to the minimum MAT.

[0017] A model transfer type can be defined based on whether the information includes information about a structure related to at least one of the models.

[0018] Based on the fact that the model transfer type is first type, the information may include information about parameters related to the at least one model.

[0019] Based on the fact that the model transfer type is second type, the information may include i) information regarding the parameters and ii) information regarding the structure.

[0020] The MAT associated with the second type can be larger than the MAT associated with the first type.

[0021] The number of model transfers may be related to whether the transfer of the at least one model based on the information is an initial transfer.

[0022] The MAT related to the initial transfer can be greater than the MAT related to the number of model transfers of two or more models.

[0023] Based on the number of the at least one model being greater than a defined number, the MAT can be determined based on the priority associated with each model.

[0024] The priority can be defined based on at least one of i) the start time of model transfer associated with each model, ii) the functionality associated with each model, iii) the model transfer type, iv) the number of model transfers, and / or v) the ID of each model.

[0025] The model transfer unit may be related to the number of models that can be transferred to the first device simultaneously.

[0026] Based on the number of the at least one model that is less than or equal to the model transfer unit, the MAT can be determined as the first MAT. Based on the number of at least one model that is greater than the model transfer unit, the MAT can be determined as the second MAT. The second MAT can be greater than the first MAT.

[0027] A first device operating in a wireless communication system according to another embodiment of the present specification includes one or more transceivers, one or more processors, and one or more memories connected to the one or more processors and storing instructions.

[0028] Based on being executed by the one or more processors, the indication is configured to cause the one or more processors to execute all steps of any one of the methods.

[0029] An apparatus according to another embodiment of the present specification includes one or more memories and one or more processors functionally connected to the one or more memories.

[0030] The one or more memories store instructions that, based on being executed by the one or more processors, are configured to cause the one or more processors to execute all steps of any one of the methods.

[0031] One or more non-transitory computer-readable media according to another embodiment of the present specification store instructions. The instructions executable by one or more processors are configured to cause the one or more processors to execute all steps of any one of the methods.

[0032] A method executed by a second device in a wireless communication system according to another embodiment of the present specification includes transmitting information regarding at least one model to a first device and determining a model application time point (MAT) associated with the at least one model.

[0033] The MAT is determined based on at least one of i) a model transfer type, ii) a number of model transfers, iii) a number of the at least one model, and / or iv) a model transfer unit.

[0034] A second device operating in a wireless communication system according to yet another embodiment of this specification includes one or more transceivers, one or more processors, and one or more memories connected to the one or more processors for storing instructions.

[0035] The instructions are characterized in that, based on the fact that they are to be executed by one or more processors, the one or more processors are configured to perform all the steps of the method. [Effects of the Invention]

[0036] In the embodiments described herein, the application point of a particular model becomes clear when the transfer / delivery of that particular model is performed. Based on the application point, it can be expected that the operation / communication functions based on the transmitted model will be executed successfully, thereby guaranteeing the reliability of the operation / communication functions.

[0037] Furthermore, it can reduce unnecessary delays and signaling overhead in the execution of operational / communication functions based on the transmitted model.

[0038] The effects described herein are not limited to those mentioned above, and other effects not mentioned can be clearly understood by a person skilled in the art to which the present invention pertains from the following description. [Brief explanation of the drawing]

[0039] [Figure 1] This provides an example of a functional framework for AI / ML models. [Figure 2] The signaling procedure according to the embodiment of this specification is shown below. [Figure 3] This is a flowchart illustrating a method performed by a first apparatus according to one embodiment of this specification. [Figure 4]This is a flowchart illustrating a method performed by a second apparatus according to another embodiment of this specification. [Figure 5] This figure shows the configurations of the first and second apparatus according to embodiments of this specification. [Modes for carrying out the invention]

[0040] Preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings. The detailed description disclosed below, together with the accompanying drawings, is intended to describe exemplary embodiments of the present invention and not to show only the embodiments in which the invention can be carried out. The following detailed description includes specific details in order to provide a complete understanding of the invention. However, those skilled in the art will see that the invention can be carried out without such specific details.

[0041] In some cases, known structures and devices may be omitted or shown in the form of block diagrams focusing on the core function of each structure and device, in order to avoid ambiguity of the concept of the present invention.

[0042] In the following, downlink (DL) refers to communication from a base station to a terminal, and uplink (UL) refers to communication from a terminal to a base station. In downlink, the transmitter can be part of the base station, and the receiver can be part of the terminal. In uplink, the transmitter can be part of the terminal, and the receiver can be part of the base station. The base station can also be described as the first communication device, and the terminal as the second communication device. The term base station (BS) can be replaced with terms such as fixed station, NodeB, eNB (evolved-NodeB), gNB (Next Generation NodeB), BTS (base transceiver system), access point (AP), network (5G network), AI system, RSU (roadside unit), vehicle, robot, drone (unmanned aerial vehicle, UAV), AR (Augmented Reality) device, and VR (Virtual Reality) device. Furthermore, a terminal can be fixed or mobile, and can be replaced with terms such as UE (User Equipment), MS (Mobile Station), UT (user terminal), MSS (Mobile Subscriber Station), SS (Subscriber Station), AMS (Advanced Mobile Station), WT (Wireless terminal), MTC (Machine-Type Communication) equipment, M2M (Machine-to-Machine) equipment, D2D (Device-to-Device) equipment, vehicle, robot, AI module, drone (Unmanned Aerial Vehicle, UAV), AR (Augmented Reality) equipment, and VR (Virtual Reality) equipment.

[0043] AIML-related explanations

[0044] The advancement of AI / ML (Artificial Intelligence / Machine Learning) technology is leading to increased intelligence and sophistication of the nodes and terminals that make up wireless communication networks.

[0045] In particular, the intelligence of networks and base stations is expected to enable the rapid optimization, derivation, and application of various network / base station determination parameter values ​​in response to various environmental parameters.

[0046] The aforementioned environmental parameters may include at least one of the following: distribution / location of base stations, distribution / location / material of buildings / furniture, location / direction / speed of terminals, and climate information. However, the aforementioned parameters are merely illustrative, and the environmental parameters may further include other environmental parameters linked to the network / base station determination parameters, in addition to the listed parameters.

[0047] The network / base station determination parameter values ​​may include at least one of the following: the transmit / receive power of each base station, the transmit power of each terminal, the precoder / beam of the base station / terminal, the time / frequency resource allocation for each terminal, and the duplex scheme of each base station. However, the above parameters are illustrative only, and the network / base station determination parameter values ​​may further include other parameters determined by the network / base station in addition to the listed parameters.

[0048] In line with this trend, many standardization organizations (e.g., 3GPP (registered trademark: hereinafter the same), O-RAN) are considering the introduction of AI / ML, and studies on this are actively underway.

[0049] While AI / ML can easily be described in a narrow sense as deep learning-based artificial intelligence, conceptually it can be divided as follows:

[0050] - Artificial Intelligence: This refers to all automation where machines can replace tasks that humans would normally perform.

[0051] - Machine Learning: Without explicitly programming rules, machines learn patterns for decision-making from data.

[0052] - Deep Learning: This method uses artificial nerve membrane-based models to perform feature extraction and decision-making from unstructured data in a single step. The algorithms rely on multi-layer networks of interconnected nodes for feature extraction and transformation, inspired by the biological nervous system, i.e., nerve membranes. Common deep learning network architectures include deep nerve membranes (DNNs), recurrent nerve membranes (RNNs), and convolutional nerve membranes (CNNs).

[0053] As mentioned above, artificial intelligence (AI) is the broadest concept within AI / ML, while deep learning is the narrowest concept within AI / ML. Machine learning (ML) can be interpreted as a concept that is narrower than artificial intelligence but broader than deep learning.

[0054] Types of AI / ML based on various criteria

[0055] - Offline vs. Online

[0056] Offline Learning

[0057] Offline learning follows a sequential procedure of database collection, training, and prediction; that is, collection and training are performed offline, and the completed program can be installed on-site and used for prediction work. This offline learning method is used in most situations.

[0058] Online Learning

[0059] - Online learning refers to a method that leverages the fact that data usable for recent learning is continuously generated via the internet, and uses this additionally generated data to intensively perform additional learning and gradually improve performance.

[0060] Classification based on the concepts of AI / ML Frameworks

[0061] - Centralized Learning

[0062] In centralized learning, training data collected from multiple different nodes is reported to a centralized node, and all data resources / storage / learning (e.g., supervised, unsupervised, reinforcement learning) are executed on a single centralized node.

[0063] - Federated Learning

[0064] Federated learning involves constructing a collective model based on data distributed across different data owners. Instead of bringing the data into the model, the AI / ML model is brought to the data source, allowing local nodes / individual devices to collect data and train their own copies of the model, eliminating the need to report the source data to a central node.

[0065] In federated learning, the parameters / weights of the AI / ML model should be returned to a centralized node to support general model training. The advantages of federated learning include increased computation speed and superior information security. Specifically, the process of uploading personal data to a central server is eliminated, preventing the leakage and misuse of personal information.

[0066] - Distributed Learning

[0067] Distributed learning represents the concept of machine learning processes being extended and distributed across a node cluster. Training models are divided and shared among multiple nodes running concurrently to speed up model training.

[0068] Classification by learning method

[0069] - Supervised Learning

[0070] Instructional learning is a machine learning task that aims to learn the mapping function from input to output when given a dataset with specified labels. The input data is called training data and has known labels or results. An example of instructional learning is as follows:

[0071] 1) Regression: Linear Regression, Logistic Regression

[0072] 2) Instance‐based Algorithms: k‐Nearest Neighbor(KNN)

[0073] 3)Decision Tree Algorithms:CART

[0074] 4) Support Vector Machines: SVM

[0075] 5) Bayesian Algorithms: Naive Bayes

[0076] 6) Ensemble Algorithms: Extreme Gradient Boosting, Bagging: Random Forest

[0077] Instructional learning can be further grouped into regression and classification problems, where classification predicts labels and regression predicts quantities.

[0078] Unsupervised Learning

[0079] Unsupervised learning is a machine learning task that aims to learn the ability to explain hidden structures from unlabeled data. The input data is unlabeled, and there are no known results. Some examples of unsupervised learning include K-means clustering, principal component analysis (PCA), nonlinear independent component analysis (ICA), and LSTM.

[0080] - Reinforcement Learning

[0081] In reinforcement learning (RL), the agent interacts with the environment through a trial-and-error process to optimize long-term goals; it is a goal-oriented learning method based on interaction with the environment. The following are some examples of RL algorithms.

[0082] 1) Q-learning

[0083] 2) Multi-armed bandit learning

[0084] 3) Deep Q Network

[0085] 4)State-Action-Reward-State-Action (SARSA)

[0086] 5)Temporal Difference Learning

[0087] 6)Actor‐critic reinforcement learning

[0088] 7) Deep deterministic policy gradient

[0089] 8) Monte-Carlo tree search

[0090] Reinforcement learning can be further grouped into model-based reinforcement learning and model-free reinforcement learning.

[0091] Model-based reinforcement learning: Using RL algorithms that employ predictive models, various dynamic states of the environment, and models in which such states lead to compensation, the probability of transitions between states is obtained.

[0092] Model-free reinforcement learning: A value- or policy-based RL algorithm that achieves the greatest future reward. In multi-agent environments / states, it is computationally less complex and does not require an accurate representation of the environment.

[0093] RL algorithms can also be classified into value-based RL vs. policy-based RL, policy-based RL vs. non-policy RL, and so on.

[0094] Representative models of deep learning

[0095] 1. FFNN (Feed‐Forward Neural Network)

[0096] FFNN consists of an input layer, a hidden layer, and an output layer.

[0097] 2. RNN (Recurrent Neural Network)

[0098] RNNs are a type of artificial neural membrane in which hidden nodes are connected to directional edges, forming a directed cycle structure. They are well-suited for processing sequentially appearing data such as speech and text.

[0099] 3. CNN(Convolution Neural Network)

[0100] CNNs are used for two purposes: to reduce model complexity and to extract good features, by applying convolutional operations, which are commonly used in the fields of video and image processing.

[0101] - Kernel or filter: A unit / structure that applies weights to inputs within a specific range / unit.

[0102] - Stride: The range of movement the kernel moves within the input.

[0103] - Feature map: The result of applying a kernel to the input.

[0104] - Padding: A value added to adjust the size of the characteristic map.

[0105] - Pooling: An operation to reduce the size of a characteristic map by downsampling it (e.g., max pooling, average pooling).

[0106] 4. Autoencoder

[0107] An autoencoder is a neural network that takes a feature vector x and outputs the same or similar vector x'. In an autoencoder, the input node and output node have the same features.

[0108] Figure 1 illustrates a functional framework for an AI / ML model.

[0109] The definitions of each term and their functional operations within the framework shown in Figure 1 are based on Table 1 below.

[0110] [Table 1]

[0111] JPEG2026508835000003.jpg32158

[0112] Data Set

[0113] Data sets used in AI / ML are divided into training data, validation data, and test data, and their definitions are as follows:

[0114] - Training data

[0115] Data Set for Training the Model

[0116] - Validation data

[0117] Dataset for validating a model that has already been trained.

[0118] Validation data is typically a dataset used to prevent overfitting of the training dataset.

[0119] Validation data is a dataset used to select the best model from among the various models learned during the training process. Therefore, validation data can be considered a dataset related to training.

[0120] - Test data

[0121] The final evaluation dataset, .Test data, is unrelated to the training process.

[0122] In the case of the aforementioned dataset, the entire training set containing a certain proportion of the aforementioned data may be used.

[0123] For example, a training set containing training data and validation data in an 8:2 or 7:3 ratio may be used.

[0124] As an example, a training set containing training data, validation data, and test data in a 6:2:2 ratio may be used.

[0125] Collaboration

[0126] Depending on whether or not AI / ML functionality is available between the base station and the terminal, the level of cooperation can be defined as shown in Table 2 below.

[0127] [Table 2]

[0128] The cooperation levels shown in Table 2 are examples and may be modified and used in ways different from those exemplified, depending on the implementation method. For example, cooperation levels that combine two or more of the exemplified cooperation levels may be defined and used.

[0129] Recently, there has been active effort to apply AI / ML (artificial intelligence / machine learning) technology to wireless communication networks. In particular, 3GPP Rel-18 has initiated a study on applying AI / ML technology to air interfaces between terminals and networks. In this study, beam management (BM), CSI acquisition, and positioning are considered as major use cases for integrating AI / ML into air interfaces. This specification proposes a method for determining the application timing of an AI / ML model during model transfer.

[0130] The embodiments described below are not limited to cases where an AI / ML model is transmitted. For example, the embodiments described below may also apply when a model that implements a communication function is transmitted. For example, the embodiments described below may be extended to apply when a model related to a communication function is transmitted. In the following, "AI / ML model" may be interpreted / substituted as "model," "(specific) model related to a communication function," "(specific) model for a communication function," "(specific) model related to a communication procedure," or "(specific) model for a communication procedure."

[0131] In this specification, " / " means "and," "or," or "and / or" depending on the context. In this specification, "terminal" and "UE" can be used interchangeably with the same or similar meanings, and "base station," "network," and "TRP" can also be used interchangeably with the same or similar meanings from the perspective of air interface.

[0132] It was agreed that the two terms "model transfer" and "model delivery" related to the model transfer operation between devices / entities in the 3GPP Rel-18 AI / ML study item would be defined as shown in Table 3 below.

[0133] Specifically, both terms refer to model transfer operations between devices / entities. "Model transfer" is limited to transfers via an air interface, while "model delivery" is a term that covers model transfers via means other than an air interface.

[0134] The embodiments described herein primarily concern the “model transfer” operation; however, this is for illustrative purposes only and does not limit the scope of application of the embodiments herein. The embodiments herein can also be applied to the “model delivery” operation. Therefore, “model transfer” may be interpreted / substituted with “model delivery” below.

[0135] [Table 3]

[0136] JPEG2026508835000006.jpg191159

[0137] JPEG2026508835000007.jpg72159

[0138] The problems that this specification aims to solve are as follows:

[0139] When the AI / ML model training entity and the inference entity are different, model transfer from the training entity to the inference entity may occur. In this case, it is necessary to agree on the point in time at which the transferred model becomes applicable. Specifically, the point in time at which the transferred model becomes applicable should be agreed upon between the entity transferring the model and the entity receiving the model, allowing for the understanding of when the relevant operations / functions can be enabled.

[0140] In this specification, we assume that the entity transmitting the model is a base station / NW (network) (any entity at the end) and the entity receiving the model is a terminal. However, this is for the sake of explanation, and as stated above, the embodiments described herein can also be applied to transfer operations based on other combinations of equipment / entities (e.g., UE -> BS (NW), UE2 -> UE1). That is, in the following embodiments, "terminal" may be replaced with a base station (NW) or a first terminal, and "base station" may be replaced with a terminal or a second terminal.

[0141] Model transfer operations can be classified into various types. Referring to the definition of "model transfer" above, model transfer operations can be classified as follows:

[0142] For example, model transfer operations can be divided into i) full model transfer operations and ii) partial model transfer operations.

[0143] As an example, model transfer operations can be divided into a) the transfer operation to model parameters in a specific model structure (referred to as parameter transfer below for convenience) and b) the transfer operation between model parameters and model structure.

[0144] As mentioned above, it is possible to have a more detailed classification system with three or more levels, rather than just two.

[0145] For example, a model transfer operation may be categorized based on at least one of the aforementioned examples (at least one of i), ii), a) and / or b). Specifically, a model transfer operation may be a transfer operation to model parameters and / or model structure related to the full model / partial model.

[0146] As mentioned above, the type associated with a model transfer (model transfer type) can be based on at least one of the following: i) the type of model (e.g., full model / partial model) and / or ii) the type of information associated with the model (e.g., model parameters / model structure / model parameters and model structure).

[0147] Regarding the start time of application of the transferred model that we intend to address in this specification (hereinafter referred to as model application time (MAT)), various model transfer types / types as described above may be considered.

[0148] Here, MAT can refer to the time it takes / is required for a device receiving a model transfer to actually apply / deploy that transferred model, based on a specific point in time related to the model transfer. The base station can expect / anticipate, based on the defined / configured MAT, that the terminal will apply / deploy the transferred model after that point in time.

[0149] For example, the specific time point mentioned above could be i) the time when the base station triggers / initiates a transfer to the terminal, ii) the time when the terminal sends an ACK in response to the transfer instruction, iii) the time when the transmission of model-related data from the base station to the terminal begins / is completed, or iv) the time when the terminal sends a message to the base station indicating that the reception of the model data has been successfully completed (e.g., an ACK for that data).

[0150] After the point in time based on the example above, the terminal can report a model deployment completion related message to the base station (for example, by allocating the relevant reporting UL resource).

[0151] The reason MAT is necessary is that after decoding, analyzing, validating, and / or testing / compiling transferred model data (e.g., structure, parameters), the model may need to be applied / deployed. The time required for such a process varies depending on the model type, as described above. For example, if we define two types of models as follows, the MAT required for Type 2 will inevitably be greater than that for Type 1. Therefore, we propose Method 1 below.

[0152] Type1: model parameter transfer

[0153] Type2: model structure + parameter transfer

[0154] Method 1

[0155] Multiple model transfer types may be defined. Each model transfer type may have a separate / different MAT defined / set.

[0156] For example, different MATs may be defined / set depending on whether or not the transfer of model structure information is included (whether or not the model structure information is transferred), such as the classification into type 1 and type 2. That is, MAT1 related to Type 1 and MAT2 related to Type 2 may be defined / set. As a concrete example, the MAT (MAT1 or MAT2) may be determined based on the model transfer type (Type 1 / Type 2).

[0157] Here, the model structure information may include one or more of the following: i) model structure information such as RNN, CNN, or autoencoder; ii) the number of (hidden) layers of the same structure; iii) the number of nodes for each layer; iv) input / output parameter configuration information; and / or v) pre- / post-processing.

[0158] For model transfers (e.g., Type 2) that include the model structure, it is desirable to apply a larger MAT due to the larger model data size, the time required to interpret and deploy the model structure, the larger model verification / testing / compilation time, and / or additional verification / testing procedures. For example, MAT2 associated with model transfer type 2 may be defined / set to be larger than MAT1 associated with model transfer type 1.

[0159] The required MAT (Master Time) may differ not only depending on the type of model, but also between the initial transfer of a model and subsequent transfers to update the model after one or more previous transfers. This is because the initial model transfer between specific devices contains more information about the model than subsequent model transfers. And / or, even with the same amount of information, the initial model transfer may require more time for validation / testing / compilation of that model (parameters) by the entity receiving the model transfer. Therefore, we propose the following:

[0160] Method 2

[0161] Different model application times (MATs) may be defined / set depending on the number of model transfers. For example, different MATs may be defined / set depending on whether it is the first model transfer or a second or subsequent model transfer for the device in question.

[0162] A terminal / device can simultaneously drive multiple AI / ML models for the same or multiple functions / use cases. For example, the multiple AI / ML models may include i) one model for CSI prediction, ii) one model for beam prediction, and / or iii) one model for CSI compression.

[0163] The transfer of the relevant model may occur simultaneously / overlapping from the same or different base stations. In this case, depending on the implementation of the terminal, model transfer-related processing (e.g., analysis, verification, compilation, etc.) must be performed sequentially, and different MATs may be defined / set according to the processing related to the relevant model transfer.

[0164] The terminal may not receive the transmissions from (some) models due to insufficient memory / processing power. Therefore, we propose the following approach.

[0165] Method 3

[0166] Multiple model transfers can occur simultaneously. In such cases, different MATs may be defined / set depending on the number of models being transferred simultaneously (and their priority). For example, different MATs may be defined / set based on the number of models being transferred simultaneously and / or the priority of each model being transferred (for example, the priority of each of the models if multiple models are being transferred simultaneously).

[0167] And / or, for model transfers to a certain number of models or more, the device may choose not to receive or expect to receive the transfer. For example, a device may choose not to receive some of the multiple models being transmitted simultaneously (according to model priority). For example, a device may not expect to receive more than a certain number of models simultaneously. A base station may transmit only a certain number of models or less simultaneously to the device in question.

[0168] With Method 3, the overall MAT of the model increases, and different MATs may be applied to each model.

[0169] For example, when multiple model transfers occur simultaneously, the overall MAT value of the models can be increased so that the terminal can receive / deploy the models sequentially. Also, when multiple model transfers occur simultaneously, a priority can be set for which model to receive / deploy, and different MAT values ​​can be applied to each model. Models with higher priority, those that are deployed / compiled preferentially, can be defined / set with smaller MAT values. Models with lower priority, those that are deployed / compiled later, can be defined / set with larger MAT values.

[0170] In the following, we will carefully examine embodiments for reducing the implementation burden on the terminal. The following embodiments may be applied in addition to the embodiments described above, or independently of the previous embodiments.

[0171] In one embodiment, if a certain number of model transfers occur, a rule may be defined / set that causes the terminal to perform only a certain number of model transfers and not receive transfers for the remaining models.

[0172] In one embodiment, the terminal may not expect a certain number of model transfers to occur.

[0173] In one embodiment, if a certain number of model transfers occur, a rule may be defined to prevent the terminal from receiving transfers for all of those models.

[0174] In one embodiment, if a certain number of model transfers occur, the terminal may be configured not to expect to receive transfers for all of those models.

[0175] The aforementioned "model priority" may be constructed / defined by considering one or more of the following elements:

[0176] - At the start of the model transfer (e.g., if the transfer of model1 starts before the transfer of model2, define / set the MAT of model2 to be a larger value than the MAT of model1).

[0177] - Related functionality (importance / urgency)

[0178] For example, when a multi-model transfer occurs, the priority order for deploying / compiling models for each function can be determined according to predefined priority rules, or according to the order set / instructed by the base station. In this case, the MAT for functionality1 and the MAT for functionality2 can be set / defined to be different. As an example, even for the same functionality, the PCell / PScell ​​model can be defined / configured to take precedence over the SCell model.

[0179] - Model transfer type (e.g., model transfer type based on whether or not it includes model structure transfer)

[0180] - Number of model transfers (e.g., whether or not it is an initial model transfer)

[0181] - Model ID order (e.g., prioritize low / high IDs)

[0182] - Prioritization of models set by the base station

[0183] In one embodiment, the MAT of a model may be defined / set / increased in proportion to the number of models being transferred simultaneously. For example, the receiving device can arbitrarily decide which model to deploy first. For example, the increase in MAT may not take into account the aforementioned priority of the models. For example, the increase in MAT may be performed based on the aforementioned priority of the models.

[0184] In one embodiment, a terminal may be specified / defined not to receive or expect to receive a partial model / full model transfer.

[0185] Depending on the implementation of the terminal, it may be possible to perform model transfer-related processing for multiple models simultaneously. For such terminals, processing can be performed without applying relaxed MAT (i.e., increased MAT) as in Method 3, as long as the number of models being transferred simultaneously does not exceed a certain limit. Therefore, in addition to or as a replacement for Method 3, we propose the following method.

[0186] Method 4

[0187] (Considering devices / terminals that can receive multiple models simultaneously), a model transfer unit (MTU) or model application unit (MAU) may be defined.

[0188] The maximum number of MTU / MAUs that a terminal / device can simultaneously receive (or perform transmission-related processing on) within a specified time can be defined / configured. The same MAT may be applied to transfer operations for a number of models less than the specified maximum number of MTU / MAUs.

[0189] For example, (similar to method 3 above) a longer MAT may be applied to model transfers exceeding the maximum MTU / MAU count. Alternatively, (depending on priority) some or all of the transfer may not be performed for model transfers exceeding the maximum MTU / MAU count.

[0190] The aforementioned "maximum number of MTU / MAUs that a terminal / device can simultaneously receive (or perform transmission-related processing on) within a specified time" may differ for each terminal.

[0191] In one embodiment, information regarding the maximum MTU / MAU number can be transmitted to the base station based on the UE Capability report. Specifically, the terminal can transmit UEcapability information, including the maximum MTU / MAU number, to the base station.

[0192] In one embodiment, the maximum number of MTU / MAU may be defined as a specific value for all terminals or for each terminal type / class. In this case, the maximum number of MTU / MAU may be defined for each function / functionality / use case or for all functions / functionalities / use cases. For example, the maximum number of MTU / MAU may be defined as 2 for all functions / functionalities / use cases. For example, the maximum number of MTU / MAU for each function / functionality / use case may be defined as 1.

[0193] In addition to the above method 4, the following embodiments may be considered (for terminals / devices that support a maximum MTU / MAU of a certain number or more) when the number of currently occupied MTU / MAUs is either zero or less than the specified number.

[0194] In one embodiment, it may be assumed that there is a specific type, a specific ID, a specific number of times, and / or a specific number of simultaneous transmissions (e.g., type 1, ID=0, 2 or more times, or 1 simultaneous transmission). In such cases, a shorter MAT may be defined / configured to apply than in other cases, because MTU / MAU-related resources (e.g., memory, hardware / software resources responsible for process power (e.g., DSP, modem, FPGA, etc.)) can be allocated to the model transfer.

[0195] In one embodiment, when applying the proposed method (for example, at least one of Methods 1 to 4), a process may be defined / configured to further communicate / report to the base station whether the model transfer for a specific / all model is complete or not, taking into consideration cases where the terminal does not perform a transfer for a specific / all model. For example, a terminal can transmit information to the base station indicating whether a model transfer has been completed. For example, based on the fact that a transfer of a specific model / all models has not been performed, a terminal can transmit information to the base station regarding an incomplete model transfer.

[0196] In one embodiment, a procedure for transmitting / reporting information related to model transfer can be defined.

[0197] For example, a terminal may transmit / report the following i) and / or ii) to the base station:

[0198] i) Information regarding the currently transferred model of the terminal / device (e.g., occupied / planned MTU / MAU count, information on the model currently being transmitted / transmitted / planned to be transmitted (e.g., size, type, etc.))

[0199] ii) Hardware / software-related status information of the terminal / device related to the model transfer (e.g., information on available memory / buffers that are occupied / planned to be occupied)

[0200] As an example, the aforementioned communication / reporting process may be defined within the model transfer procedure. Specifically, the aforementioned communication / reporting process may be executed in conjunction with the model transfer procedure. The communication / reporting process may be executed before the start of the model transfer procedure, after the start of the model transfer procedure, and / or during the progress of the model transfer procedure.

[0201] For example, the aforementioned communication / reporting process may be defined as a separate procedure from the model transfer procedure. Specifically, the communication / reporting process may be executed based on a different cycle / timeframe than the model transfer procedure. Another specific example is that the communication / reporting process may be executed independently of whether or not the model transfer procedure has started.

[0202] The report in question may be a network-initiated report based on base station configuration / instructions, or a UE-initiated report based on a specific event. For example, a specific event may refer to a change in information regarding the terminal / device's transfer model or hardware / software-related information. For example, a specific event may refer to a value / number related to the terminal / device's transfer model or hardware / software that exceeds / falls below a specific threshold.

[0203] In applying the proposed methods (e.g., Method 1, Method 2, Method 3, Method 4), the minimum required application time (minimum required application time) for each type / frequency / number of simultaneous transmissions can vary from device to device. For example, in the case of a terminal's MAT, a procedure may be performed to transmit the terminal's type / frequency / number of simultaneous transmissions minimum required MAT to the base station as part of the UEcapability report. The base station may then set relevant values ​​(e.g., application time values ​​for each type / frequency / number of simultaneous transmissions) on the terminal based on the terminal's type / frequency / number of simultaneous transmissions minimum required MAT.

[0204] In this specification, the model data / information (e.g., structure, parameters) of a model transfer may be defined as control plane information or user plane information. Furthermore, the model data / information (e.g., structure, parameters) of a model transfer may be defined as L1 (physical layer) / L2 (MAC layer) / L3 (RRC layer) messages.

[0205] In the embodiments described above, the term "model application time (MAT)" was used for explanatory purposes, but this is not intended to limit the technical concept of the embodiments described above to this term. For example, "MAT" can be replaced with time related to model transfer, minimum time related to model transfer, minimum time for the model, minimum time for applying the model, processing time related to the model, minimum processing time related to the model, time for compiling the model, minimum time for compiling the model, or time related to compiling the model.

[0206] The proposed methods described herein (e.g., Method 1, Method 2, Method 3, Method 4) can be used / applied in combination, one or more of them. For example, operations related to model transfer can be performed by a terminal / base station based on at least one of Method 1, Method 2, Method 3, and / or Method 4. For example, operations related to model transfer can be performed by a terminal / base station based on a combination of two or more of Method 1, Method 2, Method 3, and / or Method 4.

[0207] From a practical standpoint, the operation of the base station / terminal according to the above embodiment (for example, operation based on at least one of Methods 1 to 4) can be processed by the device shown in Figure 5 (for example, processors 110 and 210 in Figure 5), which will be described later.

[0208] Furthermore, the operation of the base station / terminal according to the above embodiment (for example, operation based on at least one of Methods 1 to 4) may also be stored in memory (for example, 140, 240 in Figure 5) in the form of instruction words / programs (e.g., instruction, executable code) for driving at least one processor (e.g., 110, 210 in Figure 5).

[0209] In the following sections, we will carefully examine the signaling procedure based on the embodiment described above.

[0210] Figure 2 shows a signaling procedure according to an embodiment of this specification.

[0211] Figure 2 shows an example of signaling between UE (user equipment) and NW (network) based on the proposed method described above. Here, UE / NW are merely examples and can be substituted with various devices. Figure 2 is for illustrative purposes only and does not limit the scope of this specification. Furthermore, some steps shown in Figure 2 may be omitted depending on the situation and / or configuration.

[0212] In Figure 2, any entity belonging to the NW, such as a base station, BS (Base Station), Node B, or TRP, can be represented, and the UE can also be replaced by an entity / server responsible for UE-related AI / ML operations.

[0213] The UE can perform a procedure (S205) to report terminal capability values ​​for AI / ML models and / or related functions / functionalities to the base station. In this procedure, terminal capability values ​​related to MAT / MAU / MTU can be reported, based on the proposed method (e.g., Method 1, Method 2, Method 3, Method 4).

[0214] Based on the terminal capability values ​​reported in S205, the base station can perform settings for relevant functions / parameters and / or AI / ML models (S210). In this process, settings related to MAT / MAU / MTU may be included based on the proposed methods (e.g., Method 1, Method 2, Method 3, Method 4).

[0215] Subsequently, a model transfer procedure can be performed between the base station and the terminal. The detailed process of the model transfer procedure can vary, and Figure 2 shows one embodiment of the procedure. The procedure for initializing the model transfer (S215) can consist of the base station and the terminal, or the terminal and the base station. This procedure can also be omitted. After that, the transfer of the model (data / information) can consist of the base station and the terminal (S220). Although not shown in this example, the terminal's response process to the base station's transfer initialization message and / or the terminal's response process to the base station's model transfer can be included in the model transfer procedure.

[0216] After model transfer, it can be assumed that the deployment / activation / application of the model is completed by the terminal after the MAT time defined / set based on the proposed methods of this specification (e.g., Method 1, Method 2, Method 3, Method 4), and after that time the base station provides the terminal with the relevant functions (e.g., CSI reporting, beam reporting, positioning-related signaling). The activation / trigger procedure (S225) can be performed.

[0217] As previously mentioned, the aforementioned NW / UE signaling and operation can be realized by the device described below (Figure 5). For example, the NW (or base station) may correspond to the first radio device and the UE to the second radio device, and in some cases the reverse may also be considered.

[0218] For example, the aforementioned NW / UE signaling and operation may be processed by one or more processors 110, 210 in Figure 5, and the aforementioned NW / UE signaling and operation may also be stored in memory 140, 240 in the form of instruction words / programs (e.g., instruction, executable code) to drive at least one of the processors 110, 210 in Figure 5.

[0219] The embodiments described above will now be specifically explained below with reference to Figures 3 and 4, focusing on the operational aspects of the first and second devices. The methods described below are separated for the sake of explanation, and it goes without saying that some components of one method may be substituted for others or combined with others. Hereinafter, the first device may mean an entity that receives a model in a procedure related to model transfer, and the second device may mean an entity that transfers a model in a procedure related to model transfer. The first device may be a terminal, base station, or first terminal, and the second device may be a base station, terminal, or second terminal.

[0220] For example, a base station can transmit a model (and / or information about the applicable model) to a terminal. For example, a terminal can transmit a model (and / or information about the applicable model) to a base station. For example, a second terminal can transmit a model (and / or information about the applicable model) to a first terminal.

[0221] Figure 3 is a flowchart illustrating a method performed by a first apparatus according to one embodiment of this specification.

[0222] Referring to Figure 3, a method performed by a first apparatus according to one embodiment of this specification includes the steps of receiving information about at least one model (S310) and determining the model application time (S320).

[0223] In S310, the first device receives information about at least one model from the second device.

[0224] The information relating to at least one of the aforementioned models can be based on information relating to the aforementioned model transfer.

[0225] In S320, the first device determines the Model Application Time (MAT) associated with the at least one model.

[0226] The aforementioned MAT can determine the point in time at which an action / function based on the at least one model may be triggered / started / executed.

[0227] For example, the first device can expect to receive information from the second device regarding triggers / instructions for operations / functions based on at least one model after the MAT.

[0228] The aforementioned MAT can refer to the time associated with model application / activation / preparation. This will be explained in detail below.

[0229] In one embodiment, the MAT may be a time period (e.g., a second time period) from a reference time (e.g., a first time period) to the time (e.g., a second time period) when the first device completes the validation, testing, application, deployment, or compilation of the at least one model.

[0230] For example, the MAT may be defined / set / indicated to be equal to or greater than the time from the first time point to the second time point. For example, the MAT may be defined / set / indicated to be the time including the time from the first time point to the second time point.

[0231] As an example, the reference time may be i) the time when the model transfer procedure is triggered or started by the second device, ii) the time when the first device transmits a response to the trigger of the model transfer procedure, iii) the time when the transmission of the information is started, iv) the time when the transmission of the information is completed, or v) the time when the first device transmits a response to the reception of the information.

[0232] In one embodiment, the MAT may be the time it takes or is necessary for the first device to activate, apply, or prepare a model or a function related to the model.

[0233] The MAT can be determined based on at least one of the methods 1 to 4 described above.

[0234] Specifically, the MAT may be determined based on at least one of the following: i) model transfer type, ii) number of model transfers, iii) number of at least one model, and / or iv) model transfer unit.

[0235] As an example, the model t transmission type may be defined based on whether the information includes information about a structure related to at least one model.

[0236] Based on the fact that the model transfer type is first type, the information may include information about parameters related to the at least one model.

[0237] Based on the fact that the model transmission r type is second type, the information may include i) information about the parameters and ii) information about the structure.

[0238] The MAT associated with the second type may be greater than the MAT associated with the first type. This embodiment can be based on Method 1.

[0239] As an example, the number of model transfers may relate to whether the transfer of at least one model based on the information is the initial transfer. The MAT related to the initial transfer may be greater than the MAT related to two or more model transfers. This embodiment may be based on Method 2.

[0240] As an example, the MAT may be determined based on the priority associated with each model, on the basis that the number of at least one model is greater than a defined number. This embodiment may be based on Method 3. The priority may be defined based on at least one of the following: i) the start time of model transmission associated with each model, ii) the functionality associated with each model, iii) the model transmission type, iv) the number of model transmissions, and / or v) the ID of each model.

[0241] As an example, the model transmission unit can be associated with the number of models that can be simultaneously transmitted to the first device. Based on the number of at least one model less than or equal to the model transmission unit, the MAT can be determined as the first MAT. Based on the number of at least one model greater than the model transmission unit, the MAT can be determined as the second MAT. The second MAT can be greater than the first MAT. This embodiment can be based on Method 4.

[0242] The aforementioned MAT can be defined / set / determined / instructed based on the capability of the first device. Let us look closely at the embodiments related thereto in detail.

[0243] In one embodiment, the method may further include a minimum MAT reporting step and a MAT receiving step. These will be described in detail below.

[0244] In the minimum MAT reporting step, the first device reports to the second device information regarding the minimum MAT supported by the first device. The minimum MAT reporting step may be performed prior to S310.

[0245] For example, information regarding the minimum MAT may be reported based on existing UE capability transfer procedures. Specifically, the first device can transmit capability information, including information regarding the minimum MAT supported by the first device, to the second device.

[0246] In the MAT receiving step, the first device receives information related to the MAT from the second device. The MAT receiving step may be performed before S310 or S320. The MAT may be greater than or equal to the minimum MAT.

[0247] As an example, the first device can determine the MAT based on information related to the MAT. Specifically, the MAT determined based on at least one of i) model transfer type, ii) number of model transfers, iii) number of at least one model, and / or iv) model transfer unit can be set / indicated by information related to the MAT.

[0248] As an example, one or more MATs may be set / instructed based on information related to the MAT. The first device can determine the MAT from among the one or more MATs.

[0249] In other words, the MAT can be determined from one or more of the above MATs based on at least one of the following: i) model transfer type, ii) number of model transfers, iii) number of at least one model, and / or iv) model transfer unit.

[0250] Specific examples of one or more MATs are as follows:

[0251] 1) MAT by mode transfer type

[0252] 2) MATs based on the number of models transmitted simultaneously (for example, MATs when the number of at least one model is less than or equal to the model transmission unit, MATs when the number of at least one model is greater than the model transmission unit, etc.)

[0253] 3) MAT based on model transfer type and the number of models transferred simultaneously (e.g., MAT by the number of models transferred simultaneously in Type 1 (or Type 2) model transfer)

[0254] MAT based on the mode transfer type associated with initial transfer and / or the number of models transferred simultaneously

[0255] 5) MAT based on the mode transfer type and / or the number of models transferred simultaneously in relation to two or more transfers (e.g., model update)

[0256] The one or more MATs can be based on at least one of the examples 1) to 5) described above. The operations based on S310, S320, the minimum MAT reporting step and the MAT receiving step described above can be implemented by the apparatus shown in Figure 5. For example, the first apparatus 100 / 200 can control one or more transceivers 130 / 230 and / or one or more memories 140 / 240 to perform operations based on S310, S320, the minimum MAT reporting step and the MAT receiving step.

[0257] The above-mentioned embodiment will now be described in detail in terms of the operation of the second device.

[0258] The steps S410, S420, minimum MAT reception step, and MAT transmission step described later correspond to the steps S310, S320, minimum MAT reporting step, and MAT reception step described in Figure 3. Considering this correspondence, redundant explanations are omitted. That is, the specific description of the operation of the second device described later can be replaced by the description / embodiment in Figure 3 corresponding to that operation.

[0259] Figure 4 is a flowchart illustrating a method performed by a second apparatus according to another embodiment of this specification.

[0260] Referring to Figure 4, a method performed by a second apparatus according to other embodiments of this specification includes the steps of transmitting information about at least one model (S410) and determining the time of model application (S420).

[0261] In S410, the second device transmits information about at least one model to the first device.

[0262] In S420, the second device determines the Model Application Time (MAT) associated with the at least one model.

[0263] The aforementioned MAT can determine the point in time at which an action / function based on the at least one model may be triggered / disclosed / executed.

[0264] As an example, the second device can transmit to the first device information regarding triggers / instructions for operations / functions based on at least one model after the MAT.

[0265] The method may further include a minimum MAT reception step and a MAT transmission step.

[0266] In the minimum MAT receiving step, the second device receives information from the first device regarding the minimum MAT supported by the first device. The minimum MAT receiving step may be performed prior to S410.

[0267] In the MAT transmission step, the second device transmits information related to the MAT to the first device. The MAT transmission step may be performed prior to S410 or S420.

[0268] The operations based on S410, S420, the minimum MAT reception step, and the MAT transmission step described above can be performed by the apparatus shown in Figure 5. For example, the second apparatus 100 / 200 can control one or more transceivers 130 / 230 and / or one or more memories 140 / 240 to perform the operations based on S410, S420, the minimum MAT reception step, and the MAT transmission step.

[0269] In the following section, an apparatus to which the embodiments of this specification can be applied (an apparatus that realizes the method / operation according to the embodiments of this specification) will be described with reference to Figure 5.

[0270] Figure 5 shows the configurations of the first and second apparatus according to the embodiments of this specification.

[0271] The first device 100 may include a processor 110, an antenna unit 120, a transceiver 130, and a memory 140.

[0272] The processor 110 performs baseband-related signal processing and may include a higher layer processing unit 111 and a physical layer processing unit 115. The higher layer processing unit 111 can process the operation of the MAC layer, RRC layer, or higher layers. The physical layer processing unit 115 can process the operation of the PHY layer. For example, if the first device 100 is a base station device in base station-terminal communication, the physical layer processing unit 115 can perform uplink received signal processing, downlink transmission signal processing, etc. For example, if the first device 100 is a first terminal device in terminal-terminal communication, the physical layer processing unit 115 can perform downlink received signal processing, uplink transmission signal processing, sidelink transmission signal processing, etc. In addition to performing baseband-related signal processing, the processor 110 can also control the overall operation of the first device 100.

[0273] The antenna section 120 may include one or more physical antennas, and if it includes multiple antennas, it may support MIMO transmission and reception. The transceiver 130 may include an RF (Radio Frequency) transmitter and an RF receiver. The memory 140 can store information processed by the processor 110, as well as software, an operating system, applications, etc., related to the operation of the first device 100, and may also include components such as buffers.

[0274] The processor 110 of the first device 100 can be configured to implement the operation of the base station (or the operation of the first terminal device in terminal-to-terminal communication) in the base station-to-terminal communication in the embodiments described herein.

[0275] The second device 200 may include a processor 210, an antenna unit 220, a transceiver 230, and a memory 240.

[0276] The processor 210 performs baseband-related signal processing and may include a higher layer processing unit 211 and a physical layer processing unit 215. The higher layer processing unit 211 can process the operation of the MAC layer, RRC layer, or higher layers. The physical layer processing unit 215 can process the operation of the PHY layer. For example, if the second device 200 is a terminal device in base station-terminal communication, the physical layer processing unit 215 can perform downlink received signal processing, uplink transmission signal processing, etc. For example, if the second device 200 is a second terminal device in terminal-to-terminal communication, the physical layer processing unit 215 can perform downlink received signal processing, uplink transmission signal processing, sidelink received signal processing, etc. In addition to performing baseband-related signal processing, the processor 210 can also control the overall operation of the second device 200.

[0277] The antenna unit 220 may include one or more physical antennas, and if it includes multiple antennas, it may support MIMO transmission and reception. The transceiver 230 may include an RF transmitter and an RF receiver. The memory 240 can store information processed by the processor 210, and software, operating systems, applications, etc., related to the operation of the second device 200, and may also include components such as buffers.

[0278] The processor 210 of the second device 200 can be configured to implement the operation of a terminal in base station-terminal communication (or the operation of a second terminal device in terminal-terminal communication) in the embodiments described herein.

[0279] In the operation of the first device 100 and the second device 200, the matters described in the examples of this disclosure regarding the base station and terminal in base station-terminal communication (or the first terminal and second terminal in terminal-terminal communication) can be applied in the same way, and redundant explanations are omitted.

[0280] Here, the wireless communication technologies implemented in the devices 100 and 200 of this disclosure may include not only LTE, NR, and 6G, but also Narrowband Internet of Things (NB-IoT) for low-power communication. For example, NB-IoT technology is an example of LPWAN (Low Power Wide Area Network) technology and can be implemented using standards such as LTE Cat NB1 and / or LTE Cat NB2, and is not limited to the aforementioned names.

[0281] Additionally or alternatively, the wireless communication technologies implemented in the devices 100, 200 of this disclosure can communicate based on LTE-M technology. For example, LTE-M technology is an example of LPWAN technology and is known by various names such as eMTC (enhanced Machine Type Communication). For example, LTE-M technology can be implemented using at least one of various standards such as 1) LTE CAT 0, 2) LTE Cat M1, 3) LTE Cat M2, 4) LTE non-BL (non-Bandwidth Limited), 5) LTE-MTC, 6) LTE Machine Type Communication, and / or 7) LTE M, and is not limited to the names mentioned above.

[0282] Additionally or alternatively, the wireless communication technology implemented by the apparatuses 100 and 200 of the present disclosure may include at least any one of ZigBee, Bluetooth (registered trademark), and Low Power Wide Area Network (LPWAN) considering low-power communication, and is not limited to the foregoing names. For example, ZigBee technology can generate personal area networks (PANs) related to small / low-power digital communication based on various standards such as IEEE 802.15.4 and can be called by various names.

[0283] 〔Claims at the time of international filing〕 〔Claim 1〕 In a wireless communication system, a method executed by a first device, comprising: receiving information regarding at least one model from a second device; determining a model application time (MAT) related to the at least one model; wherein the MAT i) is determined based on at least one of a model transfer type, ii) a number of model transfers, iii) the number of the at least one model, and / or iv) a model transfer unit. 〔Claim 2〕 The method according to claim 1, wherein the MAT is a time related to the time from a reference time point until the validation, test, application, deployment, or compilation of the at least one model is completed by the first device. The reference time point is i) the time point when a model transfer procedure is triggered or started by the second device; ii) the time point when the first device transmits a response to the trigger of the model transfer procedure; iii) the time point when the transmission of the information is started; iv) the time point when the transmission of the information is completed, or v) the time point when the first device transmits a response to the reception of the information, and the method according to claim 2 is characterized in that. The MAT is the time required to activate, apply, or prepare a model or a function related to the model in the first device, and the method according to claim 1 is characterized in that. Reporting to the second device information regarding the minimum MAT supported by the first device; Receiving from the second device information related to the MAT; and the method according to claim 1 further includes the steps of. The MAT is greater than or equal to the minimum MAT, and the method according to claim 5 is characterized in that. The model transfer type is defined based on whether the information includes information regarding the structure related to the at least one model, and the method according to claim 1 is characterized in that. Based on the model transfer type being a first type, the information includes information regarding parameters related to the at least one model,​​​​​​​​​​​​​​​​​​​​​​​The method according to claim 7, characterized in that, based on the model transfer type being second type, the information includes i) information relating to the parameters and ii) information relating to the structure. [Claim 9] The method according to claim 8, characterized in that the MAT related to the second type is larger than the MAT related to the first type. [Claim 10] The method according to claim 1, characterized in that the number of model transfers relates to whether the transfer of the at least one model based on the information is an initial transfer. [Claim 11] The method according to claim 10, characterized in that the MAT related to the first transfer is greater than the MAT related to two or more model transfers. [Claim 12] The method according to claim 1, characterized in that the MAT is determined based on the priority associated with each model, on the basis that the number of at least one model is greater than a defined number. [Claim 13] The aforementioned priority order is, i) The start time of the model transfer associated with each model, ii) Functionality related to each model, iii) The aforementioned model transfer type, iv) The number of model transfers, and / or The method according to claim 12, characterized in that it is defined based on at least one of the IDs of each model. [Claim 14] The method according to claim 1, characterized in that the model transfer unit (unit) is related to the number of models that are simultaneously transferred to the first device. [Claim 15] Based on the number of models less than or equal to the aforementioned model transfer unit, the MAT is determined as the first MAT. Based on the number of models greater than the aforementioned model transfer unit, the MAT is determined as the second MAT. The method according to claim 14, characterized in that the second MAT is larger than the first MAT. [Claim 16] A first device that operates in a wireless communication system, One or more transceivers, One or more processors, The system comprises one or more memories connected to one or more processors for storing instructions, The first apparatus is characterized in that, based on the fact that the instructions are performed by the one or more processors, the one or more processors are configured to perform all the steps of the method according to any one of claims 1 to 15. [Claim 17] A device One or more memory devices, The system comprises one or more memory units and one or more processors functionally connected to them, The apparatus is characterized in that the one or more memory stores instructions that, based on being executed by the one or more processors, cause the one or more processors to perform all the steps of the method described in any one of claims 1 to 15. [Claim 18] One or more non-transitory computer-readable media for storing instructions, One or more non-transitory computer-readable media, characterized in that the instructions, which can be executed by one or more processors, are configured to cause the one or more processors to perform all steps of the method according to any one of claims 1 to 15. [Claim 19] A method performed by a second device in a wireless communication system, A step of transmitting information about at least one model to the first device; The step of determining the Model Application Time (MAT) associated with at least one of the aforementioned models; The aforementioned MAT is i) Model transfer type, ii) Number of model transfers, iii) The number of the at least one model, and / or iv) A method characterized by being determined based on at least one of the model transfer units. [Claim 20] A second device operating in a wireless communication system, One or more transceivers, One or more processors, The system comprises one or more memories connected to one or more processors for storing instructions, A second apparatus characterized in that, based on the fact that the instructions are performed by the one or more processors, the one or more processors are configured to perform all the steps of the method according to claim 19.

Claims

1. A method performed by a first device in a wireless communication system, Steps include receiving information about at least one model from a second device; The process includes the step of determining a Model Application Time (MAT) associated with at least one of the aforementioned models; The aforementioned MAT is i) Model transfer type, ii) Number of model transfers iii) The number of the at least one model, and / or iv) A method characterized by being determined based on at least one of the model transfer units.

2. The method according to claim 1, characterized in that the MAT is a time related to the time from a reference time to the time when the first device completes the validation, testing, application, deployment, or compilation of the at least one model.

3. The aforementioned reference date is, i) When the model transfer procedure is triggered or initiated by the second device, ii) At the point when the first device transmits a response to the trigger of the model transfer procedure, iii) When the transmission of the aforementioned information begins, iv) When the transmission of the aforementioned information is completed, or v) The method according to claim 2, characterized in that the first device transmits a response to the reception of the information.

4. The method according to claim 1, characterized in that the MAT is the time required to activate, apply, or prepare a model or a function related to the model in the first device.

5. A step of reporting to the second device information regarding the minimum MAT supported by the first device; The method according to claim 1, further comprising the step of receiving information related to the MAT from the second device.

6. The method according to claim 5, characterized in that the MAT is greater than or equal to the minimum MAT.

7. The method according to claim 1, characterized in that the model transfer type is defined based on whether the information includes information relating to the structure of at least one model.

8. Based on the fact that the model transfer type is first type, the information includes information relating to parameters associated with the at least one model. The method according to claim 7, characterized in that, based on the model transfer type being second type, the information includes i) information relating to the parameters and ii) information relating to the structure.

9. The method according to claim 8, characterized in that the MAT related to the second type is greater than the MAT related to the first type.

10. The method according to claim 1, characterized in that the number of model transfers relates to whether the transfer of the at least one model based on the information is the initial transfer.

11. The method according to claim 10, characterized in that the MAT related to the first transfer is greater than the MAT related to two or more model transfers.

12. The method according to claim 1, characterized in that the MAT is determined based on the priority associated with each model, on the basis that the number of at least one model is greater than a defined number.

13. The aforementioned priority order is, i) The start time of the model transfer associated with each model, ii) Functions related to each model iii) The aforementioned model transfer type, iv) The number of model transfers, and / or v) The method according to claim 12, characterized in that it is defined based on at least one of the IDs of each model.

14. The method according to claim 1, characterized in that the model transfer unit (unit) is related to the number of models that are simultaneously transferred to the first device.

15. Based on the number of models less than or equal to the aforementioned model transfer unit, the MAT is determined as the first MAT. Based on the number of models greater than the aforementioned model transfer unit, the MAT is determined as the second MAT. The method according to claim 14, characterized in that the second MAT is larger than the first MAT.

16. A first device that operates in a wireless communication system, One or more transceivers, One or more processors, The system comprises one or more memories connected to one or more processors for storing instructions, The first apparatus is characterized in that, based on the fact that the instructions are performed by the one or more processors, the one or more processors are configured to perform all the steps of the method according to any one of claims 1 to 15.

17. A device One or more memory devices, The system comprises one or more memory units and one or more processors functionally connected to them, The apparatus is characterized in that the one or more memory stores instructions that, based on being executed by the one or more processors, cause the one or more processors to perform all the steps of the method according to any one of claims 1 to 15.

18. One or more non-transitory computer-readable media for storing instructions, One or more non-transitor computer-readable media, characterized in that the instructions, which can be executed by one or more processors, are configured to cause the one or more processors to perform all steps of the method according to any one of claims 1 to 15.

19. A method performed by a second device in a wireless communication system, A step of transmitting information about at least one model to the first device; The process includes the step of determining a Model Application Time (MAT) associated with at least one of the aforementioned models; The aforementioned MAT is i) Model transfer type, ii) Number of model transfers iii) The number of the at least one model, and / or iv) A method characterized by being determined based on at least one of the model transfer units.

20. A second device that operates in a wireless communication system, One or more transceivers, One or more processors, The system comprises one or more memories connected to one or more processors for storing instructions, A second apparatus characterized in that, based on the fact that the instructions are performed by the one or more processors, the one or more processors are configured to perform all the steps of the method according to claim 19.

Images