Method and device for utilizing artificial intelligence and machine learning in wireless communication system

Abstract

The present disclosure relates to a method for utilizing artificial intelligence and machine learning, the method being performed by a UE and comprising the steps of: transmitting, to a base station, a first message requesting partition information for each partition into which a dataset required for UE-side model training is divided; receiving, from the base station, a second message including the partition information; and transmitting, to the base station, a third message requesting transmission of a partition of the dataset on the basis of the partition information, wherein the third message is characterized by including at least one of information indicating a partition for which transmission is requested or information indicating a partition already held.

Description

Method and device for utilizing artificial intelligence and machine learning in wireless communication systems

[0001] The present disclosure relates to the operation of a terminal and a base station of a wireless communication system. More specifically, the present disclosure relates to an embodiment related to a terminal or base station for utilizing artificial intelligence and machine learning.

[0002] 5G mobile communication technology defines a wide frequency band to enable fast transmission speeds and new services, and can be implemented not only in frequency bands below 6 GHz ('Sub 6 GHz'), such as 3.5 gigahertz (3.5 GHz), but also in ultra-high frequency bands called millimeter waves (mmWave), such as 28 GHz and 39 GHz ('Above 6 GHz'). In addition, for 6G mobile communication technology, which is referred to as a system beyond 5G, implementation in the terahertz band (e.g., the 3 terahertz (3 THz) band at 95 GHz) is being considered to achieve transmission speeds 50 times faster and ultra-low latency reduced to one-tenth compared to 5G mobile communication technology.

[0003] In the early stages of 5G mobile communication technology, aiming to satisfy service support and performance requirements for enhanced Mobile BroadBand (eMBB), Ultra-Reliable Low-Latency Communications (URLLC), and massive Machine-Type Communications (mMTC), technologies such as beamforming and Massive MIMO to mitigate path loss and increase transmission distance in ultra-high frequency bands, support for various numerologies (such as the operation of multiple subcarrier spacings) and dynamic operation of slot formats for the efficient utilization of ultra-high frequency resources, initial access techniques to support multi-beam transmission and broadband, definition and operation of Band-Width Parts (BWP), Low Density Parity Check (LDPC) codes for high-volume data transmission, new channel coding methods such as Polar Codes for the reliable transmission of control information, and L2 pre-processing (L2 Standardization has been carried out for pre-processing, network slicing which provides a dedicated network specialized for specific services, and other methods.

[0004] Currently, discussions are underway to improve and enhance the performance of the initial 5G mobile communication technology, taking into account the services that the 5G mobile communication technology was intended to support. Additionally, standardization of the physical layer is in progress for technologies such as V2X (Vehicle-to-Everything), which helps autonomous vehicles make driving decisions and enhance user convenience based on their own location and status information transmitted by the vehicle; NR-U (New Radio Unlicensed), which aims for system operation in unlicensed bands to comply with various regulatory requirements; NR terminal low power consumption technology (UE Power Saving); Non-Terrestrial Network (NTN), which is direct terminal-satellite communication for securing coverage in areas where communication with the terrestrial network is impossible; and positioning.

[0005] In addition, standardization is underway in the field of wireless interface architecture / protocols for technologies such as the Industrial Internet of Things (IIoT) for supporting new services through linkage and convergence with other industries, Integrated Access and Backhaul (IAB) which provides nodes for expanding network service areas by integrating wireless backhaul links and access links, Mobility Enhancement including Conditional Handover and Dual Active Protocol Stack (DAPS) Handover, and 2-step Random Access (2-step RACH for NR) which simplifies random access procedures. Standardization is also underway in the field of system architecture / services for 5G baseline architectures (e.g., Service based Architecture, Service based Interface) for incorporating Network Functions Virtualization (NFV) and Software-Defined Networking (SDN) technologies, and Mobile Edge Computing (MEC), which provides services based on the location of the terminal.

[0006] When such 5G mobile communication systems are commercialized, connected devices, which are increasing explosively, will be connected to communication networks. Accordingly, it is expected that there will be a need to enhance the functionality and performance of 5G mobile communication systems and to integrate the operation of connected devices. To this end, new research is planned to be conducted on 5G performance improvement and complexity reduction, support for AI services, support for metaverse services, and drone communication using eXtended Reality (XR), Artificial Intelligence (AI), and Machine Learning (ML) to efficiently support Augmented Reality (AR), Virtual Reality (VR), and Mixed Reality (MR).

[0007] Furthermore, the advancement of these 5G mobile communication systems encompasses multi-antenna transmission technologies such as new waveforms to guarantee coverage in the terahertz band of 6G mobile communication technology, Full Dimensional MIMO (FD-MIMO), array antennas, and large-scale antennas; metamaterial-based lenses and antennas to improve terahertz band signal coverage; high-dimensional spatial multiplexing technology using OAM (Orbital Angular Momentum); and Reconfigurable Intelligent Surface (RIS) technology; as well as Full Duplex technology for enhancing frequency efficiency and system networks in 6G mobile communication technology; AI-based communication technologies that realize system optimization by utilizing satellites and AI from the design stage and internalizing end-to-end AI support functions; and the realization of services of complexity exceeding the limits of terminal computing capabilities by utilizing ultra-high-performance communication and computing resources. It could serve as a foundation for the development of next-generation distributed computing technologies.

[0008] In the present invention, various embodiments for more efficiently transmitting information related to AI (artificial intelligence) / ML (machine learning) are disclosed. The technical problems to be solved by the present invention are not limited to those mentioned above, and other technical problems not mentioned will be clearly understood by those skilled in the art from the description below.

[0009] According to one embodiment of the present disclosure, a method performed by a terminal of a wireless communication system comprises: a step of transmitting a first message to a base station requesting partition information for each partition of a dataset required for terminal-side model learning, wherein the data set is divided into one or more partitions; a step of receiving a second message from the base station containing the partition information; and a step of transmitting a third message to the base station requesting the transmission of one or more partitions of the dataset based on the partition information, wherein the third message comprises at least one of information indicating one or more partitions requesting transmission or information indicating one or more partitions already possessed.

[0010] Additionally, the first message includes at least one of information regarding a use-case to which the terminal-side model learning is applied or information regarding the type of dataset, and the second message includes at least one of information regarding the type of dataset, information regarding the total number of partitions of the dataset, information regarding the ID (identifier) ​​or index of one or more partitions, or information regarding the size of one or more partitions, and the use-case is characterized as being CSI (channel state information) compression.

[0011] Additionally, the method further comprises the steps of: receiving one or more radio resource control (RRC) segments containing partitions of the dataset from the base station; and assembling the one or more RRC segments to identify the partitions, wherein the dataset includes at least one of parameter information for a learned model or information necessary for compatibility between a terminal-side model and a network-side model.

[0012] In addition, the assembled partition is characterized by being transmitted to an object that performs upper layer or terminal-side model training.

[0013] According to one embodiment of the present disclosure, a method performed by a base station of a wireless communication system comprises: receiving a first message from a terminal requesting partition information for each partition of a dataset required for learning a terminal-side model, wherein the dataset is divided into one or more partitions; transmitting a second message to the terminal that includes the partition information; and receiving a third message from the terminal requesting the transmission of one or more partitions of the dataset according to the partition information, wherein the third message comprises at least one of information indicating one or more partitions requesting transmission or information indicating one or more partitions already possessed.

[0014] According to one embodiment of the present disclosure, a terminal of a wireless communication system comprises: at least one transceiver; at least one processor connected to the at least one transceiver so as to be communicable with the at least one transceiver; and a memory that stores an instruction connected to the at least one processor so as to be communicable with the at least one processor and executable individually or in any combination thereof, wherein the terminal transmits a first message to a base station requesting partition information for each partition of a dataset divided into one or more partitions necessary for terminal-side model learning, receives a second message from the base station containing the partition information, and transmits a third message to the base station requesting the transmission of one or more partitions of the dataset based on the partition information, wherein the third message comprises at least one of information indicating one or more partitions requesting transmission or information indicating one or more partitions already possessed.

[0015] According to one embodiment of the present disclosure, a base station of a wireless communication system comprises: at least one transceiver; at least one processor connected to the at least one transceiver so as to be communicable with the at least one transceiver; and a memory that stores an instruction connected to the at least one processor so as to be communicable with the at least one processor and executable individually or in any combination thereof, wherein the base station receives a first message requesting partition information for each partition of a dataset divided into one or more partitions required for learning a model on the terminal side from the base station, transmits a second message containing said partition information to the terminal, and receives a third message from the terminal requesting transmission of one or more partitions of the dataset according to said partition information, wherein the third message includes at least one of information indicating one or more partitions requesting transmission or information indicating one or more partitions already held.

[0016] According to one embodiment of the present disclosure, information related to AI / ML can be efficiently transmitted through signaling between each object. In addition, AI / ML-related models can be utilized more efficiently at a terminal or base station.

[0017] The effects obtainable in the present disclosure are not limited to those mentioned in the various embodiments, and other unmentioned effects will be clearly understood by those skilled in the art to which the present disclosure pertains from the description below.

[0018] FIG. 1 is a drawing illustrating the structure of a mobile communication system according to one embodiment of the present disclosure.

[0019] FIG. 2 is a diagram illustrating a wireless connection state transition in a mobile communication system according to one embodiment of the present disclosure.

[0020] FIG. 3 is a drawing illustrating a CSI compression use-case using a two-sided model according to one embodiment of the present disclosure.

[0021] FIG. 4 is a diagram illustrating signaling for a two-sided model according to one embodiment of the present disclosure.

[0022] FIG. 5 is a drawing illustrating the internal structure of a terminal according to one embodiment of the present disclosure.

[0023] FIG. 6 is a drawing illustrating the structure of a base station according to one embodiment of the present disclosure.

[0024] Hereinafter, embodiments of the present disclosure will be described in detail with reference to the attached drawings.

[0025] In describing the embodiments, technical details that are well known in the technical field to which this disclosure belongs and are not directly related to this disclosure are omitted. This is intended to convey the essence of this disclosure more clearly without obscuring it by omitting unnecessary explanations.

[0026] For the same reason, some components in the attached drawings have been exaggerated, omitted, or schematically depicted. Additionally, the dimensions of each component do not entirely reflect their actual dimensions. Identical or corresponding components in each drawing have been assigned the same reference numbers.

[0027] The advantages and features of the present disclosure, and the methods for achieving them, will become clear by referring to the embodiments described below in detail together with the accompanying drawings. However, the present disclosure is not limited to the embodiments disclosed below but may be implemented in various different forms. These embodiments are provided merely to ensure that the disclosure is complete and to fully inform those skilled in the art of the scope of the disclosure, and the present disclosure is defined only by the scope of the claims. Throughout the specification, the same reference numerals refer to the same components. Furthermore, in describing the present disclosure, if it is determined that a detailed description of related functions or configurations might unnecessarily obscure the essence of the present disclosure, such detailed description is omitted. Additionally, the terms described below are defined considering their functions in the present disclosure, and these may vary depending on the intentions or conventions of the user or operator. Therefore, their definitions should be based on the content throughout the specification.

[0028] Hereinafter, the base station is an entity that performs resource allocation for terminals and may be at least one of a gNode B, eNode B, Node B, BS (Base Station), wireless access unit, base station controller, or a node on a network. The terminal may include a UE (User Equipment), MS (Mobile Station), cellular phone, smartphone, computer, or a multimedia system capable of performing communication functions. In this disclosure, the Downlink (DL) refers to the wireless transmission path of a signal transmitted by a base station to a terminal, and the Uplink (UL) refers to the wireless transmission path of a signal transmitted by a terminal to a base station. Furthermore, although embodiments of this disclosure are described below using a 5G system as an example, embodiments of this disclosure may be applied to other communication systems having similar technical backgrounds or channel types. For example, LTE or LTE-A mobile communication and mobile communication technologies developed after 5G may be included therein. Additionally, this disclosure may be applied to other communication systems with some modifications made at the discretion of a person with skilled technical knowledge, provided that it does not deviate significantly from the scope of this disclosure. The contents of this disclosure are applicable to FDD and TDD systems.

[0029] At this point, it will be understood that each block of the process flow diagrams and combinations of the flow diagrams can be executed by computer program instructions. Since these computer program instructions can be loaded into the processor of a general-purpose computer, a special-purpose computer, or other programmable data processing equipment, the instructions executed through the processor of the computer or other programmable data processing equipment create means to perform the functions described in the flow diagram block(s). Since these computer program instructions can also be stored in computer-available or computer-readable memory that can be directed toward the computer or other programmable data processing equipment to implement the function in a specific way, the instructions stored in computer-available or computer-readable memory can also produce a manufactured item containing instruction means to perform the function described in the flow diagram block(s). Since computer program instructions can be loaded onto a computer or other programmable data processing equipment, instructions that perform a series of operation steps on the computer or other programmable data processing equipment to create a process executed by the computer can also provide steps for executing the functions described in the flowchart block(s).

[0030] Additionally, each block may represent a module, segment, or part of code containing one or more executable instructions for executing a specific logical function(s). It should also be noted that in some alternative execution examples, the functions mentioned in the blocks may occur out of order. For example, two blocks described in succession may actually be executed substantially simultaneously, or the blocks may sometimes be executed in reverse order according to their corresponding functions.

[0031] In this embodiment, the term "part" refers to a software or hardware component such as an FPGA (Field Programmable Gate Array) or an ASIC (Application Specific Integrated Circuit), and the "part" performs certain roles. However, the meaning of "part" is not limited to software or hardware. The "part" may be configured to reside in an addressable storage medium or configured to run one or more processors. Accordingly, as an example, the "part" includes components such as software components, object-oriented software components, class components, and task components, as well as processes, functions, attributes, procedures, subroutines, segments of program code, drivers, firmware, microcode, circuits, data, databases, data structures, tables, arrays, and variables. The functions provided within the components and "parts" may be combined into a smaller number of components and "parts" or further separated into additional components and "parts." In addition, the components and '~parts' may be implemented to utilize one or more CPUs within the device or secure multimedia card. Also, in the embodiment, the '~part' may include one or more processors.

[0032] FIG. 1 is a drawing illustrating the structure of a mobile communication system according to one embodiment of the present disclosure.

[0033] Referring to FIG. 1, the wireless access network of a mobile communication system (New Radio, NR) can be composed of a base station (next generation Node B, hereinafter gNB) (1-10) and an AMF (1-05, New Radio Core Network).

[0034] The user terminal (New Radio User Equipment, hereinafter NR UE or terminal) (1-15) can connect to an external network through gNB (1-10) and AMF (1-05).

[0035] A mobile communication system can be a next-generation mobile communication system, and a base station can be a next-generation base station.

[0036] According to one embodiment, the gNB (1-10) in FIG. 1 may correspond to the eNB (Evolved Node B) of an existing LTE system. The gNB (1-10) is connected to an NR UE via a wireless channel and can provide a superior service compared to the existing Node B (1-20). In a next-generation mobile communication system according to one embodiment of the present disclosure, since all user traffic is serviced through a shared channel, a device for scheduling by collecting state information such as the buffer status, available transmission power status, and channel status of UEs may be required, and this can be performed by the gNB (1-10). According to one embodiment, a single gNB (1-10) can typically control multiple cells. To implement ultra-high-speed data transmission, it may have a bandwidth greater than the existing maximum bandwidth, and beamforming technology may be additionally incorporated by using Orthogonal Frequency Division Multiplexing (OFDM) as the wireless access technology. In addition, an Adaptive Modulation & Coding (AMC) scheme can be applied to determine the modulation scheme and channel coding rate according to the channel state of the terminal.

[0037] According to one embodiment, in FIG. 1, the AMF (1-05) can perform functions such as mobility support, bearer configuration, and quality of service (QoS) configuration. The AMF (1-05) is a device responsible for various control functions as well as mobility management functions for terminals and can be connected to multiple base stations. In addition, the mobile communication system according to one embodiment of the present disclosure can be linked with an LTE system, and for example, the AMF (1-05) can be connected to the MME (1-25) through a network interface.

[0038] According to one embodiment, the MME (1-25) can be connected to an existing base station eNB (1-30). For example, in FIG. 1, a terminal supporting LTE-NR Dual Connectivity can transmit and receive data while maintaining a connection to both the gNB (1-10) and the eNB (1-30) (1-35).

[0039] FIG. 2 is a diagram illustrating a wireless connection state transition in a mobile communication system according to one embodiment of the present disclosure.

[0040] Referring to Fig. 2, a mobile communication system may have three radio resource control (RRC) states or RRC modes.

[0041] Specifically, the connection mode (RRC_CONNECTED, 2-05) may be a wireless connection state in which the terminal can transmit and receive data. The standby mode (RRC_IDLE, 2-30) may correspond to a wireless connection state in which the terminal monitors whether paging is being transmitted to it. The above two modes are wireless connection states applicable to LTE systems, and the detailed description may be the same as that of an LTE system. A mobile communication system according to one embodiment of the present disclosure may be a next-generation mobile communication system.

[0042] According to one embodiment of the present disclosure, a new inactive (RRC_INACTIVE) radio access state (2-15) may be defined in a mobile communication system. In the inactive radio access state (2-15), a UE context may be maintained between the base station and the terminal, and RAN (radio access network) based paging may be supported. The features of the inactive radio access state (2-15) may include at least one of the following:

[0043] - Cell re-selection mobility;

[0044] - CN - NR RAN connection (both C / U-planes (control plane / user plane)) has been established for UE;

[0045] - The UE AS(Access Stratum) context is stored in at least one gNB and the UE;

[0046] - Paging is initiated by NR RAN;

[0047] - RAN-based notification area is managed by NR RAN; or

[0048] - NR RAN knows the RAN-based notification area which the UE belongs to.

[0049] According to one embodiment of the present disclosure, a terminal in an INACTIVE wireless connection state (2-15) may use a specific procedure and transition to a connection mode (2-05) or a standby mode (2-30). The terminal may transition from the INACTIVE mode (2-15) to the connection mode (2-05) according to a resume procedure. Additionally, the terminal may transition from the connection mode (2-05) to the INACTIVE mode (2-15) using a Release procedure including suspend setting information (2-10). The procedure may be performed by transmitting and receiving one or more RRC messages between the terminal and the base station and may consist of one or more steps. Additionally, according to one embodiment, the terminal may transition from the INACTIVE mode (2-15) to the standby mode (2-30) through a Resume followed by a Release procedure (2-20). The transition between the connection mode (2-05) and the standby mode (2-30) may follow LTE technology. Also, according to FIG. 2, the transition between the modes can be performed through an establishment or release procedure (2-25).

[0050] FIG. 3 is a drawing illustrating a CSI compression use-case using a two-sided model according to one embodiment of the present disclosure.

[0051] A two-sided model may be for both a network-side model (e.g., a base station or core network, etc.) (e.g., an AI or ML-related model) and a terminal-side model. Or it may be a model compatible with both sides. Or it may be a model for at least one of the network and the terminal.

[0052] An AI / ML model may correspond to a structure or algorithm, etc., capable of performing specific tasks (e.g., prediction, classification, or generation) by learning data related to AI or ML, and may be a model for performing tasks related to AI or ML, not limited to the term.

[0053] The terminal (3-05) can obtain channel state information (CSI) (3-10) (e.g., through channel measurement or calculation). According to one embodiment, the CSI (3-10) may correspond to a pre-encoding CSI, and the CSI (3-10) may indicate a precoding matrix.

[0054] According to one embodiment, the terminal (3-05) can use the CSI (3-10) as an input value for inference based on a UE side AI / ML model.

[0055] According to FIG. 3, the encoder (3-15) corresponds to a UE side AI / ML model or a compressor, or may include at least one of an encoder, a UE side AI / ML model, or a compressor.

[0056] According to one embodiment, the terminal (3-05) can encode or compress the CSI (3-10) using the encoder (3-15).

[0057] As a result of the above reasoning, the terminal (3-05) can obtain the compressed CSI (3-20) and transmit it to the base station (3-25). Since the compressed CSI (3-20) may have a smaller data size than the original CSI (3-10), less energy and wireless resources may be used for the terminal (3-05) to report to the base station (3-25).

[0058] According to one embodiment, a base station (3-25) that receives the compressed CSI (3-20) can obtain the original CSI (3-40) through inference based on an NW side AI / ML model.

[0059] According to FIG. 3, the decoder (3-30) corresponds to an NW side AI / ML model or a decompressor (or restorer), or may include at least one of a decoder, an NW side AI / ML model, or a decompressor.

[0060] According to one embodiment, the base station (3-25) can decode or restore the compressed CSI (3-20) to the original CSI (3-40) based on the decoder (3-30).

[0061] According to one embodiment, the CSI in 3-10 and 3-40 may be the same or similar information.

[0062] In one embodiment of the present disclosure, for a two-sided AI / ML model and / or (AI / ML-based) CSI compression, a UE side (UE-sided or UE-side) AI / ML model (or encoder, compressor, etc.) and an NW side (NW-sided or NW-side) AI / ML model (or decoder, decompressor, etc.) between two objects (e.g., a terminal and a base station) may be compatible, aligned, or matched with each other.

[0063] To this end, an object (e.g., a network server) may train both the UE side AI / ML model (or encoder, compressor, etc.) and the NW side AI / ML model (or decoder, decompressor, etc.) to be compatible or aligned, and then transmit each trained model and / or related model parameters and / or the dataset used for training (e.g., including one or more original CSIs and / or compressed CSIs) to an object that will use them (e.g., a terminal server).

[0064] In one embodiment of the present disclosure, the aforementioned "each learned model and / or related model parameters and / or dataset used for learning" may be collectively referred to as "dataset" for convenience. Additionally, the dataset may include information or parameters necessary for compatibility between the terminal side and the network side. An object that receives this (e.g., a terminal server (4-70)) may (re)learn a model to be used (e.g., a UE side model) based on the dataset.

[0065] In one embodiment of the present disclosure, the dataset may be transmitted via a base station and / or a terminal when being transmitted from a network server (4-05) to a terminal server. However, since the dataset may be large in size (e.g., hundreds of megabytes), transmitting it via only one terminal may result in a high consumption of energy and wireless resources of the terminal or base station.

[0066] Accordingly, in one embodiment of the present disclosure, a network server or a base station may divide the dataset into one or more partitions and transmit each partition to a terminal server via different terminals. Upon receiving this, the terminal server may combine all partitions to restore the original dataset.

[0067] Considering that existing downlink RRC segmentation supports up to 4 segments and a single RRC message (including an RRC segment message) can hold up to approximately 9KB of information, the maximum size of downlink data that can be transmitted via RRC at once is 4 * 9KB = approximately 36KB. Since each partition may be larger than the approximately 36KB supported by downlink RRC segmentation even though the dataset is partitioned, improvements to downlink RRC segmentation may be necessary.

[0068] FIG. 4 is a diagram illustrating signaling for a two-sided model according to one embodiment of the present disclosure.

[0069] According to FIG. 4, each step involves a network server (4-05) learning a two-sided model (or an NW-side model among two-sided models) and transmitting a dataset according to one embodiment to a terminal server (4-70), and the terminal server receiving the dataset can learn a terminal-side model (UE-side model) (or a UE-side model among two-sided models) based on the received dataset. The terminal server and the network server can each transmit the learned UE-side model or NW-side model to a terminal or a base station, respectively, and based thereon, the terminal and / or base station can perform inference or model operation using the two-sided model. Each step may be omitted or combined.

[0070] According to FIG. 4, the network server (4-05) may be an object that learns a two-sided model. The network server (4-05) may be an object that collects necessary data (e.g., for learning a two-sided model). The network server (4-05) may include a base station, a CN (core network), or an OAM (operation, administration, maintenance), etc.

[0071] In step 4-15, the network server (4-05) can send a request to a base station (e.g., base station 1) (4-10) for data collection (e.g., necessary for training a two-sided model).

[0072] In step 4-25, the base station (e.g., base station 1) (4-10) may provide the terminal (e.g., terminal 1) (4-20) with configuration information related to data collection (e.g., resource measurement settings, reporting settings, etc.).

[0073] In step 4-30, the terminal (e.g., terminal 1) (4-20) can perform measurements and / or data collection (e.g., necessary for two-sided model learning) and can transmit / report the collected data to the base station.

[0074] In step 4-32, the base station (e.g., base station 1) (4-10) can reprocess the data received from the terminal or transmit it as is to the network server (4-05).

[0075] In step 4-35, the network server (4-05) can learn a two-sided model based on the collected data.

[0076] According to one embodiment, the base station of 4-40 (e.g., base station 2) may be the same base station as the aforementioned base station (e.g., base station 1) (4-10). According to one embodiment, the base station of 4-40 (e.g., base station 2) may be a different base station from the aforementioned base station (e.g., base station 1) (4-10).

[0077] According to one embodiment, the terminal of 4-45 (e.g., terminal 2) may be the same terminal as the aforementioned terminal (e.g., terminal 1) (4-20).

[0078] According to one embodiment, the terminal of 4-45 (e.g., terminal 2) may be a different terminal from the aforementioned terminal (e.g., terminal 2) (4-20).

[0079] In step 4-50, a base station (e.g., base station 2) (4-40) may request terminal capability information from a terminal (e.g., terminal 2) (4-45). According to one embodiment, the capability information may be requested via a UECapabilityEnquiry message or the like. The terminal may transmit the terminal capability information to the base station. According to one embodiment, the capability information may be transmitted via a UECapabilityInfomration message or the like. According to one embodiment, step 4-50 may be omitted.

[0080] The terminal capability information transmitted by the terminal (4-45) may include at least one of the following information (or information in a combined form of the following information).

[0081] - Information 1: Information on whether the terminal supports two-sided AI / ML models and / or (AI / ML-based) CSI compression-related functions

[0082] - Information 2: Information on whether the terminal supports two-sided AI / ML models and / or (AI / ML-based) CSI compression-related (enhanced) DL (Downlink) RRC segmentation features (or features related to dataset or segmented dataset transmission).

[0083] ■ Or, whether to support up to X (e.g., 20) DL (Downlink) RRC segments. Here, X may be a natural number of a predefined fixed value.

[0084] ■ Or, the maximum number of DL (Downlink) RRC segments that the terminal can support (assemble). For example, the terminal may select one of a plurality of predefined fixed values ​​(e.g., 20, 40, 80, 160) and report it to the base station.

[0085] Alternatively, whether it supports DL RRC segmentation for RRC messages used for dataset and / or partition transmission

[0086] In step 4-55, the network server (4-05) may transmit the NW side model (or decoder, decompressor, etc.) of the learned two-sided model to the base station (e.g., base station 2) (4-40). The above step may be omitted.

[0087] In step 4-60, the base station (4-40) may provide the terminal (4-45) with inference setting information related to the two-side model and / or inference setting information related to the UE side model to perform inference based on the two-side model (e.g., CSI compression). For example, the information may be provided through an RRC Reconfiguration message or an RRC Resume message.

[0088] In step 4-65, the terminal (4-45) may not be able to perform inference according to the above configuration. For example, the terminal (4-45) may not have a UE side model available / operable for the configuration. Therefore, the terminal (4-45) may report to the base station (4-40) that inference is impossible (e.g., inapplicability). The above step may be omitted. For example, the inapplicability may be transmitted via a UE Assistance information message, an RRC Reconfiguration complete message, or an RRC Resume complete message.

[0089] In step 4-75, the terminal (4-45) may request the terminal server (4-70) to transmit the UE side model (for the above settings). Alternatively, the terminal (4-45) may request the terminal server (4-70) to learn the UE side model (for the above settings).

[0090] In step 4-80, the terminal server (4-70) may not have a UE side model (for the above settings) and may start collecting data for training the UE side model. Alternatively, the terminal server (4-70) may start collecting data to update or modify an existing model.

[0091] In step 4-85, the terminal server (4-70) may request partition information of the dataset necessary to train the UE side model from the network server (4-05). The request may be transmitted via a terminal and / or a base station. The terminal and base station may include the aforementioned terminal and base station, or may be objects different from the aforementioned terminal and base station. The request may include at least one of the following information.

[0092] - Information 1. Information related to the use-case and / or functionality (e.g., CSI compression) for which the terminal server wants to train the model (or, functional information related to AI / ML)

[0093] - Information 2. Information regarding the types of datasets supported or preferred by the terminal server for model training (e.g., model parameters and / or CSI information and / or compressed CSI information).

[0094] In one embodiment of the present disclosure, at step 4-87, the request may be transmitted by the terminal server (4-70) to a base station (e.g., 4-40, etc.). In this case, the base station may perform partitioning after receiving the dataset from the network server (4-05).

[0095] In step 4-90, the network server (4-05) (or base station) may transmit partition information of the dataset required to train the UE-side model (or encoder, compressor, etc.) to the terminal server (4-70). The transmission may be made via a terminal and / or base station. At this time, the terminal and base station may be different objects from the terminal and base station mentioned above. The partition information may include at least one of the following information (or information in a combined form of the following information).

[0096] - Information 1. Information regarding the type of dataset provided by the network server (or base station) (e.g., model parameters and / or CSI information and / or compressed CSI information, etc.)

[0097] - Information 2. Information regarding the number of times the network server (or base station) partitioned the dataset (i.e., the total number of partitions).

[0098] - Information 3. Information about the ID or index of each partition

[0099] - Information 4. Information on the size of each partition

[0100] In one embodiment of the present disclosure, in step 4-92, the partition information may be information transmitted by a base station (e.g., 4-40, etc.) to a terminal server (4-70). For example, as in 4-87, the base station may be the entity that performs partitioning after receiving a dataset from the terminal server.

[0101] In step 4-95, the terminal server (4-70) may request the network server (4-05) to transmit a partition of the dataset necessary for training the UE side model. For example, the terminal server (4-70) may specify one or more partition IDs or indices (e.g., X). Alternatively, the terminal server (4-70) may specify one or more partitions (e.g., IDs or indices) that it already possesses to the network server. The request may be transmitted via a terminal and / or a base station. The terminal and the base station may be the same or different objects as the aforementioned terminal and base station.

[0102] In one embodiment of the present disclosure, at step 4-97, the request may be transmitted by the terminal server (4-70) to a base station (e.g., 4-40, etc.). In this case, the base station may perform partitioning after receiving the entire dataset from the network server (4-05).

[0103] In step 4-110, the network server (4-05) can transmit the partition for which transmission was requested or the partition that the terminal does not currently possess (e.g., partition X) to the base station (e.g., base station 2) (4-40).

[0104] In step 4-115, the base station (4-40) may generate an RRC message containing a partition (e.g., partition X) to be transmitted and / or related parameters. If the generated RRC message is larger than the maximum supported PDCP (packet data convergence protocol) SDU (service data unit) size (e.g., about 9 KB), and / or if the terminal supports the relevant (forecast) DL RRC segmentation, the partition may be divided into one or more RRC segments.

[0105] In step 4-120, the base station (4-40) can transmit one or more RRC segments to a terminal (e.g., terminal 2) (4-45).

[0106] In step 4-125, the terminal (4-45) can restore partition X by receiving one or more RRC segments for restoration and assembling them.

[0107] In step 4-130, the terminal RRC layer can pass partition X to an upper layer (e.g., application layer, etc.).

[0108] In step 4-135, the terminal (4-45) (or the terminal upper layer, application layer, etc.) can transfer partition X to the terminal server (4-70).

[0109] In step 4-145, the terminal server (4-70) can restore the original dataset by assembling one or more partitions (e.g., X, Y, etc.) received from one or more terminals. The terminal server (4-70) can learn a UE side model (or encoder, compressor, etc.) based on the restored dataset.

[0110] According to one embodiment, the base station of 4-170 (e.g., base station 3) may be the same as or different from the aforementioned base station (e.g., base station 1 (4-10) or base station 2 (4-40)).

[0111] According to one embodiment, the terminal of 4-150 (e.g., terminal 3) may be the same as or different from the aforementioned terminal (e.g., terminal 1 (4-20) or terminal 2 (4-45)).

[0112] In step 4-160, the terminal server (4-70) can transmit the trained UE side model (or, encoder, compressor, etc.) to the terminal (e.g., terminal 3) (4-150).

[0113] In step 4-165, the network server (4-05) may transmit the NW side model (or decoder, decompressor, etc.) of the learned two-sided model to the base station (e.g., base station 3) (4-170). The above step may be omitted.

[0114] In step 4-175, the base station (4-170) may provide the terminal (4-150) with inference setting information related to the two-side model and / or inference setting information related to the UE side model to perform inference based on the two-side model (e.g., CSI compression). For example, the information may be provided through an RRC Reconfiguration message or an RRC Resume message.

[0115] In step 4-180, the terminal (4-150) can perform inference according to the configuration. For example, the terminal (4-150) may have a UE side model available / operable for the configuration. Accordingly, the terminal (4-150) can report to the base station (4-170) that inference is possible (e.g., applicability). The above step may be omitted. For example, the applicability may be transmitted via a UE Assistance information message, an RRC Reconfiguration complete message, or an RRC Resume complete message.

[0116] In step 4-185, the terminal (4-150) can obtain information about the inference result (e.g., compressed CSI, etc.) using the UE side model (or, encoder, compressor, etc.).

[0117] In step 4-190, the terminal (4-150) can transmit information about the inference result using the UE side model (e.g., compressed CSI, etc.) to the base station (e.g., base station 3) (4-170).

[0118] In step 4-195, the base station (4-170) can drive an NW side model (or, decoder, decompressor, etc.) using the information received from the terminal (4-150) and obtain and utilize the result (e.g., original CSI).

[0119] In one embodiment of the present disclosure, signaling between a base station (e.g., 4-10, 4-40, 4-170) and a network server (4-05) may be performed through Xn, NAS, M-plane (Management plane) interface, or a new protocol / interface, etc. (4-200). This may be determined or differ depending on what kind of object the network server (4-05) is (e.g., base station, CN, or OAM).

[0120] In one embodiment of the present disclosure, signaling between a base station (e.g., 4-10, 4-40, 4-170) and a terminal (e.g., 4-20, 4-45, 4-150) may be performed through a CP (control plane) (e.g., RRC), a UP (user plane), or a new protocol / interface, etc. (4-205).

[0121] In one embodiment of the present disclosure, signaling between one or more terminals (4-20, 4-45, 4-150) and a terminal server may not be subject to standardization. Alternatively, signaling (4-210) may be performed without passing through a cellular network (e.g., a 3GPP network). For example, signaling may be performed via Wi-Fi and / or IP data communication. Additionally, the signaling may be performed through the implementation of a terminal vendor.

[0122] In one embodiment of the present disclosure, the number of DL RRC segments may be greater than a certain number, and accordingly, while the base station is transmitting DL RRC segments, the terminal may transition to RRC_IDLE, transition to RRC_INACTIVE, detect a Radio Link Failure (RLF), or perform a handover. In such cases, if the terminal is unable to reuse (or discards, etc.) some RRC segments received from the previous base station and connects to a new base station to receive them again starting from the first segments, it may result in energy waste and waste of wireless resources for the terminal / base station.

[0123] To resolve this problem, when a terminal connects to a new base station by performing an RRC Setup / resume / reestablishment / Reconfiguration procedure, according to one embodiment, the terminal may store RRC segments received from the previous base station. The existence of the RRC segments stored by the terminal may be indicated through an RRC Setup / resume / reestablishment / Reconfiguration Complete message, etc.

[0124] Alternatively, according to one embodiment, the terminal may indicate an RRC segment index that has been received (or has been interrupted) through the message, etc. Through this, the new base station may request and receive RRC segments from the previous base station starting from or after the corresponding RRC segment index and transmit them to the terminal.

[0125] Alternatively, according to one embodiment, a new base station may receive (request) and transmit one or more RRC segments from a previous base station that are not yet complete or are necessary to a terminal.

[0126] The UE side AI / ML model exemplified in one embodiment may be an encoder or a compressor, and may include at least one of the UE side AI / ML model, encoder, or compressor.

[0127] The NW side AI / ML model exemplified in one embodiment may be a decoder or a decompressor, and may include at least one of the NW side AI / ML model, decoder, or decompressor.

[0128] FIG. 5 is a drawing illustrating the internal structure of a terminal according to one embodiment of the present disclosure.

[0129] Referring to FIG. 5, the terminal may include an RF (Radio Frequency) processing unit (5-10), a baseband processing unit (5-20), a storage unit (5-30), and a control unit (5-40).

[0130] The RF processing unit (5-10) can perform functions for transmitting and receiving signals through a wireless channel, such as signal band conversion and amplification. For example, the RF processing unit (5-10) can up-convert a baseband signal provided by the baseband processing unit (5-20) into an RF band signal and transmit it through an antenna, and can down-convert an RF band signal received through the antenna into a baseband signal. For example, the RF processing unit (5-10) may include a transmission filter, a reception filter, an amplifier, a mixer, an oscillator, a DAC (digital to analog converter), an ADC (analog to digital converter), etc. Although only one antenna is shown in FIG. 5, the terminal may be equipped with multiple antennas. Additionally, the RF processing unit (5-10) may include multiple RF chains. Furthermore, the RF processing unit (5-10) may perform beamforming. For the above beamforming, the RF processing unit (5-10) can adjust the phase and magnitude of each of the signals transmitted and received through multiple antennas or antenna elements. In addition, the RF processing unit can perform MIMO (multi-input multi-output) and can receive multiple layers when performing MIMO operation.

[0131] The baseband processing unit (5-20) can perform a conversion function between a baseband signal and a bit sequence according to the physical layer specifications of the system. For example, when transmitting data, the baseband processing unit (5-20) can generate complex symbols by encoding and modulating the transmitted bit sequence. Additionally, when receiving data, the baseband processing unit (5-20) can restore the received bit sequence by demodulating and decoding the baseband signal provided by the RF processing unit (5-10). For example, in the case of following the orthogonal frequency division multiplexing (OFDM) method, when transmitting data, the baseband processing unit (5-20) can generate complex symbols by encoding and modulating the transmitted bit sequence, and after mapping the complex symbols to subcarriers, can construct OFDM symbols through inverse fast Fourier transform (IFFT) operations and cyclic prefix (CP) insertion. Additionally, upon receiving data, the baseband processing unit (5-20) can divide the baseband signal provided from the RF processing unit (5-10) into OFDM symbol units, restore the signals mapped to subcarriers through a fast Fourier transform (FFT) operation, and then restore the received bit sequence through demodulation and decoding.

[0132] The baseband processing unit (5-20) and the RF processing unit (5-10) can transmit and receive signals as described above. Accordingly, the baseband processing unit (5-20) and the RF processing unit (5-10) may be referred to as a transmitting unit, a receiving unit, a transmitting and receiving unit, or a communication unit. Furthermore, at least one of the baseband processing unit (5-20) and the RF processing unit (5-10) may include a plurality of communication modules to support a plurality of different wireless access technologies. Additionally, at least one of the baseband processing unit (5-20) and the RF processing unit (5-10) may include different communication modules to process signals of different frequency bands. For example, the different wireless access technologies may include wireless LAN (e.g., IEEE 802.11), cellular network (e.g., LTE), etc. In addition, the above different frequency bands may include super high frequency (SHF) bands (e.g., 2 NRHz, NRHz) and millimeter wave (e.g., 60 GHz) bands.

[0133] The storage unit (5-30) can store data such as basic programs, application programs, and setting information for the operation of the terminal. In particular, the storage unit (5-30) can store information related to a second connection node that performs wireless communication using wireless connection technology. Additionally, the storage unit (5-30) can provide the stored data upon request from the control unit (5-40).

[0134] The control unit (5-40) can control the overall operations of the terminal. For example, the control unit (5-40) can control the terminal to perform the embodiments and / or methods of the present disclosure described above. For example, the control unit (5-40) can transmit and receive signals through the baseband processing unit (5-20) and the RF processing unit (5-10). Additionally, the control unit (5-40) can write and read data to and from the storage unit (5-30). To this end, the control unit (5-40) may include at least one processor. For example, the control unit (5-40) may include a communication processor (CP) that performs control for communication and an application processor (AP) that controls upper layers such as applications, and may include a multiple connection processing unit (5-42) as illustrated in the drawing.

[0135] FIG. 6 is a drawing illustrating the structure of a base station according to one embodiment of the present disclosure.

[0136] Referring to FIG. 6, according to one example of the present disclosure, a base station may be configured to include an RF processing unit (6-10), a baseband processing unit (6-20), a backhaul communication unit (6-30), a storage unit (6-40), and a control unit (6-50).

[0137] The RF processing unit (6-10) can perform functions for transmitting and receiving signals through a wireless channel, such as signal band conversion and amplification. That is, the RF processing unit (6-10) can up-convert a baseband signal provided by the baseband processing unit (6-20) into an RF band signal and transmit it through an antenna, and can down-convert an RF band signal received through the antenna into a baseband signal. For example, the RF processing unit (6-10) may include a transmit filter, a receive filter, an amplifier, a mixer, an oscillator, a DAC, an ADC, etc. Although only one antenna is shown in the drawing, the base station may be equipped with multiple antennas. Additionally, the RF processing unit (6-10) may include multiple RF chains. Furthermore, the RF processing unit (6-10) may perform beamforming. For beamforming, the RF processing unit (6-10) may adjust the phase and magnitude of each of the signals transmitted and received through multiple antennas or antenna elements. The above RF processing unit can perform down-to-down MIMO operation by transmitting one or more layers.

[0138] The baseband processing unit (6-20) can perform a conversion function between a baseband signal and a bit sequence according to the physical layer specifications of the wireless access technology. For example, when transmitting data, the baseband processing unit (6-20) can generate complex symbols by encoding and modulating the transmitted bit sequence. Additionally, when receiving data, the baseband processing unit (6-20) can restore the received bit sequence by demodulating and decoding the baseband signal provided by the RF processing unit (6-10). For example, in the case of an OFDM method, when transmitting data, the baseband processing unit (6-20) can generate complex symbols by encoding and modulating the transmitted bit sequence, and after mapping the complex symbols to subcarriers, can construct OFDM symbols through IFFT operation and CP insertion. Additionally, upon receiving data, the baseband processing unit (6-20) can divide the baseband signal provided by the RF processing unit (6-10) into OFDM symbol units, restore the signals mapped to subcarriers through FFT operations, and then restore the received bit sequence through demodulation and decoding. The baseband processing unit (6-20) and the RF processing unit (6-10) can transmit and receive signals as described above. Accordingly, the baseband processing unit (6-20) and the RF processing unit (6-10) may be referred to as a transmitting unit, a receiving unit, a transmitting and receiving unit, a communication unit, or a wireless communication unit.

[0139] The backhaul communication unit (6-30) can provide an interface for communicating with other nodes within the network. That is, the backhaul communication unit (6-30) can convert a bit sequence transmitted from the main base station to another node, e.g., an auxiliary base station, a core network, etc., into a physical signal, and can convert a physical signal received from the other node into a bit sequence.

[0140] The storage unit (6-40) can store data such as basic programs, application programs, and configuration information for the operation of the main station. In particular, the storage unit (6-40) can store information regarding bearers assigned to connected terminals, measurement results reported from connected terminals, etc. Additionally, the storage unit (6-40) can store information that serves as a criterion for determining whether to provide or disconnect multiple connections to the terminals. Furthermore, the storage unit (6-40) can provide the stored data upon a request from the control unit (6-50).

[0141] The control unit (6-50) can control the overall operations of the main base station. For example, the control unit (6-50) can control the base station to perform the embodiments and / or methods of the present disclosure described above. For example, the control unit (6-50) can transmit and receive signals through the baseband processing unit (6-20) and the RF processing unit (6-10) or through the backhaul communication unit (6-30). Additionally, the control unit (6-50) can write and read data to and from the storage unit (6-40). To this end, the control unit (6-50) may include at least one processor and may include a multiple connection processing unit (6-52) as illustrated in the drawing.

[0142] Meanwhile, the embodiments of the present disclosure disclosed in this specification and drawings are merely specific examples provided to facilitate the explanation of the technical content of the present disclosure and to aid in understanding the present disclosure, and are not intended to limit the scope of the present disclosure. That is, it is obvious to those skilled in the art that other variations based on the technical concept of the present disclosure are possible.

[0143] In addition, each of the above embodiments may be combined and operated as needed. For example, parts of one embodiment of the present disclosure and another embodiment may be combined to operate a base station and a terminal.

[0144] In addition, the embodiments of the present disclosure are applicable to other communication systems, and other variations based on the technical concept of the embodiments may also be implemented. For example, the embodiments may be applied to LTE systems, 5G, NR systems, or 6G systems.

[0145] Therefore, the scope of the present disclosure should not be limited to the described embodiments, but should be defined by the claims set forth below as well as equivalents thereof.

Claims

1. A method performed by a terminal of a wireless communication system, A step of transmitting a first message to a base station requesting partition information for each partition in which a dataset necessary for terminal-side model training is divided into one or more partitions; A step of receiving a second message including the partition information from the base station; and The method includes the step of transmitting a third message to the base station requesting the transmission of one or more partitions of the dataset based on the partition information. A method characterized in that the above third message includes at least one of information indicating one or more partitions requesting transmission or information indicating one or more partitions already held.

2. In Paragraph 1, The first message above includes at least one of information about a use-case to which the terminal-side model learning is to be applied or information about the type of dataset, and The second message above includes at least one of information about the type of the dataset, information about the total number of partitions of the dataset, information about the ID (identifier) ​​or index of one or more partitions, or information about the size of one or more partitions. The above use case is a method characterized by CSI (channel state information) compression.

3. In Paragraph 1, A step of receiving one or more RRC (radio resource control) segments including partitions of the dataset from the base station; and The method further includes the step of assembling one or more of the above RRC segments to identify the partition, and A method characterized in that the above dataset includes at least one of parameter information for a learned model or information necessary for compatibility between a terminal-side model and a network-side model.

4. In Paragraph 3, A method characterized in that the assembled partition is transmitted to an object that performs upper layer or terminal-side model learning.

5. A method performed by a base station of a wireless communication system, A step of receiving a first message from a terminal requesting partition information for each partition in which a dataset required for training a terminal-side model is divided into one or more partitions; A step of transmitting a second message containing the partition information to the terminal; and The method includes the step of receiving a third message from the terminal requesting the transmission of one or more partitions of the dataset according to the partition information, and A method characterized in that the above third message includes at least one of information indicating one or more partitions requesting transmission or information indicating one or more partitions already held.

6. In Paragraph 5, The first message above includes at least one of information about a use-case to which the terminal-side model learning is to be applied or information about the type of dataset, and The second message above includes at least one of information about the type of the dataset, information about the total number of partitions of the dataset, information about the ID (identifier) ​​or index of one or more partitions, or information about the size of one or more partitions. The above use case is a method characterized by CSI (channel state information) compression.

7. In Paragraph 5, The method further includes the step of transmitting one or more RRC (radio resource control) segments, including partitions of the dataset, to the terminal. The above dataset includes at least one of parameter information for a learned model or information necessary for compatibility between a terminal-side model and a network-side model, and A method characterized in that the partition assembled based on the above one or more RRC segments is transmitted to an object to perform upper layer or terminal-side model learning.

8. In a terminal of a wireless communication system, At least one transceiver; At least one processor connected to the above at least one transceiver so as to be able to communicate; and The device includes a memory that stores instructions that are connected to communicate with at least one processor and can be executed individually or in any combination of the at least one processors, wherein the terminal transmits a first message to a base station requesting partition information for each partition of a dataset divided into one or more partitions necessary for terminal-side model training, receives a second message from the base station containing the partition information, and transmits a third message to the base station requesting the transmission of one or more partitions of the dataset based on the partition information. The terminal is characterized in that the above third message includes at least one of information indicating one or more partitions requesting transmission or information indicating one or more partitions already possessed.

9. In Paragraph 8, The first message above includes at least one of information about a use-case to which the terminal-side model learning is to be applied or information about the type of dataset, and The second message above includes at least one of information about the type of the dataset, information about the total number of partitions of the dataset, information about the ID (identifier) ​​or index of one or more partitions, or information about the size of one or more partitions. The above use case is a terminal characterized by CSI (channel state information) compression.

10. In Paragraph 8, The above command receives one or more radio resource control (RRC) segments containing a partition of the dataset from the base station, and assembles the one or more RRC segments to identify the partition. A terminal characterized by the above dataset including at least one of parameter information for a learned model or information necessary for compatibility between a terminal-side model and a network-side model.

11. In Paragraph 10, A terminal characterized in that the assembled partition is transmitted to an object that performs upper layer or terminal-side model learning.

12. In a base station of a wireless communication system, At least one transceiver; At least one processor connected to the above at least one transceiver so as to be able to communicate; and The memory includes a command that is connected to communicate with at least one processor and is executable individually or in any combination of the at least one processor, wherein the base station receives a first message from a terminal requesting partition information for each partition of a dataset divided into one or more partitions necessary for terminal-side model training, transmits a second message containing the partition information to the terminal, and receives a third message from the terminal requesting the transmission of one or more partitions of the dataset according to the partition information. A base station characterized by the above-mentioned third message including at least one of information indicating one or more partitions requesting transmission or information indicating one or more partitions already possessed.

13. In Paragraph 12, The first message above includes at least one of information about a use-case to which the terminal-side model learning is to be applied or information about the type of dataset, and The second message above includes at least one of information about the type of the dataset, information about the total number of partitions of the dataset, information about the ID (identifier) ​​or index of one or more partitions, or information about the size of one or more partitions. The above use case is a base station characterized by CSI (channel state information) compression.

14. In Paragraph 12, The above command causes the terminal to transmit one or more RRC (radio resource control) segments including partitions of the dataset, and A base station characterized by the above dataset including at least one of parameter information for a learned model or information necessary for compatibility between a terminal-side model and a network-side model.

15. In Paragraph 14, A base station characterized in that the partition assembled based on the above one or more RRC segments is transmitted to an object to perform upper layer or terminal-side model learning.

Images