Data collection method for ai / ML in wireless communication system

Abstract

Disclosed is a method of using AI / ML for channel state information report collection performed by a terminal. The disclosed method comprises the steps of: receiving, from a base station, integration / separation information of data for AI / ML; receiving a reference signal from the base station; determining a data collection scheme on the basis of the integration / separation information of the data for AI / ML; collecting data related to channel state information on the basis of the determined data collection scheme and the reference signal; and inputting the collected data to an AI / ML model.

Description

Data Collection Methods for AI / ML in Wireless Communication Systems

[0001] The present disclosure relates to a method for collecting data for AI / ML in a wireless communication system. The present disclosure discloses a method for collecting data for AI / ML in a wireless communication system supporting sub-band full duplex (SBFD) communication and applying the collected data to an AI / ML model.

[0002] A communication system may include a core network, base stations (e.g., macro base stations, small base stations, relays, etc.), terminals, etc. Communication between the base stations and terminals may be performed based on various radio access technologies (RATs) (e.g., 4G communication technology, 5G communication technology, WiBro (wireless broadband) technology, WLAN (wireless local area network) technology, WPAN (wireless personal area network) technology, etc.).

[0003] Fifth-generation mobile communications considered wide bandwidth allocation in the millimeter wave band to enable high-speed data transmission. However, utilizing the high frequencies of the millimeter wave band resulted in high path loss and degraded uplink performance of mobile terminals with limited battery power. To overcome these limitations of the millimeter wave band, beamforming technology was introduced, which utilizes multiple antennas to increase antenna gain. However, the characteristics of mobile terminals necessitated a new technology called beam management. Realistically, precisely steering a narrow beam with high gain to a moving terminal presents challenges, limiting the ability to provide sufficient coverage for mobile communication systems.

[0004] 3GPP selected the sub-band full duplex system (SBFD) and AI / ML as study items in 5G Release 18 and conducted research and discussions.

[0005] Sub-band Full Duplex (SBFD) supports simultaneous transmission and reception of uplink and downlink. According to SBFD, uplink and downlink can be allocated by subband in the frequency domain. SBFD is referred to as cross division duplex (XDD). In XDD, 'X' stands for time or frequency. SBFD differs from the existing TDD and FDD in that it separates the uplink and downlink using both the time domain and the frequency domain. The frame structure in SBFD mode can be very flexible compared to the existing TDD and FDD. In addition, by utilizing SBFD, wireless communication systems can improve uplink throughput, uplink transmission delay, and enhance coverage through uplink repetition.

[0006] AI / ML technologies are being used to streamline network operations and optimize service quality by leveraging artificial intelligence and machine learning. 3GPP is prioritizing the adoption of AI / ML in scenarios such as beam management, positioning, channel state information prediction, and channel state information compression for 5G, expecting improvements in system throughput, resource utilization, and coverage.

[0007] According to at least one embodiment, a method for a terminal or base station to collect data for AI / ML in a mobile communication system supporting SBFD and AI / ML is disclosed.

[0008] According to at least one embodiment, a method is disclosed for utilizing AI / ML using collected data by a terminal or base station in a mobile communication system supporting SBFD and AI / ML.

[0009] In one aspect, a method for collecting data for AI / ML performed by a terminal is disclosed. The disclosed method comprises the steps of: receiving integration / separation information for data for AI / ML from a base station; receiving a reference signal from the base station; determining a data collection method based on the integration / separation information for data for AI / ML; collecting data based on the determined data collection method and the reference signal; and inputting the collected data into an AI / ML model.

[0010] According to at least one embodiment, the terminal (120-1) and / or the base station (110-1) may determine to integrate / separately collect data of the SBFD time interval and data of the Non-SBFD time interval based on the integration / separation information of data for AI / ML. If the data of the SBFD time interval is separated, the separated data may be input into the AI / ML model of the SBFD time interval to improve the performance of the AI / ML model of the SBFD time interval. Similarly, if the data of the Non-SBFD time interval is separated, the separated data may be input into the AI / ML model of the Non-SBFD time interval to improve the performance of the AI / ML model of the Non-SBFD time interval. If the difference between the channel environment of the SBFD time interval and the channel environment of the Non-SBFD time interval is not large, the terminal (120-1) and / or the base station (110-1) may acquire data without distinguishing between the SBFD time interval and the Non-SBFD time interval. This simplifies the process of AI / ML inference, such as channel state information inference, beam management, interference information inference, and location information inference, and saves resources. According to at least one embodiment, the terminal (120-1) and / or the base station (110-1) can monitor the performance of the AI / ML model and decide whether to reuse it, thereby preventing waste of time and resources required for retraining other AI / ML models. According to at least one embodiment, the AI / ML function can be improved by allowing the terminal (120-1) and / or the base station (110-1) to store the AI / ML model in advance and use the stored AI / ML model when necessary while the validity period has not expired.

[0011] FIG. 1 is a diagram illustrating a wireless communication system according to an exemplary embodiment.

[0012] Figure 2 is a block diagram exemplarily showing the configuration of a communication node constituting a communication system.

[0013] Figure 3 is a diagram illustrating an example of a 5G network architecture.

[0014] Figure 4 is a diagram showing the legacy frame structure in TDD mode.

[0015] FIG. 5 is a diagram illustrating a frame structure in SBFD (subband full duplex) mode according to an exemplary embodiment.

[0016] Figure 6 is a block diagram showing the entire cycle of AI / ML functions in a mobile communication system.

[0017] FIG. 7 is a flowchart illustrating a method for collecting data for AI / ML in a mobile communication system according to an exemplary embodiment.

[0018] Figure 8 is a flowchart exemplarily showing the process of performing step S110 of Figure 7.

[0019] Figure 9 is a diagram showing examples in which Associated ID is included in IE.

[0020] Fig. 10 is a flowchart exemplarily showing the process of performing step S130 of Fig. 7.

[0021] Figure 11 is a diagram illustrating an example of a data collection section.

[0022] Figures 12 to 14 are conceptual diagrams showing examples of how a terminal (120-1) collects data based on an Associated ID.

[0023] Fig. 15 is a flowchart exemplarily showing the process of performing step S110 of Fig. 7.

[0024] Fig. 16 is a flowchart exemplarily showing the process of performing step S130 of Fig. 7.

[0025] Figure 17 is a diagram exemplifying data integration / separation information for AI / ML.

[0026] Figure 18 is a diagram exemplifying data integration / separation information for AI / ML.

[0027] Figure 19 is a diagram exemplifying data integration / separation information for AI / ML.

[0028] Figure 20 is a flowchart exemplarily showing the process of performing step S110 of Figure 7.

[0029] Figure 21 is a flowchart exemplarily showing the process of performing step S130 of Figure 7.

[0030] Figure 22 is a diagram exemplifying a resource allocation configuration for CSI measurement in a mobile communication system.

[0031] Figure 23 is a diagram illustrating an example of a CSI resource configuration for measuring SBFD time intervals in a communication system.

[0032] Figure 24 is a diagram illustrating an example of a CSI resource configuration for measuring a non-SBFD time interval in a communication system.

[0033] Figure 25 is a diagram illustrating an example of a method for distinguishing between SBFD time intervals and non-SBFD time intervals through CSI resource configuration in a communication system, in which the distinction is made in the frequency domain.

[0034] Figure 26 is a diagram exemplarily showing the pattern cycles of downlink and uplink in a mobile communication system.

[0035] Figure 27 is a flowchart illustrating a case where the terminal side proactively performs the management function of the AI / ML model.

[0036] Figure 28 is a flowchart illustrating a case where the base station proactively performs the management function of the AI / ML model.

[0037] Figure 29 is a flowchart illustrating a method of utilizing an AI / ML model for information inference of an uplink channel.

[0038] Figure 30 is a diagram illustrating an IE including information indicating an Associated ID.

[0039] Figure 31 is a flowchart illustrating an example of a management method for AI / ML.

[0040] Figure 32 is a flowchart illustrating an example of a management method for AI / ML.

[0041] In one aspect, a method for collecting data for AI / ML performed by a terminal is disclosed. The disclosed method comprises the steps of: receiving integration / separation information for data for AI / ML from a base station; receiving a reference signal from the base station; determining a data collection method based on the integration / separation information for data for AI / ML; collecting data based on the determined data collection method and the reference signal; and inputting the collected data into an AI / ML model.

[0042] The integration / separation information of data for the above AI / ML may include information on whether data in the SBFD time interval and non-SBFD data are separated or integrated.

[0043] The integration / separation information of data for the above AI / ML may include information explicitly indicating whether data in the SBFD time interval and data in the non-SBFD time interval are separated or integrated.

[0044] The integration / separation information of data for the above AI / ML may include information implicitly indicating whether data in the SBFD time interval and data in the non-SBFD time interval are separated or integrated.

[0045] If the integration / separation information of data for the AI / ML indicates separation of data in the SBFD time period and data in the Non-SBFD time period, the terminal can separate data collected in the SBFD time period and input the separated data into the AI / ML model of the SBFD time period.

[0046] If the integration / separation information of data for the AI / ML indicates separation of data for the SBFD time period and data for AI / ML for the Non-SBFD time period, the terminal can separate data collected in the Non-SBFD time period and input the separated data into the AI / ML model of the Non-SBFD time period.

[0047] If the integration / separation information for the above AI / ML data indicates integration of data for the SBFD time period and data for AI / ML for the Non-SBFD time period, the terminal can collect data without distinguishing between the SBFD time period and the Non-SBFD time period, and input the collected data into the AI / ML model.

[0048] The integration / separation information of data for the above AI / ML includes information on the associated ID of the AI / ML model, and the associated ID can be assigned according to the data collection configuration of the AI / ML model.

[0049] The terminal can determine whether to separate or integrate data of the SBFD time period and data of the Non-SBFD time period based on the data collection configuration of the AI / ML model corresponding to the Associated ID.

[0050] The terminal can collect data based on the data collection configuration of the AI / ML model corresponding to the Associated ID, and input the collected data into the AI / ML model corresponding to the Associated ID.

[0051] The integration / separation information of data for the above AI / ML is transmitted by a broadcast message, and the broadcast message may include information explicitly indicating whether data in the SBFD time interval and data in the Non-SBFD time interval are separated or integrated.

[0052] The integration / separation information of data for the above AI / ML is transmitted by DCI (Downlink Control Information), and the DCI may include information indicating at least one AI / ML model and control information related to the at least one AI / ML model.

[0053] The integration / separation information of data for the above AI / ML is transmitted by a MAC CE (MAC Control Element), and the MAC CE may include information indicating at least one AI / ML model and activation / deactivation information of the at least one AI / ML model.

[0054] The integration / separation information of data for the above AI / ML is transmitted through at least one of the resource configuration information of the reference signal and the resource configuration information of the channel status report, and at least one of the resource configuration information of the reference signal and the resource configuration information of the channel status report can implicitly indicate whether data of the SBFD time interval and data of the Non-SBFD time interval are separated or integrated.

[0055] If the reference signal resources according to the resource configuration information of the above reference signal are allocated only to the DL sub-band of the SBFD time interval, the resource configuration of the above reference signal may implicitly indicate separating data of the SBFD time interval.

[0056] The method may further include the steps of transmitting performance feedback of the first AI / ML model to a base station; receiving an instruction to reuse the first AI / ML model from the base station; and reusing the first AI / ML model according to the instruction to reuse the first AI / ML model.

[0057] In another aspect, a method performed by a base station is disclosed. The disclosed method comprises the steps of transmitting data integration / separation information for AI / ML; transmitting a reference signal; and receiving a channel state information report derived by inputting data collected according to a data collection method determined based on the data integration / separation information for AI / ML into an AI / ML model.

[0058] In another aspect, a method performed by a base station is disclosed. The disclosed method comprises the steps of: transmitting data integration / separation information for AI / ML; receiving a reference signal from a terminal; determining a data collection method based on the data integration / separation information for AI / ML; collecting data based on the determined data collection method and the reference signal; and inputting the collected data into an AI / ML model.

[0059] The present invention is susceptible to various modifications and embodiments. Specific embodiments are illustrated and described in detail in the drawings. However, this is not intended to limit the present invention to specific embodiments, but rather to encompass all modifications, equivalents, and alternatives falling within the spirit and technical scope of the present invention.

[0060] While terms such as "first" and "second" may be used to describe various components, these components should not be limited by these terms. These terms are used solely to distinguish one component from another. For example, without departing from the scope of the present invention, a first component could be referred to as a "second component," and similarly, a second component could also be referred to as a "first component." The term "and / or" encompasses any combination of multiple related items or any one of multiple related items.

[0061] When a component is referred to as being "connected" or "connected" to another component, it should be understood that it may be directly connected or connected to that other component, but that there may be other components intervening. Conversely, when a component is referred to as being "directly connected" or "connected" to another component, it should be understood that there are no other components intervening.

[0062] The terminology used in this application is only used to describe specific embodiments and is not intended to limit the present invention. The singular expression includes the plural expression unless the context clearly indicates otherwise. In this application, it should be understood that the terms "comprise" or "have" indicate the presence of a feature, number, step, operation, component, part, or combination thereof described in the specification, but do not exclude in advance the possibility of the presence or addition of one or more other features, numbers, steps, operations, components, parts, or combinations thereof.

[0063] Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. Terms defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant technology, and will not be interpreted in an idealized or overly formal sense unless explicitly defined herein.

[0064] Hereinafter, with reference to the attached drawings, preferred embodiments of the present invention will be described in more detail. In order to facilitate an overall understanding in describing the present invention, identical reference numerals will be used for identical components in the drawings, and redundant descriptions of identical components will be omitted.

[0065] Table 1 shows the abbreviations used in this disclosure.

[0066] 약어전체 이름3GPP3rd Generation Partnership ProjectACKAcknowledgementAFApplication FunctionAIArtificial IntelligenceAMFAccess and Mobility Management FunctionAUSFAuthentication Server FunctionBWPBandwidth PartC-RNTICell RNTICSIChannel State InformationCSI-RSChannel State Information Reference SignalCLICross link InterferenceCEControl ElementDCIDownlink Control InformationIEInformation elementMACMedium Access ControlMLMachine LearningNSSFNetwork Slicing Selection FunctionNEFNetwork Exposure FunctionNRFNF Repository FunctionPCFPolicy Control FunctionPDCCHPhysical Downlink Control ChannelPDSCHPhysical Downlink Shared ChannelPDUProtocol Data UnitPRACHPhysical Random Access ChannelPT-RSPhase Tracking Reference SignalPUCCHPhysical Uplink Control ChannelPUSCHPhysical Uplink Shared ChannelRARandom AccessRACHRandom Access ChannelRANRadio Access NetworkRBResource BlockRRCRadio Resource ControlRSRPReference Signal Received PowerRSRQReference Signal Received QualityRSSIReceived Signal StrengthIndicatorSBFDSub-Band Full DuplexSISelf InterferenceSIBSystem information blockUEUser EquipmentDLDownlinkULUplinkRARRandom Access ResponseRRMRadio Resource ManagementRRCRadio Resource ControlRSRPReference Signal Received PowerQoSQuality of ServiceDRBData Radio BearerDAPSDual Active Protocol StackTDDTime Division DuplexingFDDFrequency Division DuplexingSSBSS BlockCQIChannel Quality IndicatorPMIPrecoding Matrix IndicatorLILayer IndicatorRIRank IndicatorQCLQuasi coLocation

[0067] FIG. 1 is a diagram illustrating a wireless communication system (100) according to an exemplary embodiment.

[0068] Referring to FIG. 1, a wireless communication system (100) may be composed of a plurality of communication nodes (110-1, 110-2, 110-3, 120-1, 120-2, 120-3, 120-4). Here, a communication node means a node capable of transmitting and receiving signals in the wireless communication system (100), and each of the plurality of communication nodes (110-1, 110-2, 110-3, 120-1, 120-2, 120-3, 120-4) may support at least one communication protocol. For example, each of the plurality of communication nodes (110-1, 110-2, 110-3, 120-1, 120-2, 120-3, 120-4) may support cellular communication (e.g., long term evolution (LTE), advanced LTE-A, 5G NR, 5G-Advanced, etc. as defined in the 3rd generation partnership project (3GPP) standard).

[0069] For example, each of the plurality of communication nodes (110-1, 110-2, 110-3, 120-1, 120-2, 120-3, 120-4) can support a communication protocol based on CDMA (code division multiple access), a communication protocol based on WCDMA (wideband CDMA), a communication protocol based on TDMA (time division multiple access), a communication protocol based on FDMA (frequency division multiple access), a communication protocol based on OFDM (orthogonal frequency division multiplexing), a communication protocol based on OFDMA (orthogonal frequency division multiple access), a communication protocol based on SC (single carrier)-FDMA, a communication protocol based on NOMA (non-orthogonal multiple access), a communication protocol based on SDMA (space division multiple access), SBFD (sub-band full duplex), AI / ML, etc.

[0070] A plurality of communication nodes (110-1, 110-2, 110-3, 120-1, 120-2, 120-3, 120-4) may include a plurality of base stations (110-1, 110-2, 110-3) and a plurality of terminals (120-1, 120-2, 120-3, 120-4). Each of the base stations (110-1, 110-2, 110-3) may form a cell. The cell may include a small cell, a macro cell, a pico cell, a femto cell, etc., but the embodiment is not limited thereto. For example, the first terminal (120-1) and the second terminal (120-2) may belong to the coverage of the first base station (110-1), the third terminal (120-3) may belong to the coverage of the second base station (110-2), and the fourth terminal (120-4) may belong to the coverage of the third base station (110-3).

[0071] The plurality of base stations (110-1, 110-2, 110-3) may be referred to as gNodeB (gNB), NodeB, evolved NodeB, base transceiver station (BTS), radio base station, radio transceiver, access point, access node, road side unit (RSU), radio remote head (RRH), transmission point (TP), transmission and reception point (TRP), relay node, etc. The plurality of base stations (110-1, 110-2, 110-3) may form a radio access network (RAN). The radio access network may be connected to a core network.

[0072] Each of the plurality of terminals (120-1, 120-2, 120-3, 120-4) may be referred to as a terminal, an access terminal, a mobile terminal, a station, a subscriber station, a mobile station, a portable subscriber station, a node, a device, etc.

[0073] The communication protocol support ranges of the plurality of base stations (110-1, 110-2, 110-3) may be different from each other. For example, some of the plurality of base stations (110-1, 110-2, 110-3) may support the SBFD mode, while others may not support the SBFD mode. Similarly, the communication protocol support ranges of the plurality of terminals (120-1, 120-2, 120-3, 120-4) may also be different from each other. For example, some of the plurality of terminals (120-1, 120-2, 120-3, 120-4) may support the SBFD mode, while others may not support the SBFD mode.

[0074] The communication protocol support ranges of the plurality of base stations (110-1, 110-2, 110-3) may differ from each other. For example, some of the plurality of base stations (110-1, 110-2, 110-3) may support AI / ML functions, while others may not support AI / ML functions. Similarly, the communication protocol support ranges of the plurality of terminals (120-1, 120-2, 120-3, 120-4) may also differ from each other. For example, some of the plurality of terminals (120-1, 120-2, 120-3, 120-4) may support AI / ML functions, while others may not support AI / ML functions.

[0075] For convenience, the first terminal (120-1) among the terminals is referred to as terminal (120-1) below, and the first base station (110-1) among the base stations is referred to as base station (110-1).

[0076] Fig. 2 is a block diagram exemplarily showing the configuration of a communication node (200) constituting a communication system. At least some of the communication nodes (110-1, 110-2, 110-3, 120-1, 120-2, 120-3) shown in Fig. 1 may correspond to the communication node (200) shown in Fig. 2.

[0077] Referring to FIG. 2, a communication node (200) may include at least one processor (210), a memory (220), and a transmission / reception device (230) that is connected to a network and performs communication. In addition, the communication node (200) may further include an input interface device (240), an output interface device (250), a storage device (260), etc. Each component included in the communication node (200) may be connected by a bus (270) and communicate with each other.

[0078] The processor (210) can execute program commands stored in at least one of the memory (220) and the storage device (260). The processor (210) may refer to a central processing unit (CPU), a graphics processing unit (GPU), or a dedicated processor in which methods according to embodiments of the present invention are performed. Each of the memory (220) and the storage device (260) may be configured with at least one of a volatile storage medium and a non-volatile storage medium. For example, the memory (220) may be configured with at least one of a read-only memory (ROM) and a random access memory (RAM).

[0079] Figure 3 is a diagram illustrating an example of a 5G network architecture.

[0080] Referring to FIG. 3, the 5G network architecture may include NSSF providing a network slicing control solution, NEF providing a network function opening solution, NRF providing an NF interworking control solution within a 5G network, PCF providing a data packet flow policy control solution, UDM providing a user information and policy management solution, AMF providing a network access and mobility management solution, SMF providing a terminal / network session management solution, UPF providing a base station-to-base station user packet routing and terminal connectivity provision solution, a radio access network (RAN), and user equipment (UE).

[0081] The N1 reference point is defined to transmit signaling between the UE and the AMF, the N2 reference point is defined to connect between the RAN's access node and the AMF, and the N3 reference point is defined to connect between the RAN's access node and the UPF.

[0082] Figure 4 is a diagram showing the legacy frame structure in TDD mode.

[0083] Referring to FIG. 4, a time interval (40) may be defined in the time domain. The time interval may also be referred to as a time unit. The time interval (40) may correspond to any one of a slot, a symbol, and a subframe. As another example, the time interval (40) may be defined to include multiple slots. As another example, the time interval (40) may be defined to include multiple symbols. As another example, the time interval (40) may be defined to include multiple subframes.

[0084] In TDD mode, downlink resources may be allocated in the first to fourth time intervals (40, 41, 42, 43). Downlink signals may be transmitted in the first to fourth time intervals (40, 41, 42, 43). Uplink resources may be allocated in the fifth time interval (44). Downlink signals may be transmitted in the fifth time interval (44). According to the legacy frame structure illustrated in FIG. 4, even when the terminal (120-1) needs to transmit an uplink signal between the first to fourth time intervals (40, 41, 42, 43), the terminal (120-1) may wait until the fifth time interval (44) to which the uplink resources are allocated arrives before transmitting the uplink signal. This may cause a delay (latency) until the terminal (200-1) transmits the uplink signal.

[0085] FIG. 5 is a diagram illustrating a frame structure in SBFD (subband full duplex) mode according to an exemplary embodiment.

[0086] Referring to FIG. 5, in the SBFD mode, uplink resources and downlink resources can be allocated simultaneously in the first to fourth time intervals (40, 41, 42, 43). In the first to fourth time intervals (40, 41, 42, 43), UL subbands for uplink transmission and DL subbands for downlink transmission can be allocated simultaneously. The time interval in which UL subbands and DL subbands are allocated simultaneously to enable subband full-duplex communication can be referred to as an SBFD time interval.

[0087] Although not shown in FIG. 5, there may be a time period to which downlink is allocated prior to the first to fourth time intervals (40, 41, 42, 43) corresponding to the SBFD time period. Through the downlink allocated prior to the SBFD time period, the base station can transmit scheduling information for the SBFD time period and SBFD subband to the terminal.

[0088] In FIG. 5, the first to fourth time intervals (40, 41, 42, and 43) may correspond to the SBFD time domain. The SBFD time domain may be defined to include at least one slot. As another example, the SBFD time domain may be defined to include at least one symbol. As another example, the SBFD time domain may be defined to include at least one subframe.

[0089] Each of the UL subband and the DL subband may include at least one RB. For example, each of the UL subband and the DL subband may include RBs arranged contiguously. As another example, each of the UL subband and the DL subband may include RBs arranged non-contiguously in at least part.

[0090] Each of the UL subbands and the DL subbands may include at least one BWP. For example, each of the UL subbands and the DL subbands may include at least one legacy BWP. As another example, each of the UL subbands and the DL subbands may include at least one BWP separately defined to support SBFD.

[0091] As described above, SBFD subbands can be configured within a TDD mode carrier for SBFD mode operation. The SBFD subbands can include UL subbands and DL subbands. In FIG. 5, the UL subband is exemplarily shown as being located in the middle of the carrier. However, the embodiment is not limited thereto. For example, the UL subband can be located at the edge of the carrier. In another example, the UL subband can be located at a location other than the exact center of the carrier.

[0092] Additionally, a guard period may be placed between the SBFD time domain and the non-SBFD time domain to which SBFD is not applied. Each of the SBFD time domain and the non-SBFD time domain may include at least one symbol.

[0093] The time and frequency positions of the SBFD subbands can be configured and indicated semi-statically. Alternatively, the time and frequency positions of the SBFD subbands can be configured and indicated dynamically.

[0094] In SBFD mode, the efficiency of wireless resource use can be improved by flexibly distributing uplink and downlink resources. In addition, the delay can be reduced because the time that the terminal (120-1) waits for uplink signal transmission and the time that the terminal (120-1) waits for downlink signal reception are reduced. In addition, in SBFD mode, the terminal (120-1) repeatedly transmits UL signals and repeatedly receives DL signals in multiple time intervals (40, 41, 42, 43), thereby increasing the transmission gain of the UL and DL signals, and expanding the coverage of the first base station (110-1) and the first terminal (120-1).

[0095] Figure 6 is a block diagram illustrating the entire cycle of AI / ML functions in a mobile communication system. The AI / ML function blocks illustrated in Figure 6 may be implemented by at least one of the communication nodes of the mobile communication system. For example, the AI / ML function blocks illustrated in Figure 6 may be implemented by at least one of a base station and a terminal.

[0096] When the terminal (120-1) is connected to the service area of ​​the base station (110-1) that supports AI / ML functions, the terminal (120-1) or the base station can activate at least some of the function blocks (610, 612, 614, 616, 618) of FIG. 6.

[0097] The terminal (120-1) can collect data related to a wireless channel by activating the data collection block (610). The terminal (120-1) can collect data related to wireless channel information from reference signals (CSI-RS, SSB, PT-RS, PRS, downlink reference signal, etc.). When the base station (110-1) activates the data collection block (610), data related to wireless channel information can be collected from reference signals (SRS, PRS, PT-RS) and channel status reports.

[0098] The data collected by the data collection block (610) can be used as data for learning / monitoring / inference. The model learning block (612) can train an AI / ML model using the data collected by the data collection block (610). The model storage block (618) can store the trained AI / ML model. The stored model can be used by the inference block (616). The inference block (616) can infer information necessary for operating a wireless communication system based on the data collected by the data collection block (610).

[0099] The inference block (616) can report the inference result to the AI / ML management block (614). The AI / ML management block (614) can perform overall functions related to training of the AI / ML model, selection of the AI / ML model, activation / deactivation of the AI / ML model, and control of other function blocks. The AI / ML management block (614) can monitor the performance of the AI / ML model to maintain the performance of the AI / ML and provide feedback on the training result to the model training block (612). If the performance of the monitored AI / ML model is low and retraining is required, the AI / ML management block (614) can request retraining from the model training block (612). The AI / ML management block (614) can determine activation / deactivation of the AI / ML model and model selection / change. The AI / ML management block (614) performs management instructions such as fallback to a conventional communication method that does not use the AI / ML function.

[0100] Hereinafter, a time interval to which the SBFD mode is applied is referred to as an 'SBFD time interval', and a time interval to which the SBFD mode is not applied is referred to as a 'Non-SBFD time interval'. An SBFD time interval may include at least one symbol to which the SBFD mode is applied. As another example, an SBFD time interval may include at least one slot to which the SBFD mode is applied. The slot may be replaced by a subframe or a frame. A Non-SBFD time interval may include at least one symbol to which the SBFD is not applied. As another example, a Non-SBFD time interval may include at least one slot to which the SBFD mode is not applied. The slot may be replaced by a subframe or a frame.

[0101] FIG. 7 is a flowchart illustrating a method for collecting data for AI / ML in a mobile communication system according to an exemplary embodiment.

[0102] Referring to FIG. 7, the base station (110-1) may transmit to the terminal (120-1) integration / separation information of data for AI / ML. The integration / separation information of data for AI / ML may include information on whether data in the SBFD time interval and data in the Non-SBFD time interval are to be integrated and collected or collected separately. The integration / separation information of data for AI / ML may include information that explicitly / directly indicates whether data in the SBFD time interval and data in the Non-SBFD time interval are to be integrated or separated. As another example, the integration / separation information of data for AI / ML may include information that implicitly indicates whether data in the SBFD time interval and data in the Non-SBFD time interval are to be integrated and collected or collected separately.

[0103] The base station (110-1) can decide whether to manage the data of the SBFD time period and the data of the Non-SBFD time period in an integrated manner or separately. Here, the data of the SBFD time period may be data derived based on the measurement result of the reference signal acquired in the SBFD time period. The data of the Non-SBFD time period may be data derived based on the measurement result of the reference signal acquired in the Non-SBFD time period. The data derived based on the measurement result of the reference signal may include, but is not limited to, CQI, PMI, LI, RI, L1-RSRP, phase noise information, interference measurement information, etc.

[0104] Typically, wireless channel conditions can differ between SBFD and non-SBFD time periods. For example, during SBFD, communication nodes receive and transmit simultaneously, which may result in a smaller number of transmit antennas. Consequently, the strength of the transmitted signal may be weaker during SBFD compared to non-SBFD time periods. Furthermore, in SBFD mode, CLI and SI are more severe, and channel variability can also be relatively greater.

[0105] When the moving speed of the terminal (120-1) is fast, when the distance between the terminal (120-1) and the base station is long, or when the terminal (120-1) is in an NLoS (Non Line of Sight) environment, there may be a large difference between the wireless channel environment in the SBFD time period and the wireless channel environment in the Non-SBFD time period.

[0106] If there is a large difference in the wireless channel environment between the SBFD time interval and the Non-SBFD time interval, a model trained by data collected in the SBFD time interval may not be suitable for inferring channel state information, positioning, interference information, beam management, etc. in the Non-SBFD time interval. In addition, a model trained by data collected in the Non-SBFD time interval may not be suitable for inferring channel state information, positioning, interference information, beam management, etc. in the SBFD time interval. Therefore, the base station (110-1) can manage the data collected in the SBFD time interval and the data collected in the Non-SBFD time interval separately. In addition, the base station (110-1) can manage the data collected in the SBFD time interval and the data collected in the Non-SBFD time interval so that they are input to different AI / ML models, respectively.

[0107] The base station (110-1) and / or the terminal (120-1) can learn / manage / use the AI / ML model of the SBFD time period and the AI / ML model of the Non-SBFD time period by distinguishing between them. Here, the AI / ML model of the SBFD time period may refer to an AI / ML model that performs channel state information inference, positioning, beam management, interference information inference, etc. in the SBFD time period. The AI / ML model of the Non-SBFD time period may refer to an AI / ML model that performs channel state information inference, positioning, beam management, interference information inference, etc. in the Non-SBFD time period.

[0108] For example, the base station (110-1) and / or the terminal (120-1) can use both the AI / ML model for the SBFD time interval and the AI / ML model for the Non-SBFD time interval. As another example, the base station (110-1) and / or the terminal (120-1) can use only one of the AI / ML model for the SBFD time interval and the AI / ML model for the Non-SBFD time interval. As another example, the base station (110-1) and / or the terminal (120-1) can use an AI / ML model that does not distinguish between time intervals. Here, the AI / ML model that does not distinguish between time intervals may mean an AI / ML model that performs channel state information inference, positioning, beam management, interference information inference, etc. based on collected data without distinguishing between SBFD time intervals and Non-SBFD time intervals.

[0109] The base station (110-1) and / or the terminal (120-1) can separate or distinguish the data collected in the SBFD time interval from the data collected in the Non-SBFD time interval. For example, the base station (110-1) and / or the terminal (120-1) can utilize the data collected in the SBFD time interval (e.g., reference signal measurement results) for learning or inference of an AI / ML model in the SBFD time interval, and can utilize the data collected in the Non-SBFD time interval (e.g., reference signal measurement results) for learning or inference of an AI / ML model in the Non-SBFD time interval. As another example, the base station (110-1) and / or the terminal (120-1) can utilize the data collected in only one of the SBFD time interval and the Non-SBFD time interval for learning or inference of an AI / ML model in either the SBFD time interval or the Non-SBFD time interval.

[0110] When the moving speed of the terminal (120-1) is slow, when the distance between the terminal (120-1) and the base station is close, or when the terminal (120-1) is in a LoS (Line of Sight) environment, the difference between the wireless channel environment in the SBFD time period and the wireless channel environment in the Non-SBFD time period may be small.

[0111] When the difference in wireless channel environment between SBFD and non-SBFD time periods is small, it may be efficient to use a single AI / ML model without distinguishing between time periods. Accordingly, the base station (110-1) can integrate and manage data from SBFD and non-SBFD time periods.

[0112] The base station (110-1) and / or the terminal (120-1) can learn / manage / use an AI / ML model without distinguishing between SBFD time intervals and Non-SBFD time intervals. For example, the base station (110-1) and / or the terminal (120-1) can use the same AI / ML model in SBFD time intervals and Non-SBFD time intervals. The base station (110-1) and / or the terminal (120-1) can utilize the collected data without distinguishing between SBFD time intervals and Non-SBFD time intervals for learning or inference of an AI / ML model that does not distinguish between time intervals. The base station (110-1) and / or the terminal (120-1) can learn / manage / use an AI / ML model for general time intervals without distinguishing between SBFD time intervals and Non-SBFD time intervals. The general time intervals can include both SBFD time intervals and Non-SBFD time intervals. The above general time interval may include either only one of the SBFD time interval and the Non-SBFD time interval.

[0113] The base station (110-1) and / or the terminal (120-1) can use data collected in the SBFD time interval and data collected in the Non-SBFD time interval for learning or inference of an AI / ML model without distinguishing between them. The base station (110-1) and / or the terminal (120-1) can use data collected during a general time interval for learning or inference of an AI / ML model without distinguishing between the SBFD time interval and the Non-SBFD time interval. The general time interval can include both the SBFD time interval and the Non-SBFD time interval. The general time interval can also include only one of the SBFD time interval and the Non-SBFD time interval.

[0114] At step S120, the base station (110-1) can transmit a reference signal to the terminal (120-1). The reference signal can include a CSI-RS, SSB, PT-RS, PRS, downlink reference signal, etc.

[0115] At step S130, the terminal (120-1) may acquire data for AI / ML based on the integration / separation information for AI / ML data. The data for AI / ML may include at least one of data for inferring channel state information of the AI / ML model, data for positioning, data for beam management, and data for interference management. However, the embodiment is not limited thereto.

[0116] If the integration / separation information for data for AI / ML explicitly / directly or implicitly indicates separation of data for AI / ML, the terminal (120-1) can distinguish between data collected in the SBFD time interval and data collected in the non-SBFD time interval. The terminal (120-1) may acquire data only in either the SBFD time interval or the non-SBFD time interval.

[0117] If the integration / separation information for data for AI / ML explicitly / directly or implicitly indicates the integration of data for AI / ML, the terminal (120-1) can manage the data collected in the SBFD time interval and the data collected in the non-SBFD time interval without distinguishing between them. The terminal (120-1) can acquire data without distinguishing between the SBFD time interval and the non-SBFD time interval.

[0118] At step S140, the terminal (120-1) can report channel status information.

[0119] If the integration / separation information for data for AI / ML explicitly / directly or implicitly indicates separation of data for AI / ML, the terminal (120-1) can distinguish between the channel state information of the SBFD time interval and the channel state information of the non-SBFD time interval. The terminal (120-1) can report each of the channel state information of the SBFD time interval and the channel state information of the non-SBFD time interval. The terminal (120-1) can report only one of the channel state information of the SBFD time interval and the channel state information of the non-SBFD time interval.

[0120] Channel state information in the SBFD time interval can be determined based on data collected in the SBFD time interval. The AI / ML model in the SBFD time interval can infer channel state information in the SBFD time interval by inputting data collected in the SBFD time interval. Channel state information in the non-SBFD time interval can be determined based on data collected in the non-SBFD time interval. The AI / ML model in the non-SBFD time interval can infer channel state information in the non-SBFD time interval by inputting data collected in the SBFD time interval.

[0121] While Figure 7 exemplifies AI / ML inferring channel state information, the embodiment is not limited thereto. For example, the AI / ML model may be utilized for positioning, beam management, interference management, and more.

[0122] If the integration / separation information for data for AI / ML explicitly / directly or implicitly indicates integration of data for AI / ML, the terminal (120-1) may not distinguish between data in the SBFD time interval and data in the Non-SBFD time interval. The terminal (120-1) may report channel state information of a general time interval without distinction between the SBFD time interval and the Non-SBFD time interval. The general time interval may include both the SBFD time interval and the Non-SBFD time interval. The general time interval may include only one of the SBFD time interval and the Non-SBFD time interval.

[0123] Channel state information for a typical time interval can be determined based on data acquired by the terminal (120-1) without distinguishing between SBFD and non-SBFD time intervals. In this case, the AI / ML model can be used without distinguishing between SBFD and non-SBFD time intervals.

[0124] Below, we describe specific examples of data integration / separation information for AI / ML.

[0125] Figure 8 is a flowchart exemplarily showing the process of performing step S110 of Figure 7.

[0126] Referring to FIG. 8, at step S112-1, the base station (110-1) can determine an associated ID of an AI / ML model. The associated ID may also be referred to as an associated identifier.

[0127] Associated IDs can represent the configuration of data collection for training AI / ML models. Associated IDs can be used to ensure the consistency of the AI / ML model's training and inference environments.

[0128] Terminal (120-1) can verify the configuration for data collection from the Associated ID and collect data according to the configuration. Terminal (120-1) can use the data collected according to the Associated ID to train or develop an AI / ML model.

[0129] AI / ML models have characteristics called overfitting and generalization. Overfitting occurs when a model performs well on a specific dataset and deteriorates on other datasets. Conversely, generalization refers to the ability to maintain performance across diverse datasets. There are limitations to the generalization of AI / ML models. In other words, AI / ML model performance can only be guaranteed for data collected within a relatively limited data collection configuration. Therefore, wireless communication networks can define data collection configurations that guarantee AI / ML model performance and assign Associated IDs to each data collection configuration.

[0130] A wireless communication network can define a set (range) of generalizable datasets that guarantee the performance of AI / ML models and determine a data collection configuration to match them. The wireless communication network can define various data collection configurations and assign an Associated ID to each data collection configuration. The wireless communication network can train or develop AI / ML models based on the data collection configuration corresponding to each Associated ID.

[0131] The method for assigning Associated IDs may vary depending on the use case of the AI / ML model. For example, for AI / ML models used for beam management, data collection configurations may be determined based on the same downlink spatial domain transmission filter, a reference signal in quasi-colocation (QCL) state, or the antenna panels used for transmission and reception. Different Associated IDs may be assigned to each data collection configuration. In the example of CSI compression and prediction, data collection configurations may be determined based on SNR, CQI, PMI, RI, which indicate channel conditions, and the antenna panels used for transmission and reception.

[0132] The data collection configuration corresponding to the Associated ID may be determined based on the time interval during which the data is collected. For example, at least one Associated ID may correspond to a data collection configuration in which data is collected in the SBFD time interval. In this case, the base station (110-1) and / or the terminal (120-1) may train, develop, and use an AI / ML model using data collected only in the SBFD time interval from the Associated ID. As another example, at least one Associated ID may correspond to a data collection configuration in which data is collected in the Non-SBFD time interval. In this case, the base station (110-1) and / or the terminal (120-1) may train, develop, and use an AI / ML model using data collected only in the Non-SBFD time interval from the Associated ID. As another example, at least one Associated ID may correspond to a data collection configuration in which data is collected in a general time interval without distinguishing between the SBFD and Non-SBFD time intervals. In this case, the base station (110-1) and / or the terminal (120-1) can train, develop, and use an AI / ML model using the collected data without distinguishing between SBFD time intervals and Non-SBFD time intervals from the Associated ID.

[0133] Each AI / ML model can be assigned a Model ID. An Associated ID can correspond to a single Model ID. Alternatively, an Associated ID can correspond to multiple Model IDs. Associated IDs and Model IDs can have a many-to-many relationship.

[0134] For example, if the downlink spatial domain transmission filter is configured with multiple (N) numbers, the Associated ID for beam management can also be configured with multiple (N) numbers, and the AI / ML model used for the corresponding AI / ML beam management can be configured with N numbers corresponding thereto. Similarly, if the base station (110-1) configures multiple (N) sets of reference signals / resources that are not in the QCL state, the AI / ML model and associated ID used for the corresponding AI / ML beam management can be configured with N numbers corresponding thereto. On the other hand, if the AI / ML model has excellent generalization characteristics and its performance is guaranteed in all datasets, it can be configured and operated with a small number (less than N) of AI / ML models.

[0135] In step S114-1, the base station (110-1) can configure information indicating the Associated ID.

[0136] The base station (110-1) can configure information indicating the Associated ID in the RRC message. The base station (110-1) can include information indicating the Associated ID in various IEs. For example, the base station (110-1) can include information indicating the Associated ID in at least one of an IE for configuring CSI-RS resources, an IE for configuring SRS resources, and an IE for configuring CSI reporting resources. The base station (110-1) can include information indicating the Associated ID in an IE for configuring resources of reference signals other than CSI-RS or in any other IE.

[0137] FIG. 9 is a diagram illustrating examples of IEs including an Associated ID. The Associated ID may be included in at least one of the IEs (NZP-CSI-RS-ResourceSet, CSI-SSB-ResourceSet, NZP-CSI-RS-Resource) for indicating a resource configuration of a CSI-RS. The Associated ID may also be included in an IE (CSI-ReportConfig) for indicating a CSI report resource configuration. The Associated ID may also be included in an IE (SRS-ResourceSet or SRS-Resource) for indicating an SRS resource configuration.

[0138] The base station (110-1) may include one Associated ID in an RRC message. As another example, the base station (110-1) may include multiple Associated IDs in an RRC message. The base station (110-1) may also include multiple Associated IDs in a single IE.

[0139] The base station (110-1) can use the Associated ID to instruct the terminal (120-1) to integrate / separate data for AI / ML. For example, the base station (110-1) can instruct the terminal (120-1) to separate data for AI / ML by transmitting the Associated ID corresponding to the configuration for collecting data in the SBFD time interval.

[0140] Referring again to FIG. 8, the base station (110-1) may transmit Associated ID indication information. The base station (110-1) may transmit an RRC message including Associated ID indication information.

[0141] The terminal (120-1) can collect data only in the SBFD time period or can separate the data collected in the SBFD time period and use it for the development, training, and inference of an AI / ML model. For example, the base station (110-1) can instruct the separation of data for AI / ML by transmitting an Associated ID corresponding to a configuration for collecting data in a non-SBFD time period to the terminal (120-1). The terminal (120-1) can collect data only in the SBFD time period or can separate the data collected in the SBFD time period and use it for the development, training, and inference of an AI / ML model.

[0142] Fig. 10 is a flowchart exemplarily showing the process of performing step S130 of Fig. 7.

[0143] In step S132-1, the terminal (120-1) can receive Associated ID indication information. The terminal (120-1) can receive an RRC message including Associated ID indication information. The terminal (120-1) can confirm at least one of the CSI-RS resource configuration, SSB resource configuration, SRS resource configuration, and CSI reporting resource configuration from the RRC message. The terminal (120-1) can confirm at least one Associated ID from the RRC message.

[0144] In step S134-1, the terminal (120-1) may determine a data collection method from at least one Associated ID. For example, the terminal (120-1) may collect data only in the SBFD time interval. As another example, the terminal (120-1) may collect data only in the Non-SBFD time interval. As another example, the terminal (120-1) may collect data in both the SBFD time interval and the Non-SBFD time interval, but may manage the data collected in the SBFD time interval and the data collected in the Non-SBFD time interval separately. As another example, the terminal (120-1) may collect data without distinguishing between the SBFD time interval and the Non-SBFD time interval. In step S140 of FIG. 7, the terminal (120-1) may include the Associated ID in the channel state information report. The base station (110-1) can determine from the Associated ID included in the channel state information report by the terminal (120-1) whether the channel state information is about the channel state of the SBFD time interval, the channel state of the Non-SBFD time interval, or the channel state of the general time interval that does not distinguish between the SBFD time interval and the Non-SBFD time interval.

[0145] Figure 11 is a diagram illustrating an example of a data collection section.

[0146] Assuming that the TDD pattern is DXXXU, the TDD pattern may include a Non-SBFD time interval (72) and an SBFD time interval (74).

[0147] Figures 12 to 14 are conceptual diagrams showing examples of how a terminal (120-1) collects data based on an Associated ID.

[0148] Referring to FIG. 12, a base station (110-1) may transmit a reference signal including information on an Associated ID (A) corresponding to a data collection configuration for collecting data in a SBFD time interval to a terminal (120-1). The base station (110-1) may transmit the reference signal to the terminal (120-1) in the SBFD time interval. For example, the base station (110-1) may transmit at least one of a CSI-RS, an SSB, a PT-RS, and a PRS in the SBFD time interval (e.g., the SBFD time interval (74) of FIG. 11). The terminal (120-1) may receive the reference signal through a DL subband in the SBFD time interval. The terminal (120-1) may determine a data collection method based on the Associated ID (A). The terminal (120-1) may collect data in the SBFD time interval based on the Associated ID (A). The terminal (120-1) can collect data based on a reference signal received in the SBFD time interval. The terminal (120-1) can distinguish the data collected in the SBFD time interval from the data collected in the Non-SBFD time interval. The terminal (120-1) may not collect data in the Non-SBFD time interval and may collect data only in the SBFD time interval. The terminal (120-1) can utilize the data collected in the SBFD time interval for training, development, and inference of the AI / ML model corresponding to the Associated ID (A). The terminal (120-1) can use the data collected in the SBFD time interval as input data of the AI / ML model corresponding to the Associated ID (A). The AI / ML model corresponding to the Associated ID (A) can be used to infer a result value suitable for the use case of the SBFD time interval.

[0149] Referring to FIG. 13, a base station (110-1) may transmit a reference signal including an Associated ID (B) corresponding to a data collection configuration for collecting data in a Non-SBFD time interval to a terminal (120-1). The base station (110-1) may transmit the reference signal to the terminal (120-1) in a Non-SBFD time interval (e.g., the Non-SBFD time interval (72) of FIG. 11). For example, the base station (110-1) may transmit at least one of a CSI-RS, an SSB, a PT-RS, and a PRS in the Non-SBFD time interval. The terminal (120-1) may receive the reference signal in the Non-SBFD time interval. The terminal (120-1) may determine a data collection method based on the Associated ID (B). The terminal (120-1) may collect data in the Non-SBFD time interval based on the Associated ID (B). The terminal (120-1) can collect data based on a reference signal received in the Non-SBFD time interval. The terminal (120-1) can distinguish the data collected in the Non-SBFD time interval from the data collected in the SBFD time interval. The terminal (120-1) may not collect data in the SBFD time interval and may collect data only in the Non-SBFD time interval. The terminal (120-1) can utilize the data collected in the Non-SBFD time interval for training, development, and inference of the AI / ML model corresponding to the Associated ID (B). The terminal (120-1) can use the data collected in the Non-SBFD time interval as input data of the AI / ML model corresponding to the Associated ID (B). The AI / ML model corresponding to the Associated ID (B) can be used to infer a result value suitable for the use case of the Non-SBFD time interval.

[0150] Referring to FIG. 14, a base station (110-1) can transmit a reference signal including information on an Associated ID (C) corresponding to a configuration for collecting data without distinguishing between a Non-SBFD time interval and an SBFD time interval to a terminal (120-1). The terminal (120-1) can receive the reference signal without distinguishing between a Non-SBFD time interval and an SBFD time interval. The terminal (120-1) can determine a data collection method based on the Associated ID (C). The terminal (120-1) can collect data without distinguishing between a Non-SBFD time interval and an SBFD time interval based on the Associated ID (C). The terminal (120-1) can utilize the collected data for training, development, and inference of an AI / ML model corresponding to the Associated ID (C). The terminal (120-1) can use the collected data as input data of an AI / ML model corresponding to the Associated ID (C). The AI / ML model corresponding to the Associated ID(C) can be used to infer results suitable for use cases of general time intervals without distinction between Non-SBFD time intervals and SBFD time intervals.

[0151] Although not shown in FIGS. 12 to 14, the terminal (120-1) may receive two or more Associated IDs.

[0152] For example, terminal (120-1) can receive Associated ID (A) of Fig. 12 and Associated ID (B) of Fig. 13. In this case, terminal (120-1) can collect data in both SBFD time intervals and Non-SBFD time intervals. Terminal (120-1) can separate / distinguish data collected in SBFD time intervals and data collected in Non-SBFD time intervals. Terminal (120-1) can use data collected in SBFD time intervals as input data of AI / ML model corresponding to Associated ID (A), and can use data collected in Non-SBFD time intervals as input data of AI / ML model corresponding to Associated ID (Y).

[0153] Terminal (120-1) can receive Associated ID (A) of Fig. 12 and Associated ID (C) of Fig. 14. In this case, terminal (120-1) can use the data collected in the SBFD time interval data and the non-SBFD time interval data as input data of the AI / ML model corresponding to Associated ID (C). Terminal (120-1) can separate the data collected in the SBFD time interval data and use it as input data of the AI / ML model corresponding to Associated ID (A).

[0154] Fig. 15 is a flowchart exemplarily showing the process of performing step S110 of Fig. 7.

[0155] Referring to FIG. 15, in step S112-2, the base station (110-1) may determine whether to integrate / separate data for AI / ML. The base station (110-1) may determine whether to integrate or separate data for AI / ML in the SBFD time period and data for AI / ML in the non-SBFD time period.

[0156] In step S114-2, the base station (110-1) may configure data integration / separation information for AI / ML. For example, the base station (110-1) may configure the data integration / separation information for AI / ML to indicate integration or separation of data for AI / ML.

[0157] In step S116-2, the base station (110-1) can transmit integration / separation information of data for AI / ML.

[0158] Fig. 16 is a flowchart exemplarily showing the process of performing step S130 of Fig. 7.

[0159] Referring to FIG. 16, in step S132-2, the terminal (110-1) can receive integration / separation information for data for AI / ML. The terminal (110-1) can determine integration or separation of data for AI / ML based on the instruction.

[0160] If the integration / separation information for data for AI / ML in step S134-2 indicates separation of data for AI / ML, the terminal (110-1) can separate / distinguish data collected in the SBFD time interval and data collected in the Non-SBFD time interval. The terminal (110-1) may collect data only in either the SBFD time interval or the Non-SBFD time interval. In step S140 of FIG. 7, the terminal (110-1) can use the data collected in the SBFD time interval as input data for the AI / ML model of the SBFD time interval, and can use the data collected in the Non-SBFD time interval as input data for the AI / ML model of the Non-SBFD time interval. As another example, the terminal (110-1) can use the data collected only in the SBFD time interval as input data for the AI / ML model of the SBFD time interval. Terminal (110-1) can use data collected only in the Non-SBFD time period as input data for an AI / ML model in the Non-SBFD time period.

[0161] If the integration / separation information for data for AI / ML in step S134-2 indicates integration of data for AI / ML, the terminal (110-1) can integrate and manage the data collected in the SBFD time interval and the data collected in the Non-SBFD time interval. In step S140 of FIG. 7, the terminal (110-1) can integrate the data collected in the SBFD time interval and the data collected in the Non-SBFD time interval and utilize them as input data for the AI / ML model of the general time interval. In step S140 of FIG. 7, the terminal (120-1) can include the integration / separation information for data for AI / ML in the channel state information report. The base station (110-1) can determine, from the integration / separation information for data for AI / ML included in the channel state information report by the terminal (120-1), whether the corresponding channel state information is about the channel state of the SBFD time interval or about the channel state of the general time interval that does not distinguish between the SBFD time interval and the Non-SBFD time interval.

[0162] Figure 17 is a diagram exemplifying data integration / separation information for AI / ML.

[0163] Referring to FIG. 17, the base station (110-1) may broadcast a message including information on integration / separation of data for AI / ML in step S116-2 of FIG. 15. The broadcasted message may be, for example, an SIB. However, the embodiment is not limited thereto. The broadcasted message may include explicit information (e.g., SBFD-CSI-Acquisition-r19 of FIG. 17) on whether data for AI / ML is to be integrated or separated within the service area of ​​the base station (110-1). The broadcasted message may additionally include SBFD-related information. The above SBFD-related information may include at least one of the number of uplink subbands, the location of uplink frequency resources (start RB and end RB or start RB and number of used RBs), frequency hopping-related information (inter-slot hopping or intra-slot hopping, frequency hopping offset list), the number of downlink subbands, and the location of downlink frequency resources (start RB and end RB or start RB and number of used RBs).

[0164] Figure 18 is a diagram exemplifying data integration / separation information for AI / ML.

[0165] In step S114-2 of FIG. 15, the base station (110-1) can include data integration / separation information for AI / ML in the DCI. The base station (110-1) can use a new RNTI (referred to as AI-RNTI for convenience herein). The base station can assign an AI-RNTI to each terminal or group of terminals. The base station (110-1) can scramble the DCI using each AI-RNTI. The terminal (120-1) can extract the control command of the base station (110-1) transmitted to it from the DCI using the AI-RNTI assigned to it. The control command of the base station (110-1) can include a control command related to AI / ML.

[0166] The base station (110-1) can transmit a control command for AI model management to the terminal (120-1) through DCI. As shown in FIG. 18, the DCI can include a field (181) including DCI format information, a field (182) including identification information of the AI / ML model, and a field (183) including a control command transmitted to the terminal (120-2) in relation to the AI / ML model. FIG. 18 exemplarily shows a case where the field (182) includes a model ID of the AI / ML model. However, the field (182) may also include the Associated ID described above. As another example, the field (182) may include identification information for a functional group to which the AI / ML model belongs. The field (183) may include control information and / or management information such as activation / deactivation / retraining / discarding / model transfer / model transmission of the AI / ML model identified by the field (182).

[0167] Figure 19 is a diagram exemplifying data integration / separation information for AI / ML.

[0168] In step S114-2 of FIG. 15, the base station (110-1) may include integration / separation information for data for AI / ML in the MAC CE (Medium Access Control Control Element). The base station (110-1) may transmit a command for activation / deactivation of the AI / ML model to the terminal (120-1) through the MAC CE. The MAC CE may include a field (191) including identification information of the AI / ML model and a field (192) including an activation / deactivation command of the AI / ML model. In FIG. 19, the case where the field (191) includes the model ID of the AI / ML model is exemplarily shown. However, the field (191) may also include the Associated ID described above. As another example, the field (191) may include identification information for a functional group to which the AI / ML model belongs. The MAC CE may indicate activation of one or more AI / ML models. For example, if an AI / ML model for channel estimation of the SBFD time interval is activated by the MAC CE, the terminal (120-1) can separate / distinguish the data collected in the SBFD time interval. As another example, if the AI / ML model activated by the MAC CE is only an AI / ML model for channel estimation of a general time interval that does not distinguish between the SBFD time interval and the Non-SBFD time interval, the terminal (120-1) can integrate and manage the collected data without distinguishing between the SBFD time interval and the Non-SBFD time interval. The MAC CE can instruct activation / deactivation of one AI / ML model. As another example, the MAC CE can instruct activation / deactivation of multiple AI / ML models.

[0169] The embodiment is not limited to what is described above. The base station (110-1) may also include integration / separation information of data for AI / ML in the RRC message.

[0170] Figure 20 is a flowchart exemplarily showing the process of performing step S110 of Figure 7.

[0171] Step S112-3 of FIG. 20 is similar to step S112-2 of FIG. 15, and thus, description thereof is omitted. Referring to FIG. 20, in step S114-3, the base station (110-1) may determine a reference signal resource configuration such that integration / separation information for data for AI / ML is implicitly indicated. As another example, the base station (110-1) may determine a channel state information reporting resource configuration such that integration / separation information for data for AI / ML is implicitly indicated.

[0172] For example, if the base station (110-1) decides to separate data for AI / ML in the SBFD time interval, the reference signal resources (CSI-RS resources, SSB resources, SRS resources, PRS resources, PT-RS resources, etc.) or channel state information reporting resources (CSI reporting resources, etc.) may be arranged only in the DL subbands in the SBFD time interval. If the base station (110-1) decides to separate data for AI / ML in the Non-SBFD time interval, the reference signal resources or channel state information reporting resources may be arranged only in the Non-SBFD time interval. If the base station (110-1) decides not to separate data for AI / ML in the SBFD time interval and data for AI / ML in the Non-SBFD time interval, the reference signal resources or channel state information reporting resources may be arranged only in the general time interval that does not distinguish between the Non-SBFD time interval and the SBFD time interval.

[0173] Figure 21 is a flowchart exemplarily showing the process of performing step S130 of Figure 7.

[0174] Referring to FIG. 21, in step S132-3, the terminal (120-1) can check the reference signal resource configuration (CSI-RS resource configuration, SSB resource configuration, SRS resource configuration, etc.) or the channel state information reporting resource configuration (CSI reporting resource configuration, etc.). The terminal (120-1) can check the data separation / integration information for AI / ML implicitly indicated from the reference signal resource configuration or the channel state information reporting resource configuration.

[0175] In step S134-3, the terminal (120-1) can collect data based on data separation / integration information for AI / ML implicitly indicated from the reference signal resource configuration or channel state information reporting resource configuration.

[0176] For example, if the reference signal resource or the channel state information reporting resource is deployed only in the DL subband of the SBFD time interval, the terminal (120-1) can separate / distinguish the data collected in the SBFD time interval and utilize it as input for the AI / ML model. If the reference signal resource or the channel state information reporting resource is deployed only in the Non-SBFD time interval, the terminal (120-1) can separate / distinguish the data collected in the Non-SBFD time interval and utilize it as input for the AI / ML model. If the reference signal resource or the channel state information reporting resource is deployed in a general time interval that does not distinguish between the SBFD time interval and the Non-SBFD time interval, the terminal (120-1) can integrate the data without distinguishing between the SBFD time interval and the Non-SBFD time interval and utilize it as input for the AI / ML model.

[0177]

[0178] Figure 22 is a diagram exemplifying a resource allocation configuration for CSI measurement in a mobile communication system.

[0179] As shown in the drawing, in order to measure channel state information, the base station (110-1) can allocate some of the PDCCH or PDSCH resources to CSI-RS and transmit a reference signal for channel state information measurement to the terminal. This CSI-RS resource allocation can be configured in a periodic, aperiodic, or semi-persistent manner depending on the configuration of the base station. In addition, since the base station (110-1) can allocate CSI-RS resources only to the DL band, in the non-SBFD time interval, CSI-RS resources can be allocated to the range of the entire channel bandwidth, as shown in the left figure of FIG. 22. On the other hand, in the SBFD time interval, CSI-RS resources can be allocated only to PRBs existing within the DL subband. Therefore, the CSI-RS resources configured at the boundary between the DL subband and the UL subband can only utilize the portion of the PRB existing within the DL subband.

[0180]

[0181] Figure 23 is a diagram illustrating an example of a CSI resource configuration for measuring SBFD time intervals in a communication system.

[0182] In FIGS. 23, 24, and 25 below, “D” represents a DL slot, “U” represents a UL slot, and “X” represents a SBFD slot.

[0183] Assuming that the TDD pattern of the base station is DXXXU as in the example in the drawing, if the CSI resources configured by the base station to the terminal are configured only for the Non-SBFD time period (DL period in the example in the drawing), the terminal can implicitly consider the CSI-RS as an instruction to collect data related to the channel state information of the Non-SBFD time period from the base station. The data collected through this reporting configuration can be utilized for AI / ML models of the Non-SBFD time period.

[0184]

[0185] Figure 24 is a diagram exemplarily showing a CSI resource configuration for measuring a Non-SBFD time interval in a communication system.

[0186] Assuming that the TDD pattern of the base station is DXXXU as in the example in the drawing, if the CSI resources configured by the base station to the terminal are configured for both non-SBFD time intervals (DL intervals in the example in the drawing) and SBFD time intervals, the terminal can consider the CSI-RS as an implicit instruction to collect data from the base station for the SBFD time interval. The data collected through this reporting configuration can be utilized for AI / ML models for the SBFD time interval.

[0187] Figure 25 is a diagram illustrating an example of a method for distinguishing between SBFD time intervals and non-SBFD time intervals through CSI resource configuration in a communication system, in which the distinction is made in the frequency domain.

[0188] As in the example in the drawing, if the CSI resource configured by the base station to the terminal is not configured in the uplink subband, the terminal can consider the corresponding CSI-RS as an implicit instruction to collect data in the SBFD time interval from the base station. The data collected through this reporting configuration can be utilized for AI / ML models in the SBFD time interval. Conversely, if the CSI resource configured by the base station to the terminal is configured in the uplink subband, the terminal can consider the corresponding CSI-RS as an implicit instruction to collect data in the non-SBFD time interval from the base station. The data collected through this reporting configuration can be utilized for AI / ML models in the non-SBFD time interval. The terminal can utilize the SBFD-related information broadcast by the base station to identify the location of the corresponding uplink subband.

[0189] Figure 26 is a diagram exemplarily showing pattern periods of downlink and uplink in a mobile communication system.

[0190] In the case of time division multiplexing in mobile communication systems, the downlink and uplink are used in the same frequency band by dividing time. In the case of 5G mobile communication systems, this downlink and uplink allocation pattern is propagated to the service area of ​​the base station through the Dl-Ul-TransmissionPeriodicty information block, and the terminal can recognize the downlink reception and uplink transmission timing based on the information. In addition, the base station can set multiple Dl-Ul-TransmissionPeriodicty values ​​as needed. In the case of the SBFD period, it was agreed upon as the sum of the Dl-Ul-TransmissionPeriodicty periods in the 3GPP standardization meeting. If the base station decides to process CSI information separately into non-SBFD time intervals and SBFD time intervals, it needs to match the TDD pattern when performing periodic CSI reporting. For example, if Dl-Ul-TransmissionPeriodicty (pattern 1) is time and Dl-Ul-TransmissionPeriodicty (pattern 2) is time, non-SBFD time interval measurement data must be transmitted in non-SBFD slots, and therefore must be transmitted at a period of Dl-Ul-TransmissionPeriodicty (pattern 1) + Dl-Ul-TransmissionPeriodicty (pattern 2). Similarly, SBFD time interval measurement data must be transmitted in SBFD slots, and therefore must be transmitted at a period of . The above can be equally applied to CSI reporting as well as CSI resource allocation. Based on this, the base station can adjust the CSI reporting period of the UE to separate the data measured in the SBFD time interval and the non-SBFD time interval. In the example of 5G, the period can be the CSI-ReportPeriodicityAndOffset message in the reportConfigType of the CSI-ReportConfig message.Similarly, the base station can adjust the CSI resource period of the terminal to measure data separately in SBFD and non-SBFD time intervals. In the case of 5G, this period can be the CSI-ResourcePeriodicityAndOffset message.

[0191] Fig. 27 is a flowchart illustrating a case where the management function of an AI / ML model is proactively performed on the terminal (120-1) side. The description of steps S110, S120, S130, and S140 of Fig. 27 is similar to the description of steps S110, S120, S130, and S140 of Fig. 7, and therefore is omitted.

[0192] Referring to FIG. 27, the terminal (120-1) can utilize the data collected in step S130 as input data for an AI / ML model. The terminal (120-1) can train or develop an AI / ML model using the data collected in step S130. The terminal (120-1) can input the data collected in step S130 into the AI / ML model and output a result value according to the use case through an inference function. The terminal (120-1) can monitor the performance of the AI / ML model by monitoring the inference result of the AI / ML model. The terminal (120-1) can decide to reuse, discard, replace, etc. the AI / ML model based on the performance monitoring result. The terminal (120-1) can decide to fallback to a legacy model that does not use the AI / ML model based on the performance monitoring result.

[0193] Fig. 28 is a flowchart illustrating a case where the management function of the AI / ML model is proactively performed on the base station (110-1) side. The description of steps S110, S120, S130, and S140 of Fig. 28 is similar to the description of steps S110, S120, S130, and S140 of Fig. 7, and therefore is omitted.

[0194] Referring to FIG. 28, at step S160, the base station (110-1) can monitor the performance of the AI / ML model by monitoring the inference results of the AI / ML model. Based on the performance monitoring results, the base station (110-1) can decide whether to reuse, discard, or replace the AI / ML model. Based on the performance monitoring results, the base station (110-1) can decide to fallback to a legacy model that does not use the AI / ML model.

[0195] Unlike what is shown in FIGS. 27 and 28, each of the terminal (120-1) and the base station (110-1) may perform the management function of the AI / ML model.

[0196] Figures 7 through 28 primarily describe cases where data for AI / ML models is collected via downlink reference signals. Even when data for AI / ML models is collected via uplink reference signals, integration / separation information for AI / ML data can be utilized.

[0197] Figure 29 is a flowchart illustrating a method for utilizing an AI / ML model for an uplink channel. In describing the embodiment of Figure 29, any overlap with Figures 1 through 28 will be omitted.

[0198] Referring to FIG. 29, in step S210, the base station (110-1) may transmit data integration / separation information for AI / ML. Similar to step S110, the base station (110-1) may transmit an Associated ID. As another example, the base station (110-1) may transmit data integration / separation information for AI / ML through DCI or MAC CE. As another example, the base station (110-1) may implicitly indicate data integration / separation information for AI / ML through SRS resource configuration.

[0199] The base station (110-1) can include information indicating the Associated ID in an RRC message. FIG. 30 is a diagram illustrating an IE including information indicating the Associated ID. Referring to FIG. 30, the base station (110-1) can include information indicating the Associated ID in an IE (SRS-ResourceSet) related to the resource configuration of the SRS, which is an uplink reference signal. The base station (110-1) can also include information indicating the Associated ID in another IE related to the SRS.

[0200] Referring again to FIG. 29, at step S230, the base station (110-1) can collect data based on the integration / separation information of data for AI / ML. The execution method of step S230 may be similar to the execution method of step S130 of FIG. 7. At step S240, the base station (110-1) can utilize the collected data as input data for the AI / ML model.

[0201] Figure 31 is a flowchart illustrating an example of a management method for AI / ML.

[0202] Referring to FIG. 31, the base station (110-1) can proactively perform AI / ML management. In step S310, the base station (110-1) can transmit a first AI / ML model usage instruction to the terminal (120-1). The base station (110-1) can transmit the first AI / ML model usage instruction using at least one of the Associated ID, Model ID, and other integration / separation information for AI / ML data described above.

[0203] The first AI / ML model can be any one of an AI / ML model for the SBFD time interval, an AI / ML model for the non-SBFD time interval, or an AI / ML model for the general time interval. For convenience, the following description assumes that the first AI / ML model is an AI / ML model for the general time interval that does not distinguish between the SBFD time interval and the non-SBFD time interval.

[0204] At step S312, the terminal (120-1) can transmit performance feedback of the first AI / ML model or the inference result of the first AI / ML model to the base station (110-1). The terminal (120-1) can generate performance feedback of the first AI / ML model or the inference result of the first AI / ML model using the collected data without distinguishing between SBFD time intervals and non-SBFD time intervals.

[0205] The terminal (120-1) can generate performance feedback of the first AI / ML model by comparing the inference result of the first AI / ML model with the measurement result measured by the terminal (120-1). For example, the terminal (120-1) can measure at least one beam using a predefined set of monitoring resources. Based on the reception performance of the beams, the terminal (120-1) can select K beams (K is an arbitrary natural number) with the best reception performance.

[0206] Terminal (120-1) can infer the reception performance of a beam using the first AI / ML model. For example, the first AI / ML model can select M beams for terminal (120-1) based on the inference results for at least one beam.

[0207] The terminal (120-1) can generate performance feedback of the first AI / ML model by comparing the K beams derived based on the beam measurement results with the M beams inferred by the first AI / ML model. For example, if there is an intersection between the K beam sets and the M beam sets, the terminal (120-1) can determine that the performance of the first AI / ML model is good. If there is no intersection between the K beam sets and the M beam sets, the terminal (120-1) can determine that the performance of the first AI / ML model is bad. The terminal (120-1) can generate performance feedback of the first AI / ML model based on at least one of the presence or absence of an intersection between the K beam sets and the M beam sets and the size of the intersection.

[0208] Although the above description exemplifies the generation of AI / ML performance feedback by the terminal (120-1) on a model-by-model basis, the embodiment is not limited thereto. For example, the terminal (120-1) may also generate performance feedback for related AI / ML model(s) on a functionality-by-functional basis. The terminal (120-1) may also generate performance feedback for AI / ML model(s) corresponding to the relevant ID.

[0209] Terminal (120-1) can optionally generate performance feedback for the first AI / ML model. For example, terminal (120-1) may omit generating performance feedback for the first AI / ML model, taking into account the reliability of the performance feedback for the first AI / ML model.

[0210] For example, if the interval between the time at which the terminal (120-1) measures a given parameter and the time at which the first AI / ML model infers the given parameter is greater than or equal to a threshold value, the terminal (120-1) may determine that the reliability of the performance feedback based on the measurement and inference results is low. In this case, the terminal (120-1) may not generate performance feedback.

[0211] As another example, the first base station (110-1) may not utilize low-reliability performance feedback. For example, since the first base station (110-1) performs resource scheduling for the terminal (120-1), it can know the measurement resources utilized by the terminal (120-1). If the time interval between the time when the first AI / ML model of the terminal (120-1) infers a predetermined parameter and the time when the terminal (120-1) measures the predetermined parameter is greater than or equal to a threshold value, the first base station (110-1) may not evaluate the first AI / ML model using the performance feedback obtained from the terminal (120-1). In this case, the first base station (110-1) may ignore the performance feedback without utilizing it.

[0212] At step S314, the base station (110-1) can monitor the performance of the first AI / ML model.

[0213] If the monitoring result in step S316 is higher than the reference value, the base station (110-1) may decide to reuse the first AI / ML model.

[0214] At step S318, the base station (110-1) may transmit a reuse instruction for the first AI / ML model to the terminal (120-1). The terminal (120-1) may extend the usage period of the first AI / ML model. The terminal (120-1) may input the collected data into the first AI / ML model without distinguishing between SBFD and non-SBFD time periods as before.

[0215] In the following, the second AI / ML model is assumed to be an AI / ML model for the SBFD time interval. This is an example for convenience of explanation, and the second AI / ML model may also be an AI / ML model for the non-SBFD time interval.

[0216] At step S320, the base station (110-1) may store a second AI / ML model. The second AI / ML model may be a model trained based on data previously collected during the SBFD time interval. The terminal (120-1) may train / develop the second AI / ML model and transmit it to the base station (110-1). Alternatively, the base station (110-1) or a network node connected to the base station (110-1) may train / develop the second AI / ML model for the terminal (120-1). The base station (110-1) may store the second AI / ML model.

[0217] In step S322, the base station (110-1) may decide to use the second AI / ML model. For example, if there is a lot of interference in the channel or the moving speed of the terminal (120-1) increases, the channel environment of the SBFD time period may be different from the channel environment of the non-SBFD time period. In this case, the variation of the channel state information measured by the terminal (120-1) by receiving the reference signal may increase, and the base station (110-1) may decide to use the second AI / ML model of the SBFD time period. The base station (110-1) may determine to use the second AI / ML model during the effective time (t) of the second AI / ML model. val ) can be used to check whether the validity period (t) of the second AI / ML model has elapsed. val ) has not elapsed, the base station (110-1) can transmit an instruction to use the second AI / ML model to the terminal (120-1).

[0218] At step S324, the terminal (120-1) can transmit performance feedback of the second AI / ML model or the inference result of the first AI / ML model to the base station (110-1). The terminal (120-1) can generate performance feedback of the second AI / ML model or the inference result of the first AI / ML model using data collected during the SBFD time interval.

[0219] Terminal (120-1) can generate performance feedback of the second AI / ML model by comparing the inference result of the second AI / ML model with the measurement result measured by terminal (120-1). For example, terminal (120-1) can measure at least one beam using a predefined set of monitoring resources. Terminal (120-1) can select K beams (K is an arbitrary natural number) with the best reception performance based on the reception performance of the beams.

[0220] Terminal (120-1) can infer the reception performance of a beam using the second AI / ML model. For example, the second AI / ML model can select M beams for terminal (120-1) based on the inference results for at least one beam.

[0221] The terminal (120-1) can generate performance feedback of the first AI / ML model by comparing the K beams derived based on the beam measurement results with the M beams inferred by the second AI / ML model. For example, if there is an intersection between the K beam sets and the M beam sets, the terminal (120-1) can determine that the performance of the second AI / ML model is good. If there is no intersection between the K beam sets and the M beam sets, the terminal (120-1) can determine that the performance of the second AI / ML model is bad. The terminal (120-1) can generate performance feedback of the second AI / ML model based on at least one of the presence or absence of an intersection between the K beam sets and the M beam sets and the size of the intersection.

[0222] Although the above description exemplifies the generation of AI / ML performance feedback by the terminal (120-1) on a model-by-model basis, the embodiment is not limited thereto. For example, the terminal (120-1) may also generate performance feedback for related AI / ML model(s) on a functionality-by-functional basis. The terminal (120-1) may also generate performance feedback for AI / ML model(s) corresponding to the relevant ID.

[0223] Terminal (120-1) can optionally generate performance feedback for the second AI / ML model. For example, terminal (120-1) may omit generating performance feedback for the second AI / ML model, taking into account the reliability of the performance feedback for the second AI / ML model.

[0224] For example, if the interval between the time when the terminal (120-1) measures a given parameter and the time when the second AI / ML model infers the given parameter is greater than a threshold value, the terminal (120-1) may determine that the reliability of the performance feedback based on the measurement and inference results is low. In this case, the terminal (120-1) may not generate performance feedback.

[0225] As another example, the first base station (110-1) may not utilize unreliable performance feedback. For example, since the first base station (110-1) performs resource scheduling for the terminal (120-1), it can know the measurement resources utilized by the terminal (120-1). If the time interval between the time when the second AI / ML model of the terminal (120-1) infers a predetermined parameter and the time when the terminal (120-1) measures the predetermined parameter is greater than or equal to a threshold value, the first base station (110-1) may not evaluate the second AI / ML model using the performance feedback obtained from the terminal (120-1). In this case, the first base station (110-1) may ignore the performance feedback without utilizing it.

[0226] At step S326, the base station (110-1) can monitor the performance of the second AI / ML model.

[0227] If the monitoring result in step S328 is higher than the reference value, the base station (110-1) may decide to reuse the second AI / ML model.

[0228] At step S330, the base station (110-1) can transmit a reuse instruction of the second AI / ML model to the terminal (120-1). The terminal (120-1) can extend the usage period of the second AI / ML model.

[0229] Figure 32 is a flowchart illustrating an example of a management method for AI / ML.

[0230] Referring to Fig. 32, the terminal (120-1) can proactively perform AI / ML management. In describing the embodiment of Fig. 32, any overlapping portions with Fig. 31 are omitted.

[0231] Referring to FIG. 31, at step S410, the base station (110-1) can transmit an instruction to use the first AI / ML model to the terminal (120-1).

[0232] At step S412, the terminal (120-1) can monitor the performance of the first AI / ML model.

[0233] If the monitoring result in step S414 is higher than the reference value, the terminal (120-1) may decide to reuse the first AI / ML model.

[0234] At step S416, the terminal (120-1) can transmit a request for reuse of the first AI / ML model to the base station (110-1).

[0235] At step S418, the base station (110-1) may transmit a reuse instruction for the first AI / ML model to the terminal (120-1). The terminal (120-1) may extend the usage period of the first AI / ML model. The terminal (120-1) may input the collected data into the first AI / ML model without distinguishing between SBFD and non-SBFD time periods as before.

[0236] At step S420, the terminal (120-1) may store a second AI / ML model. The second AI / ML model may be a model trained based on data collected in advance during the SBFD time interval.

[0237] At step S422, the base station (110-1) may decide to use the second AI / ML model. The base station (110-1) may determine the validity time (t) of the second AI / ML model. val ) can be used to check whether the validity period (t) of the second AI / ML model has elapsed. val ) has not elapsed, the base station (110-1) can transmit an instruction to use the second AI / ML model to the terminal (120-1).

[0238] At step S424, the terminal (120-1) can monitor the performance of the second AI / ML model.

[0239] If the monitoring result is higher than the reference value in step S426, the terminal (120-1) may decide to reuse the second AI / ML model.

[0240] At step S428, the terminal (120-1) can transmit a request for reuse of the second AI / ML model to the base station (110-1).

[0241] At step S430, the base station (110-1) can transmit a reuse instruction of the second AI / ML model to the terminal (120-1). The terminal (120-1) can extend the usage period of the second AI / ML model.

[0242] Hereinafter, with reference to FIGS. 1 to 31, a method for collecting data for AI / ML and a method for using AI / ML in other mobile communication systems according to exemplary embodiments have been described. According to at least one embodiment, based on integration / separation information of data for AI / ML, a terminal (120-1) and / or a base station (110-1) may determine integration / separation of data of an SBFD time interval and data of a Non-SBFD time interval. When data of an SBFD time interval is separated, the separated data may be input into an AI / ML model of the SBFD time interval to improve the performance of the AI / ML model of the SBFD time interval. Similarly, when data of a Non-SBFD time interval is separated, the separated data may be input into an AI / ML model of the Non-SBFD time interval to improve the performance of the AI / ML model of the Non-SBFD time interval. If the difference between the channel environment of the SBFD time interval and the channel environment of the Non-SBFD time interval is not large, the terminal (120-1) and / or the base station (110-1) can acquire data without distinguishing between the SBFD time interval and the Non-SBFD time interval. This simplifies the AI / ML inference process, such as channel state information inference, beam management, interference information inference, and location information inference, and saves resources. According to at least one embodiment, the terminal (120-1) and / or the base station (110-1) can monitor the performance of the AI / ML model and decide whether to reuse it, thereby preventing waste of time and resources required for retraining other AI / ML models. According to at least one embodiment, the AI / ML function can be improved by allowing the terminal (120-1) and / or the base station (110-1) to pre-store the AI / ML model and use the stored AI / ML model when necessary while the validity period has not expired.

[0243]

[0244] Based on the description of the various embodiments of the present disclosure above, it will be apparent to those skilled in the art that the methods and / or processes of the present invention, and their steps, can be implemented by hardware, software, or any combination of hardware and software suitable for a specific application. The hardware may include a general-purpose computer and / or a dedicated computing device, or a specific computing device or a specific aspect or component of a specific computing device. The processes may be implemented by one or more processors, such as a microprocessor, a controller, such as a microcontroller, an embedded microcontroller, a microcomputer, an arithmetic logic unit (ALU), a digital signal processor, such as a programmable digital signal processor, or other programmable devices, having internal and / or external memory. Additionally, or alternatively, the processes may be implemented in an application specific integrated circuit (ASIC), a programmable gate array, such as a field programmable gate array (FPGA), a programmable logic unit (PLU), or Programmable Array Logic (PAL), or any other device capable of executing instructions and responding thereto, or any other device or combination of devices that may be configured to process electronic signals. The processing device may execute an operating system (OS) and one or more software applications running on the operating system. Furthermore, the processing device may access, store, manipulate, process, and generate data in response to the execution of the software.For ease of understanding, the processing unit is sometimes described as being used alone. However, those skilled in the art will appreciate that the processing unit may include multiple processing elements and / or multiple types of processing elements. For example, the processing unit may include multiple processors, or a processor and a controller. Additionally, other processing configurations, such as parallel processors, are also possible.

[0245] Software may include computer programs, codes, instructions, or a combination of one or more of these, which may configure a processing device to perform a desired operation or, independently or collectively, command the processing device. The software and / or data may be permanently or temporarily embodied in any type of machine, component, physical device, virtual equipment, computer storage media or devices, or transmitted signal waves, for interpretation by the processing device or for providing instructions or data to the processing device. The software may also be distributed over networked computer systems and stored or executed in a distributed manner. The software and data may be stored on one or more machine-readable recording media.

[0246] Furthermore, the objects of the technical solution of the present invention or the parts contributing to prior arts may be implemented in the form of program instructions that can be executed through various computer components and recorded on a machine-readable medium. The machine-readable medium may include program instructions, data files, data structures, etc., alone or in combination. The program instructions recorded on the machine-readable recording medium may be those specially designed and configured for the embodiment or may be known and available to those skilled in the art of computer software. Examples of the machine-readable recording medium include magnetic media such as hard disks, floppy disks, and magnetic tapes, optical media such as CD-ROMs, DVDs, and Blu-rays, magneto-optical media such as floptical disks, and hardware devices specially configured to store and execute program instructions such as ROMs, RAMs, and flash memories. Examples of program instructions include, but are not limited to, machine code, byte code, as well as high-level language code that can be executed by a computer using an interpreter, etc., which can be created using a structured programming language such as C, an object-oriented programming language such as C++, or a high-level or low-level programming language (assembler, hardware description languages, and database programming languages ​​and technologies), which can be stored and compiled or interpreted for execution on any of the aforementioned devices, as well as a heterogeneous combination of processors, processor architectures, or combinations of different hardware and software, or any other machine capable of executing program instructions.

[0247] Accordingly, in one aspect of the present invention, when the methods and combinations thereof described above are performed by one or more computing devices, the methods and combinations thereof may be implemented as executable code that performs each step. In another aspect, the methods may be implemented as systems that perform the steps, and the methods may be distributed in various ways across devices, or all functions may be integrated into a single dedicated, standalone device or other hardware. In yet another aspect, the means for performing the steps associated with the processes described above may comprise any of the hardware and / or software described above. All such sequential combinations and arrangements are intended to fall within the scope of the present disclosure.

[0248] For example, the hardware device described above may be configured to operate as one or more software modules to perform the operations of the embodiments, and vice versa. The hardware device may include a processor, such as an MPU, a CPU, a GPU, or a TPU, coupled with a memory, such as a ROM / RAM, for storing program instructions and configured to execute the instructions stored in the memory, and may include a communication unit capable of transmitting and receiving signals with an external device. In addition, the hardware device may include a keyboard, a mouse, or other external input devices for receiving instructions written by developers.

[0249] Although the present invention has been described above with specific details such as specific components and limited embodiments and drawings, these are provided only to help a more general understanding of the present invention, and the present invention is not limited to the above embodiments, and a person having ordinary knowledge in the technical field to which the present invention pertains can make various modifications and variations from this description.

[0250] Therefore, the spirit of the present invention should not be limited to the embodiments described above, and all modifications equivalent to or equivalent to the appended claims of this disclosure, as well as those equivalent to or equivalent to the claims, are considered to fall within the scope of the spirit of the present invention. For example, appropriate results may be achieved even if the described techniques are performed in a different order from the described method, and / or components of the described systems, structures, devices, circuits, etc. are combined or combined in a different form from the described method, or are replaced or substituted by other components or equivalents.

[0251] Such equivalent or equivalent modifications would include, for example, logically equivalent methods capable of producing the same results as those obtained by carrying out the method according to the present invention, and the spirit and scope of the present invention should not be limited by the examples set forth above, but should be understood in the broadest sense permissible by law.

Claims

In terms of data collection methods for AI / ML, A step of receiving integration / separation information of data for AI / ML from a base station; A step of receiving a reference signal from the base station; A step of determining a data collection method based on the integration / separation information of data for the above AI / ML; A step of collecting data based on a determined data collection method and the reference signal; A method for collecting data for AI / ML, including the step of inputting the collected data into an AI / ML model. In the first paragraph, A data collection method for AI / ML, wherein the integration / separation information for the above AI / ML includes information on whether data in the SBFD time interval and non-SBFD data are separated or integrated. In the second paragraph, A data collection method for AI / ML, wherein the data integration / separation information for the above AI / ML includes information explicitly indicating whether data in the SBFD time interval and data in the non-SBFD time interval are separated or integrated. In the second paragraph, A data collection method for AI / ML, wherein the data integration / separation information for the above AI / ML includes information implicitly indicating whether data in the SBFD time interval and data in the non-SBFD time interval are separated or integrated. In the second paragraph, A data collection method for AI / ML in which, when the integration / separation information for the above AI / ML indicates separation of data in the SBFD time period and data in the non-SBFD time period, the terminal separates the data collected in the SBFD time period and inputs the separated data into the AI / ML model in the SBFD time period. In the second paragraph, If the integration / separation information for the above AI / ML indicates the separation of data for the SBFD time interval and data for AI / ML for the Non-SBFD time interval, The above terminal is a data collection method for AI / ML that separates data collected in the Non-SBFD time period and inputs the separated data into an AI / ML model of the Non-SBFD time period. In the second paragraph, If the integration / separation information for the above AI / ML indicates integration of data for the SBFD time interval and data for AI / ML for the Non-SBFD time interval, A data collection method for AI / ML in which the terminal collects data without distinguishing between the SBFD time interval and the Non-SBFD time interval and inputs the collected data into an AI / ML model. In the first paragraph, The integration / separation information of data for the above AI / ML includes information on the associated ID of the AI / ML model. The above Associated ID is a data collection method for AI / ML that is assigned according to the data collection configuration of the AI / ML model. In paragraph 8, A data collection method for AI / ML, wherein the terminal determines whether to separate or integrate data of the SBFD time period and data of the Non-SBFD time period based on the data collection configuration of the AI / ML model corresponding to the Associated ID. In paragraph 9, A data collection method for AI / ML, wherein the terminal collects data related to the channel status information based on a data collection configuration of an AI / ML model corresponding to the Associated ID, and inputs the collected data into an AI / ML model corresponding to the Associated ID. In the first paragraph, The integration / separation information of data for the above AI / ML is delivered by broadcasting messages. A data collection method for AI / ML, wherein the above broadcast message includes information explicitly indicating whether data in the SBFD time interval and data in the non-SBFD time interval are to be separated or integrated. In the first paragraph, The integration / separation information of data for the above AI / ML is transmitted by DCI (Downlink Control Information). A data collection method for AI / ML, wherein the DCI includes information indicating at least one AI / ML model and control information related to the at least one AI / ML model. In the first paragraph, The integration / separation information of data for the above AI / ML is transmitted by MAC CE (MAC Control Element). A data collection method for AI / ML, wherein the MAC CE includes information indicating at least one AI / ML model and activation / deactivation information of the at least one AI / ML model. In the first paragraph, The integration / separation information of the data for the above AI / ML is transmitted through at least one of the resource configuration information of the reference signal and the resource configuration information of the channel status report. A data collection method for AI / ML, wherein at least one of the resource configuration information of the above reference signal and the resource configuration information of the above channel status report implicitly indicates whether data of the SBFD time interval and data of the Non-SBFD time interval are separated or integrated. In the first paragraph, A data collection method for AI / ML, wherein the resource configuration of the reference signal implicitly indicates separating data of the SBFD time period, when the reference signal resources according to the resource configuration information of the above reference signal are allocated only to the DL sub-band of the SBFD time period. In the first paragraph, A step of transmitting performance feedback of the first AI / ML model to the base station; A step of receiving a reuse instruction of the first AI / ML model from the base station; and A data collection method for AI / ML, further comprising a step of reusing the first AI / ML model according to instructions for reusing the first AI / ML model. In a method performed by a base station, Step of transmitting integration / separation information of data for AI / ML; Step of transmitting a reference signal; A method comprising a step of receiving a channel status information report derived by inputting collected data into an AI / ML model according to a data collection method determined based on integration / separation information of data for the above AI / ML. In a method performed by a base station, Step of transmitting integration / separation information of data for AI / ML; A step of receiving a reference signal from a terminal; A step of determining a data collection method based on the integration / separation information of data for the above AI / ML; A step of collecting data based on a determined data collection method and the reference signal; A method comprising the step of inputting collected data into an AI / ML model.

Images