Model training method and device for communication based on artificial intelligence and / or machine learning

Abstract

The embodiment of the invention provides a model training method and device for communication based on artificial intelligence and / or machine learning (AI / ML). A terminal receives data for AI / ML model training from a base station, and performs AI / ML model training based on the received data. After performing the AI / ML model training, the terminal transmits a first message to the base station indicating that data collection ends.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to wireless communications applicable to 5G NR, 5G-Advanced, and 6G. BACKGROUND

[0002] As the times develop, more and more communication devices need greater communication traffic, and require improved wireless broadband communication than the existing LTE system, that is, the next generation 5G system. In such a next generation 5G system called NewRAT, communication scenarios are divided into enhanced mobile broadband (eMBB) / ultra-reliability and low-latency communication (URLLC) / massive machine-type communications (mMTC), etc.

[0003] Among them, eMBB is a next-generation mobile communication scenario with high spectral efficiency, high user experience data rate, high peak data rate, etc., URLLC is a next-generation mobile communication scenario with ultra-reliability, ultra-low latency, ultra-high availability, etc. (for example, V2X, emergency services, remote control), and mMTC is a next-generation mobile communication scenario with low cost, low energy consumption, short data packets, and massive connectivity features (for example, IoT). SUMMARY

[0004] TECHNICAL PROBLEM The embodiments of the present disclosure provide a method and apparatus related to data collection of a terminal and a network when a terminal side performs AI / ML model training for efficient execution of beam management using AI / ML in a wireless communication system.

[0005] TECHNICAL SOLUTION An embodiment of the present disclosure provides a method in a wireless communication system, a terminal receives data for AI / ML model training from a base station, and performs AI / ML model training based on the received data. After performing the AI / ML model training, the terminal transmits a first message indicating the end of data collection to the base station.

[0006] In addition, an embodiment of the present disclosure provides a method in a wireless communication system, a base station transmits data for AI / ML model training to a terminal. After performing the AI / ML model training, the base station receives a first message indicating the end of data collection from the terminal.

[0007] Furthermore, one embodiment of this specification provides a communication device in a wireless communication system, comprising: at least one processor; and at least one memory storing instructions and operatively electrically connected to the at least one processor, wherein operations performed based on the instructions executed by the at least one processor include: receiving data for AI / ML model training from a base station, and performing AI / ML model training based on the received data. After performing AI / ML model training, a first message indicating the end of data collection is sent to the base station.

[0008] Furthermore, one embodiment of this specification provides a base station in a wireless communication system, comprising: at least one processor; and at least one memory storing instructions and operatively electrically connected to the at least one processor, wherein operations performed based on the instructions executed by the at least one processor include: transmitting data for AI / ML model training to a terminal. After performing AI / ML model training, receiving a first message from the terminal indicating the end of data collection.

[0009] The base station may send configuration information to the terminal for receiving the data, and the terminal may receive the configuration information, wherein the configuration information may include event information associated with the end of the data collection, and the terminal may send the first message based on the event information.

[0010] The event information may include at least one of threshold information associated with the performance of the AI / ML model and time information for terminating the training of the AI / ML model.

[0011] The first message can be used for data transmission deactivation.

[0012] Furthermore, the terminal may send a second message to the base station requesting the data transmission, and the base station may receive the second message. Based on the second message, the base station may send the configuration information to the terminal, and the terminal may receive the configuration information.

[0013] The second message may include at least one of time information and data volume information for terminating the training of the AI / ML model.

[0014] Beneficial effects According to the disclosure in this specification, in AI / ML model training performed by the terminal, it is possible to prevent prolonged unnecessary model training, thereby reducing overall system overhead. Attached Figure Description

[0015] Figure 1 A diagram illustrating a wireless communication system.

[0016] Figure 2 This shows the structure of the radio frame used by NR.

[0017] Figures 3a to 3c An example diagram illustrating an exemplary architecture for wireless communication services.

[0018] Figure 4 The time slot structure of an NR frame is shown.

[0019] Figure 5 This shows an example of subframe types in NR.

[0020] Figure 6 The structure of the self-contained time slot is shown.

[0021] Figure 7 An example of initial beam measurement and selection in NR is shown.

[0022] Figure 8 This illustrates an example of the initial access process between a terminal and a base station in NR.

[0023] Figure 9 An example of candidate beam configuration in NR is shown.

[0024] Figures 10a to 10c The three processes used for beam management in NR are shown.

[0025] Figures 11a to 11c This shows an example of the beam reporting process in NR.

[0026] Figures 12a to 12b This demonstrates beam measurement and spatial domain beamprediction using AI / ML.

[0027] Figure 13 This demonstrates temporal domain beam prediction using AI / ML.

[0028] Figure 14 This specification illustrates a method of operating a terminal according to an embodiment of the present specification.

[0029] Figure 15 This specification illustrates a method of operating a terminal according to another embodiment.

[0030] Figure 16 This specification illustrates a method of operating a terminal according to another embodiment.

[0031] Figures 17a to 17b This describes the process of a terminal and a base station according to an embodiment of this specification.

[0032] Figure 18This specification illustrates a process for a terminal and base station for event-based model training according to an embodiment of this specification.

[0033] Figure 19 This specification illustrates a process for a terminal and base station for event-based model training according to another embodiment of the present specification.

[0034] Figure 20 This specification illustrates a process for a terminal and base station for event-based model training according to another embodiment of the present specification.

[0035] Figure 21 This specification illustrates a process for a terminal and base station for event-based model training according to another embodiment of the present specification.

[0036] Figures 22 to 23 This illustrates the process of terminating model training in a terminal and base station according to an embodiment of this specification.

[0037] Figure 24 An apparatus according to one embodiment of this specification is shown.

[0038] Figure 25 A block diagram illustrating the configuration of a terminal according to an embodiment of this specification.

[0039] Figure 26 A block diagram illustrating the configuration of the processor disclosed in this specification is shown.

[0040] Figure 27 To show in detail Figure 24 The transceiver or transceiver of the first device shown Figure 25 Block diagram of the transceiver section of the device shown. Detailed Implementation

[0041] It should be noted that the technical terms used in this specification are only used to describe specific embodiments and are not intended to limit the content of this specification. Furthermore, unless otherwise defined in this invention, the technical terms used in this specification should be interpreted as meaning commonly understood by one of ordinary skill in the art to which this specification pertains, and should not be interpreted as having an overly broad or overly narrow meaning. Additionally, when the technical terms used in this specification are inappropriate and fail to accurately express the content and ideas of this specification, they should be replaced with technical terms that a person of skill in the art can correctly understand. Furthermore, the general terms used in this specification should be interpreted according to their predefined content or context, and should not be interpreted as having an overly narrow meaning.

[0042] Furthermore, unless the context explicitly states otherwise, the singular expressions used in this specification include the plural expressions. In this application, terms such as “constituting” or “having” should not be construed as necessarily including all of the multiple constituent elements or steps described in the specification, but should be interpreted as not necessarily excluding some of the constituent elements or steps, or possibly including additional constituent elements or steps.

[0043] Furthermore, the terms including ordinal numbers such as 1, 2, etc., used in this specification may be used to describe various constituent elements, but the constituent elements shall not be limited by the terms. The terms are used only for the purpose of distinguishing one constituent element from another. For example, without departing from the scope of the claims, a first constituent element may be named a second constituent element, and similarly, a second constituent element may be named a first constituent element.

[0044] When it is mentioned that a constituent element is "connected" or "continued" to another constituent element, it can mean that the element is directly connected or continues to the other constituent element, or that there are other constituent elements in between. Conversely, when it is mentioned that a constituent element is "directly connected" or "directly continued" to another constituent element, it should be understood that there are no other constituent elements in between.

[0045] The embodiments are described in detail below with reference to the accompanying drawings. Identical or similar components are assigned the same reference numerals, and repeated descriptions thereof are omitted, regardless of the drawing numbers. Furthermore, in describing the invention, detailed descriptions of relevant prior art are omitted when it is determined that such descriptions may obscure the gist of this specification. It should also be noted that the drawings are only used to facilitate a clearer understanding of the content and ideas of this specification and should not be construed as limiting the content and ideas of this specification. The content and ideas of this specification should be interpreted as extending beyond the drawings to all modifications, equivalents, and substitutions.

[0046] In this specification, "A or B" can mean "A only", "B only", or "both A and B". In other words, in this specification, "A or B" can be interpreted as "A and / or B". For example, in this specification, "A, B, or C" can mean "A only", "B only", "C only", or "any combination of A, B, and C".

[0047] The forward slash ( / ) or comma used in this specification can mean "and / or". For example, "A / B" can mean "A and / or B". Therefore, "A / B" can mean "A only", "B only", or "both A and B". For example, "A, B, C" can mean "A, B or C".

[0048] In this specification, "at least one of A and B" may mean "only A", "only B" or "both A and B". Additionally, in this specification, the expressions "at least one of A or B" or "at least one of A and / or B" can be interpreted in the same way as "at least one of A and B".

[0049] Additionally, in this specification, "at least one of A, B and C" can mean "A only", "B only", "C only", or "any combination of A, B and C". Furthermore, "at least one of A, B or C" or "at least one of A, B and / or C" can mean "at least one of A, B and C".

[0050] Furthermore, the parentheses used in this specification may mean "for example." Specifically, when identified as "Control Information (PDCCH)," "PDCCH (physical downlink control channel)" may be mentioned as an example of "Control Information." In other words, "Control Information" in this specification is not limited by "PDCCH," and "PDCCH" may be mentioned as an example of "Control Information." Additionally, when identified as "Control Information (i.e., PDCCH)," "PDCCH" may be mentioned as an example of "Control Information."

[0051] In this specification, the technical features described individually in a single drawing can be implemented either independently or simultaneously.

[0052] The accompanying drawings exemplarily illustrate a UE (User Equipment), but the UE shown may also be replaced by terms such as Terminal or Mobile Equipment. Furthermore, the UE can be a portable device such as a laptop, mobile phone, PDA (Personal Digital Assistant), smartphone, or multimedia device, or a non-portable device such as a PC or in-vehicle device.

[0053] Below is an example of a UE acting as a wireless communication device (e.g., a wireless communication device, a wireless apparatus, or a wireless equipment). The operations performed by the UE can be performed by any wireless communication device. It can also be referred to as a wireless communication device, a wireless communication apparatus, a wireless apparatus, or a wireless equipment, etc.

[0054] The term "base station" as used below generally refers to a fixed location for communication with wireless devices. It can also be used as a broad term including eNodeB (evolved-NodeB), eNB (evolved-NodeB), BTS (Base Transceiver System), Access Point, gNB (Next generation NodeB), RRH (remote radiohead), TP (transmission point), RP (reception point), relay, etc.

[0055] This specification uses LTE systems, LTE-A systems, and NR systems to describe embodiments, but these embodiments can also be applied to any communication system to which the definitions apply.

[0056] Wireless Communication Systems Benefiting from the success of LTE (long term evolution) / LTE-Advanced (LTE-A) for fourth-generation mobile communication, fifth-generation (so-called 5G) mobile communication, as the next generation, has been commercialized and is undergoing further research.

[0057] The International Telecommunication Union (ITU) defines fifth-generation mobile communication as providing a maximum data transmission speed of 20Gbps and a minimum sensor speed of 100Mbps or more at any location. Its official name is "IMT-2020".

[0058] The ITU proposes three major use cases: eMBB (enhanced Mobile Broadband), mMTC (massive Machine Type Communication), and URLLC (Ultra Reliable and Low Latency Communications).

[0059] URLLC addresses application scenarios requiring high reliability and low latency. Examples include autonomous driving, factory automation, and augmented reality services, all of which demand high reliability and low latency (e.g., less than 1ms). Currently, 4G (LTE) latency is statistically between 21-43ms (best 10%) and 33-75ms (median). This is insufficient to support services requiring less than 1ms latency. Then, eMBB application scenarios involve those requiring mobile ultra-wideband.

[0060] In other words, the fifth-generation mobile communication system supports higher capacity than the current 4G LTE, increasing the density of mobile broadband users and supporting D2D (Device to Device), high stability, and MTC (Machine-type communication). To better realize the Internet of Things (IoT), 5G development also aims for lower standby time and lower power consumption than 4G mobile communication systems. For this 5G mobile communication, new radio access technologies (New RAT or NR) can be proposed.

[0061] NR frequency bands can be defined as frequency ranges of two types (FR1 and FR2). The numerical values ​​of the frequency ranges can vary; for example, the frequency ranges of the two types (FR1 and FR2) can be shown in Table 1 below. For ease of description, FR1 in the frequency ranges used in NR systems can refer to "sub-6GHz range," and FR2 can refer to "above 6GHz range," and can be referred to as millimeter wave (mmW).

[0062] [Table 1] The frequency range of an NR system can be varied. For example, FR1, as shown in Table 1, can include a frequency band from 410 MHz to 7125 MHz. That is, FR1 can include frequency bands above 6 GHz (or 5850, 5900, 5925 MHz, etc.). For example, the frequency bands above 6 GHz (or 5850, 5900, 5925 MHz, etc.) included in FR1 can include unlicensed bands. Unlicensed bands can be used for various purposes, such as vehicle communications (e.g., autonomous driving).

[0063] On the other hand, 3GPP-based communication standards define downlink physical channels corresponding to resource elements that carry information from the upper layers, and downlink physical signals corresponding to resource elements used by the physical layer but not carrying information from the upper layers. For example, the physical downlink shared channel (PDSCH), physical broadcast channel (PBCH), physical multicast channel (PMCH), physical control format indicator channel (PCFICH), physical downlink control channel (PDCCH), and physical hybrid ARQ indicator channel (PHICH) are defined as downlink physical channels, while reference signals and synchronization signals are defined as downlink physical signals. A reference signal (RS), also known as a pilot, refers to a predefined signal with a specific waveform known to both the gNB and the UE. Examples include cell-specific RS, UE-specific RS (UE-RS), positioning RS (PRS), and channel state information RS (CSI-RS), which are defined as downlink reference signals. The 3GPP LTE / LTE-A standard defines uplink physical channels corresponding to resource elements carrying information from the upper layers, and uplink physical signals corresponding to resource elements used by the physical layer but not carrying information from the upper layers. For example, the physical uplink shared channel (PUSCH), physical uplink control channel (PUCCH), and physical random access channel (PRACH) are defined as uplink physical channels. The standard also defines demodulation reference signals (DMRS) for uplink control / data signals and sounding reference signals (SRS) for uplink channel measurements.

[0064] In this specification, PDCCH (Physical Downlink Control CHannel), PCFICH (Physical Control Format Indicator CHannel), PHICH (Physical Hybrid Automatic Retransmit Request Indicator CHannel), and PDSCH (Physical Downlink Shared CHannel) refer to the sets of time-frequency resources or resource elements carrying DCI (downlink control information), CFI (Control Format Indicator), downlink ACK, NACK (ACK nowlegement / Negative ACK), and downlink data, respectively. Similarly, PUCCH (Physical Uplink Control CHannel), PUSCH (Physical Uplink Shared CHannel), and PRACH (Physical Random Access CHannel) refer to the sets of time-frequency resources or resource elements carrying UCI (Uplink Control Information), uplink data, and random access signals, respectively.

[0065] Figure 1 A diagram illustrating a wireless communication system.

[0066] Reference Figure 1 It is known that the wireless communication system includes at least one base station (BS). The BS is divided into gNodeB (or gNB) 20a and eNodeB (or eNB) 20b. The gNB 20a supports fifth-generation mobile communication. The eNB 20b supports fourth-generation mobile communication, namely LTE (long term evolution).

[0067] Each base station 20a and 20b provides communication services for a specific geographical area (usually called a cell) 20-1, 20-2, and 20-3. A cell can be further divided into multiple areas (called sectors).

[0068] A UE (User Equipment) typically belongs to a cell, which is called the serving cell. The base station that provides communication services to the serving cell is called the serving base station (BS). Wireless communication systems are cellular systems, therefore, there are other cells adjacent to the serving cell. These other cells are called neighboring cells. The base station that provides communication services to the neighboring cells is called the neighboring base station (BS). The serving cell and neighboring cells are determined relative to the UE.

[0069] Below, downlink refers to communication between base station 20 and UE10, and uplink refers to communication between UE10 and base station 20. In the downlink, the transmitter can be part of base station 20, and the receiver can be part of UE10. In the uplink, the transmitter can be part of UE10, and the receiver can be part of base station 20.

[0070] On the other hand, wireless communication systems can be broadly categorized into FDD (frequency division duplex) and TDD (time division duplex) methods. In FDD, uplink and downlink transmissions occupy different frequency bands. In TDD, uplink and downlink transmissions occupy the same frequency band but occur at different times. The channel response in TDD is essentially reciprocal. Within a given frequency region, the downlink and uplink channel responses are almost identical. Therefore, in TDD-based wireless communication systems, the downlink channel response has the advantage of being derived from the uplink channel response. In TDD, uplink and downlink transmissions are time-divided across the entire frequency band, thus the base station's downlink transmission and the UE's uplink transmission cannot occur simultaneously. In TDD systems where uplink and downlink transmissions are divided into subframe units, they are executed in different subframes.

[0071] Figure 2 This shows the structure of the radio frame used by NR.

[0072] In NR, uplink and downlink transmissions are composed of frames. A radio frame is 10ms long and is defined as two 5ms half-frames (HF). A half-frame is defined as five 1ms subframes (SF). A subframe is divided into more than one time slot, the number of time slots within a subframe depending on the subcarrier spacing (SCS). Each time slot includes 12 or 14 OFDM(A) symbols depending on the cyclic prefix (CP). With standard CP, each time slot includes 14 symbols. With extended CP, each time slot includes 12 symbols. These symbols can include OFDM symbols (or CP-OFDM symbols) and SC-FDMA symbols (or DFT-s-OFDM symbols).

[0073] Supports diverse parameter sets (numerology) In NR systems, with the development of wireless communication technology, multiple parameter sets (numerologies) can also be provided to terminals. For example, when the SCS is 15kHz, it supports a wide area in the traditional cellular band; when the SCS is 30kHz / 60kHz, it supports dense-urban areas, lower latency, and wider carrier bandwidth; and when the SCS is 60kHz or higher, it supports a bandwidth greater than 24.25GHz to overcome phase noise.

[0074] The parameter set can be defined based on the CP (cycle prefix) length and the subcarrier spacing (SCS). A cell can provide multiple parameter sets to the terminal. When μ represents the parameter set index, the CP length corresponding to each subcarrier spacing can be shown in the table below.

[0075] [Table 2] For a standard CP, when the index of the parameter set is represented as μ, the number of OFDM symbols per slot (N) slot symb ), Number of time slots per frame (N) frame,μ slot ) and the number of time slots in each subframe (N) subframe,μ slot As shown in the table below.

[0076] [Table 3] For extended CP, when the index of the parameter set is represented as μ, the number of OFDM symbols per slot (N) slot symb ), Number of time slots per frame (N) frame,μ slot ) and the number of time slots in each subframe (N) subframe,μ slot As shown in the table below.

[0077] [Table 4] In NR systems, OFDM(A) parameter sets (e.g., SCS, CP length, etc.) can be set differently between multiple cells merged into a single terminal. Therefore, the absolute time intervals of time resources (e.g., SF, time slots, or TTI) consisting of the same number of symbols (collectively referred to as TU (Time Unit) for convenience) can be set differently between the merged cells.

[0078] Figures 3a to 3c This is an example diagram illustrating an exemplary architecture for wireless communication services.

[0079] Reference Figure 3a The UE connects to LTE / LTE-A based cells and NR based cells via DC (dual connectivity).

[0080] The NR-based cell is connected to the core network used for the original fourth-generation mobile communication, namely EPC (Evolved Packet Core).

[0081] Reference Figure 3b Unlike Figure 3a Cells based on LTE / LTE-A are connected to the core network used for fifth-generation mobile communication, namely the 5G core network.

[0082] Based on Figure 3a and Figure 3b The service model shown is called NSA (non-standalone).

[0083] Reference Figure 3c The UE only connects to NR-based cells. This service method based on this architecture is called SA (standalone).

[0084] On the other hand, in the aforementioned NR, it is possible to consider using downlink subframes for receiving data from the base station and using uplink subframes for transmitting data to the base station. This approach can be applied to both paired and unpaired spectrum. A pair of spectrums means that two carrier spectrums are included for downlink and uplink operation. For example, in a pair of spectrums, a carrier may include a pair of downlink and uplink frequency bands.

[0085] Figure 4 The time slot structure of an NR frame is shown.

[0086] A time slot comprises multiple symbols in the time domain. For example, a time slot contains 14 symbols for a normal CP, while a time slot contains 12 symbols for an extended CP. A carrier comprises multiple subcarriers in the frequency domain. An RB (Resource Block) is defined in the frequency domain as multiple (e.g., 12) consecutive subcarriers. A BWP (Bandwidth Part) is defined in the frequency domain as multiple consecutive (physical, P) RBs, which can correspond to a set of parameters (e.g., SCS, CP length, etc.). A terminal can construct up to N (e.g., 4) BWPs in both the downlink and uplink. Downlink or uplink transmissions can be performed through active BWPs, and at a given time, only one BWP established for the terminal can be active. In the resource grid, each element is called a resource element (RE), which can be mapped to a complex number of symbols.

[0087] Figure 5 This shows an example of subframe types in NR.

[0088] Figure 5 The TTI (transmission time interval) shown can be referred to as a subframe or time slot for NR (or new RAT). Figure 5 Subframes (or time slots) can be used in NR (or new RAT) TDD systems to minimize data transmission latency. For example... Figure 5 As shown, a subframe (or time slot) comprises 14 symbols. The first few symbols of a subframe (or time slot) can be used for the downlink (DL) control channel, and the last few symbols can be used for the uplink (UL) control channel. The remaining symbols can be used for either DL or UL data transmission. Based on this subframe (or time slot) structure, downlink and uplink transmissions can proceed sequentially within a single subframe (or time slot). Therefore, downlink data can be received within a subframe (or time slot), and uplink acknowledgment responses (ACK / NACK) can also be transmitted within the same subframe (or time slot).

[0089] This type of subframe (or time slot) structure can be called a self-contained subframe (or time slot).

[0090] Specifically, the first N symbols within a time slot can be used to transmit the DL control channel (hereinafter referred to as the DL control area), and the last M symbols within a time slot can be used to transmit the UL control channel (hereinafter referred to as the UL control area). N and M are integers greater than or equal to 0. The resource area (hereinafter referred to as the data area) between the DL control area and the UL control area can be used for either DL data transmission or UL data transmission. For example, in the DL control area, the physical downlink control channel (PDCCH) can be transmitted, and in the DL data area, the physical downlink shared channel (PDSCH) can be transmitted. In the UL control area, the physical uplink control channel (PUCCH) can be transmitted, and in the UL data area, the physical uplink shared channel (PUSCH) can be transmitted.

[0091] Using this subframe (or time slot) structure offers the advantage of reducing the time required to retransmit data that has received errors, thereby minimizing the final data transmission standby time. In this self-contained subframe (or time slot) structure, a time gap is required during the transition from transmit mode to receive mode or vice versa. Therefore, in the subframe structure, a portion of the OFDM symbols during the transition from DL to UL can be set as a guard period (GP).

[0092] Figure 6 The structure of the self-contained time slot is shown.

[0093] In the NR system, a frame is characterized by a self - contained structure that can fully include a DL control channel, DL or UL data, a UL control channel, etc. within one time slot. For example, the first N symbols within the time slot can be used to transmit a DL control channel (hereinafter referred to as the DL control region), and the last M symbols within the time slot can be used to transmit a UL control channel (hereinafter referred to as the UL control region). N and M are integers greater than or equal to 0 respectively. The resource region (hereinafter referred to as the data region) between the DL control region and the UL control region can be used for DL data transmission or for UL data transmission. As an example, the following configurations can be considered. Each interval is listed in chronological order.

[0094] 1. DL only configuration 2. UL only configuration 3. Mixed UL - DL configuration - DL region + GP (Guard Period) + UL control region - DL control region + GP + UL region DL region: (i) DL data region, (ii) DL control region + DL data region UL region: (i) UL data region, (ii) UL data region + UL control region PDCCH can be transmitted in the DL control region, and PDSCH can be transmitted in the DL data region. PUCCH can be transmitted in the UL control region, and PUSCH can be transmitted in the UL data region. DCI (Downlink Control Information), such as DL data scheduling information, UL data scheduling information, etc., can be transmitted in the PDCCH. UCI (Uplink Control Information), such as ACK / NACK (Positive Acknowledgement / Negative Acknowledgement) information for DL data, CSI (Channel State Information) information, SR (Scheduling Request), etc., can be transmitted in the PUCCH. GP can provide a time difference during the process of the base station and the terminal switching from the transmission mode to the reception mode or from the reception mode to the transmission mode. Within a sub - frame, a part of the symbols at the time point of switching from DL to UL can be set as GP.

[0095] <Beam Management in NR> The current 3GPP NR beam management method can be divided into the initial access phase and the connection establishment phase. The terminal performing the initial access procedure sets the initial transmit / receive (Tx / Rx) beam through the random access procedure, namely the RACH (Random Access Channel) procedure.

[0096] Figure 7 An example of initial beam measurement and selection in NR is shown.

[0097] Reference Figure 7 To provide base station transmit beam (gNB Txbeam) configuration to terminals (UE1 / UE2) without cellular connectivity, the base station periodically and repeatedly transmits SSBs (Synchronization Signal Blocks) mapped with beams in different directions. Furthermore, the SSBs can be transmitted at 20ms intervals within 5ms. Specifically, the default value for initial cell selection can be 20ms.

[0098] Terminals can select a suitable (qualified) SSB by performing signal measurements on periodically transmitted SSBs, and inform the base station of the relevant information of the selected Tx beam by sending a PRACH (Physical Random Access Channel) preamble mapped to the corresponding SSB. For example, based on signal strength measurements, terminals located at different locations, i.e., UE1, selects the SSB with SSB index 3, while UE2 selects the SSB with SSB index 9. UE1 and UE2 respectively send the PRACH preamble corresponding to their selected SSBs. Here, it is assumed that each SSB is beamformed in a specific direction.

[0099] Figure 8 This illustrates an example of the initial access process between a terminal and a base station in NR.

[0100] Reference Figure 8 After the terminal (UE) is powered on (S801), the UE receives the cellular-related parameter information (e.g., PRACH information corresponding to each SSB) required for the initial access phase through the system information message sent by the base station (gNB) (S802). Here, the system information message includes MIB (Master Information Block) and SIB1 (System Information Block 1) which includes cellular common information.

[0101] After obtaining the system information message, the terminal receives periodically transmitted SSBs from the base station (S803). Furthermore, the terminal measures the RSRP (Reference Signal Received Power) of the received SSBs. From the RSRP measurements of N SSBs (beams), the terminal selects the SSB (beam) with the highest / qualified value (S804).

[0102] The terminal then sends the RA (Random Access) preamble (S805) belonging to the PRACH resource corresponding to the selected SSB (beam) to the base station. In this way, the terminal can inform the base station of the selected initial beam information.

[0103] The base station receives the RA (Random Access) preamble belonging to the PRACH resource corresponding to the selected SSB (beam) from the terminal, and in response to it, sends the RAR (Random Access Response) to the terminal using the selected SSB (beam) (S806).

[0104] Furthermore, for base stations that do not know the location / beam information of the terminal initially entering the cell, i.e., the terminal performing the CBRA (Content-Based Random Access) procedure, up to 64 beams can be configured in a cell-commonly manner for beam configuration of connectionless terminals. The terminal then performs sequential measurements of all beams to find the optimal beam at its location. As the number of beams within the cell increases, this not only leads to time delays in beam selection and cellular connection but may also increase terminal power consumption due to the large number of beams measured.

[0105] To mitigate these issues, base stations can determine the approximate location / beam of the initial access terminal by mapping a wide beam to the SSB, and then configure the terminal with a narrow beam through beam refinement after accessing the cell. However, while a narrow beam provides a high data rate, it is sensitive to terminal movement or environmental changes, making it prone to connection drops. To address this, base stations allocate CSI resources (CSI-RS / SSB) mapped with candidate beams to the terminal in a UE-specific manner, enabling the terminal to continuously measure surrounding beam strength and report the results to the base station. This can be configured from the base station through CSI resource configuration and CSI report configuration.

[0106] Figure 9 An example of candidate beam configuration in NR is shown.

[0107] Terminals receiving beam reporting configurations perform reporting based on the base station's configuration by measuring the reference signal (RS) assigned to them. This follows the 3GPP-defined CSI framework. However, this UE-specific CSI configuration method has a problem: as the number of terminals within a cell increases, the RS resources allocated to each terminal also increase dramatically. To mitigate this resource overhead, the base station can choose to... Figure 9 The method shown assigns the same candidate beams, or CSI resources, to terminals located in similar locations. This can be called UE group-specific CSI resource configuration. However, if terminals with different mobility share the same resources, a problem arises where new candidate beam resources need to be assigned to terminals leaving that resource area. If the minimum number of candidate beams is assigned to reduce resource overhead for terminals with high / medium mobility, the terminals will frequently undergo RRC reconfiguration. Candidate beam reconfiguration via RRC introduces relatively large latency, which may lead to beam disconnection. To mitigate this problem, the base station can utilize candidate beams by appropriately increasing the number of beams belonging to the CSI resource set. However, from the terminal's perspective, there may be a trade-off between increased measurement burden due to the increased number of beams.

[0108] Figures 10a to 10c The three processes used for beam management in NR are shown.

[0109] Beam management in NR can be divided into three processes in terms of the physical layer definition. Figure 10a Process 1 (P1) is shown. Figure 10b Process 2 (P2) is shown. Figure 10cProcess 3 (P3) is illustrated. P1 involves searching for a transmit / receive beam pair (Tx / Rx beam pair) while simultaneously performing TRP (Transmit / Receive Point) beam sweeping and UE beam sweeping, similar to the beam configuration method for the terminal performing the initial access procedure described above. The terminal, in connected mode, senses that the configured beam will be swept through the candidate beams (i.e., CSI resource set) configuration from the base station and first performs a signal strength measurement on the TRP beam. After selecting the terminal's TRP beam via P2, the base station repetitions the selected beam via P3. The terminal can select the UE beam while performing UE beam sweeping. In this operation, the choice of which beam the UE selects can be determined by the terminal. This operation can be applied to both downlink (DL) and uplink (UL).

[0110] Figures 11a to 11c This shows an example of the beam reporting process in NR.

[0111] Beam sweeping is a method that implicitly maps beam information to RS resource information by configuring specific candidate beams, i.e., CSI resource sets, and informing the terminal of reference signal (RS) resource information through the base station. That is, instead of informing the terminal of the actual beam index, it uses RS resource indicators (RIs) to implicitly map the index information to the RS information, allowing the base station to perceive the mapped beam information. This is configured using the 3GPP CSI framework. The terminal measures the RS strength of the resources configured by the base station and reports the RSRP information for the best four beams (RIs) to the base station. The method of reporting this measurement result also follows the base station's RRC configuration; 3GPP defines one of three configuration methods.

[0112] - Periodic reporting - Aperiodic reporting - Semi-persistent reporting Figure 11aThis illustrates a periodic CSI reporting method triggered by RRC configuration. Specifically, the terminal receives an RRC configuration message from the base station. This message includes the configuration of CSI-related RS resources and reporting methods, specifically CSI resource set information and information indicating that CSI reporting is periodic (S1101a). Subsequently, the terminal receives periodically transmitted RS messages based on the received RRC configuration message (S1102a and S1105a), and measures the beam signal strength based on the received RS messages (S1103a and S1106a). Furthermore, the terminal periodically reports the measurement results (values) to the base station (S1104a and S1107a).

[0113] Figure 11b This illustrates a non-periodic CSI reporting method. Even if CSI-related RS resources and reporting methods are configured via RRC configuration messages, the terminal will not perform beam measurement via RS without a trigger message (or information) from the lower layer. Specifically, the terminal receives an RRC configuration message (S1101b) from the base station containing configuration information for CSI-related RS resources and reporting methods, namely CSI resource set information and information that CSI reporting is non-periodic. CSI report triggering is implemented via MAC (Media Access Control) CE (Control Unit) or DCI (Downlink Control Information). The terminal receives CSI report trigger information including a trigger indication from the base station via MAC CE or DCI (S1102b), and receives a one-time RS transmission based on the received trigger indication (S1103b). Here, the transmission of RS regarding the CSI resource set can be sent after a specific time (e.g., X time slots) following the transmission of the CSI report trigger information. Then, the terminal measures the signal strength of the beam based on the received RS (S1104b). Furthermore, the terminal reports the measurement result (value) to the base station all at once (S1105b). Here, the CSI report can be sent a specific time (e.g., Y time slots) after the CSI report trigger information is received.

[0114] Figure 11cThis illustrates a semi-persistent reporting method, serving as an intermediate between periodic and non-periodic reporting methods. A terminal receiving an RRC configuration message regarding CSI-related RS resources and reporting methods periodically performs CSI reporting only when activated by MAC CE, and before receiving a deactivation message (or information). Specifically, the terminal receives an RRC configuration message (S1101c) from the base station containing configurations of CSI-related RS resources and reporting methods, namely CSI resource set information and information that CSI reporting is semi-persistent. CSI report activation is achieved through MAC CE. The terminal receives CSI report activation information, including an activation indication, from the base station via MAC CE (S1102c and S1110c), and receives periodically transmitted RS (S1103c, S1106c, S1111c, and S1114c) based on the received activation indication. It also measures the beam signal strength based on the received RS (S1104c, S1107c, S1112c, and S1115c). Furthermore, the terminal periodically reports the measurement results (values) to the base station (S1105c, S1108c, S1113c, and S1116c). If, after CSI reporting is activated, the terminal receives CSI report deactivation information, including a deactivation indication, from the base station via MAC CE (S1109c), the terminal stops CSI reporting.

[0115] Recently, 3GPP has been considering applying AI / ML models to improve the latency and terminal power consumption of this beam search / measurement, and has begun research to discuss its feasibility and potential spec impact.

[0116] The table below discusses a list of terms applicable to AI / ML.

[0117] Table 6 3GPP has decided to study the impact of data collection on the UE-side for performing AI model training related to beam management procedures, as well as the beam result reporting method on the specifications.

[0118] Furthermore, traditional beam management in NR systems leads to increased system overhead and terminal power consumption as the number of beams and terminals increases. Moreover, for terminals in the initial cellular access phase, the selection of the initial beam after measuring all beams can cause cellular access latency. To mitigate these issues, a scheme utilizing an AI / ML model to predict the overall beam strength through partial beam measurements has been proposed, but its detailed process or scheme is currently undefined. This specification aims to propose a scheme that enables NW to efficiently collect beam information from terminals for training relevant models when implementing an AI / ML-based beam management scheme.

[0119] Figures 12a to 12b This demonstrates beam measurement and spatial domain beamprediction using AI / ML.

[0120] Currently, 3GPP RAN (radio access network) WG1 (working group 1) has begun research on "AI / ML for beam management" and agreed to discuss spatial DL beam prediction (BM-Case 1) and temporal DL beam prediction (BM-Case 2) as sub-use cases. This involves predicting the beam strength of set A by measuring the beams belonging to set B. For the spatial DL beam prediction case... Figures 12a to 12b It is shown, and in Figure 12a This shows the case where set B is a subset of set A. Figure 12bConsider a set B consisting of wide beams and a set A consisting of narrow beams, i.e., sets composed of different beams. For the case of temporal DL beam prediction, in addition to the spatial DL beam prediction cases i) where set B is a subset of set A, and ii) where sets A and B are different sets, consider the case iii) where sets A and B consist of the same set. Temporal DL beam prediction predicts future beam information based on past beam measurement information; therefore, it is possible to consider predicting the entire beam based on spatial DL beam prediction and then applying it to the case iii) where sets A and B consist of the same set. For this reason, it is expected that the spatial DL beam prediction cases i) where set B is a subset of set A and ii) where sets A and B are different sets will be used as basic beam prediction schemes.

[0121] Figure 13 This demonstrates temporal domain beam prediction using AI / ML.

[0122] Temporal beam prediction in BM-Case 2 is defined as the operation of predicting the beam pattern at a specific point in the near future (i.e., the output) based on past beam measurement results (i.e., input). Figure 13 As shown. In this case, in addition to the above-mentioned i) the case where set B is a subset of set A, and ii) the case where set A and set B are different sets, the beam used for input and the beam set derived as output can also be considered in the case where set A and set B are composed of the same set.

[0123] Furthermore, beam management utilizing AI / ML models requires a model training process. For beam management model training to be performed at the user terminal (UE), the terminal needs to receive the configuration of the DL RS (Downlink Reference Signal) used for training from the base station. RAN1 has agreed to consider the following two options as methods for receiving this configuration.

[0124] Option 1. Data collection initiated / triggered by configuration from NW (Network). Option 2. Request from UE for data collection Terminals can begin training using the two options mentioned above, but there is currently no specific discussion on how the base station should provide data for training, or how long data collection by the terminal should continue. Typically, model training can terminate when the model's performance reaches the target value. However, despite continuous training, model performance may still fail to reach a satisfactory level, or for some reason, continued training itself may actually degrade overall system performance. Terminals or NWs that determine this situation need to terminate model training or release related configuration procedures, but this has not yet been discussed.

[0125] Based on the above, this specification aims to propose a configuration procedure for data transmission and a scheme for determining data transmission termination when performing beam management using an AI / ML model for efficient UE-side model training.

[0126] Furthermore, this specification aims to propose a scheme for a data collection procedure for model training, which includes the following steps: when a terminal (UE) needs to collect model training-related data to perform communication using an AI / ML model, it requests a base station to send data for data collection, or receives a message from the base station informing it of the start of data transmission for model training; the terminal uses resources configured from the base station to receive the sent data, thereby performing model training through data collection; and when a specific event is met, it sends a data termination message to the base station to terminate data collection. More specifically, the steps for determining the termination time of model training data collection, as well as the related message definitions and terminal / base station operations, are defined.

[0127] Figure 14 This specification illustrates a method of operating a terminal according to an embodiment of the present specification.

[0128] Reference Figure 14 The terminal receives data for AI / ML model training from the base station (S1401) and performs AI / ML model training based on the received data (S1402). After performing AI / ML model training, the terminal sends a first message to the base station indicating the end of data collection (S1403). Here, the first message can be used to deactivate the data transmission during AI / ML model training.

[0129] Figure 15 This specification illustrates a method of operating a terminal according to another embodiment.

[0130] exist Figure 15 In Figure 14 Based on the embodiments described above, a data transmission request for AI / ML model training sent by the terminal to the base station is further illustrated.

[0131] That is, the terminal that decides to start collecting data for AI / ML model training sends a request message (second message) to the base station to request the transmission of the data required for the data collection (S1500). Here, the second message may include at least one of the following i) to iii).

[0132] i) Functionality / Model ID ii) An indicator that specifies the data collection purpose (e.g., model training) corresponding to the feature / model ID. iii) Information on the minimum / maximum / preferred time (timer or time instance) and / or data volume / size required for data collection for model training. Upon receiving the second message from the terminal, the base station should begin sending data to the terminal based on the received information. This may be because the resource information used to send the data requested by the terminal has been pre-configured for the terminal during the function / model identification process, or it may be due to the addition of configuration messages from the base station to the terminal upon receiving the second message. That is, the sending of the second message from the terminal to the base station may actually mean that the terminal is requesting configuration of resources for data transmission for model training, or, if already configured, requesting activation of those resources. The terminal can send a request message, i.e., the second message, to the base station and begin model training accordingly. Specifically, after sending the second message requesting data transmission for AI / ML model training, the terminal receives data for AI / ML model training from the base station (S1501) and performs AI / ML model training based on the received data (S1502). After performing AI / ML model training, the terminal sends a first message to the base station indicating the end of data collection (S1503).

[0133] Alternatively, it can be defined as follows: after the terminal sends a request message (i.e., the second message) to the base station, the base station sends a response message to the second message, and the terminal expects to receive the response message.

[0134] Figure 16 This specification illustrates a method of operating a terminal according to another embodiment.

[0135] exist Figure 16 In Figure 14 Based on the embodiments described above, an additional illustration shows the terminal receiving a data transmission start notification from the base station for AI / ML model training.

[0136] That is, the terminal receives a data transmission start notification from the base station for AI / ML model training (S1600). Then, it receives data for AI / ML model training from the base station (S1601) and performs AI / ML model training based on the received data (S1602). After performing AI / ML model training, the terminal sends a first message to the base station indicating the end of data collection (S1603).

[0137] The notification that data transmission for AI / ML model training has begun from the base station can be sent and received without the aforementioned... Figure 15 In the case of the second message from the terminal requesting the data transmission of AI / ML model training, as described in the document, the sending and receiving are performed.

[0138] Furthermore, notification of the start of data transfer for AI / ML model training can be achieved through configuration messages associated with data for a specific model training. For example, if the activation of data transfer resources for a specific model training is achieved through this configuration message, it can signify the start of data transfer for model training on the terminal.

[0139] Figures 17a to 17b This describes the process of a terminal and a base station according to an embodiment of this specification.

[0140] Figures 17a to 17b The process is illustrated according to an embodiment of data collection for model training, wherein... Figure 17a This illustrates the process of initiating data collection from a terminal request. Figure 17b This illustrates the process of starting data collection from base station configuration.

[0141] exist Figure 17a In this context, it is assumed that the terminal can be configured with the following three models.

[0142] Model 1: BM - Case 1, Set A has 64 items, Set B has 8 items → Function ID #1 Model 2: BM - Case 1, Set A has 126 items, Set B has 8 items → Function ID #2 Model 3: BM - Case 1, Set A / B 64 → Function ID #3 Reference Figure 17a The terminal sends AI / ML related capability information to the base station, i.e., the NW (network) (S1701a), and thereby receives AI / ML related configuration from the NW (network) (S1702a). For example, the terminal sends capability information including function IDs #1 and #2, and the base station that receives this information sends AI / ML related configuration to the terminal based on the information received from the terminal.

[0143] The above process can be referred to as a function identification process or a model identification process. Here, the terminal is a terminal with the capability to perform AI / ML-based beam management. Its AI / ML-related capability information may include a functionality / model ID and related additional information, which is then sent to the base station. Based on the AI / ML-related capability information received from the terminal, the base station sends the Set A / B beam information (e.g., CSI-RS resource configuration) required for model training to the terminal, thereby allowing the terminal to receive at least one AI / ML-related configuration. Figure 17a The following AI / ML related configurations are shown in the image.

[0144] -Function ID #1: CSI-RS Resource Sets #1, #2 -Function ID #2: CSI-RS Resource Sets #3, #4 In this context, a data request message can indicate a request to activate / start a configuration for a configured function / model. (See also...) Figure 17a The terminal sends a data request for model training (S1703a) that includes information about a configuration in the configured function / model. As an example, the data request message for model training may include and send information about "Function ID #1: CSI-RS Resource Set #1".

[0145] Alternatively, it can be defined as follows: after the terminal sends a data request message, the base station sends a response message to the data request message, and the terminal expects to receive this response message. In this case, the base station sends an acknowledgment message (S1704a) as a response to the data request message sent by the terminal.

[0146] As an alternative, for terminals that have sent AI / ML related capability information to the base station but have not received relevant configuration information from the base station, a configuration message informing the terminal of resource information used for data transmission can be sent from the base station to the terminal in response to a data request message from that terminal. In this case, the data request message sent by the terminal may mean that the terminal is requesting configuration of transmission resources for data used for model training. (See also...) Figure 17b The terminal sends a data request message (S1703b) containing AI / ML related capability information to the base station (NW, network) for model training, and thereby receives a configuration message (S1704b) containing AI / ML related configuration from the NW. For example, the terminal sends capability information including function ID #1, and the base station that receives this information can send a configuration message including information about CSI-RS resource set #1 to the terminal based on the information received from the terminal.

[0147] Subsequently, the base station, i.e., the NW (network), activates the resources for data transmission used for model training. Specifically, it activates CSI-RS resource set #1 (S1705a, S1705b) and performs the corresponding data transmission (S1706a, S1706b). Here, the corresponding data transmission means repeatedly sending RS information about CSI-RS resource set #1. The terminal performs model training by collecting data sent from the base station, and when model training is complete, it indicates to the base station that model training has ended (S1707a, S1707b). In this case, the message indicating the end of model training may include the corresponding function ID information, i.e., information about function ID #1.

[0148] This specification further states that when the terminal performing model training satisfies the following events (1~4), it is decided to terminate the data collection for the model.

[0149] 1. During model training, determine when the model performance exceeds a certain threshold. 2. The situation where, after model training begins, the model performance does not exceed a certain threshold within a certain time period (Timer); 3. The base station terminates data provision to the terminal; and / or 4. The arrival of the data collection time period (timer) requested or configured by the terminal. When applying events 1 and 2 above, the performance threshold and the timer value applied to the time period / duration can be configured by the base station, and preferably configured in units of the indicated function / model received from the terminal. Here, the same or different parameters can be applied as thresholds for events 1 and 2. If different parameters are defined, the threshold defined for event 2 preferably has a smaller value than the threshold defined for event 1. Among the parameters mentioned above, the value corresponding to the timer can also replace the value contained in the terminal's data request message.

[0150] The aforementioned event parameters can be configured by the base station to the terminal during the function / model identification process between the terminal and the base station, or they can be configured as a response to the terminal sending a data request message via a confirm / configuration message received from the base station. Alternatively, they can be included in the configuration message if the base station sends a configuration message for data used for terminal model training without sending a data request message from the terminal.

[0151] Additionally, the timers defined in events 2 and 4 can be defined with the same or different parameters.

[0152] The timer mentioned above can be defined differently based on various factors that can measure the total amount of data, such as data volume / size / time instance. The application of other factors that have the same meaning as the amount of data / time required to train the entire model as described above can be considered as the same technique as that presented in this specification.

[0153] A terminal that decides to terminate model training based on event 1 or event 2 may send a message to the base station in order to inform the base station that the data transmission being provided to the terminal is no longer needed. This message may include at least one of the following information.

[0154] - Functionality / Model ID; - Indicator information that tells the reason why model training terminated.

[0155] Here, the reason for the termination of model training can be a successful model training message in event 1, or a failed model training message in event 2.

[0156] Alternatively, a model training termination message can also be defined as being sent from the base station to the terminal based on event 3 or event 4. In the case of event 4, the termination of model training can be implicitly identified without message transmission, and data collection and provisioning operations can be deactivated. Furthermore, the aforementioned termination message can be transmitted from the base station to the terminal or from the terminal to the base station as needed.

[0157] The following describes the operations of the terminal and base station that were trained using the data collection model based on events 1 to 4 above.

[0158] Figure 18 This specification illustrates a process for a terminal and base station for event-based model training according to an embodiment of this specification.

[0159] Figure 18 This illustrates the process between the terminal and the base station when at least Event 1 is configured.

[0160] Event 1: During model training, the model performance is determined to exceed a certain threshold. The following is for reference Figure 18 Describe the terminal operation in detail.

[0161] The terminal receives a message from the base station that includes a data transmission start notification for model training or data-related configuration information (S1802). The message may include event parameters (e.g., event 1) for terminating model training and / or related threshold parameters (e.g., threshold value). As an example, the configuration / confirmation message may include and receive information about function ID #1, event 1 information, and the threshold information for event 1.

[0162] Before the terminal receives a message including a notification that data transmission for model training has started or data-related configuration information, it may optionally send a message to the base station requesting activation of data transmission for model training (S1801). For example, the data request message for model training may include and send information about function ID #1.

[0163] Subsequently, the terminal receives data transmitted from the base station (S1804) and performs model training based on the received data. Here, receiving data means repeatedly receiving RSs about the activated CSI-RS resource set #1. While performing model training, the terminal can periodically evaluate the performance of the model. When the model's performance value reaches the threshold received from the base station, i.e., when event 1 is satisfied, the terminal decides to terminate model training.

[0164] When the decision is made to terminate model training, the terminal sends a message to the base station informing it that model training has been terminated (S1805). For example, the model training termination message may include and send information about function ID #1.

[0165] The following is for reference Figure 18 A detailed description of base station operation.

[0166] The base station sends a message to the terminal that includes a notification of the start of data transmission for model training or data-related configuration information (S1802). This message may include event parameters (e.g., event 1) for terminating model training and / or related threshold parameters (e.g., threshold value). As an example, the configuration / acknowledgment message may include and send information about function ID #1, event 1 information, and the threshold information for event 1.

[0167] Before the base station sends a message including a notification that data transmission for model training has started or data-related configuration information, it may optionally receive a message from the terminal requesting activation of data transmission for model training (S1801). For example, the data request message for model training may include and receive information about function ID #1.

[0168] Subsequently, the base station activates the data transmission resources used for model training. Specifically, it activates CSI-RS resource set #1 (S1803) and performs the corresponding data transmission (S1804). Here, the corresponding data transmission means repeatedly transmitting RS regarding the activated CSI-RS resource set #1. In the case of model training for beam management, beams can be periodically transmitted via the configured CSI-RS.

[0169] When it is decided to terminate the model training of the terminal, a message informing the terminal that the model training has been terminated is received (S1805). For example, the model training termination message may include and receive information about function ID #1. When the terminal receives the message informing it that the model training has been terminated, the base station deactivates the data transmission resources used for model training (S1806). That is, the base station deactivates CSI-RS resource set #1 used for model training and stops the corresponding data transmission.

[0170] Figure 19 This specification illustrates a process for a terminal and base station for event-based model training according to another embodiment of the present specification.

[0171] Figure 19 This illustrates the process of the terminal and base station when at least Event 2 is configured.

[0172] Event 2: After model training begins, the model performance does not exceed a certain threshold within a certain time period (Timer). Event 2 can be operated in the same way as Event 1 with a specific timer. If Event 1 is configured to operate with a timer, the performance is checked during the timer's operation to see if a threshold is met. If the threshold is met, model training can be terminated even before the timer expires, and when the timer expires, the terminal can notify the base station that model training has been terminated.

[0173] The following is for reference Figure 19 Describe the terminal operation in detail.

[0174] The terminal receives a message from the base station that includes a notification of the start of data transmission for model training or data-related configuration information (S1902). This message may include event parameters (e.g., event 2) for terminating model training, related timer parameters (e.g., timer value), and threshold parameters (e.g., threshold value). As an example, the configuration / acknowledgment message may include and receive information about function ID #1, event 2 information, timer information for event 2, and threshold information.

[0175] Before the terminal receives a message that includes a notification that data transmission for model training has started or data-related configuration information, it may optionally send a message to the base station requesting activation of data transmission for model training (S1901). For example, the data request message for model training may include and send information about function ID #1.

[0176] Subsequently, the terminal receives data transmitted from the base station (S1904) and performs model training based on the received data. Here, receiving data means repeatedly receiving RSs regarding the activated CSI-RS resource set #1.

[0177] The terminal starts the timer based on previously received timer information. The timer can be started either i) when it receives a message from the base station notifying it of the start of data transmission for model training or a message containing data-related configuration information, or ii) from the moment it receives the first data resource configured / sent from the base station.

[0178] During the timer's execution, the terminal periodically evaluates the model's performance. When the model's performance value reaches the previously received threshold (event 1), the terminal decides to terminate model training. Furthermore, if the threshold received from the base station is not reached during the timer's execution, and the timer expires (event 2), the terminal also decides to terminate model training.

[0179] When the decision is made to terminate model training, the terminal sends a message to the base station informing it that model training has been terminated (S1905). For example, a model training termination message including information about function ID #1 can be sent.

[0180] The following is for reference Figure 19 A detailed description of base station operation.

[0181] The base station sends a message to the terminal that includes a notification of the start of data transmission for model training or data-related configuration information (S1902). This message may include event parameters (e.g., event 2) for terminating model training, related timer parameters (e.g., timer value), and threshold parameters (e.g., threshold value). As an example, the configuration / acknowledgment message may include and send information about function ID #1, event 2 information, timer information for event 2, and threshold information.

[0182] Before the base station sends a message including a notification that data transmission for model training has started or data-related configuration information, it may optionally receive a message from the terminal requesting activation of data transmission for model training (S1901). For example, the data request message for model training may include and receive information about function ID #1.

[0183] Subsequently, the base station activates the data transmission resources used for model training. Specifically, it activates CSI-RS resource set #1 (S1903) and performs the corresponding data transmission (S1904). Here, the corresponding data transmission means repeatedly transmitting RS regarding the activated CSI-RS resource set #1. In the case of model training for beam management, beams can be periodically transmitted via the configured CSI-RS.

[0184] When it is decided to terminate the model training of the terminal, a message informing the terminal that the model training has been terminated is received (S1905). For example, the model training termination message may include and receive information about function ID #1. When the terminal receives the message informing it that the model training has been terminated, the base station deactivates the data transmission resources used for model training (S1906). That is, the base station deactivates the CSI-RS resource set #1 used for model training and stops the corresponding data transmission.

[0185] Figure 20 This specification illustrates a process for a terminal and base station for event-based model training according to another embodiment of the present specification.

[0186] Figure 20 This illustrates the process between the terminal and the base station when at least Event 3 is configured.

[0187] -Event 3: Situation where the base station terminates data supply to the terminal The following is for reference Figure 20 Describe the terminal operation in detail.

[0188] The terminal receives a message from the base station that includes a notification of the start of data transmission for model training or data-related configuration information (S2002). This message may include event parameters (e.g., event 1, event 2, and / or event 3) for terminating model training, relevant timer parameters (e.g., timer values), and threshold parameters (e.g., threshold values). As an example, the configuration / acknowledgment message may include and receive information about function ID #1, event 1 / 2 / 3 information, timer information for event 2, and threshold information.

[0189] Before the terminal receives a message that includes a notification that data transmission for model training has started or data-related configuration information, it may optionally send a message to the base station requesting activation of data transmission for model training (S2001). For example, the data request message for model training may include and send information about function ID #1.

[0190] Subsequently, the terminal receives data transmitted from the base station (S2004) and performs model training based on the received data. Here, receiving data means repeatedly receiving RSs related to the activated CSI-RS resource set #1.

[0191] The terminal starts the timer based on previously received timer information. The timer can be started either i) when it receives a message from the base station notifying it of the start of data transmission for model training or a message containing data-related configuration information, or ii) from the moment it receives the first data resource configured / sent from the base station.

[0192] During the timer's execution, the terminal periodically evaluates the model's performance. When the model's performance value reaches a previously received threshold, satisfying event 1, the terminal decides to terminate model training. Similarly, during the timer's execution, when event 3 is satisfied due to receiving a message from the base station informing that the resource configuration for data collection sent for the terminal has been released / deactivated, the terminal also decides to terminate model training. For example, the terminal decides to terminate model training by receiving a data transmission deactivation message (S2006) from the base station that includes function ID #1 information.

[0193] Furthermore, if the threshold received from the base station is not reached during the timer's operation, and event 2 is satisfied when the timer expires, the terminal also decides to terminate model training.

[0194] The following is for reference Figure 20 A detailed description of base station operation.

[0195] The base station sends a message to the terminal that includes a notification of the start of data transmission for model training or data-related configuration information (S2002). This message may include event parameters for terminating model training (e.g., event 1, event 2, and / or event 3), relevant timer parameters (e.g., timer values), and threshold parameters (e.g., threshold values). As an example, the configuration / acknowledgment message may include and send information about function ID #1, event 1 / 2 / 3 information, timer information for event 2, and threshold information.

[0196] Before the base station sends a message including a notification that data transmission for model training has started or data-related configuration information, it may optionally receive a message from the terminal requesting activation of data transmission for model training (S2001). For example, the data request message for model training may include and receive information about function ID #1.

[0197] Subsequently, the base station activates the data transmission resources used for model training. Specifically, it activates CSI-RS resource set #1 (S2003) and performs the corresponding data transmission (S2004). Here, the corresponding data transmission means repeatedly sending RS messages about the activated CSI-RS resource set #1. In the case of model training for beam management, beams can be periodically transmitted via the configured CSI-RS.

[0198] If the base station decides to release / deactivate the CSI-RS resources configured for terminal model training, it sends a message to the terminal indicating the termination of model training (or a message informing the terminal of the release / deactivation of data-related resources) and stops the requested data transmission. For example, the base station decides to deactivate CSI-RS resource set #1 (S2005) and sends a data transmission deactivation message to the terminal including function ID #1 information (S2006).

[0199] Figure 21 This specification illustrates a process for a terminal and base station for event-based model training according to another embodiment of the present specification.

[0200] Figure 21 This illustrates the process between the terminal and the base station when Event 4 is configured.

[0201] Event 4: Cases during data collection (timer) when a terminal request or configuration is received. Event 4 can be applied independently without the presence of Events 1 through 3. That is, the time duration is set by the terminal or base station, and model training is performed only within that time period.

[0202] The following is for reference Figure 21 Describe the terminal operation in detail.

[0203] The terminal sends a message to the base station requesting activation of data transmission for model training (S2101). For example, the data request message for model training may include and send information about function ID #1 and the time information (e.g., timer) required for the corresponding model training.

[0204] Subsequently, the terminal receives a message from the base station that includes a notification of the start of data transmission for model training or data-related configuration information (S2102). This message may include timing information for continuous data transmission (e.g., a timer). As an example, the configuration / acknowledgment message may include and receive information about function ID #1 and timer information for event 4.

[0205] The timer value can be included in a message sent from the terminal to the base station or from the base station to the terminal. If both messages contain a timer value, the timer value included in the message sent from the base station to the terminal is preferred.

[0206] Subsequently, the terminal receives data transmitted from the base station (S2104) and performs model training based on the received data. Here, receiving data means repeatedly receiving RSs related to the activated CSI-RS resource set #1.

[0207] The terminal starts the aforementioned timer. The timer can be started i) when it receives a message from the base station notifying it of the start of data transmission for model training or a message containing data-related configuration information, or ii) from the moment it receives the first data resource configured / sent from the base station.

[0208] During the timer's operation, the terminal performs training on the model. When the timer expires and event 4 is satisfied, the terminal decides to terminate the model training (S2105).

[0209] The following is for reference Figure 21 A detailed description of base station operation.

[0210] The base station receives a message from the terminal requesting activation of data transmission for model training (S2101). For example, the data request message for model training may include and receive information about function ID #1 and the time information (e.g., timer) required for the corresponding model training.

[0211] Subsequently, the base station sends a message to the terminal including a notification that data transmission for model training has begun or data-related configuration information (S2102). This message may include timing information for continuous data transmission (e.g., a timer). As an example, the configuration / acknowledgment message may include and send information about function ID #1 and timer information for event 4.

[0212] The timer value can be included in a message sent from the terminal to the base station or from the base station to the terminal. If both messages contain a timer value, the timer value included in the message sent from the base station to the terminal is preferred.

[0213] Subsequently, the base station activates the data transmission resources used for model training. That is, it activates CSI-RS resource set #1 (S2103) and performs the corresponding data transmission (S2104). Here, the corresponding data transmission means repeatedly sending RS about the activated CSI-RS resource set #1. In the case of model training for beam management, beams can be periodically transmitted through the configured CSI-RS.

[0214] The base station starts the aforementioned timer. The timer can be started i) when a message is sent to the terminal to notify the start of data transmission for model training or when data-related configuration information is sent, or ii) from the moment the first data resource configured by the base station is sent.

[0215] During the timer's operation, the base station repeatedly (or periodically) transmits the data according to its configuration. When the timer expires, i.e., event 4 is met, the base station decides to terminate data transmission. That is, the base station performs deactivation of CSI-RS resource set #1 (S2106).

[0216] Figures 22 to 23 This illustrates the process of terminating model training in a terminal and base station according to an embodiment of this specification.

[0217] Figure 22 This illustrates the process of model training termination caused by the terminal. Figure 23 This illustrates the process of model training termination caused by the base station.

[0218] Reference Figures 22 to 23 The terminal sends a message to the base station requesting activation of data transmission for model training (S2201, S2301). For example, the data request message for model training may include and send information about function ID #1 and its corresponding time period information (e.g., timer).

[0219] Subsequently, the base station sends a message to the terminal that includes a notification that data transmission for model training has begun or data-related configuration information (S2202, S2302). As an example, the configuration / acknowledgment message may include and send information about function ID #1.

[0220] The base station activates the data transmission resources used for model training, i.e., activates CSI-RS resource set #1 (S2203, S2303), and performs the corresponding data transmission (S2204, S2304). Here, the corresponding data transmission means repeatedly sending RS about the activated CSI-RS resource set #1. In the case of model training for beam management, beams can be periodically transmitted through the configured CSI-RS.

[0221] The terminal described in this specification, when sending a data request message to a base station to collect data for model training, may send the amount / size of the data it wishes to collect or the minimum time required to collect that data. Furthermore, even without the configuration regarding the aforementioned events described in this specification, model training can be terminated based on the time or amount of data contained in the data request message sent by the terminal to the base station (or by including the time or amount of data in the configuration message from the base station informing the terminal of the start of data transmission).

[0222] First, the base station identifies the information received from the terminal as one of the terminal's assistance information and can continuously provide data until the time or amount of data received from the terminal is satisfied. However, this may mean that, depending on the base station's implementation, data transmission may be terminated by the base station at any time even if the aforementioned information is not satisfied. That is, as... Figures 22 to 23As shown, either the terminal that starts model training or the base station that starts data transmission can determine, based on their respective internal decisions, that model training will be terminated (S2205, S2305). Afterwards, the terminal can send a model training termination message to the base station (S2206), or the base station can inform the terminal that data transmission will be stopped (S2306). The base station that receives the model training termination message from the terminal performs the deactivation of the corresponding data transmission resources (S2207), while the terminal that receives the data release / deactivation message from the base station performs model training termination (S2308). The base station that sends the data release / deactivation message to the terminal then performs the deactivation of the corresponding data transmission resources (S2307).

[0223] Secondly, if data transmission for model training begins, the base station should provide data to the terminal at least before the time or amount of data received from (or sent to) the terminal is satisfied, in order to provide the data requested by the terminal. That is, the base station and the terminal should perform data transmission / collection before the time or amount of data is satisfied, and after the time or amount of data is satisfied, model training can be terminated via a model training termination message or a data release / deactivation message, depending on the implementation of the base station or the terminal. Alternatively, the base station and the terminal may also identify the time or amount of data configured from the terminal as the termination point for data collection and provision, meaning that model training must be terminated if the configured time or amount of data is reached. This indicates that the time duration / instance or data size / volume sent by the terminal or received from the base station can be used as a numerical value to determine the time and can be used in the same way as the timers defined in the operation based on the configured events presented in this specification.

[0224] The following describes the use of time instances instead of the timers defined in this specification.

[0225] A time instance can be defined as a count parameter (COUNT). The terminal or base station can include a maximum count value (MAX_CNT) in the data request or data-related configuration message. This serves the same purpose as the timer mentioned earlier. First, consider the case where MAX_CNT replaces the timer in the aforementioned events. The terminal performs model training while periodically receiving datasets from the base station. Each time the terminal receives a periodic dataset, it increments the COUNT value by 1. If the COUNT value has not reached the MAX_CNT value, the terminal can evaluate the model's performance. If the performance exceeds a threshold, it sends a model training termination message to the base station and terminates model training. For events 2, 3, and 4, the execution can also be replaced by checking if the COUNT value reaches MAX. Second, for cases where there are no events, but the terminal sends a timer to the base station or the base station sends a timer to the terminal, the MAX_CNT value can be set instead of the timer value, thus defining the same operation.

[0226] Alternatively, data request / termination messages can be defined using PHY (Physical Layer), MAC (Media Access Control), CE (Control Unit), or RRC (Radio Resource Control) messages. If defined via PHY, a data request PUCCH can be defined for each function / model of the terminal, and the same PUCCH can also be used when the terminal sends a termination signal. This is preferably applicable to situations where training occurs frequently, such as online learning, and in the case of PUCCH definition, the parameter information presented in this specification is preferably configured in advance processes such as function / model identification.

[0227] The contents disclosed in this specification can be used independently or in any combination. Furthermore, although this specification is based on a 5G NR system, it is unrelated to any specific wireless communication technology, and all situations where the concepts applicable to this specification are included within its scope.

[0228] Figure 24 An apparatus according to one embodiment of this specification is shown.

[0229] Reference Figure 24 The wireless communication system may include a first device 100a and a second device 100b.

[0230] The first device 100a may be a base station, network node, transmission terminal, receiving terminal, wireless device, wireless communication equipment, vehicle, vehicle equipped with autonomous driving function, connected car, unmanned aerial vehicle (UAV), AI (Artificial Intelligence) module, robot, AR (Augmented Reality) device, VR (Virtual Reality) device, MR (Mixed Reality) device, holographic device, public safety device, MTC device, IoT device, medical device, fintech device (or financial device), security device, climate / environment device, 5G service related device, or other devices related to the Fourth Industrial Revolution.

[0231] The second device 100b can be a base station, network node, transmission terminal, receiving terminal, wireless device, wireless communication equipment, vehicle, vehicle equipped with autonomous driving function, connected car, unmanned aerial vehicle (UAV), AI (Artificial Intelligence) module, robot, AR (Augmented Reality) device, VR (Virtual Reality) device, MR (Mixed Reality) device, holographic device, public safety device, MTC device, IoT device, medical device, fintech device (or financial device), security device, climate / environment device, 5G service related device, or other devices related to the Fourth Industrial Revolution.

[0232] The first device 100a may include: at least one processor, such as processor 1020a; at least one memory, such as memory 1010a; and at least one transceiver, such as transceiver 1031a. The processor 1020a may perform the aforementioned functions, steps, and / or methods. The processor 1020a may execute more than one protocol. For example, the processor 1020a may execute more than one layer of a wireless interface protocol. The memory 1010a may be connected to the processor 1020a and store various forms of information and / or instructions. The transceiver 1031a may be connected to the processor 1020a and control the transmission and reception of wireless signals.

[0233] The second device 100b may include at least one processor, such as processor 1020b; at least one or more memory devices, such as memory 1010b; and at least one transceiver, such as transceiver 1031b. The processor 1020b may perform the aforementioned functions, steps, and / or methods. The processor 1020b may implement more than one protocol. For example, the processor 1020b may implement more than one layer of a wireless interface protocol. The memory 1010b may be connected to the processor 1020b and store various forms of information and / or instructions. The transceiver 1031b may be connected to the processor 1020b and control the transmission and reception of wireless signals.

[0234] The memory 1010a and / or the memory 1010b can be connected internally or externally to the processor 1020a and / or the processor 1020b, or they can be connected to other processors via various technologies such as wired or wireless connections.

[0235] The first device 100a and / or the second device 100b may have more than one antenna. For example, antenna 1036a and / or antenna 1036b may be configured to transmit and receive wireless signals.

[0236] Figure 25 A block diagram illustrating the configuration of a terminal according to an embodiment of this specification.

[0237] in particular Figure 25 This is to show the foregoing in more detail. Figure 24 A diagram of the device.

[0238] The device includes a memory 1010, a processor 1020, a transceiver unit 1031, a power management module 1091, a battery 1092, a display device 1041, an input unit 1053, a speaker 1042 and a microphone 1052, a SIM (subscriber identification module) card, and one or more antennas.

[0239] Processor 1020 can be configured to implement the functions, steps, and / or methods described and presented in this specification. The radio interface protocol layer can be implemented in processor 1020. Processor 1020 may include an ASIC (application-specific integrated circuit), other chipsets, logic circuits, and / or data processing devices. Processor 1020 can be an application processor (AP). Processor 1020 may include at least one of a DSP (digital signal processor), CPU (central processing unit), GPU (graphics processing unit), and modem (modulator and demodulator). Processor 1020 can, for example, be a SNAPDRAGON™ series processor manufactured by Qualcomm®, an EXYNOS™ series processor manufactured by Samsung®, an A-series processor manufactured by Apple®, a HELIO™ series processor manufactured by MediaTek®, an ATOM™ series processor manufactured by Intel®, a KIRINT™ series processor manufactured by HiSilicon®, or a corresponding next-generation processor.

[0240] The power management module 1091 manages the power of the processor 1020 and / or the transceiver unit 1031. The battery 1092 supplies power to the power management module 1091. The display device 1041 outputs the results processed by the processor 1020. The input unit 1053 receives inputs to be used by the processor 1020. The input unit 1053 can be displayed on the display device 1041. A SIM card is an integrated circuit used to identify and authenticate the IMSI (International Mobile Subscriber Identity) of network users in mobile devices such as mobile phones and computers, and to securely store the associated keys. Phonebook information can also be stored in many SIM cards.

[0241] Memory 1010 is operatively integrated with processor 1020 to store various information for enabling processor 1010 to operate. Memory 1010 may include ROM (read-only memory), RAM (random access memory), flash memory, memory card, storage medium, and / or other storage devices. When embodiments are implemented in software, the techniques described herein can be implemented as modules (e.g., steps, functions, etc.) that perform the functions described herein. Modules may be stored in memory 1010 and operated by processor 1020. Memory 1010 may be implemented internally to processor 1020. Alternatively, memory 1010 may be implemented externally to processor 1020 and may be communicatively connected to processor 1020 by various means known in the art.

[0242] Transceiver unit 1031 is operatively coupled to processor 1020 to transmit and / or receive wireless signals. Transceiver unit 1031 includes a transmitter and a receiver. Transceiver unit 1031 may include baseband circuitry for processing wireless frequency signals. The transceiver unit controls one or more wires to transmit and / or receive wireless signals. To initiate communication, processor 1020 sends instruction information to transceiver unit 1031 to transmit wireless signals, such as those constituting voice communication data. An antenna performs the function of transmitting and receiving wireless signals. When receiving wireless signals, transceiver unit 1031 can transmit signals and convert them into baseband for processing by processor 1020. The processed signals can be converted into auditory or visual information output through speaker 1042.

[0243] Speaker 1042 outputs the sound-related results processed by processor 1020. Microphone 1052 receives the sound-related inputs that processor 1020 will use.

[0244] Users may input command information such as a phone number by pressing a button (or touching) on ​​the input unit 1053 or through voice activation based on the microphone 1052. The processor 1020 receives and processes this command information to perform appropriate functions such as making a phone call. Operational data can be retrieved from the SIM card or memory 1010. Additionally, the processor 1020 can display the command or operation information on the display device 1041 for user convenience.

[0245] Figure 26 A block diagram illustrating the configuration of the processor disclosed in this specification is shown.

[0246] Reference Figure 26As can be seen, implementing the processor 1020 disclosed in this specification may include multiple circuits to implement the functions, steps, and / or methods described and proposed in this specification. For example, the processor 1020 may include a first circuit 1020-1, a second circuit 1020-2, and a third circuit 1020-3. Additionally, although not shown, the processor 1020 may include more circuits. Each circuit may include multiple transistors.

[0247] The processor 1020 may also be referred to as an ASIC (application-specific integrated circuit) or an AP (application processor), and may include at least one of a DSP (digital signal processor), a CPU (central processing unit), or a GPU (graphics processing unit).

[0248] Figure 27 To show in detail Figure 24 The transceiver or transceiver of the first device shown Figure 25 Block diagram of the transceiver section of the device shown.

[0249] Reference Figure 27 The transceiver unit 1031 includes a transmitter 1031-1 and a receiver 1031-2. The transmitter 1031-1 includes a Discrete Fourier Transform (DFT) unit 1031-11, a subcarrier mapper 1031-12, an IFFT unit 1031-13, a CP insertion unit 1031-14, and a wireless transmission unit 1031-15. The transmitter 1031-1 may also include a modulator. Additionally, it may include, for example, a scramble unit (not shown), a modulation mapper (not shown), a layer mapper (not shown), and a layer permutator (not shown), which may be configured before the DFT unit 1031-11. That is, to prevent an increase in the peak-to-average power ratio (PAPR), the transmitter 1031-1 first passes the information through the DFT 1031-11 before mapping the signal to the subcarriers. The signal spread (or synonymously precoded) by the DFT section 1031-11 is subcarrier mapped by the subcarrier mapper 1031-12, and then formed into a signal on the time axis by the IFFT (Inverse Fast Fourier Transform) section 1031-13.

[0250] The DFT unit 1031-11 performs a DFT on the input symbols to output complex-valued symbols. For example, if the input symbols are Ntx (but Ntx is a natural number), the DFT size is Ntx. The DFT unit 1031-11 can be called a transform precoder. The subcarrier mapper 1031-12 maps the complex symbols to subcarriers in the frequency domain. The complex symbols can be mapped to resource elements corresponding to resource blocks allocated for data transmission. The subcarrier mapper 1031-12 can be called a resource element mapper. The IFFT unit 1031-13 performs an IFFT on the input symbols to output a baseband signal for data as a time-domain signal. The CP insertion unit 1031-14 copies the latter part of the baseband signal for data and inserts it into the former part of the baseband signal for data. By inserting CP, ISI (Inter-Symbol Interference) and ICI (Inter-Carrier Interference) can be prevented, and orthogonality can be maintained in multiplexed channels.

[0251] On the other hand, receiver 1031-2 includes a wireless receiving unit 1031-21, a CP removal unit 1031-22, an FFT unit 1031-23, and an equalization unit 1031-24. The wireless receiving unit 1031-21, CP removal unit 1031-22, and FFT unit 1031-23 of receiver 1031-2 perform the opposite functions of the wireless transmitting unit 1031-15, CP insertion unit 1031-14, and IFF unit 1031-13 in transmitter 1031-1. Receiver 1031-2 may also include a demodulator.

[0252] The preferred embodiments have been described above by way of example, but the disclosure of this specification is not limited to these specific embodiments. Therefore, it can be modified, altered or improved into various forms within the scope of the spirit and claims of this specification.

[0253] In the exemplary system described above, the method is described as a series of steps or blocks based on the flowchart, but is not limited to the order of the described steps. Some steps may be performed in a different order or simultaneously. Furthermore, those skilled in the art will understand that the steps shown in the flowchart are not exclusive, and may include other steps or may delete one or more steps from the flowchart without affecting the scope of the claims.

[0254] The claims described in this specification can be combined in various ways. For example, the technical features of the method claims described in this specification can be combined and implemented by an apparatus, and the technical features of the apparatus claims described in this specification can be combined and implemented by a method. Furthermore, the technical features of the method claims and the apparatus claims described in this specification can be combined and implemented by an apparatus, and the technical features of the method claims and the apparatus claims described in this specification can be combined and implemented by a method.

Claims

1. A method for operating a terminal in a wireless communication system, comprising: The steps for receiving data for training AI / ML (Artificial Intelligence / Machine Learning) models; The steps of training the AI / ML model are performed based on the data; as well as The step of sending a first message indicating the end of data collection after performing the training of the AI / ML model.

2. The method for operating a terminal in a wireless communication system according to claim 1, characterized in that, The method further includes the step of receiving configuration information for receiving the data, wherein the configuration information includes event information associated with the end of the data collection, and the terminal sends the first message based on the event information.

3. The method for operating a terminal in a wireless communication system according to claim 1, characterized in that, The event information includes at least one of threshold information associated with the performance of the AI / ML model and time information for terminating the training of the AI / ML model.

4. The method for operating a terminal in a wireless communication system according to claim 1, characterized in that, The first message is used for data transmission deactivation.

5. The method for operating a terminal in a wireless communication system according to claim 2, characterized in that, The method further includes the step of sending a second message requesting the transmission of the data; wherein... The configuration information is based on the second message received.

6. The method for operating a terminal in a wireless communication system according to claim 5, characterized in that, The second message includes at least one of time information and data volume information for terminating the training of the AI / ML model.

7. A method for operating a base station in a wireless communication system, comprising: The steps for sending data for training AI / ML (Artificial Intelligence / Machine Learning) models; as well as The step of receiving a first message indicating the end of data collection after performing the AI / ML model training.

8. The method according to claim 7, characterized in that, The method further includes the step of sending configuration information for the data transmission, wherein the configuration information includes event information associated with the end of the data collection, and the first message is received based on the event information.

9. The method for operating a base station in a wireless communication system according to claim 7, characterized in that, The event information includes at least one of threshold information associated with the performance of the AI / ML model and time information for terminating the training of the AI / ML model.

10. The method for operating a base station in a wireless communication system according to claim 7, characterized in that, The first message is used for data transmission deactivation.

11. The method for operating a base station in a wireless communication system according to claim 8, characterized in that, The method further includes the step of receiving a second message requesting the transmission of the data.

12. The method for operating a base station in a wireless communication system according to claim 11, characterized in that, The second message includes at least one of time information and data volume information for terminating the training of the AI / ML model.

13. A communication device in a wireless communication system, comprising: At least one processor; as well as At least one memory that stores instructions and is operatively electrically connected to the at least one processor, wherein, when executed by the at least one processor, the instructions cause the communication device to perform the following operations: Receive data for training AI / ML (Artificial Intelligence / Machine Learning) models; Based on the data, perform the training of the AI / ML model; and After the AI / ML model training is performed, a first message indicating the end of data collection is sent.

14. The communication device in the wireless communication system according to claim 13, characterized in that, When the instructions are executed by the at least one processor, the communication device also performs the following operations: Receive configuration information for receiving the data. The configuration information includes event information associated with the end of the data collection, and the first message is sent based on the event information.

15. The communication device in the wireless communication system according to claim 13, characterized in that, The event information includes at least one of threshold information associated with the performance of the AI / ML model and time information for terminating the training of the AI / ML model.

16. The communication device in the wireless communication system according to claim 13, characterized in that, The first message is used for data transmission deactivation.

17. The communication device in the wireless communication system according to claim 14, characterized in that, When the instructions are executed by the at least one processor, the communication device also performs the following operations: Send a second message requesting the transmission of the data. The configuration information is based on the second message received.

18. The communication device in the wireless communication system according to claim 17, characterized in that, The second message includes at least one of time information and data volume information for terminating the training of the AI / ML model.

Images