Joint learning

Abstract

Disclosed in the present specification is a communication method for an NF related to AI and ML. The method may comprise the steps of: receiving a request message please support information from an AI / ML AS; sending a subscription request message requesting to subscribe to an event related to the MA PDU session information to the SMF; receiving MA PDU session information from the SMF; and determining support information.

Description

Technical Field

[0001] This specification relates to mobile communications. Background Art

[0002] The 3rd Generation Partnership Project (3GPP) Long Term Evolution (LTE) is a technology for implementing high-speed packet communications. Many solutions have been proposed for the LTE goals, including those aimed at reducing user and provider costs, improving service quality, and expanding and improving coverage and system capacity. 3GPP LTE requires lower cost per bit, increased service availability, flexible use of frequency bands, simple structure, open interfaces, and appropriate power consumption of terminals as high-level requirements.

[0003] Work has begun in the International Telecommunication Union (ITU) and 3GPP to develop requirements and specifications for new radio (NR) systems. 3GPP must identify and develop the technical components required to successfully standardize the new RAT to meet both the urgent market needs and the longer-term requirements set forth by the ITU Radiocommunication Sector (ITU-R) International Mobile Telecommunications (IMT)-2020 process in a timely manner. In addition, NR should be able to use any frequency band in the range of at least up to 100 GHz, which can be used for wireless communications even in the more distant future.

[0004] The goal of NR is to be a single technology framework that addresses all use cases, requirements, and deployment scenarios, including enhanced mobile broadband (eMBB), massive machine type communications (mMTC), ultra-reliable and low-latency communications (URLLC), etc. NR should be forward compatible in nature.

[0005] There is an ongoing discussion about Federated Learning (FL). However, until now, there has been no efficient way to perform the behaviors involved in FL. Summary of the invention

[0006] Technical Solution

[0007] In one aspect, a method for performing communication with an NF associated with an AI and a ML is provided. The method may include: receiving a request message requesting support information from an AI / ML AS; sending a request message requesting to subscribe to an event associated with MA PDU session information to an SMF; receiving the MA PDU session information from the SMF; and determining the support information based on the MA PDU session information.

[0008] In another aspect, a device for implementing the above method is provided.

[0009] In one aspect, a method for performing communication by a UE is provided. The method may include: sending a PDU session establishment request message for establishing a PDU to an SMF; and receiving a PDU session establishment accept message from the SMF.

[0010] In another aspect, a device for implementing the above method is provided. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 An example of a communication system to which an implementation of the present disclosure is applied is shown.

[0012] Figure 2 An example of a wireless device to which an implementation of the present disclosure is applied is shown.

[0013] Figure 3 An example of a UE to which an implementation of the present disclosure is applied is shown.

[0014] Figure 4 An example of a 5G system architecture to which implementations of the present disclosure are applied is shown.

[0015] Figure 5 and Figure 6 An example of a PDU session establishment process to which an implementation of the present disclosure is applied is shown.

[0016] Figure 7 An example of generating an MA PDU session is shown.

[0017] Figure 8 An example of applying ATSSS rules to an MA PDU session is shown.

[0018] Fig. 9 An example of split AI / ML according to the disclosed embodiments of the present specification is shown.

[0019] Fig.10 An example of AI / ML model deployment according to the disclosed implementation of this specification is shown.

[0020] Fig.11 An example of AI / ML model deployment according to the disclosed implementation of this specification is shown.

[0021] Fig.12 is an example of a process for providing 5GS auxiliary information for joint learning behavior according to an embodiment of the present disclosure.

[0022] Fig.13a and Fig.13b is an example process for establishing a VN group by an AF according to one embodiment of the present disclosure.

[0023] Fig.14is an example of the structure of an MA PDU session according to one embodiment of the present disclosure.

[0024] Fig.15a and Fig.15b An example of a process for providing 5GS support information for federated learning operations is illustrated.

[0025] Fig.16 An example of a process for providing support information based on MA PDU session information according to one embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0026] The following techniques, devices and systems can be applied to various wireless multiple access systems. Examples of multiple access systems include code division multiple access (CDMA) systems, frequency division multiple access (FDMA) systems, time division multiple access (TDMA) systems, orthogonal frequency division multiple access (OFDMA) systems, single carrier frequency division multiple access (SC-FDMA) systems, and multi-carrier frequency division multiple access (MC-FDMA) systems. CDMA can be implemented by radio technologies such as Universal Terrestrial Radio Access (UTRA) or CDMA2000. TDMA can be implemented by radio technologies such as Global System for Mobile Communications (GSM), General Packet Radio Service (GPRS), or Enhanced Data Rate for GSM Evolution (EDGE). OFDMA can be implemented by radio technologies such as Institute of Electrical and Electronics Engineers (IEEE) 802.11 (Wi-Fi), IEEE 802.16 (WiMAX), IEEE 802.20, or Evolved UTRA (E-UTRA). UTRA is part of Universal Mobile Telecommunications System (UMTS). 3rd Generation Partnership Project (3GPP) Long Term Evolution (LTE) is part of Evolved UMTS (E-UMTS) using E-UTRA. 3GPP LTE adopts OFDMA in the downlink (DL) and SC-FDMA in the uplink (UL). Evolutions of 3GPP LTE include LTE-Advanced (LTE-A), LTE-A Pro and / or 5G New Radio (NR).

[0027] For ease of description, the implementation of the present disclosure is mainly described with respect to a 3GPP-based wireless communication system. However, the technical features of the present disclosure are not limited thereto. For example, although the following detailed description is given based on a mobile communication system corresponding to a 3GPP-based wireless communication system, aspects of the present disclosure that are not limited to a 3GPP-based wireless communication system are applicable to other mobile communication systems.

[0028] For terms and techniques not specifically described among the terms and techniques used in the present disclosure, reference may be made to wireless communication standard documents published prior to the present disclosure.

[0029] In the present disclosure, "A or B" may mean "only A", "only B", or "both A and B". In other words, "A or B" in the present disclosure may be interpreted as "A and / or B". For example, "A, B, or C" in the present disclosure may mean "only A", "only B", "only C", or "any combination of A, B, and C".

[0030] In the present disclosure, a slash ( / ) or a comma (,) may mean "and / or". For example, "A / B" may mean "A and / or B". Thus, "A / B" may mean "only A", "only B", or "both A and B". For example, "A, B, C" may mean "A, B, or C".

[0031] In the present disclosure, "at least one of A and B" may mean "only A", "only B", or "both A and B". In addition, the expression "at least one of A or B" or "at least one of A and / or B" in the present disclosure may be interpreted as being the same as "at least one of A and B".

[0032] In addition, in the present disclosure, "at least one of A, B, and C" may mean "only A", "only B", "only C", or "any combination of A, B, and C". In addition, "at least one of A, B, or C" or "at least one of A, B and / or C" may mean "at least one of A, B, and C".

[0033] In addition, the brackets used in the present disclosure may mean "for example". Specifically, when it is shown as "control information (PDCCH)", "PDCCH" may be proposed as an example of "control information". In other words, the "control information" in the present disclosure is not limited to "PDCCH", and "PDCCH" may be proposed as an example of "control information". In addition, even when it is shown as "control information (ie, PDCCH)", "PDCCH" may be proposed as an example of "control information".

[0034] The technical features described separately in one drawing of the present disclosure may be implemented separately or simultaneously.

[0035] Although not limited thereto, the various descriptions, functions, processes, suggestions, methods and / or operational flowcharts of the present disclosure disclosed herein may be applied to various fields requiring wireless communication and / or connection between devices (e.g., 5G).

[0036] Hereinafter, the present disclosure will be described in more detail with reference to the accompanying drawings. Unless otherwise specified, the same reference numerals in the following drawings and / or descriptions may refer to the same and / or corresponding hardware blocks, software blocks and / or functional blocks.

[0037] Figure 1 An example of a communication system to which an implementation of the present disclosure is applied is shown.

[0038] Figure 1 The 5G usage scenarios shown in are only exemplary, and the technical features of the present disclosure can be applied to Figure 1 Other 5G usage scenarios not shown.

[0039] The three main requirement categories for 5G include (1) enhanced mobile broadband (eMBB) category, (2) massive machine type communication (mMTC) category, and (3) ultra-reliable and low-latency communication (URLLC) category.

[0040] Reference Figure 1 , the communication system 1 includes wireless devices 100a to 100f, a base station (BS) 200, and a network 300. Figure 1 A 5G network is shown as an example of a network of the communication system 1, but implementations of the present disclosure are not limited to the 5G system and may be applied to future communication systems other than the 5G system.

[0041] The BS 200 and the network 300 may be implemented as wireless devices, and a specific wireless device may operate as a BS / network node relative to other wireless devices.

[0042] The wireless devices 100a to 100f represent devices that perform communication using a radio access technology (RAT) (e.g., 5G NR or LTE) and may be referred to as communication / radio / 5G devices. The wireless devices may include, but are not limited to, a robot 100a, vehicles 100b-1 and 100b-2, an extended reality (XR) device 100c, a handheld device 100d, a home appliance 100e, an Internet of Things (IoT) device 100f, and an artificial intelligence (AI) device / server 400. For example, a vehicle may include a vehicle with a wireless communication function, an autonomous driving vehicle, and a vehicle capable of performing communication between vehicles. A vehicle may include an unmanned aerial vehicle (UAV) (e.g., a drone). The XR device may include an augmented reality (AR) / virtual reality (VR) / mixed reality (MR) device and may be implemented in the form of a head mounted device (HMD), a head up display (HUD) installed in a vehicle, a television, a smart phone, a computer, a wearable device, a home appliance device, a digital signage, a vehicle, a robot, and the like. Handheld devices may include smart phones, smart boards, wearable devices (e.g., smart watches or smart glasses), and computers (e.g., notebooks). Home appliances may include TVs, refrigerators, and washing machines. IoT devices may include sensors and smart meters.

[0043] In the present disclosure, the wireless devices 100a to 100f may be referred to as user equipment (UE). The UE may include, for example, a cellular phone, a smart phone, a laptop computer, a digital broadcast terminal, a personal digital assistant (PDA), a portable multimedia player (PMP), a navigation system, a tablet personal computer (PC), a tablet PC, an ultrabook, a vehicle, a vehicle with an autonomous driving function, a connected car, a UAV, an AI module, a robot, an AR device, a VR device, an MR device, a holographic device, a public safety device, an MTC device, an IoT device, a medical device, a FinTech device (or a financial device), a security device, a weather / environmental device, a device related to 5G services, or a device related to the field of the fourth industrial revolution.

[0044] The wireless devices 100a to 100f may be connected to the network 300 via the BS 200. AI technology may be applied to the wireless devices 100a to 100f, and the wireless devices 100a to 100f may be connected to the AI ​​server 400 via the network 300. The network 300 may be configured using a 3G network, a 4G (e.g., LTE) network, a 5G (e.g., NR) network, and a super 5G network. Although the wireless devices 100a to 100f may communicate with each other via the BS 200 / network 300, the wireless devices 100a to 100f may perform direct communication (e.g., side link communication) with each other without passing through the BS 200 / network 300. For example, the vehicles 100b-1 and 100b-2 may perform direct communication (e.g., vehicle-to-vehicle (V2V) / vehicle-to-everything (V2X) communication). An IoT device (e.g., a sensor) may perform direct communication with other IoT devices (e.g., sensors) or other wireless devices 100a to 100f.

[0045] Wireless communication / connection 150a, 150b, and 150c may be established between wireless devices 100a to 100f and / or between wireless devices 100a to 100f and BS 200 and / or between BS 200. In this document, wireless communication / connection may be established through various RATs (e.g., 5G NR) such as uplink / downlink communication 150a, sidelink communication (or device-to-device (D2D) communication) 150b, inter-base station communication 150c (e.g., relay, integrated access and backhaul (IAB)), etc. The wireless devices 100a to 100f and BS 200 / wireless devices 100a to 100f may send / receive radio signals to each other through wireless communication / connection 150a, 150b, and 150c. For example, wireless communication / connection 150a, 150b, and 150c may send / receive signals through various physical channels. To this end, various configuration information configuration processes for sending / receiving radio signals, various signal processing processes (e.g., channel coding / decoding, modulation / demodulation, and resource mapping / demapping), and at least a portion of a resource allocation process may be performed based on various proposals of the present disclosure.

[0046] NR supports multiple parameter sets (and / or multiple subcarrier spacing (SCS)) to support various 5G services. For example, if the SCS is 15kHz, wide areas can be supported in traditional cellular bands, and if the SCS is 30kHz / 60kHz, dense cities, lower latency, and wider carrier bandwidths can be supported. If the SCS is 60kHz or higher, bandwidths greater than 24.25GHz can be supported to overcome phase noise.

[0047] The NR frequency band may be defined as two types of frequency ranges, namely, frequency range 1 (FR1) and frequency range 2 (FR2). The numerical value of the frequency range may be changed. For example, the two types of frequency ranges (FR1 and FR2) may be as shown in Table 1 below. For ease of explanation, in the frequency range used in the NR system, FR1 may mean "below 6 GHz range", FR2 may mean "above 6 GHz range", and may be referred to as millimeter wave (mmW).

[0048] [Table 1]

[0049] Frequency range name Corresponding frequency range Subcarrier spacing FR1 450MHz-6000MHz 15kHz, 30kHz, 60kHz FR2 24250MHz-52600MHz 60kHz, 120kHz, 240kHz

[0050] As described above, the numerical value of the frequency range of the NR system can be changed. For example, FR1 may include a frequency band of 410 MHz to 7125 MHz, as shown in Table 2 below. That is, FR1 may include a frequency band of 6 GHz (or 5850 MHz, 5900 MHz, 5925 MHz, etc.) or higher. For example, a frequency band of 6 GHz (or 5850 MHz, 5900 MHz, 5925 MHz, etc.) or higher included in FR1 may include an unlicensed frequency band. The unlicensed frequency band may be used for various purposes, such as for communication of vehicles (e.g., autonomous driving).

[0051] [Table 2]

[0052] Frequency range name Corresponding frequency range Subcarrier spacing FR1 410MHz-7125MHz 15kHz, 30kHz, 60kHz FR2 24250MHz-52600MHz 60kHz, 120kHz, 240kHz

[0053] Here, the radio communication technology implemented in the wireless device in the present disclosure may include narrowband IoT (NB-IoT) technology for low-power communication as well as LTE, NR and 6G. For example, NB-IoT technology may be an example of a low-power wide area network (LPWAN) technology, which may be implemented in specifications such as LTE Cat NB1 and / or LTE Cat NB2, and may not be limited to the above names. Additionally and / or alternatively, the radio communication technology implemented in the wireless device in the present disclosure may communicate based on LTE-M technology. For example, LTE-M technology may be an example of LPWAN technology, and may be referred to as various names such as enhanced MTC (eMTC). For example, LTE-M technology may be implemented in at least one of various specifications such as 1) LTE Cat 0, 2) LTE Cat M1, 3) LTE Cat M2, 4) LTE non-bandwidth limited (non-BL), 5) LTE-MTC, 6) LTE machine type communication and / or 7) LTE M, and may not be limited to the above names. Additionally and / or alternatively, the radio communication technology implemented in the wireless device in the present disclosure may include at least one of ZigBee, Bluetooth, and / or LPWAN considering low-power communication, and may not be limited to the above names. For example, ZigBee technology may generate a personal area network (PAN) associated with low-power / low-power digital communication based on various specifications such as IEEE 802.15.4, and may be referred to as various names.

[0054] Figure 2 An example of a wireless device to which an implementation of the present disclosure is applied is shown.

[0055] exist Figure 2 In the embodiment, the first wireless device 100 and / or the second wireless device 200 may be implemented in various forms according to use cases / services. For example, {the first wireless device 100 and the second wireless device 200} may correspond to Figure 1At least one of {wireless devices 100a to 100f and BS 200}, {wireless devices 100a to 100f and wireless devices 100a to 100f} and / or {BS 200 and BS 200}. The first wireless device 100 and / or the second wireless device 200 may be configured by various elements, devices / components and / or modules.

[0056] The first wireless device 100 may include at least one transceiver (eg, transceiver 106 ), at least one processing chip (eg, processing chip 101 ), and / or one or more antennas 108 .

[0057] The processing chip 101 may include at least one processor (eg, the processor 102 ) and at least one memory (eg, the memory 104 ). Additionally and / or alternatively, the memory 104 may be placed outside the processing chip 101 .

[0058] The processor 102 may control the memory 104 and / or the transceiver 106, and may be adapted to implement the descriptions, functions, processes, suggestions, methods, and / or operational flow charts described in the present disclosure. For example, the processor 102 may process the information in the memory 104 to generate first information / signals, and then send a radio signal including the first information / signals through the transceiver 106. The processor 102 may receive a radio signal including second information / signals through the transceiver 106, and then store information obtained by processing the second information / signals in the memory 104.

[0059] The memory 104 may be operatively connected to the processor 102. The memory 104 may store various types of information and / or instructions. The memory 104 may store firmware and / or software code 105 that implements codes, commands, and / or command sets that, when executed by the processor 102, perform the descriptions, functions, processes, suggestions, methods, and / or operational flow charts disclosed in the present disclosure. For example, the firmware and / or software code 105 may implement instructions that, when executed by the processor 102, perform the descriptions, functions, processes, suggestions, methods, and / or operational flow charts disclosed in the present disclosure. For example, the firmware and / or software code 105 may control the processor 102 to execute one or more protocols. For example, the firmware and / or software code 105 may control the processor 102 to execute one or more layers of a radio interface protocol.

[0060] In this document, the processor 102 and the memory 104 may be part of a communication modem / circuit / chip designed to implement a RAT (e.g., LTE or NR). The transceiver 106 may be connected to the processor 102 and transmit and / or receive radio signals through one or more antennas 108. Each transceiver 106 may include a transmitter and / or a receiver. The transceiver 106 may be used interchangeably with a radio frequency (RF) unit. In the present disclosure, the first wireless device 100 may represent a communication modem / circuit / chip.

[0061] The second wireless device 200 may include at least one transceiver (eg, transceiver 206 ), at least one processing chip (eg, processing chip 201 ), and / or one or more antennas 208 .

[0062] The processing chip 201 may include at least one processor (eg, the processor 202 ) and at least one memory (eg, the memory 204 ). Additionally and / or alternatively, the memory 204 may be placed outside the processing chip 201 .

[0063] The processor 202 may control the memory 204 and / or the transceiver 206, and may be adapted to implement the descriptions, functions, processes, suggestions, methods, and / or operational flow charts described in the present disclosure. For example, the processor 202 may process the information in the memory 204 to generate third information / signals, and then send a radio signal including the third information / signals through the transceiver 206. The processor 202 may receive a radio signal including fourth information / signals through the transceiver 106, and then store the information obtained by processing the fourth information / signals in the memory 204.

[0064] The memory 204 may be operatively connected to the processor 202. The memory 204 may store various types of information and / or instructions. The memory 204 may store firmware and / or software code 205 that implements codes, commands, and / or command sets that, when executed by the processor 202, perform the descriptions, functions, processes, suggestions, methods, and / or operational flow charts disclosed in the present disclosure. For example, the firmware and / or software code 205 may implement instructions that, when executed by the processor 202, perform the descriptions, functions, processes, suggestions, methods, and / or operational flow charts disclosed in the present disclosure. For example, the firmware and / or software code 205 may control the processor 202 to execute one or more protocols. For example, the firmware and / or software code 205 may control the processor 202 to execute one or more layers of a radio interface protocol.

[0065] In this document, the processor 202 and the memory 204 may be part of a communication modem / circuit / chip designed to implement a RAT (e.g., LTE or NR). The transceiver 206 may be connected to the processor 202 and transmit and / or receive radio signals through one or more antennas 208. Each transceiver 206 may include a transmitter and / or a receiver. The transceiver 206 may be used interchangeably with an RF unit. In the present disclosure, the second wireless device 200 may represent a communication modem / circuit / chip.

[0066] In the following, the hardware elements of the wireless devices 100 and 200 will be described in more detail. One or more protocol layers may be implemented by (but not limited to) one or more processors 102 and 202. For example, one or more processors 102 and 202 may implement one or more layers (e.g., functional layers such as a physical (PHY) layer, a medium access control (MAC) layer, a radio link control (RLC) layer, a packet data convergence protocol (PDCP) layer, a radio resource control (RRC) layer, and a service data adaptation protocol (SDAP) layer). One or more processors 102 and 202 may generate one or more protocol data units (PDUs), one or more service data units (SDUs), messages, control information, data, or information according to the descriptions, functions, processes, suggestions, methods, and / or operation flow charts disclosed in the present disclosure. One or more processors 102 and 202 may generate one or more protocol data units (PDUs) and / or one or more service data units (SDUs) according to the descriptions, functions, processes, suggestions, methods, and / or operational flowcharts disclosed in the present disclosure. One or more processors 102 and 202 may generate messages, control information, data, or information according to the descriptions, functions, processes, suggestions, methods, and / or operational flowcharts disclosed in the present disclosure. One or more processors 102 and 202 may generate signals (e.g., baseband signals) including PDUs, SDUs, messages, control information, data, or information according to the descriptions, functions, processes, suggestions, methods, and / or operational flowcharts disclosed in the present disclosure, and provide the generated signals to one or more transceivers 106 and 206. One or more processors 102 and 202 can receive signals (e.g., baseband signals) from one or more transceivers 106 and 206 according to the descriptions, functions, processes, suggestions, methods and / or operational flow charts disclosed in the present disclosure, and obtain PDUs, SDUs, messages, control information, data or information.

[0067] One or more processors 102 and 202 may be referred to as controllers, microcontrollers, microprocessors or microcomputers. One or more processors 102 and 202 may be implemented by hardware, firmware, software or a combination thereof. As an example, one or more application specific integrated circuits (ASICs), one or more digital signal processors (DSPs), one or more digital signal processing devices (DSPDs), one or more programmable logic devices (PLDs) or one or more field programmable gate arrays (FPGAs) may be included in one or more processors 102 and 202. For example, one or more processors 102 and 202 may be configured by a group of communication control processors, application processors (APs), electronic control units (ECUs), central processing units (CPUs), graphics processing units (GPUs) and memory control processors. One or more memories 104 and 204 may be connected to one or more processors 102 and 202 and store various types of data, signals, messages, information, programs, codes, instructions and / or commands. The one or more memories 104 and 204 may be configured by random access memory (RAM), dynamic RAM (DRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EPROM), flash memory, volatile memory, non-volatile memory, hard disk drive, register, buffer memory, computer-readable storage medium, and / or a combination thereof. The one or more memories 104 and 204 may be located inside and / or outside of the one or more processors 102 and 202. The one or more memories 104 and 204 may be connected to the one or more processors 102 and 202 through various technologies such as wired or wireless connections.

[0068] One or more transceivers 106 and 206 can send user data, control information and / or radio signals / channels mentioned in the description, function, process, suggestion, method and / or operation flow chart disclosed in the present disclosure to one or more other devices. One or more transceivers 106 and 206 can receive user data, control information and / or radio signals / channels mentioned in the description, function, process, suggestion, method and / or operation flow chart disclosed in the present disclosure from one or more other devices. For example, one or more transceivers 106 and 206 can be connected to one or more processors 102 and 202 and send and receive radio signals. For example, one or more processors 102 and 202 can perform control so that one or more transceivers 106 and 206 can send user data, control information or radio signals to one or more other devices. One or more processors 102 and 202 can perform control so that one or more transceivers 106 and 206 can receive user data, control information or radio signals from one or more other devices.

[0069] One or more transceivers 106 and 206 may be connected to one or more antennas 108 and 208. Additionally and / or alternatively, one or more transceivers 106 and 206 may include one or more antennas 108 and 208. One or more transceivers 106 and 206 may be adapted to transmit and receive user data, control information, and / or radio signals / channels mentioned in the description, functions, processes, suggestions, methods, and / or operational flow charts disclosed in the present disclosure through one or more antennas 108 and 208. In the present disclosure, one or more antennas 108 and 208 may be multiple physical antennas or multiple logical antennas (e.g., antenna ports).

[0070] One or more transceivers 106 and 206 may convert received user data, control information, radio signals / channels, etc. from RF band signals into baseband signals so as to process the received user data, control information, radio signals / channels, etc. using one or more processors 102 and 202. One or more transceivers 106 and 206 may convert user data, control information, radio signals / channels, etc. processed using one or more processors 102 and 202 from baseband signals into RF band signals. To this end, one or more transceivers 106 and 206 may include (analog) oscillators and / or filters. For example, one or more transceivers 106 and 206 may up-convert OFDM baseband signals into OFDM signals through their (analog) oscillators and / or filters under the control of one or more processors 102 and 202, and transmit the up-converted OFDM signals at a carrier frequency. One or more transceivers 106 and 206 may receive the OFDM signal at a carrier frequency and down-convert the OFDM signal to an OFDM baseband signal through its (analog) oscillator and / or filter under the control of one or more processors 102 and 202 .

[0071] although Figure 2 140 , but the wireless devices 100 and 200 may also include additional components. The additional components 140 may be configured differently depending on the type of the wireless devices 100 and 200. For example, the additional components 140 may include at least one of a power supply unit / battery, an input / output (I / O) device (e.g., an audio I / O port, a video I / O port), a drive device, and a computing device. The additional components 140 may be connected to one or more processors 102 and 202 via various technologies such as wired or wireless connections.

[0072] In an implementation of the present disclosure, a UE may operate as a transmitting device in an uplink (UL) and as a receiving device in a downlink (DL). Within an implementation of the present disclosure, a BS may operate as a receiving device in the UL and as a transmitting device in the DL. In the following, for ease of description, it is mainly assumed that the first wireless device 100 acts as a UE and the second wireless device 200 acts as a BS. For example, a processor 102 connected to, installed on, or started in the first wireless device 100 may be adapted to perform UE behavior according to an implementation of the present disclosure, or to control a transceiver 106 to perform UE behavior according to an implementation of the present disclosure. A processor 202 connected to, installed on, or started in the second wireless device 200 may be adapted to perform BS behavior according to an implementation of the present disclosure, or to control a transceiver 206 to perform BS behavior according to an implementation of the present disclosure.

[0073] In this disclosure, a BS is also referred to as a Node B (NB), an eNodeB (eNB), or a gNB.

[0074] Figure 3 An example of a UE to which an implementation of the present disclosure is applied is shown.

[0075] Reference Figure 3 , UE 100 may correspond to Figure 2 A first wireless device 100 is provided.

[0076] UE 100 includes a processor 102 , a memory 104 , a transceiver 106 , one or more antennas 108 , a power management module 141 , a battery 142 , a display 143 , a keypad 144 , a subscriber identity module (SIM) card 145 , a speaker 146 , and a microphone 147 .

[0077] The processor 102 may be adapted to implement the descriptions, functions, processes, suggestions, methods and / or operational flow charts disclosed in the present disclosure. The processor 102 may be adapted to control one or more other components of the UE 100 to implement the descriptions, functions, processes, suggestions, methods and / or operational flow charts disclosed in the present disclosure. The layers of the radio interface protocol may be implemented in the processor 102. The processor 102 may include an ASIC, other chipsets, logic circuits and / or data processing devices. The processor 102 may be an application processor. The processor 102 may include at least one of a DSP, a CPU, a GPU, a modem (modulator and demodulator). The processor 102 may be implemented in a processor 102. MANUFACTURED BY SNAPDRAGON TM Series processors, Manufactured by EXYNOS TMSeries processors, A series processors manufactured by Made by HELIO TM Series processors, ATOM manufactured TM An example of processor 102 is found in the ATTACH ® series of processors or a corresponding next generation processor.

[0078] The memory 104 is operably connected to the processor 102 and stores various information for operating the processor 102. The memory 104 may include ROM, RAM, flash memory, memory card, storage medium and / or other storage devices. When the embodiment is implemented in software, the technology described herein can be implemented with modules (e.g., processes, functions, etc.) that execute the descriptions, functions, processes, suggestions, methods and / or operational flow charts disclosed in this disclosure. The modules can be stored in the memory 104 and executed by the processor 102. The memory 104 can be implemented within the processor 102 or outside the processor 102, in which case those memories can be communicatively connected to the processor 102 via various means known in the art.

[0079] The transceiver 106 is operably coupled to the processor 102 and transmits and / or receives radio signals. The transceiver 106 includes a transmitter and a receiver. The transceiver 106 may include a baseband circuit for processing radio frequency signals. The transceiver 106 controls one or more antennas 108 to transmit and / or receive radio signals.

[0080] The power management module 141 manages the power of the processor 102 and / or the transceiver 106. The battery 142 supplies power to the power management module 141.

[0081] The display 143 outputs the result processed by the processor 102. The keyboard 144 receives input to be used by the processor 102. The keyboard 144 may be shown on the display 143.

[0082] The SIM card 145 is an integrated circuit designed to securely store the International Mobile Subscriber Identity (IMSI) number and its associated keys, which are used to identify and authenticate subscribers on mobile telephony devices such as mobile phones and computers. Contact information associated with many SIM cards can also be stored.

[0083] The speaker 146 outputs sound-related results processed by the processor 102. The microphone 147 receives sound-related input to be used by the processor 102.

[0084] Figure 4 An example of a 5G system architecture to which implementations of the present disclosure are applied is shown.

[0085] The 5G system (5GS) architecture includes the following network functions (NFs).

[0086] -Authentication Server Function (AUSF)

[0087] - Access and Mobility Management Function (AMF)

[0088] -Data Network (DN), e.g. operator service, Internet access or third-party service

[0089] - Unstructured Data Storage Function (UDSF)

[0090] -Network Exposure Function (NEF)

[0091] -Intermediate NEF (I-NEF)

[0092] -Network Repository Function (NRF)

[0093] - Network Slice Selection Function (NSSF)

[0094] -Policy Control Function (PCF)

[0095] -Session Management Function (SMF)

[0096] -Unified Data Management (UDM)

[0097] - Unified Data Repository (UDR)

[0098] -User Plane Function (UPF)

[0099] -UE Radio Capability Management Function (UCMF)

[0100] - Application Function (AF)

[0101] -User Equipment (UE)

[0102] - (Radio) Access Network ((R)AN)

[0103] -5G-Equipment Identification Register (5G-EIR)

[0104] -Network Data Analysis Function (NWDAF)

[0105] -Billing function (CHF)

[0106] In addition, the following network functions can be considered.

[0107] -Non-3GPP Interworking Function (N3IWF)

[0108] -Trusted Non-3GPP Gateway Function (TNGF)

[0109] -Wired Access Gateway Function (W-AGF)

[0110] Figure 4 Depicted is the 5G system architecture in a non-roaming scenario, represented using reference points showing how the various network functions interact with each other.

[0111] exist Figure 4 In the figure, for the sake of clarity of the point-to-point diagram, UDSF, NEF and NRF are not depicted. However, all depicted network functions can interact with UDSF, UDR, NEF and NRF as needed.

[0112] For the sake of clarity, Figure 4 The UDR and its connections to other NFs (e.g., PCF) are not depicted in the figure. Figure 4 The NWDAF and its connections with other NFs (e.g., PCF) are not depicted.

[0113] The 5G system architecture includes the following reference points:

[0114] -N1: Reference point between UE and AMF.

[0115] -N2: Reference point between (R)AN and AMF.

[0116] -N3: Reference point between (R)AN and UPF.

[0117] -N4: Reference point between SMF and UPF.

[0118] -N6: Reference point between UPF and data network.

[0119] -N9: Reference point between two UPFs.

[0120] The following reference points illustrate the interactions that exist between NF services in a NF.

[0121] -N5: Reference point between PCF and AF.

[0122] -N7: Reference point between SMF and PCF.

[0123] -N8: Reference point between UDM and AMF.

[0124] -N10: Reference point between UDM and SMF.

[0125] -N11: Reference point between AMF and SMF.

[0126] -N12: Reference point between AMF and AUSF.

[0127] -N13: Reference point between UDM and AUSF.

[0128] -N14: Reference point between two AMFs.

[0129] -N15: Reference point between PCF and AMF in non-roaming scenario, reference point between PCF and AMF in visited network in roaming scenario.

[0130] - N16: Reference point between two SMFs (between the SMF in the visited network and the SMF in the home network in roaming case).

[0131] -N22: Reference point between AMF and NSSF.

[0132] In some cases, several NFs may need to be associated with each other to serve a UE.

[0133] The PDU session establishment process is described. Please refer to Section 4.3.2 of 3GPP TS23.502V16.3.0 (2019-12).

[0134] Figure 5 and Figure 6 An example of a PDU session establishment process to which an implementation of the present disclosure is applied is shown.

[0135] A PDU session establishment may correspond to:

[0136] -PDU session establishment process initiated by UE.

[0137] - UE initiated PDU session switching between 3GPP and non-3GPP.

[0138] - UE initiated PDU session handover from EPS to 5GS.

[0139] -Network triggered PDU session establishment process.

[0140] A PDU session may be (a) associated with a single access type (i.e., 3GPP access or non-3GPP access) at a given time, or (b) associated with multiple access types (i.e., one 3GPP access and one non-3GPP access) at the same time. A PDU session associated with multiple access types is called a multi-access PDU (MA PDU) session, and it may be requested by a UE with access service steering, switching, splitting (ATSSS) capability.

[0141] Figure 5 and Figure 6 Specifies the procedure for establishing a PDU session associated with a single access type at a given time.

[0142] Figure 5 and Figure 6The procedure shown assumes that the UE is already registered with the AMF, so unless the UE is emergency registered, the AMF has already retrieved the user subscription data from the UDM.

[0143] First, describe Figure 5 process.

[0144] (1) Step 1: To establish a new PDU session, the UE generates a new PDU session ID.

[0145] The UE initiates the UE-requested PDU session establishment process by transmitting a NAS message containing a PDU session establishment request message in an N1 SM container. The PDU session establishment request message includes the PDU session ID, the requested PDU session type, the requested session and service continuity (SSC) mode, 5GSM capabilities, protocol configuration options (PCO), SM PDU DN request container, UE integrity protection maximum data rate, etc.

[0146] If the PDU session establishment is a request to establish a new PDU session, the request type indicates "initial request", and if the request refers to an existing PDU session handover between 3GPP access and non-3GPP access or refers to a PDU session handover from an existing packet data network (PDN) connection in the EPC, the request type indicates "existing PDU session". If the PDU session establishment is a request to establish a PDU session for emergency services, the request type indicates "emergency request". If the request refers to an existing PDU session for emergency services handover between 3GPP access and non-3GPP access or refers to a PDU session handover from an existing PDN connection for emergency services in the EPC, the request type indicates "existing emergency PDU session".

[0147] The UE includes the S-NSSAI from the allowed NSSAIs for the current access type. If a mapping of allowed NSSAIs is provided to the UE, the UE shall provide both the S-NSSAI of the visited PLMN (VPLMN) from the allowed NSSAI and the corresponding S-NSSAI of the HPLMN from the mapping of allowed NSSAIs.

[0148] (2) Step 2: AMF selects SMF. If the request type indicates "initial request" or the request is due to a handover from EPS or from a non-3GPP access served by a different AMF, the AMF stores an association of the S-NSSAI, Data Network Name (DNN), PDU Session ID, SMF ID, and Access Type of the PDU Session.

[0149] If the request type is "initial request" and if an old PDU session ID indicating an existing PDU session is also contained in the message, the AMF selects an SMF and stores an association of the new PDU session ID, the S-NSSAI, the selected SMF ID and the access type of the PDU session.

[0150] If the request type indicates "existing PDU session", the AMF selects the SMF based on the SMF-ID received from the UDM. The AMF updates the access type stored for the PDU session.

[0151] If the request type indicates that "existing PDU session" refers to an existing PDU session that is moved between 3GPP access and non-3GPP access, then if the serving PLMN S-NSSAI of the PDU session is present in the allowed NSSAI of the target access type, the PDU session establishment procedure may be performed in the following cases:

[0152] - The SMF ID corresponding to the PDU Session ID and the AMF belong to the same PLMN;

[0153] - The SMF ID corresponding to the PDU Session ID belongs to the HPLMN;

[0154] Otherwise, the AMF shall reject the PDU session establishment request with an appropriate rejection cause.

[0155] The AMF shall reject the request from an emergency registered UE and the request type indicates neither "Emergency Request" nor "Existing Emergency PDU Session".

[0156] (3) Step 3: If the AMF does not have an association with the SMF for the PDU Session ID provided by the UE (e.g., when the request type indicates "initial request"), the AMF invokes a Create SM Context Request procedure (e.g., Nsmf_PDUSession_CreateSMContext request). If the AMF already has an association with the SMF for the PDU Session ID provided by the UE (e.g., when the request type indicates "existing PDU Session"), the AMF invokes an Update SM Context Request procedure (e.g., Nsmf_PDUSession_UpdateSMContext request).

[0157] AMF sends the S-NSSAI of the serving PLMN from the allowed NSSAI to SMF. For roaming scenarios in Local Breakout (LBO), AMF also sends the mapping of the corresponding S-NSSAI of the HPLMN from the allowed NSSAI to SMF.

[0158] The AMF ID uniquely identifies the GUAMI of the UE serving the AMF of the UE. The AMF forwards the PDU Session ID together with the N1 SM container containing the PDU Session Establishment Request message received from the UE. If available at the AMF, the General Public Subscription Identifier (GPSI) shall be included.

[0159] When a UE in limited service state has registered for emergency services without providing a SUPI, the AMF provides a PEI instead of a SUPI. In the case where a UE in limited service state has registered for emergency services with a SUPI but has not been authenticated, the AMF indicates that the SUPI is not authenticated. The SMF determines that the UE is not authenticated when the SMF does not receive a SUPI for the UE or when the AMF indicates that the SUPI is not authenticated.

[0160] The AMF may include the PCF ID in the Nsmf_PDUSession_CreateSMContext request. The PCF ID identifies the home PCF (H-PCF) in the non-roaming case and the visited PCF (V-PCF) in the LBO roaming case.

[0161] (4) Step 4: If the session management subscription data for the corresponding SUPI, DNN and S-NSSAI of the HPLMN is not available, the SMF may retrieve the session management subscription data from the UDM and the subscription will be notified when the subscription data is modified.

[0162] (5) Step 5: Based on the request received in step 3, the SMF sends a create SM context response message (e.g., Nsmf_PDUSession_CreateSMContext response) or an update SM context response message (e.g., Nsmf_PDUSession_UpdateSMContext response) to the AMF.

[0163] If in step 3 the SMF receives the Nsmf_PDUSession_CreateSMContext request and the SMF is able to process the PDU session establishment request, the SMF creates an SM context and responds to the AMF by providing the SM context ID.

[0164] When the SMF decides not to accept the establishment of a PDU session, the SMF rejects the UE request via NAS SM signaling including the relevant SM rejection cause by responding to the AMF with a Nsmf_PDUSession_CreateSMContext response. The SMF also indicates to the AMF that the PDU session ID will be considered released, the SMF proceeds to step 20 below, and the PDU session establishment process stops.

[0165] (6) Step 6: Optional secondary authentication / authorization may be performed.

[0166] (7a) Step 7a: If dynamic Policy and Charging Control (PCC) is to be used for the PDU Session, the SMF may perform PCF selection.

[0167] (7b) Step 7b: The SMF may perform the SM policy association establishment procedure to establish an SM policy association with the PCF and obtain the default PCC rules for the PDU session.

[0168] (8) Step 8: SMF selects one or more UPFs.

[0169] (9) Step 9: The SMF may execute the SMF-initiated SM policy association modification process to provide information related to the policy control request triggering conditions that have been met.

[0170] (10) Step 10: If the request type indicates "initial request", the SMF may initiate an N4 session establishment procedure with the selected UPF. Otherwise, the SMF may initiate an N4 session modification procedure with the selected UPF.

[0171] In step 10a, the SMF may send an N4 session establishment / modification request to the UPF and provide packet detection, enforcement and reporting rules to be installed on the UPF for the PDU session. In step S10b, the UPF may confirm by sending an N4 session establishment / modification response.

[0172] (11) Step 11: SMF sends an N1N2Message transfer message (e.g., Namf_Communication_N1N2MessageTransfer) to AMF.

[0173] The N1N2Message transmission message may include the N2 SM information. The N2 SM information carries the information that the AMF should forward to the (R)AN, which may include:

[0174] -CN tunnel information: the core network address of the N3 tunnel corresponding to the PDU session;

[0175] - one or more Quality of Service (QoS) profiles and corresponding QoS Flow IDs (QFIs);

[0176] -PDU Session ID: indicates to the UE the association between (R)AN resources and the PDU session for that UE;

[0177] - S-NSSAI with the value of the serving PLMN (i.e., HPLMNS-NSSAI, or, in case of LBO roaming, VPLMNS-NSSAI);

[0178] - User plane security implementation information determined by the SMF;

[0179] - If the User Plane Security Implementation Information indicates that integrity protection is "preferred" or "required (or mandatory)", the SMF also includes the UE integrity protection maximum data rate as received in the PDU Session Establishment Request message;

[0180] - Redundant Sequence Number (RSN) parameter.

[0181] The N1N2Message transfer message may include an N1 SM container. The N1 SM container contains a PDU Session Establishment Accept message that the AMF should provide to the UE. The PDU Session Establishment Accept message includes the S-NSSAI from the allowed NSSAI. For the LBO roaming scenario, the PDU Session Establishment Accept message includes the S-NSSAI of the allowed NSSAI from the VPLMN, and it also includes the corresponding S-NSSAI of the mapped HPLMN from the allowed NSSAI received by the SMF in step 3.

[0182] If necessary, multiple QoS rules, QoS flow levels, QoS parameters of QoS flows associated with these QoS rules and QoS profiles may be included in the PDU Session Setup Accept message within the N1 SM container and in the N2 SM information.

[0183] If the PDU Session Establishment fails anywhere between step 5 and step 11, the N1N2Message transmission message shall include the N1 SM container with the PDU Session Establishment Reject message and shall not include any N2 SM information. (R)AN sends a NAS message containing the PDU Session Establishment Reject message to the UE. In this case, steps 12 to 17 are skipped.

[0184] (12) Step 12: The AMF sends a NAS message to the (R)AN containing the PDU Session ID and PDU Session Establishment Accept message targeted at the UE and the N2 SM information received from the SMF in the N2 PDU Session Request message.

[0185] (13) Step 13: The (R)AN may initiate an AN-specific signaling exchange with the UE related to the information received from the SMF. For example, in the case of NG-RAN, an RRC connection reconfiguration may occur where the UE establishes the necessary NG-RAN resources related to the QoS rules of the PDU Session Request received in step 12.

[0186] The (R)AN forwards the NAS message (PDU Session ID, N1 SM container (PDU Session Setup Accept message)) provided in step 12 to the UE. The (R)AN shall only provide the NAS message to the UE if the AN-specific signaling exchanged with the UE includes (R)AN resource addition associated with the received N2 command.

[0187] If the N2 SM information is not included in step 11, the following steps 14 to 16b and step 17 are omitted.

[0188] Now, the description in Figure 5 After the process Figure 6 process.

[0189] (14) Step 14: (R)AN sends an N2 PDU session response message to AMF. The N2 PDU session response message may include PDU session ID, cause, N2 SM information (PDU session ID, AN tunnel information, list of accepted / rejected QFIs, user plane implementation policy notification), etc.

[0190] (15) Step 15: AMF sends an SM context update request message (e.g., Nsmf_PDUSession_UpdateSMContext request) to SMF. AMF forwards the N2 SM information received from (R)AN to SMF.

[0191] (16a) Step S16a: SMF initiates the N4 session modification process with UPF. SMF provides AN tunnel information and corresponding forwarding rules to UPF.

[0192] (16b) Step S16b: UPF provides an N4 session modification response to SMF.

[0193] After this step, the UPF may deliver to the UE any DL packets that have been buffered for this PDU Session.

[0194] (16c) Step 16c: The SMF may register with the UDM for a given PDU Session if the SMF has not yet registered for that PDU Session.

[0195] (17) Step 17: SMF sends an update SM context response message (e.g., Nsmf_PDUSession_UpdateSMContext response) to AMF.

[0196] After this step, AMF forwards the relevant events subscribed by SMF.

[0197] (18) Step 18: If at any time during this process, after step 5, the PDU session establishment is unsuccessful, the SMF may notify the AMF by calling Nsmf_PDUSession_SMContextStatusNotify(release). If any PDU session address (e.g., IP address) is allocated, the SMF may also release any N4 session created and, if any, release the association with the PCF. In this case, step 19 is skipped.

[0198] (19) Step 19: In case of PDU session type IPv6 or IPv4v6, the SMF may generate an IPv6 router advertisement and send it to the UE.

[0199] (20) Step 20: SMF may perform SMF-initiated SM policy association modification.

[0200] (21) Step 21: If the PDU session establishment fails after step 4, then if the SMF no longer handles the UE's PDU session, the SMF can cancel the subscription to the modification of the session management subscription data.

[0201] <Multiple Access (MA) PDU Session>

[0202] In the prior art, an MA PDU session is a session that can be served by using both 3GPP access and non-3GPP access using one PDU session.

[0203] Figure 7 An example of generating a MAPDU session is shown.

[0204] exist Figure 7 In , an MA PDU session is a PDU session and has a separate session tunnel for each access. One is established on a 3GPP access and the other PDU session is established on an untrusted non-3GPP access (e.g., WLAN AN).

[0205] Since the MA PDU session is a session, the MA PDU session has the following characteristics.

[0206] (i) A DNN;

[0207] (ii) a UPF anchor (UPF-A);

[0208] (iii) a PDU type (e.g., IPv6);

[0209] (iv) a session IP address;

[0210] (v) an SSC mode;

[0211] (vi) A HPLMN S-NSSAI.

[0212] The MA PDU session enables a multipath data link between the UE and UPF-A. This can be implemented below the IP layer.

[0213] A MA PDU session can be established by one of the following procedures.

[0214] (i) It can be established through two separate PDU session establishment procedures. This is called a separate establishment.

[0215] (ii) It can be established through one MA PDU session establishment process. That is, one session establishment request can be used to establish MA PDU sessions in two accesses at the same time. This is called binding establishment.

[0216] After the MA PDU session is established, session management (SM) signaling related to the MA PDU session may be sent and received through random access.

[0217] A. Separate establishment of MA PDU session

[0218] An MA PDU session may be established by two separate PDU session establishment procedures. For example, a UE may establish an MA PDU session on a 3GPP access and then perform a PDU session establishment procedure on a non-3GPP access to add the non-3GPP access to the MAPDU session created on the 3GPP access. The request type in the setup request message for adding the second access may be set to "MA PDU Request".

[0219] B. Binding Establishment

[0220] A MA PDU session can be established for both 3GPP access and non-3GPP access simultaneously through one procedure. Such a procedure may be referred to as a UE-requested MA PDU session establishment procedure. The above procedure may be useful when the UE intends to establish an MA PDU session while the UE has registered with the 5GC through two accesses. Instead of performing two separate PDU session establishment procedures, the UE can establish an MA PDU session by performing one MAPDU session establishment procedure.

[0221] Figure 8 An example of applying ATSSS rules to a MAPDU session is shown.

[0222] Reference Figure 8If the SMF wants to move the IP flow sent to the non-3GPP access to the 3GPP access through the 3GPP access while a multi-access (MA) PDU session is established, an updated ATSSS (Access Service Steering, Switching and Splitting) rule can be sent.

[0223] <Services based on artificial intelligence (AI) / machine learning (ML)>

[0224] In 3GPP, plans to support AL / ML-based services in 5GS are being discussed.

[0225] For example, with the assistance of 5GS, methods are being discussed to enable AI / ML services and transmission to support the distribution, delivery, and training of AI / ML models for various applications, such as image / speech recognition, robot control, and automobiles.

[0226] AI / ML operations can include, for example, three main types:

[0227] -Split of AI / ML operations between AI / ML endpoints

[0228] -AI / ML model / data distribution and sharing on 5G systems

[0229] -Distributed / Federated Learning (FL) on 5G Systems

[0230] To provide intelligent transport support for application-layer AI / ML tasks, the following examples illustrate how AI / ML service providers can leverage 5GS as a platform target.

[0231] 1. Objective 1: Study possible architectures and functional extensions to support application-layer AI / ML tasks.

[0232] a. To support applied AI / ML operations together with AI / ML model vendors, it supports monitoring of network resource utilization associated with UE in 5G systems.

[0233] b.5GS information exposure is extended to 5GC NF to expose UE and / or network health and performance predictions (e.g., location, QoS, load, congestion, etc.). Supports whether and how this information should be exposed to UE and / or authorized third parties to support application AI / ML operations.

[0234] c. Enhance the external parameter configuration of 5GC based on applying AI / ML operations (e.g., expected UE activity behavior, expected UE mobility, etc.).

[0235] d. Study enhancements to other 5GC features that can be used to support applied AI / ML operations.

[0236] 2. Objective 2: Study possible QoS and policy enhancements to support AI / ML operational services while supporting normal (non-AI / ML) 5GS user services.

[0237] 3. Objective 3: Study whether and how 5GS can provide assistance to AF and UE to manage FL operations and model distribution / redistribution (i.e., FL member selection, group performance monitoring, appropriate network resource allocation and guarantees). These studies can be conducted to facilitate collaborative application AI / ML-based federated learning operations between application clients and application servers running on UEs.

[0238] All UEs participating in a given application AI / ML task can receive services through the same S-NSSAI as the AF.

[0239] The following examples are examples of three major AI / ML operations.

[0240] Artificial intelligence (AI) / machine learning (ML) is being used in various applications across industries. In mobile communication systems, mobile devices (e.g., smartphones, cars, robots) are increasingly replacing traditional algorithms (e.g., speech recognition, image recognition, video processing) with AI / ML models to implement applications.

[0241] The 5G system can support at least three types of AI / ML tasks. The following examples show three types of AI / ML tasks:

[0242] -Split of AI / ML operations between AI / ML endpoints;

[0243] -AI / ML model / data distribution and sharing on 5G systems;

[0244] -Distributed / federated learning on 5G systems;

[0245] Fig. 9 The example of FIG. 1 shows an example of AI / ML job splitting between AI / ML endpoints.

[0246] The following figures are created to explain specific examples of this specification. Since the names of specific devices or the names of specific signals / messages / fields described in the figures are presented as examples, the technical features of this specification are not limited to the specific names used in the following figures.

[0247] Fig. 9 An example of split AI / ML according to the disclosed embodiments of the present specification is shown.

[0248] Splitting AI / ML inference methods can be done as follows Fig. 9The AI / ML job / model is split into several parts depending on the current job and environment. The computationally intensive and energy-intensive parts can be offloaded to the network endpoint, while the privacy-sensitive and latency-sensitive parts can remain on the terminal device. The device can execute the task / model until a specific part / layer and then send the intermediate data to the network endpoint. The network endpoint can execute the remaining parts / layers and feed the inference results back to the device.

[0249] Fig.10 The examples in show examples of AI / ML model / data distribution and sharing via 5G systems.

[0250] The following figures are created to explain specific examples of this specification. Since the names of specific devices or the names of specific signals / messages / fields described in the figures are presented as examples, the technical features of this specification are not limited to the specific names used in the following figures.

[0251] Fig.10 An example of AI / ML model deployment according to the disclosed implementation of this specification is shown.

[0252] AI / ML model distribution methods can be as follows Fig.10 As shown. The multifunctional mobile terminal may need to switch AI / ML models as the operation and environment change. The condition for adaptive model selection may be that the model to be selected is available for the mobile device. However, due to the fact that AI / ML models are becoming more and more diverse and the storage resources of the UE are limited, it may be decided not to preload all candidate AI / ML models. Online model deployment (i.e., downloading new models) may be required, through which the AI / ML model can be deployed from the NW endpoint to the device as needed to adapt to the changing AI / ML tasks and environment. To this end, it may be necessary to continuously monitor the model performance at the UE.

[0253] Fig.11 The example shows an example of distributed / federated learning over a 5G system.

[0254] The following figures are created to explain specific examples of this specification. Since the names of specific devices or the names of specific signals / messages / fields described in the figures are presented as examples, the technical features of this specification are not limited to the specific names used in the following figures.

[0255] Fig.11 An example of AI / ML model deployment according to the disclosed implementation of this specification is shown.

[0256] like Fig.11 As shown, an example of a federated learning (FL) system can be depicted.

[0257] The cloud server can train the global model by aggregating the partially trained local models on each terminal device. Within each training iteration, the UE can perform training based on the model downloaded from the AI ​​server using local training data. The UE can then report the intermediate training results to the cloud server via the 5G UL channel. The server can aggregate the intermediate training results of the UE and update the global model. The updated global model is then deployed back to the UE, and the UE can perform training for the next iteration.

[0258] The present disclosure may be implemented in one or more combinations, such as a combination including at least one of those described below. Although each drawing shows an embodiment of each disclosure, the embodiments of the drawings may be implemented in combination with each other.

[0259] The description of the solution proposed by the present disclosure may include any combination of one or more actions / components / steps described below. The following methods described below may be performed or used in combination or complementarily.

[0260] For example, 5GS support for FL operations can be discussed, for example, 5GS assistance for federated learning operations.

[0261] On the other hand, in 5G, support for services based on artificial intelligence (AI) / machine learning (ML) is being discussed. Application functions (AFs) such as application servers (AS) can perform federated learning (FL) operations. However, there is no way for the 5G system (5GS) to support these AFs to perform FL operations. For example, there is no specific discussion on how the 5GS provides auxiliary information to these AFs, nor is there any discussion on how the AFs request auxiliary information from the 5GS.

[0262] For example, it is necessary to discuss how 5GS will provide supporting information to assist AF in selecting UEs for FL behavior. It is also necessary to discuss what information 5GS will receive from AF and what information 5GS will provide to AF. In addition, in order for 5GS to support FL behavior for AI / ML applications, it is necessary to discuss how to collect information within 5GS based on the application's request (e.g., preferred time of day for performing FL behavior, area for FL behavior) and provide auxiliary information to AF, such as a list of candidate UEs for FL behavior and a recommended time of day for FL behavior.

[0263] It may be discussed whether and how 5GS provides support to AF and / or UE to manage FL operations and model distribution / redistribution (e.g., selecting FL members, monitoring group performance, and allocating and ensuring appropriate network resources). For reference, in various descriptions of the present disclosure, AF may be used interchangeably with application server (AS). This may facilitate joint learning operations based on collaborative application AI / ML between application clients and application servers running on UEs.

[0264] To support FL behavior of AF and UE, the following examples can be discussed:

[0265] Support for UE selection for FL behavior can be discussed. For example, whether, how, and what information the 5GC provides to the AF to assist the AF in selecting and managing the UE group that will be part of the FL behavior can be discussed. Note that the AF can control and manage FL group management. For example, whether, how, and what information from the 5GC is needed to support the AF in selecting and managing the UE group that will be part of the FL behavior can be discussed;

[0266] Performance monitoring / exposure may be discussed. For example, monitoring and exposing UE or UE group performance (e.g., aggregated QoS parameters) related to FL behavior may be discussed. For example, whether and which existing or new monitoring events (e.g., QoS, location, load, congestion) are needed to capture specific system performance and predictions for traffic associated with AI / ML operations for FL behavior may be discussed.

[0267] FL performance can be discussed. For example, when the application server receives local ML model training information from other UEs to perform global model updates, it can be discussed how to support AF to improve FL performance between UEs (e.g., manage latency deviation).

[0268] Various solutions have been proposed to assist FL operation of AI / ML applications in 5GS. For example, it has been proposed to provide FL-related information by collecting information in 5GS based on a request of an application (e.g., preferred FL operation time slots (e.g., preferred time of day for FL operation), regions for FL operation, and QoS information). For example, based on a request of an application, information in 5GS is collected, and information such as a list of suitable FL candidate UEs or a recommended FL time slot is proposed. In another example, it has been proposed to provide network-related information (e.g., QoS, etc.) about a UE requested by an application.

[0269] As in the previously described examples, the network function (NF) that collects information from the 5GS and generates auxiliary information to be delivered to the AI / ML application server may be a network data analysis function (NWDAF). Alternatively, such a NF may be a new NF (e.g., an AIML NF / NEF). In various examples of the present disclosure, an AIML NF / NEF may represent an NF that generates auxiliary information related to AIML. AIML NF may also be used interchangeably with AIML NF or NEF. Here, the new NF may act as a separate standalone NF, or it may be an NF integrated with an existing NF, NEF, or PCF.

[0270] If NWDAF is responsible for generating auxiliary information, a new analysis needs to be defined. For example, the new analysis can take as input a request for auxiliary information from an AI / ML application server and information collected from 5GC. Based on the new analysis, NWDAF can collect information from 5GC and, in response, provide auxiliary information to the application server as an output analysis.

[0271] A new NF (e.g., a separate standalone NF or an NF integrated with an existing NF) may also be used as a NF to generate auxiliary information. In this case, the new NF may receive a request for auxiliary information from an application server based on a new service operation. Then, in response, the new NF may provide the auxiliary information to the application server by collecting information from the 5GC.

[0272] Refer to the following Fig.12 The described example is an example of a method using a NF integrated with a NEF (referred to as an AIML NF). In this example, the AIML NF collects information from the network to provide a recommended UE list and a recommended FL time zone, region, etc. For example, the AIML NF can provide a recommended UE list and a recommended FL time zone, region, etc. that can satisfy the conditions (QoS, region, etc.) for the UE list provided by the application.

[0273] An example involving AF-related 5GS auxiliary information for joint learning behavior is described.

[0274] This example provides a solution around 5GS support for federated learning behaviors.

[0275] The following is an overview of the proposed solution to provide 5GS assistance information to the AF for joint learning behavior:

[0276] -The solution introduces an AIML NF that can be deployed with NEF and used for external exposure to support 5G AIML services. For example, the AIML NF can receive requests from the AF (i.e., AI / ML application server) and respond / notify the AF of 5GS assistance information for AIML behavior.

[0277] -The AI / ML application server requests auxiliary information together with input parameters for FL operation. For example, the input parameters may include a list of UEs, QoS references, geographic areas, a minimum number of UEs for FL operation, a preferred AIML duration, whether to request a list of candidate UEs for FL operation, a preferred time of day for AIML operation, radio access technology, signal quality, etc. For example, some FL operations may need to be performed during a specific time period, or some FL operations may not need to be performed immediately. In this case, the AI / ML application server can provide a preferred AIML task duration. In addition, depending on the size of the training data, the AI / ML application server can provide an estimated AIML task duration. By considering the preferred and expected AIML operation duration and predicting network conditions, such as user data congestion time, the AIML NF / NEF can derive a recommended AIML operation duration.

[0278] -The AI / ML application server may subscribe to (or subscribe to) AIML assistance information for FL behavior. For example, if the threshold of the member UE's QoS reference or QoS requirement cannot be met due to changes in network conditions, or if a new candidate UE is expected to meet the threshold, the AI / ML application server may attempt to update the FL member UE for the FL behavior after initiating the FL behavior. In this case, the AIML NF / NEF provides the AI / ML application server with updated assistance information (e.g., newly recommended UEs) so that the AI / ML application server can reselect FL member UEs that can meet the QoS requirements based on the assistance information.

[0279] Based on the subscription correlation ID, the AI / ML application server can manage the subscription (or multiple subscriptions). The AI / ML application server can use the subscription correlation ID to update the subscription (e.g., new UE list, QoS requirements) or unsubscribe the AIML NF / NEF.

[0280] -AIML NF / NEF can interact with other NFs such as NWDAF, UDM, SMF, UPF, OAM to process requests from AI / ML application servers. Based on the collected information, AIML NF / NEF generates auxiliary information for FL operation (e.g., a list of candidate UEs in the recommended time slot), thereby providing the auxiliary information for FL operation requested by the AI / ML application server.

[0281] Reference Fig.12 The example in describes an example process for providing 5GS auxiliary information for federated learning behavior.

[0282] Fig.12A process for providing 5GS assistance information to an AI / ML application server for federated learning behavior is shown.

[0283] The following figures are intended to illustrate specific embodiments of the present disclosure. The names of specific devices or specific signals / messages / fields shown in the figures are for illustrative purposes only, and the technical features of this specification are not limited to the specific names used in the following figures.

[0284] Fig.12 is an example of a process for providing 5GS auxiliary information for joint learning behavior according to an embodiment of the present disclosure.

[0285] 1. The AI / ML application server can request / subscribe to 5G assistance information with specific conditions for FL operation by calling Nnef_AIMLAssistanceInfo_Request / Nnef_AIMLAssistanceInfo_Subscribe. The AI / ML application server can request FL assistance information for the UE list included in the request. It can also indicate that the UE is authorized by the AI / ML application server (application side) to perform FL operation.

[0286] The input parameters of the Nnef_AIMLAssistanceInfo_Request service operation are as follows:

[0287] -AF identifier.

[0288] - List of UEs (GPSI, External Group ID or IP address).

[0289] - AF-Service-Identifier.

[0290] - External application identifier.

[0291] -QoS reference or QoS requirement.

[0292] - Whether to request a list of candidate UEs (if indicated, the AIML NF / NEF will provide a list of candidate UEs in the response).

[0293] - Time when auxiliary information is required (indicates to the AIML NF / NEF the latest time that the analytics consumer expects to receive supporting data provided by the AIML NF / NEF).

[0294] - Expected AIML behavior time (estimated time per AI / ML application server based on training information such as training model, training data size)

[0295] - Preferred AIML behavior time period (the time period during which the AI / ML application server wants to perform AIML operations).

[0296] - Whether to request a recommended (suggested) AIML operation time period (if indicated, the AIML NF / NEF provides the recommended AIML operation time period in the response).

[0297] -Traffic volume used to download global models and upload training results.

[0298] -Traffic volume used to download global models and upload training results.

[0299] -DNN and S-NSSAI.

[0300] - The expected UE movement trajectory (eg, planned movement path) of each UE in the UE list.

[0301] - Area of ​​interest (for AIML operations of UEs within the area of ​​interest).

[0302] - Minimum number of UEs required for FL behavior.

[0303] - Preferred Radio Access Technology (Access type / RAT type (e.g. WLAN, 5G) if the AI / ML application server expects the UE to maintain FL behavior).

[0304] - For WLAN, preference is given to signal quality (requirements of UE radio signal for FL operation, e.g., average received signal strength indication (RSSI) and round trip time (RTT)).

[0305] -UE traffic usage (expected UE traffic usage or remaining traffic range for FL behavior).

[0306] The additional input parameters for the Nnef_AIMLAssistanceInfo_Subscribe service operation are as follows:

[0307] - In case of AI / ML application server subscribing to AIML NF / NEF, notification target address (+ notification correlation ID).

[0308] - Reporting parameters (e.g., periodic reporting, reporting thresholds (QoS reference or QoS requirement)).

[0309] 2. If the request from the AI / ML application server is approved, the AIML NF / NEF calls the Nudm_SDM_Get service operation to check from the UDM stored in the UE subscription information whether the UE provided by the AI / ML application server is authorized by the 5GC to support AIML operations and whether each UE agrees to participate in FL operations. After receiving a successful response, the AIML NF / NEF can subscribe to notifications about modifications to the subscription data by using Nudm_SDM_Subscribe.

[0310] 3. For AIML operation, for UEs authorized by 5GC, the AIML NF / NEF can derive auxiliary information by determining appropriate analysis information and input parameters to be collected based on the input requested in step 1. For example, the auxiliary information can be a list of candidate UEs that can satisfy the request of the AI / ML application server or a recommended coverage area in a recommended time slot.

[0311] To request / subscribe to analysis information from NWDAF, the input from step 1 can be used as input parameters. For example, QoS criteria, desired AIML operation time period, and preferred AIML operation time period can be used to request QoS sustainability analysis at that point in time. In addition, for NWDAF, "time when auxiliary information is required" can be used as input for "time when analysis information is required". Global model download and training result upload traffic can be used to estimate the time spent on global model download and training result upload.

[0312] The AIML NF / NEF may request / subscribe to analysis information from the NWDAF in order to collect the requested analysis information of the UE (e.g., UE communication analysis, UE mobility analysis, user data congestion analysis, QoS persistence analysis, WLAN performance analysis per UE). If some input parameters are not provided by the AI / ML application server, the AIML NF / NEF may derive them based on local configuration or based on analysis such as UE mobility analysis and WLAN performance analysis per UE. If "preferred AIML operation time period" and "time when auxiliary information is required" are not provided as input parameters, it may be considered that the AIML operation will be performed immediately.

[0313] 4. NWDAF collects data for the requested analysis ID.

[0314] 5. NWDAF derives the requested analysis.

[0315] 6. NWDAF provides analysis response to AIML NF / NEF.

[0316] 7. The AIML NF / NEF collects all the analysis information from the NWDAF and takes into account factors such as QoS, user data congestion, UE mobility, preferred AIML time period for FL operation, preferred radio access technology, preferred signal quality and UE service usage to derive auxiliary information that can satisfy the request from the AI / ML application server.

[0317] 8. The AIML NF / NEF responds / notifies the AI / ML application server of the derived auxiliary information.

[0318] The output parameters of the Nnef_AIMLAssistanceInfo_Request service operation can be as follows:

[0319] -A collection of 5GS assistance information requested by the AI / ML application server. Each 5GS assistance information may contain:

[0320] -A collection of 5GS assistance information requested by the AI / ML application server. Each 5GS assistance information may contain:

[0321] - List of candidate UEs for FL operation.

[0322] - Recommended time period for performing AIML operations.

[0323] - Recommended regions for AIML operations.

[0324] -Recommended access type / RAT type for performing AIML operations.

[0325] Each of the above information may be provided in priority order to help the AI / ML application server select a UE.

[0326] The list of candidate UEs for the recommended FL operation may differ according to the recommended time period and / or recommended area for the AIML operation.

[0327] If a "Time when auxiliary information is required" is provided in step 1, the AIML NF / NEF takes this into account when responding to the AI / ML application server.

[0328] Based on the received 5GS assistance information, the AI / ML application server can select and manage the group of UEs and determine the start time of the FL operation.

[0329] 9-13. If in step 1 and step 3, the AI / ML application server and the AIML NF / NEF subscribe to the AIML auxiliary information, when the NWDAF generates a new analysis, the AIML NF / NEF notifies the AI / ML application server of the new auxiliary information.

[0330] After the FL operation starts, due to changed network conditions, the AIML NF / NEF may notify new AIML assistance information (e.g., new recommended UEs), and the AI / ML application server may reselect FL members based on this information.

[0331] according to Fig.12 In the example above, UDM can use new subscription information to determine whether the UE is authorized for 5G AI / ML operations. AIML NF can collaborate with NEF at the same location. AIML NF / NEF can use new service operations to support application AI / ML operations. NWDAF can support WLAN performance analysis per UE.

[0332] In the following, reference will be made to Fig.13a and Fig.13b An example of providing auxiliary information is described as follows. Fig.13a and Fig.13b In the following example, a new analysis used by the NWDAF may be defined. According to the following example, the NWDAF may receive a request from an AI / ML application server as input for the analysis, and provide auxiliary information to the AI / ML application server as output. In the following example, the AF may send a member selection flag. If the member selection flag is 1, the NWDAF of the 5GC may provide a recommended UE list as the output of the new analysis. If the member selection flag is 0, the NWDAF may not recommend a list of members to the AIML application server. Alternatively, the NWDAF may collect information about the member list delivered by the application server (such as QoS information or network performance), and collect information from the 5GC by subscribing to a notification service for data related to other NFs (such as UPF and SMF). Based on the collected information, the NWDAF may deliver auxiliary information to the AI / ML application server as analysis output.

[0333] Reference Fig.13a and Fig.13b The example in describes an example of a 5G VN group that utilizes federated learning supported by 5GC.

[0334] The following is an example of a 5GS Assistance to Federated Learning Operation.

[0335] In federated learning (FL), multiple rounds of training and interaction are required between the central server and edge nodes (e.g., edge servers or UEs). Federated learning can be regarded as a type of group learning that does not expose the privacy of edge nodes. In order to improve the efficiency of federated learning, it is necessary to ensure a more reliable connection between edge nodes and the central server, as well as the consistency of communication characteristics and requirements of each edge node, such as resources, bandwidth, latency, and unified training start and end.

[0336] The following example proposes configuring the edge nodes and central server of the federated learning into a 5G virtual network (VN) group to facilitate the unified management of members within the LAN group, resource allocation, access and mobility related policies, and coordination of session management related policies. Here, VN can be an example of a technology for enabling a 5G network to support group communication.

[0337] The NWDAF will provide the AF with a proposal for establishing a VN group and related policies, and if the AF needs to select members, it may also include a list of proposed FL members. This process requires multiple interactions with other 5GC NFs, so the AF is deeply involved.

[0338] Reference Fig.13a and Fig.13b , describes an example of the VN group establishment process of AF. Fig.13a and Fig.13b The VN group establishment process initiated by the AF is illustrated.

[0339] The following figures are intended to illustrate specific embodiments of the present disclosure. The specific device names or specific signal / message / field names shown in the figures are for illustrative purposes only, and the technical features of this specification are not limited to the specific names used in the following figures.

[0340] Fig.13a and Fig.13b is an example process for establishing a VN group by an AF according to one embodiment of the present disclosure.

[0341] 1.AF subscribes to receive analytics via NEF by calling the Nnef_AnalyticsExposure_Subscribe service operation. Parameters include:

[0342] 1.AF can subscribe to receive analytics through NEF by calling Nnef_AnalyticsExposure_Subscribe service operation. The message sent by AF can contain the following parameters:

[0343] -Analysis ID.

[0344] -Member selection flag: A flag indicating whether member selection is required.

[0345] -Analyze filter information, including:

[0346] - Member list (list of SUPIs or group IDs if the member selection flag is 1), candidate member list (list of SUPIs or group IDs if the member selection flag is 1), or all UEs (if the member selection flag is 1).

[0347] - [Optional] Maximum and / or minimum number of members (if the member selection flag is 1).

[0348] - Area of ​​interest.

[0349] -S-NSSAI or DNN.

[0350] - Desired time period.

[0351] - Information about the ML model to be trained (e.g., model size).

[0352] -Performance requirements (e.g. bandwidth, time delay, etc.).

[0353] 2. Based on the subscription from AF, NEF subscribes to NWDAF by calling the Nnwdaf_AnalyticsSubscription_Subscribe service operation.

[0354] 3. If the member selection flag is 1 and the user consent is authorized, the NWDAF can start selecting members from the UE list provided by the AF. The NWDAF can perform "UE mobility analysis", "abnormal behavior related network data analysis", "network performance analysis", "NF load analysis", "user data congestion analysis", "DN performance analysis" and other analysis IDs to select appropriate members.

[0355] 4. NWDAF collects relevant analysis data and performs analysis. Based on the analysis results, NWDAF can recommend a VN group and then derive recommended policy-related information, which may include VN group data, PCC rule-related information, and access and mobility-related rules (e.g., URSP). If the member selection flag is 0, NWDAF can subscribe to QoS information, network performance, and other relevant data from the UPF, SMF, and other relevant NFs on the member list. Whether NWDAF recommends establishing a VN group can take into account the capabilities of the SMF for the VN group.

[0356] 5. NWDAF may notify NEF of the proposed strategy for establishing a VN group (including PCC rules and access and mobility related policies), as well as the list of selected members matching the analysis filter information (if the member selection flag is 1) by calling the Nnwdaf_AnalyticsSubscription_Notify service operation.

[0357] 6. NEF notifies AF of the analytics information by calling the Nnef_AnalyticsExposure_Notify service operation.

[0358] 7. If the member selection flag is 1, the AF may modify the proposed member list based on local policy. At this point, the AF may cancel the previous subscription and send a new subscription to the NWDAF to perform group monitoring on the updated member list.

[0359] 8. Based on the subscription from AF, NEF subscribes to NWDAF by calling Nnwdaf_AnalyticsSubscription_Subscribe service operation. NWDAF can subscribe QoS information, network performance and other information to UPF, SMF and other related NFs on the updated member list.

[0360] 9. The AF may send the VN group data defined in clause 4.15.6.3b of 3GPP TS23.502 V17.5.0 via the UDR using the procedure specified in clause 4.15.6.7 of TS23.502 V17.5.0. Based on the VN group data, the PCF may generate URSP rules and deliver them to the UE using the procedure specified in clause 4.2.4.3 of TS23.502 V17.5.0.

[0361] 10. The AF may send the dedicated PCC rules for the VN group defined in 6.3 of TS 23.503 V17.5.0 to the PCF via the UDR using the procedure specified in clause 4.15.6.7 of TS 23.502 V17.5.0.

[0362] 11. AF can trigger the PDU session establishment process.

[0363] 12. A UE that has previously received the URSP rules may establish a PDU session and join (or subscribe to) a VN group using the procedure in clause 4.3.2.2.1 of TS 23.502 V17.5.0.

[0364] 13. The VN group with established AF can start joint learning with the selected UE.

[0365] 14. During the learning period, AF can send FL progress and resource requirements to NWDAF via NEF. UPF, SMF and other related NFs can send QoS information, network performance and other related information to NWDAF.

[0366] 15. The NWDAF may modify the proposed membership and policies, including VN group data, PCC rules, or access and mobility related policies (e.g., URSP).

[0367] 16. NWDAF notifies NEF of the updated recommended member list (if the member selection flag is "1") and the updated recommended strategy by calling the Nnwdaf_AnalyticsSubscription_Notify service operation.

[0368] 17. NEF may notify AF of the analytics information by calling Nnef_AnalyticsExposure_Notify service operation. If AF modifies the suggested member list based on local policy, AF may cancel the previous subscription and send a new subscription on the updated member list.

[0369] 18. The AF may modify the VN group data and / or members via the UDR using the procedures specified in clause 4.15.6.7 of TS23.502 V17.5.0. The PCF may modify the URSP rules based on the VN group data and transmit them to the UE using the procedures specified in clause 4.2.4.3 of TS23.502 V17.5.0.

[0370] 19. The AF may modify the PCC rules for a set of VNs using the procedure specified in paragraph 4.15.6.7 of TS23.502 V17.5.0. The FL operation may include multiple rounds of training. Steps 14 to 19 may be repeated as many times as required by the AF to NWDAF subscription requirements.

[0371] according to Fig.13a and Fig.13b For example, NWDAF can support application federated learning related analysis (e.g., new analysis ID). NEF can support receiving application federated learning related subscription requests.

[0372] There is an ongoing discussion about Federated Learning (FL). However, what has not been discussed in the past is how to perform FL-related operations efficiently.

[0373] Terminals supporting MA PDU sessions utilize both accesses (3GPP and non-3GPP). When selecting UEs to meet the QoS for AIML FL operation, it is necessary to have a way to prioritize UEs with MA PDU sessions.

[0374] For example, in 5GC, MAPDU session information can be used as information for selecting UEs that can meet the QoS of AIML FL behavior for the terminal. Terminals with MA PDU sessions can have 3GPP and non-3GPP access. Therefore, compared with terminals that use only a single PDU session, terminals with MA PDU sessions can flexibly meet the QoS of the terminal even if the performance of the access network changes. Therefore, the AI / ML application server may want to give priority to selecting terminals that support MA PDU sessions. Even if the AI / ML application server does not explicitly request information about terminals with MA PDU sessions, 5GC may give priority to terminals with MAPDU sessions as a list of candidate UEs for FL. Therefore, in order to support FL operations of AI / ML applications in 5GC, MA PDU session related information needs to be provided to the AF.

[0375] In the following, an example of utilizing an MA PDU session is described.

[0376] A multi-access PDU (MA PDU) session is a PDU session that can simultaneously connect to two access networks (3GPP access and non-3GPP access (e.g., untrusted, trusted non-3GPP access, wired 5G access)) and send and receive services between the terminal and the data network. MA PDU sessions are defined to support access service steering, switching, and splitting (ATSSS) technology, which enables the simultaneous use of 3GPP and non-3GPP access networks to send and receive services between the terminal and the network.

[0377] For MA PDU sessions, two tunnels are created, one for carrying services over 3GPP access and one for carrying services over non-3GPP access. Each tunnel is independently isolated and only one of them can be used to forward services. The choice of which access tunnel to use to send the service is determined by policies determined by the network. Based on these policies, the SMF can generate ATSSS rules and N4 rules to control the service and send them to the terminal and UPF respectively. Based on this, the terminal can decide to which access the UL service should be sent, and the UPF can decide to which access the DL service should be sent.

[0378] The following figures are intended to illustrate specific embodiments of the present disclosure. The specific device names or specific signal / message / field names shown in the figures are for illustrative purposes only, and the technical features of this specification are not limited to the specific names used in the following figures.

[0379] Fig.14 is an example of the structure of an MA PDU session according to one embodiment of the present disclosure.

[0380] The UE may support ATSSS and the UE may want to enable MA PDU sessions. In this case, the UE may generate a PDU session establishment request including ASSSS capability and indicate the request type as "MA PDU request" in the UL NAS transport message when sending the request.

[0381] If the AMF supports MA PDU sessions, the AMF may select an SMF that supports MA PDU sessions when creating a PDU session. The AMF may then send a request message to the SMF by passing an "MA PDU Request" indication to the SMF and including the access type in the N1 SM container. In addition, if the UE is registered to two accesses, the AMF may notify the SMF of additional access type information.

[0382] Based on the session management subscriber information, the SMF can determine whether to allow the MA PDU session. If a dynamic PCC is used for the MA PDU session, the SMF can send an SM policy control create message with an "MA PDU request" indication to the PCF, and the SMF can send the ATSSS capability of the MAPDU session. The SMF can send the currently used access type, RAT type, and additional access type information to the PCF. The PCF then decides whether to allow the MA PDU session based on the network operator policy and subscriber information.

[0383] A terminal with an MA PDU session can have both 3GPP access and non-3GPP access, so that even if one access becomes unavailable, it can communicate with the other access, and the bandwidth of the two accesses can be aggregated. The terminal and / or the network can perform round-trip time (RTT) measurements and packet loss rate (PLR) measurements for each access network. Based on measurements such as RTT and PLR measurements, the terminal and / or the network can determine congestion on a specific access and select an access that can maximize the service transmission rate of the corresponding uplink and downlink to send the service.

[0384] Therefore, compared with a terminal using a single PDU session, a terminal supporting an MA PDU session can flexibly meet the QoS of the terminal even when the performance of the access network changes. Therefore, the AI / ML application server may want to give priority to selecting a terminal supporting an MA PDU session. In addition, even if the AI / ML application server does not explicitly request information about a terminal with an MA PDU session, the 5GC may give priority to a terminal with an MAPDU session as a list of candidate UEs for FL.

[0385] Various examples in the present disclosure describe how 3GPP 5GC can provide a list of candidate UEs having MA PDU session information of a terminal to an AI / ML application server, or how to provide a list of candidate UEs to an AI / ML application server taking into account the MA PDU session information as information to support application AI / ML FL operations.

[0386] Various examples of the present disclosure describe how to provide 5GS assistance information (including MAPDU session information) to an AF supporting AI and ML services. The method may include a combination of one or more actions / configurations / steps according to the following various examples.

[0387] The behaviors described in the various examples disclosed in this specification can be supported by the following NFs:

[0388] - A new NF that supports applying AI / ML behaviors (e.g., AIML NF); or

[0389] - A NF that integrates the above new or existing NFs (NEF, PCF) with functionality that supports application AI / ML behaviors (e.g., AIML NF / NEF); or

[0390] -NWDAF.

[0391] The SMF or PCF may provide the UE's MA PDU session information to the NWDAF, AIML NF / NEF, AIML NF, etc. The MA PDU session information may indicate whether the terminal is using an MA PDU session, access information for which an MA PDU session has been established, and / or whether a single access PDU session can be changed to an MA PDU session.

[0392] For example, MA PDU session information may be interpreted as information related to the UE's ability to simultaneously use 3GPP access and non-3GPP access for application AI / ML behavior / service support. For example, whether the UE is using an MA PDU session may be interpreted as whether the UE is simultaneously using 3GPP access and non-3GPP access to support application AI / ML operations / services.

[0393] Various examples of the present disclosure describe the process of operating based on the new service operation on the AIML NF / NEF. In various examples of the present disclosure, the NWDAF can support the behavior of the AIML NF / NEF. In this case, the input / output parameters of the AIMLNF / NEF described below can be supported by the input / output of the new analysis in the NWDAF. If another NF supports the AIMLNF / NEF behavior, the new service operation provided by the NF can support the AIML NF / NEF behavior and the parameters associated with the behavior.

[0394] In the various examples disclosed herein, AI / ML services may be used interchangeably with AI / ML-based services, AI / ML businesses, AI / ML applications, AI / ML transmissions, AI / ML operations, AI / ML operation businesses, applied AI / ML operations, applied AI / ML operation businesses, etc.

[0395] In various examples disclosed herein, the terms user equipment (UE) and terminal may be used interchangeably.

[0396] In various examples disclosed herein, the terms application function (AF) and application server (AS) may be used interchangeably.

[0397] In various examples disclosed herein, analytics information may be used interchangeably with analytics data, network analytics data, network analytics information, and the like.

[0398] In the following, we focus on what is suggested by various examples of the present disclosure.For network data analysis according to the prior art, see TS23.288 V17.5.0.

[0399] According to various examples of the present disclosure, Fig.15a and Fig.15b FIG. 4 shows an example process of a 3GPP network delivering 5GC auxiliary information (including MA PDU session information of a terminal) to an AI / ML application server.

[0400] The following figures are intended to illustrate specific embodiments of the present disclosure. The specific device names or specific signal / message / field names shown in the figures are for illustrative purposes only, and the technical features of this specification are not limited to the specific names used in the following figures.

[0401] Fig.15a and Fig.15b An example of a process for providing 5GS assistance information for federated learning operations is shown.

[0402] Fig.15a and Fig.15b A process for providing 5GS assistance information (including MA PDU session information) for joint learning behavior is shown.

[0403] depending on Fig.15a and Fig.15b The contents proposed in the Fig.12 For example, you can modify Fig.12 Step 1 of , so that the request from the AI / ML application includes the MA PDU session information. For example, you can modify Fig.12Step 2 of the MAPDU session information to enable the AIMF NF / NEF to obtain subscription related information from the UDM in the MAPDU session information. For example, you can add Fig.15a Step 3 and Fig.15b Step 10 of the procedure to obtain the MA PDU session information from the SMF or PCF. Fig.15a and Fig.15b Step 9 and step 15 are used to provide information including MAPDU session information to the AI / ML application server.

[0404] 1. The AI / ML application server can request assistance information in order to use AI / ML specific behaviors (e.g., FL behaviors). For example, the AI / ML application server can send a request message for assistance information or a subscription request message to subscribe to a service that provides assistance information to the AIML NF / NEF. For example, the AI / ML application server can request / subscribe to 5G assistance information for FL operations under certain conditions by calling Nnef_AIMLAssistanceInfo_Request / Nnef_AIMLAssistanceInfo_Subscribe. The AI / ML application server can request FL assistance information for the UE list included in the request. It can also indicate that the UE is authorized by the AI / ML application server (application side) to perform FL operations.

[0405] The Nnef_AIMLAssistanceInfo_Request service operation has the following input parameters. For example, the request message sent by the AI / ML application server may include one or more of the following information, as shown in the following example:

[0406] -AF identifier.

[0407] - List of UEs (GPSI, External Group ID or IP address).

[0408] - AF-Service-Identifier.

[0409] - External application identifier.

[0410] -QoS reference or QoS requirement.

[0411] - Whether to request a list of candidate UEs (if indicated, the AIML NF / NEF will provide a list of candidate UEs in the response).

[0412] - Time when auxiliary information is required (indicates to the AIML NF / NEF the latest time that the analytics consumer expects to receive supporting data provided by the AIML NF / NEF).

[0413] - Expected AIML behavior time (estimated time for each AI / ML application server based on training information such as training model, training data size)

[0414] - Preferred AIML behavior time period (the time period during which the AI / ML application server wants to perform AIML operations).

[0415] - Whether to request the recommended AIML operation time period (if indicated, the AIML NF / NEF will provide the recommended AIML operation time period in the response).

[0416] -Traffic volume used to download global models and upload training results.

[0417] -Traffic volume used to download global models and upload training results.

[0418] -DNN and S-NSSAI.

[0419] - The expected UE movement trajectory (eg, planned movement path) of each UE in the UE list.

[0420] - Area of ​​interest (for AIML operation of UEs in the area of ​​interest).

[0421] - Minimum number of UEs required for FL operation.

[0422] - Preferred Radio Access Technology (Access Type / RAT Type (e.g. WLAN, 5G, Multi-Access) if the AI / ML Application Server expects the UE to maintain FL operation). For example, the AI / ML Application Server may send a request message containing "Multi-Access" to the AIML NF / NEF to indicate that it prefers UEs with MA PDU sessions for FL operation.

[0423] - For WLAN, preference is given to signal quality (requirements of UE radio signal for FL operation, e.g., average received signal strength indication (RSSI) and round trip time (RTT)).

[0424] -UE traffic usage (UE traffic usage or expected range of remaining traffic for FL operation).

[0425] - Preference for multi-access UEs. For example, the AI / ML application server may include preference information for multi-access UEs to indicate whether it prefers UEs with MA PDU sessions.

[0426] -Multiple access PDU session information of UE. The AI / ML application server may include "MAPDU session information of UE" in the request message to request auxiliary information including MAPDU session information from the AIML NF / NEF.

[0427] The AI / ML application server may send a request message to the 5GC requesting auxiliary information. The AI / ML application server may include information related to the MA PDU session in the input parameters used to obtain the 5GC auxiliary information. Based on the input parameters used to obtain the 5GC auxiliary information, the ML application server may indicate its preference for UEs connected with two access types (for which MA PDU sessions are created / allowed / supported) and request the MA PDU session information for each UE. For example, the AI / ML application server may include in the request message information indicating its preference for UEs connected with two access types (for which MA PDU sessions are created / allowed / supported) and / or requesting the MA PDU session information for each UE. Alternatively, the AI / ML application server may request a list of UEs with MA PDU sessions, UEs allowed to have MAPDU sessions, UEs supporting MAPDU sessions, or UEs connected with two access types as recommended candidate UEs. The request from the AI / ML application server may be sent to the AIML NF / NEF.

[0428] The additional input parameters for the Nnef_AIMLAssistanceInfo_Subscribe service operation are as follows:

[0429] - Notification target address (+ Notification Correlation ID) (if AI / ML Application Server subscribes to AIML NF / NEF).

[0430] - Reporting parameters (e.g., periodic reporting, reporting thresholds (QoS reference or QoS requirement).

[0431] 2-3. If the request from the AI / ML application server is approved, the AIML NF / NEF calls the Nudm_SDM_Get service operation to check from the UDM stored in the UE subscription information whether the UE provided by the AI / ML application server is authorized by the 5GC to support AIML operations and whether each UE agrees to participate in FL operations. After receiving a successful response, the AIML NF / NEF can subscribe to notifications about modifications to the subscription data by using Nudm_SDM_Subscribe.

[0432] The AIML NF / NEF may decide to collect information related to the MA PDU session for the FL candidate UE. For example, if there is a request from the AI / ML application server for information related to the MA PDU session for the FL candidate UE, the AIML NF / NEF may decide to collect information related to the MA PDU session for the UE. Alternatively, even if such a request is not explicitly made, the AIML NF / NEF may decide to collect information related to the MA PDU session for the FL candidate UE based on preset information. The AIML NF / NEF may subscribe to the notification service (event exposure) of each UDM / SMF / PCF using an event ID to obtain (or receive) information related to the MA PDU session from the UDM / SMF / PCF of the UE. New event IDs associated with the MA PDU session information may be defined, and the AIML NF / NEF may receive information related to the MA PDU session from the UDM / SMF / PCF by sending these new event IDs. The AIML NF / NEF may subscribe to receive information related to the MA PDU session from the UDM / SMF / PCF either once or periodically or whenever the information changes.

[0433] The MA PDU Session Information event may include the following information for a specific DNN / S-NSSAI. The MA PDU Session Information event may be an event based on a new event ID associated with the previously described MA PDU Session Information. For example, the MA PDU Session Information event may be an event that provides MA PDU Session Information to the AIML NF / NEF. The MA PDU Session Information may include one or more of the following information examples.

[0434] -Whether UE supports MAPDU session: AIML NF / NEF can receive from SMF or PCF whether UE supports MA PDU session.

[0435] -Whether the UE allows the MA PDU session (e.g., information related to whether the UE accepts the MA PDU session): The AIML NF / NEF may receive information related to whether the MA PDU session is allowed from the SMF or UDM. For example, whether the MA PDU session is allowed may be stored in the "Indication of whether to allow the establishment of the MA PDU session" in the ATSSS information included in the SM subscription data.

[0436] -Whether the UE's MAPDU session is currently established: AIML NF / NEF can receive information related to whether the UE's MAPDU session is currently established from SMF or PCF.

[0437] The AIML NF / NEF may obtain information from the SMF and / or PCF, such as the following examples. For example, the AIML NF / NEF may obtain the following information from the PCF by utilizing the access type change event of the PCF. For example, the AIML NF / NEF may obtain the MAPDU session information by extending the access type change event of the SMF to include the access type change event of the PCF to obtain the MAPDU session information from the SMF:

[0438] - Information about which access type the UE's MAPDU session is connected to (e.g. 3GPP, non-3GPP, both);

[0439] AIML NF / NEF can obtain additional access types via AdditionalAccessInfo in the access type change event of PCF. "AdditionalAccessInfo" is not currently defined in the access type change event of SMF, but it can be extended like PCF. For example, AdditionalAccessInfo parameter can be newly defined in the access type change event used by SMF. Based on the AdditionalAccessInfo parameter, additional access type information can be provided. For example, additional access type information can be added to the default access type information to indicate that it is a multi-access PDU session; and / or

[0440] The conventional access type AdditionalAccessInfo does not indicate whether the user plane resources of the MAPDU session are actually allocated. For example, when the terminal is registered in both accesses, an MAPDU session can be generated. In this case, when the terminal enters the CM-IDLE state in the 3GPP access due to the inactivity timer, the user plane resources can be released. Therefore, a new service can be used to indicate whether the actual user plane resources have been established. Especially for non-3GPP access, when the terminal enters the IDLE state, the network may not be able to establish user plane resources, so the actual establishment of user plane resources may be important. For example, the AS can know whether the actual user plane resources have been established.

[0441] The AIML NF / NEF may know whether the PDU session of the DNN / S-NSSAI used by the UE for FL operation supports the MA PDU session, whether it is connected to the MA PDU session, and / or information related to which access type the PDU session is connected to.

[0442] 4. AIML NF / NEF may request analysis information from NWDAF to derive auxiliary information. To this end, AIML NF / NEF may send a request message to NWDAF to request analysis information, or AIML NF / NEF may send a subscription request message to NWDAF to subscribe to the service of sending analysis information.

[0443] For example, for AIML operation, for UEs authorized by 5GC, the AIML NF / NEF can derive auxiliary information by determining appropriate analysis information and input parameters to be collected based on the input requested in step 1. For example, the auxiliary information can be a list of candidate UEs that can satisfy the request of the AI / ML application server or a recommended coverage area in a recommended time slot.

[0444] To request / subscribe to analysis information from NWDAF, the input from step 1 can be used as input parameters. For example, QoS criteria, desired AIML operation time period, and preferred AIML operation time period can be used to request QoS sustainability analysis at that point in time. In addition, for NWDAF, "time when auxiliary information is required" can be used as input for "time when analysis information is required". Global model download and training result upload traffic can be used to estimate the time spent on global model download and training result upload.

[0445] The AIML NF / NEF may request / subscribe to analysis information from the NWDAF to collect the requested analysis information of the UE (e.g., UE communication analysis, UE mobility analysis, user data congestion analysis, QoS sustainability analysis, WLAN performance analysis per UE). If some input parameters are not provided by the AI / ML application server, the AIML NF / NEF may derive them based on local configuration or based on analysis such as UE mobility analysis and WLAN performance analysis per UE. If "preferred AIML operation time period" and "time when assistance information is required" are not provided as input parameters, it may be assumed that the AIML operation is performed immediately.

[0446] 5. NWDAF may collect data for the requested analysis ID.

[0447] 6. NWDAF can derive the requested analysis.

[0448] 7. NWDAF can provide analysis response to AIML NF / NEF.

[0449] 8. AIML NF / NEF can collect all analytical information from NWDAF and derive auxiliary information to satisfy the request of AI / ML application server by taking into account QoS, user data congestion, UE mobility, preferred AIML time zone for FL operation, preferred radio access technology, preferred signal quality and UE traffic usage, etc.

[0450] 9. AIML NF / NEF can respond / notify the derived auxiliary information to the AI / ML application server.

[0451] Based on the MA PDU session information and other network information (e.g., network analysis information from NWDAF, etc.), the AIML NF / NEF can generate auxiliary information to be delivered to the AI / ML application server. The AIML NF / NEF can then forward the auxiliary information to the AI / ML application server.

[0452] The output parameters of the Nnef_AIMLAssistanceInfo_Request service operation can be as follows:

[0453] -A collection of 5GS assistance information requested by the AI / ML application server. Each 5GS assistance information may contain:

[0454] -A collection of 5GS assistance information requested by the AI / ML application server. Each 5GS assistance information may include:

[0455] - List of candidate UEs for FL operation.

[0456] - Recommended time period for performing AIML operations.

[0457] - Recommended regions for AIML operations.

[0458] -Recommended access type / RAT type for performing AIML operations.

[0459] -Multi-access PDU session information (eg, the MA PDU session information may be MA PDU session information for one or more UEs included in the candidate list).

[0460] For the complete list of UEs received from the AI / ML application server, the AIML NF / NEF may send the MA PDU session information of all UEs or the MA PDU session information of specified candidate UEs to the AI / ML application server. For example, the MA PDU session information may include whether the UE supports MAPDU sessions, whether the UE allows MA PDU sessions, whether the UE's MA PDU session is connected, and / or the access type to which the UE is currently connected. Alternatively, the AIML NF / NEF may not send the MAPDU session information directly to the AI / ML application server, but when the AIML NF / NEF selects candidate UEs, the AIML NF / NEF may consider the MAPDU session information when recommending a list of UEs (e.g., giving priority to UEs with MA PDU sessions).

[0461] Even if the AIML NF / NEF does not send the MA PDU session information directly to the AI / ML application server, the AIML NF / NEF may consider the MA PDU session information to recommend a list of candidate UEs. In this case, the AIML NF / NEF may also consider the preferred radio access technology information sent by the AI / ML application server in step 1. For example, the AIML NF / NEF may generate a list of candidate UEs based on the AI / ML application server, as shown in the following example (if the AI / ML application server does not send the preferred radio access technology, the AIML NF / NEF may determine the preferred radio access technology according to the AIML NF / NEF internal settings or according to a preset, and perform an action to generate a list of candidate UEs). In the following, reference is made to the first to third examples:

[0462] As a first example, the AI / ML application server may request the AIML NF / NEF to select 40 UEs from a list of 100 UEs and send information to the AIML NF / NEF that the preferred radio access technology is multi-access. In step 8, the AIML NF / NEF may select more UEs than requested by the AI / ML application server by considering the MAPDU session information and analysis information (e.g., QoS, user data congestion, UE mobility, etc.) obtained from the NWDAF, and after selecting more UEs that can meet the requirements of the AI / ML application server, perform the following operations 1) to 4). For example, the AIML NF / NEF may first select more UEs (50 UEs) than the 40 UEs that can meet the requirements of the AI / ML application server. Then, the AIML NF / NEF may perform any of the following actions in Examples 1) to 4) for these 50 UEs:

[0463] 1) The AIML NF / NEF may first select UEs (e.g., 10 UEs) whose PDU sessions for AI / ML operations are created as MA PDU sessions and connected with both 3GPP access and non-3GPP access;

[0464] 2) AIML NF / NEF is connected to 3GPP access or non-3GPP access, but uses MA PDU session, so it can select UEs that can also connect to other accesses (10 UEs);

[0465] 3) AIML NF / NEF can select UEs (10 UEs) for which MAPDU sessions are allowed (even if MAPDU sessions have not been created) based on subscriber information;

[0466] 4) There may be 20 UEs for which only single access PDU sessions are supported / allowed. For these UEs, the AIML NF / NEF can select 10 UEs that can best meet the requirements of the AI / ML application server by considering the analysis collected in step 8, without considering the MA PDU session information.

[0467] The AIML NF / NEF may select 40 UEs based on the same behaviors as 1) to 4) above.

[0468] It should be noted that the above operations 1) to 4) are only illustrative and the scope of the present disclosure is not limited by the above operations. Based on both the MAPDU session information and the analysis information obtained from the NWDAF, the AIML NF / NEF can select a UE that can meet the requirements of the / ML application server.

[0469] As a second example, consider the case where the AI / ML application server requests 40 candidate UEs out of 100 UEs from the AIML NF / NEF, requesting that the preferred radio access technology is multi-access, followed by Wi-Fi. In this case, the AIML NF / NEF can select 40 UEs that can meet the requirements of the AI / ML application server from 50 UEs in the following manner:

[0470] 1) AIML NF / NEF prioritizes and selects UEs connected with both 3GPP access and non-3GPP access (10 UEs), where the PDU session for AI / ML operation is created as MA PDU session;

[0471] 2) AIML NF / NEF selects UEs (10 UEs) connected with non-3GPP access, connected with MAPDU session and then connected with 3GPP access;

[0472] 3) AIML NF / NEF can select UEs (20 UEs) that are connected to non-3GPP access and only support / allow single access PDU sessions. For example, there may be less than 20 UEs connected to non-3GPP access and only 10 UEs support / allow single access PDU sessions. In this case, AIML NF / NEF can select the remaining 10 UEs by first selecting UEs with MAPDU sessions (5 UEs) among those UEs connected to 3GPP access, and then selecting UEs with single PDU sessions (5 UEs) connected to 3GPP access.

[0473] In the third example, the AI / ML application server requested 40 UEs, but the number of UEs (30) that can meet the AI / ML application server requirements (e.g., QoS) is less than 40. In this case, the AIML NF / NEF may perform the actions in the following examples 1) and 2):

[0474] 1) AIML NF / NEF can select up to 30 UEs that can meet the requirements of the application server regardless of the MAPDU session information or preferred radio access technology;

[0475] 2) The AIML NF / NEF may then first select the UEs closest to the AI / ML application server requirement as the remaining 10 UEs, regardless of the MAPDU session information or preferred radio access technology. If there are UEs with the same proximity as the requirement, the AIML NF / NEF may consider the MA PDU session information and prioritize UEs with both 3GPP and non-3GPP connections based on the established MA PDU sessions. The AIML NF / NEF may then select a UE with which it has established a MA PDU session and is connected to only one access (3GPP or non-3GPP). Next, the AIML NF / NEF may select a UE with a single PDU session.

[0476] Each of the above information may be provided in priority order to help the AI / ML application server select a UE.

[0477] The list of candidate UEs for the recommended FL behavior may vary according to the recommended time period and / or recommended area for AIML operation.

[0478] In response to the AI / ML application server, if the “time when auxiliary information is needed” is provided in step 1, the AIML NF / NEF can take it into account.

[0479] Based on the received 5GS assistance information, the AI / ML application server can select and manage UE groups and determine the start time of FL behavior.

[0480] 10. The AIML NF / NEF may have subscribed to the notification service of the UDM, SMF and / or PCF. In this case, the AI / ML application server can receive MA PDU session information from the UDM, SMF and / or PCF while the FL behavior of the application is ongoing. For example, when the FL behavior of the application is ongoing, the MA PDU session information of some or all UEs in the list of UEs requested by the AI / ML application server may change. Then, the UDM, SMF and / or PCF may periodically notify the AIML NF / NEF of the UE's MA PDU session information. Alternatively, if the terminal's MA PDU session information has changed, the UDM, SMF and / or PCF may notify the AIML NF / NEF of the UE's MA PDU session information.

[0481] 11-15. If the AI / ML application server and AIML NF / NEF subscribe (or have subscribed) to AIML auxiliary information in steps 1 and 4, then when NWDAF generates a new analysis, AIML NF / NEF can notify the AI / ML application server of the new auxiliary information. Note that when NWDAF performs a new analysis, steps 11 to 15 can be performed in the same manner as steps 5 to 9 described above.

[0482] In step 14, the AIML NF / NEF may generate new auxiliary information. Taking into account the updated MA PDU session information, the AIML NF / NEF may provide the new auxiliary information to the AI / ML application server in the same manner as in step 9. For example, a UE that is only connected to one (3GPP or non-3GPP) access may be connected to two accesses by additionally connecting to another access. In this case, the AIML NF / NEF may include the UE in the additionally recommended UEs. In another example, a UE connected to two accesses may be disconnected from one access. In this case, taking into account the access preference, the AIML NF / NEF may exclude the UE from the recommended UE list. Alternatively, if UEs that are not connected to an MA PDU session are connected to an MA PDU session, the AIML NF / NEF may consider adding these UEs as candidate UEs.

[0483] For example, the AIML NF / NEF may send the list of UEs together with the MAPDU session information to the AI / ML application server. Based on the new MAPDU session information and the information collected from other NFs (e.g., NWDAF), the AIML NF / NEF may generate new auxiliary information.

[0484] After the FL operation starts, due to changed network conditions, the AIML NF / NEF may notify new AIML assistance information (e.g., new recommended UEs), and the AI / ML application server may reselect FL members based on this information.

[0485] The following figures are intended to illustrate specific embodiments of the present disclosure. The specific device names or specific signal / message / field names shown in the figures are for illustrative purposes only, and the technical features of this specification are not limited to the specific names used in the following figures.

[0486] Fig.16 An example of a process for providing auxiliary information based on MA PDU session information according to one embodiment of the present disclosure is shown.

[0487] It should be noted that Fig.16 The processes shown in the examples are illustrative only and are not intended to limit the scope of this disclosure. Fig.16 Limitations of the examples in .

[0488] For example, for Fig.16 Example, you can also apply Figure 12 to Figure 15b For example, even if Fig.16 Although no behavior, content, etc. are directly described in the examples, any behavior, content, etc. described in the various examples of the present disclosure may also be applied.

[0489] Note that Fig.16 The PCF and UDM are omitted in the examples, but the PCF and UDM operations can also be performed according to Fig.16 The example is executed during the process.

[0490] exist Fig.16 In the example, step S1604 and step S1605 may be performed by the PCF.

[0491] In step S1601, the UE may send a PDU session establishment request message to the SMF.

[0492] In step S1602, the SMF may send a PDU session establishment accept message to the UE. Fig.16 In the figure, only one UE is shown, but this is only for illustrative purposes. The SMF may perform the operations according to steps S1601 and S1602 with one or more UEs.

[0493] In step S1603, the application server may send a request message to the AIML NF (or NEF). The request message may include information requesting a UE list and MA PDU session information.

[0494] For example, the request message may also include information related to the preferred radio access technology. For example, the preferred radio access technology information may also include one of WLAN, 5G, or multi-access.

[0495] For example, the request message may also include information of UEs preferring MA PDU sessions, UEs supporting MA PDU sessions, UEs allowing MA PDU sessions, and UEs connected with both access types.

[0496] In step S1604, the AIML NF (or NEF) may send a subscription request message to the SMF. The subscription request message may be a request for subscribing to an event of receiving the MA PDU session information of the UE in the UE list received from the application server.

[0497] For example, the MA PDU session information may include one or more of the following information: information related to whether the UE supports MAPDU sessions, information related to whether the UE allows MA PDU sessions, information related to whether an MA PDU session has been established for the UE, and / or information about which access type the UE's MA PDU session (if established) is associated with.

[0498] In step S1605, the SMF may send the MA PDU session information to the AIML NF (or NEF). For example, the join request message may include a list of UEs. The SMF may send the MA PDU session information associated with the UEs included in the UE list.

[0499] In step S1606, the AIML NF (or NEF) may determine the assistance information based on the MA PDU session information. The assistance information may include a list of candidate UEs for FL behavior. The AIML NF (or NEF) may determine the assistance information based on the MA PDU session information, as previously described in various examples.

[0500] In step S1607, the AIML NF (or NEF) may send auxiliary information to the application server. Based on the auxiliary information, the application server may determine the UE to participate in the FL behavior, and may request network resources for the determined UE from the 5GS (e.g., a 5G network including network nodes such as SMF, UPF, PCF).

[0501] According to various embodiments of the present disclosure, the NF of 5GC (e.g., AIML NF, NWDAF, etc.) can collect the MA PDU session information of the terminal from SMF / PCF / UDM (e.g., "whether the terminal supports MA PDU sessions, whether the terminal allows MA PDU sessions, whether the MAPDU session of the terminal is currently established, and to which access type (3GPP, non-3GPP, both) the MA PDU session of the terminal is connected").

[0502] According to various embodiments of the present disclosure, the NF of 5GC (AIML NF, NWDAF, etc.) can provide the AF with the obtained MAPDU session information of the terminal.

[0503] According to various embodiments of the present disclosure, the NF (AIMLNF, NWDAF, etc.) of 5GC may provide the AF with a list of candidate UEs for FL in consideration of the obtained MAPDU session information of the terminal.

[0504] According to various embodiments of the present disclosure, the various operations described in the present disclosure may be performed by a new NF (e.g., AIMLNF) to support application AI / ML operations. Alternatively, the various behaviors described in the present disclosure may be supported by a new NF or a NF or NWDAF whose functions are integrated with an existing NF (NEF, PCF) (e.g., AIML NF / NEF).

[0505] For example, the various behaviors described in this disclosure may be supported by a new NF of AI / ML or an NF integrated with an existing NF. In this case, the new NF or integrated NF may operate based on the new service operation. Alternatively, if the various operations described in this disclosure are supported by NWDAF, the NWDAF may operate based on the new analysis.

[0506] According to various embodiments of the present disclosure, when requesting auxiliary information from 5GC for member selection for FL operation, the AI / ML application server may provide a list of UEs. The AI / ML application server may also request the MA PDU session information of the UE. Alternatively, the AI / ML application server may request a list of candidate UEs, giving priority to UEs with MA PDU sessions, UEs allowed to have MA PDU sessions, UEs supporting MA PDU sessions, or UEs connected to both access types. The request from the AI / ML application server may be forwarded to the AIML NF / NEF.

[0507] According to various embodiments of the present disclosure, the AIML NF / NEF determines whether any request received from the AI / ML application server is a request for MA PDU session information. If there is a request for MA PDU session information of the UE from the AI / ML application server, or even if there is no explicit request, the AIML NF / NEF may decide to collect MA PDU session information for FL candidate UEs based on preset information.

[0508] According to various embodiments of the present disclosure, the AIML NF may obtain (e.g., receive) information related to the MAPDU session of each UE in one or more of the following examples from the UDM / SMF / PCF:

[0509] - Information related to whether the UE supports MA PDU sessions;

[0510] - Information related to whether the UE's MA PDU session is allowed;

[0511] - Information related to whether the UE's MA PDU session is currently established;

[0512] - Information about which access type (3GPP, non-3GPP, both) the UE's MA PDU session is connected to.

[0513] According to various embodiments of the present disclosure, the AIML NF / NEF may collect information (such as network analysis) from other NFs (e.g., NWDAF, ...). The AIML NF / NEF may then generate auxiliary information to be passed to the AI / ML application server, and may pass the auxiliary information to the AI / ML application server. For example, for a UE list received from the AI / ML application server, the AIML NF / NEF may forward a complete UE list or a recommended UE list to the AI / ML application server, including MAPDU session information for each UE. Alternatively, the AIML NF / NEF may not forward the MA PDU session information to the AI / ML application server, but may forward a list of recommended UEs to the AI / ML application server, which list takes into account the MA PDU session information of the UE.

[0514] According to various embodiments of the present disclosure, if the AIML NF / NEF subscribes to the notification service of the SMF / PCF, if the MA PDU session information of the terminal has changed, the AIML NF / NEF can be notified of the changed MA PDU session information of the terminal. Therefore, if the auxiliary information to be forwarded to the AI / ML application server changes, the AIML NF / NEF can forward the new auxiliary information to the AI / ML application server.

[0515] According to various embodiments of the present disclosure, the AI / ML application server may select a UE to participate in FL based on the UE's MA PDU session information, and may determine the time to start the FL operation and / or the time to perform the FL operation. The AI / ML application server may then request network resources for FL for the selected UE from the 5GS.

[0516] The present disclosure may have various effects.

[0517] For example, the AI / ML application server may request MAPDU session related information from the 5GS for FL operation. The 5GC may send the MAPDU session information together with the information in the 5G system (5GS) to the AI / ML application server. Based on the received information, the application server may give priority to terminals with MA PDU session advantages when selecting UEs to participate in FL, and may request network resources for FL for the selected UEs from the 5GS.

[0518] For example, by giving priority to terminals using MAPDU sessions for FL operations, FL operations can be used effectively. For example, when the AI / ML application server requests information required for FL operations from the 5GS, the AI / ML application server can request information related to the MAPDU session together. The AIML NF / NEF of the 5GS can then collect information in the 5GS, including the MA PDU session information, and provide the AI / ML application server with the MA PDU session information of the UEs participating in the FL. Alternatively, the AIML NF / NEF in the 5GS can provide the AI / ML application server with a list of UEs that consider the MA PDU session information. Based on the received information, the application server can then select a UE to join the FL and request network resources for the FL from the 5GS.

[0519] The effects that can be obtained from the specific examples of the present disclosure are not limited to those listed above. For example, a person of ordinary skill in the relevant field can understand or infer various technical effects from the present disclosure. Therefore, the specific effects of the present disclosure are not limited to those explicitly set forth herein, but can include various effects that can be understood or inferred from the technical features of the present disclosure.

[0520] For reference, the operation of the terminal (eg, UE) described in this specification may be performed by Figures 1 to 4 For example, a terminal (eg, UE) may be Figure 2The first device 100 or the second device 200 of the present invention may be a first device 100 or a second device 200. For example, the operation of the terminal (e.g., UE) described herein may be processed by one or more processors 102 or 202. The operation of the terminal described herein may be stored in one or more memories 104 or 204 in the form of instructions / programs (e.g., instructions, executable codes) that may be executed by one or more processors 102 or 202. The one or more processors 102 or 202 control the one or more memories 104 or 204 and the one or more transceivers 106 or 206, and may perform the operation of the terminal (e.g., UE) described herein by executing the instructions / programs stored in the one or more memories 104 or 204.

[0521] In addition, the instructions for performing the operation of the terminal (e.g., UE) described in the present disclosure of this specification may be stored in a non-volatile computer-readable storage medium having the instructions recorded thereon. The storage medium may be included in one or more memories 104 or 204. Furthermore, the instructions recorded in the storage medium may be executed by one or more processors 102 or 202 to perform the operation of the terminal (e.g., UE) described in the present disclosure of this specification.

[0522] For reference, the operations of the network nodes (e.g., AMF, SMF, UPF, PCF, AI / ML application server (AS), AIML NF, AIML NF / NEF, NEF, UDM, DN, NWDAF, UDR, new NF, etc.) or base stations (e.g., NG-RAN, gNB, etc.) described herein may be performed by the following description. Figures 1 to 3 For example, a network node or base station may be Figure 2 The first device 100 or Figure 2 The second device 200 of the present invention. For example, the operation of the network node or base station described herein may be processed by one or more processors 102 or 202. The operation of the terminal described herein may be stored in one or more memories 104 or 204 in the form of instructions / programs (e.g., instructions, executable codes) that can be executed by one or more processors 102 or 202. The one or more processors 102 or 202 may perform the operation of the network node or base station described herein by controlling the one or more memories 104 or 204 and the one or more transceivers 106 or 206 and executing the instructions / programs stored in the one or more memories 104 or 204.

[0523] In addition, instructions for executing the operations of the network node or base station described in the present disclosure of this specification may be stored in a non-volatile (or non-transitory) computer-readable storage medium. The storage medium may be included in one or more memories 104 or 204. And, the instructions recorded in the storage medium are executed by one or more processors 102 or 202 to perform the operations of the network node or base station.

[0524] In the above, the preferred embodiments have been exemplarily described, but the present disclosure of this specification is not limited to such specific embodiments, and thus, modifications, changes, or improvements may be made.

[0525] In the above exemplary system, the method is described as a series of steps or blocks based on the flowchart, but is not limited to the order of the steps described, and some steps may occur in a different order or simultaneously with other steps as described above. In addition, those skilled in the art will understand that the steps shown in the flowchart are not exclusive, and other steps may be included, or one or more steps of the flowchart may be deleted without affecting the scope of the rights.

[0526] The claims described herein may be combined in various ways. For example, the technical features of the method claims of this specification may be combined and implemented as a device, and the technical features of the device claims of this specification may be combined and implemented as a method. In addition, the technical features of the method claims of this specification and the technical features of the device claims may be combined and implemented as a device, and the technical features of the method claims of this specification and the technical features of the device claims may be combined and implemented as a method.

Claims

1. A method for performing communication, the method being performed by a network function NF related to artificial intelligence AI and machine learning ML, the method comprising the following steps: Receive a request message for auxiliary information from the AI / ML application server AS, The request message includes a UE list including one or more user equipments UE and information related to requesting multi-access MA protocol data unit PDU session information for the one or more UEs; Sending a subscription request message for events related to the MA PDU session information to a session management function SMF; receiving the MA PDU session information from the SMF; and Auxiliary information is determined based on the MA PDU session information.

2. The method according to claim 1, further comprising the steps of: The auxiliary information is sent to the AI / ML AS.

3. The method according to claim 1 or 2, in, The assistance information includes a list of candidate UEs for joint learning FL operation.

4. The method according to any one of claims 1 to 3, in, The assistance information includes a list of candidate UEs for the joint learning FL operation and MA PDU session information related to at least one UE included in the list of candidate UEs.

5. The method according to any one of claims 1 to 4, in, The MA PDU session information includes one or more of the following items: Information related to whether each of the one or more UEs supports the MA PDU session, information related to whether each of the one or more UEs allows the MA PDU session, information related to whether an MA PDU session has been established for each of the one or more UEs, and / or information related to which access type and / or RAT type the one or more UEs are connected to if the MA PDU session has been established.

6. The method according to any one of claims 1 to 5, in, The step of determining the auxiliary information further comprises the following steps: Based on the MA PDU session, at least one UE having the MA PDU session is first included in a list of candidate UEs.

7. The method according to any one of claims 1 to 6, in, The request message also includes preferred radio access technology information, The preferred radio access technology information includes an access type and / or a RAT type related to the radio access technology.

8. The method according to any one of claims 1 to 7, in, The request message also includes information related to preferences for UEs having an MA PDU session, UEs supporting the MA PDU session, UEs allowing the MA PDU session, and / or UEs associated with both access types / RAT types.

9. A network function NF related to artificial intelligence AI and machine learning ML, wherein: The NF includes: one or more transceivers; one or more processors; and one or more memories storing instructions and operably coupled to the one or more processors, Based on the instructions being executed by the one or more processors, operations including the method according to any one of claims 1 to 7 are performed.

10. A method for performing communication, the method being performed by a session management function SMF and comprising the following steps: receiving a PDU session establishment request message for establishing a protocol data unit PDU from a user equipment UE; Sending a PDU session establishment accept message to the UE; receiving, from a network function NF associated with artificial intelligence AI and machine learning ML, a subscription request message for an event associated with multi-access MA PDU session information; as well as sending the MA PDU session information associated with one or more UEs including the UE to the NF, The NF uses the MA PDU session information to determine auxiliary information for the AI / ML application server AS.

11. The method according to claim 10, in, The MA PDU session information includes one or more of the following items: Information related to whether each of the one or more UEs supports the MA PDU session, information related to whether each of the one or more UEs allows the MA PDU session, information related to whether an MA PDU session has been established for each of the one or more UEs, and / or information related to which access type and / or RAT type the one or more UEs are connected to if the MA PDU session has been established.

12. The method according to any one of claims 9 to 11, wherein: If the MAPDU session information is changed, the changed MAPDU session information is sent to the NF.

13. A session management function SMF for performing communication, wherein: The SMF includes: one or more transceivers; one or more processors; and one or more memories storing instructions and operably coupled to the one or more processors, Based on the instructions being executed by the one or more processors, operations including the method according to any one of claims 9 to 12 are performed.

14. A method for performing communication, the method being performed by a user equipment UE and comprising the following steps: Sending a protocol data unit (PDU) session establishment request message for establishing a protocol data unit (PDU) session to a session management function (SMF); as well as receiving a PDU session establishment accept message from the SMF, Wherein, the PDU session is a multi-access MA PDU session, In which, when the SMF receives a subscription request message for events related to MAPDU session information from a network function NF related to artificial intelligence AI and machine learning ML, the SMF uses the PDU session establishment acceptance message to send the MA PDU session information related to one or more UEs including the UE to the NF.

15. A user equipment (UE) configured to operate in a wireless communication system, the UE comprising: one or more transceivers; one or more processors; as well as one or more memories storing instructions and operably coupled to the one or more processors, Based on the instructions being executed by the one or more processors, operations including the method according to claim 14 are performed.

Images