Communication system, base station, and communication terminal

Abstract

A communication system according to the present invention comprises a base station and a communication terminal connected to the base station, and the base station determines a setting to be used in communication processing performed by the communication terminal using a model in which training has been performed, and notifies the communication terminal of the determined setting, the training uses information acquired from a communication terminal connected to the base station and another adjacent base station, and a communication terminal that receives a notification of the setting from the first base station, which is the base station for which the setting has been determined, performs communication processing with respect to the first base station or communication processing with respect to the second base station using the setting.

Description

Technical Field

[0001] This disclosure relates to wireless communication technology. Background Technology

[0002] Within the 3GPP (3rd Generation Partnership Project), the standardization organization for mobile communication systems, the fifth-generation (hereinafter sometimes referred to as "5G") radio access system (e.g., Non-Patent Document 2) is discussed as a successor to Long Term Evolution (LTE) and Long Term Evolution Advanced (LTE-A), one of the fourth-generation radio access systems (see Non-Patent Document 1). The technology for the 5G radio band is called "New Radio Access Technology" ("New Radio" is abbreviated as "NR"). The NR system is discussed based on the LTE and LTE-A systems.

[0003] For example, in Europe, the organization METIS is summarizing the requirements for 5G (see Non-Patent Document 3). In 5G wireless access systems, for LTE systems, assuming a system capacity 1000 times greater, a data transmission speed 100 times greater, a data processing latency of 1 / 5, and a simultaneous connection capacity of 100 times greater, further reductions in power consumption and device cost can be listed as requirements (see Non-Patent Document 3).

[0004] To meet these requirements, discussions on 5G standards are ongoing within 3GPP (see Non-Patent Literature 4-23).

[0005] As an access method for NR, the downlink direction uses OFDM (Orthogonal Frequency Division Multiplexing), while the uplink direction uses OFDM and DFT-s-OFDM (Discrete Fourier Transform-spread-OFDM). Furthermore, similar to LTE and LTE-A, the 5G system does not include line switching; it uses only packet communication.

[0006] In NR, higher frequencies can be used compared to LTE to increase transmission speed and reduce processing latency.

[0007] In NR, which sometimes uses frequencies higher than LTE, a narrower beam-shaped transmit / receive range is formed (beamforming) and the direction of the beam is changed (beam scanning), thereby ensuring cell coverage through capability mapping.

[0008] use Figure 1 To explain the decisions regarding the frame structure of the NR system in 3GPP as described in Non-Patent Document 1 (Chapter 5). Figure 1 This is an explanatory diagram showing the structure of the wireless frame used in an NR communication system. Figure 1 In NR, a radio frame is 10 ms long. The radio frame is divided into 10 equal-sized subframes. The NR frame structure supports one or more numberologies, i.e., one or more subcarrier spacings (SCS). In NR, a subframe is 1 ms long, and a time slot consists of 14 symbols, regardless of the subcarrier spacing. Furthermore, the number of time slots in a subframe is one when the subcarrier spacing is 15 kHz; the number of time slots in other subcarrier spacings increases proportionally to the subcarrier spacing (see Non-Patent Document 11 (3GPP TS38.211)).

[0009] Non-Patent Document 2 (Chapter 5) and Non-Patent Document 11 record decisions made in 3GPP related to channel structure in NR systems.

[0010] The Physical Broadcast Channel (PBCH) is a channel used for downlink transmission from a base station (hereinafter sometimes referred to as a "base station") to a mobile terminal device (hereinafter sometimes referred to as a "mobile terminal") or other communication terminal device (hereinafter sometimes referred to as a "communication terminal" or "terminal"). The PBCH is transmitted together with the downlink synchronization signal.

[0011] In NR, the downlink synchronization signal consists of a primary synchronization signal (P-SS) and a secondary synchronization signal (S-SS). The synchronization signal is transmitted from the base station as a synchronization signal burst (hereinafter sometimes referred to as an SS burst), at a specified period for a specified duration. An SS burst consists of synchronization signal blocks (hereinafter sometimes referred to as SS blocks) for each beam of the base station.

[0012] During the duration of an SS burst, the base station changes its beam to transmit SS blocks for each beam. An SS block consists of P-SS, S-SS, and PBCH.

[0013] The Physical Downlink Control Channel (PDCCH) is the downlink transmission channel from the base station to the communication terminal. The PDCCH transmits Downlink Control Information (DCI). The DCI includes resource allocation information for the Downlink Shared Channel (DL-SCH), one of the transmission channels described later; resource allocation information for the Paging Channel (PCH), another transmission channel described later; and HARQ (Hybrid Automatic Repeat reQuest) information related to the DL-SCH. Additionally, the DCI sometimes includes Uplink Scheduling Grant. The DCI sometimes includes response signals for uplink transmissions, namely Ack (Acknowledgement) / Nack (Negative Acknowledgement). Furthermore, to allow for flexible DL / UL handover within time slots, the DCI sometimes includes Slot Format Indication (SFI). PDCCH or DCI is also known as the L1 / L2 control signal.

[0014] In NR, there are time-domain and frequency-domain regions that can serve as candidates for containing PDCCH. This region is called the Control Resource Set (CORESET). The communication terminal monitors the CORESET to acquire the PDCCH.

[0015] The Physical Downlink Shared Channel (PDSCH) is the downlink transmission channel from the base station to the communication terminal. The PDSCH maps to the Downlink Shared Channel (DL-SCH) used as the transport channel and the PCH used as the transport channel.

[0016] The Physical Uplink Control Channel (PUCCH) is the uplink transmission channel from the communication terminal to the base station. PUCCH transmits Uplink Control Information (UCI). UCI includes response signals for downlink transmissions, such as Ack / Nack, CSI (Channel State Information), and Scheduling Requests (SRs). CSI is composed of RI (Rank Indicator), PMI (Precoding Matrix Indicator), and CQI (Channel Quality Indicator) reports. RI refers to the rank information of the channel matrix in MIMO (Multiple Input Multiple Output). PMI refers to the information of the precoding matrix used in MIMO. CQI is quality information indicating the quality of received data or the quality of the communication line. UCI is sometimes transmitted via PUSCH (described later). PUCCH or UCI is also referred to as L1 / L2 control signals.

[0017] The Physical Uplink Shared Channel (PUSCH) is the uplink transmission channel from the communication terminal to the base station. The PUSCH maps the Uplink Shared Channel (UL-SCH) as one of the transmission channels.

[0018] The Physical Random Access Channel (PRACH) is an uplink transmission channel from a communication terminal to a base station. PRACH transmits the random access preamble.

[0019] Downlink reference signals (RS) are symbols known in NR (Normally Injectable) communication systems. There are four types of downlink reference signals: UE-specific reference signals (DM-RS), phase tracking reference signals (PT-RS), positioning reference signals (PRS), and channel state information reference signals (CSI-RS). As physical layer measurements for communication terminals, there are measurements of the received power (RSRP) and received quality (RSRQ) of the reference signals.

[0020] The uplink reference signal is also a known symbol in NR communication systems. Three types of uplink reference signals are defined: Demodulation Reference Signal (DM-RS), Phase Tracking Reference Signal (PT-RS), and Sounding Reference Signal (SRS).

[0021] The transport channel described in Non-Patent Document 2 (Chapter 5) will be explained. The broadcast channel (BCH) in the downlink transport channel is broadcast to the entire coverage area of ​​its base station (cell). The BCH is mapped to the physical broadcast channel (PBCH).

[0022] HARQ-based retransmission control is applied to the Downlink Shared Channel (DL-SCH). The DL-SCH can broadcast to the entire coverage area of ​​the base station (cell). The DL-SCH supports dynamic or semi-static resource allocation. Semi-static resource allocation is also known as semi-persistent scheduling. To reduce the power consumption of communication terminals, the DL-SCH supports discontinuous reception (DRX). The DL-SCH is mapped to the Physical Downlink Shared Channel (PDSCH).

[0023] The Paging Channel (PCH) supports DRX of communication terminals to reduce power consumption. The PCH is requested to broadcast over the entire coverage area of ​​the base station (cell). The PCH is mapped to physical resources such as the Physical Downlink Shared Channel (PDSCH) that can be dynamically used for traffic.

[0024] HARQ-based retransmission control is applied to the Uplink Shared Channel (UL-SCH) in the uplink transport channel. UL-SCH supports dynamic or quasi-static resource allocation. Quasi-static resource allocation is also known as Configured Grant. UL-SCH is mapped to the Physical Uplink Shared Channel (PUSCH).

[0025] The Random Access Channel (RACH) is restricted to control information. RACH is subject to collision risks. RACH is mapped to the Physical Random Access Channel (PRACH).

[0026] The following explains HARQ. HARQ is a technique that improves the communication quality of a transmission line by combining Automatic Repeat Request (ARQ) and Forward Error Correction. HARQ has the following advantages: even for transmission lines where communication quality changes, retransmission can effectively enable error correction. In particular, during retransmission, the quality can be further improved by combining the initial received result with the retransmitted result.

[0027] Here's an example illustrating the retransmission method. When the receiving side cannot correctly decode the received data—in other words, when a CRC (Cyclic Redundancy Check) error occurs (CRC=NG)—a "Nack" is sent from the receiving side to the sending side. The sending side, upon receiving the "Nack," retransmits the data. When the receiving side can correctly decode the received data—in other words, when no CRC error occurs (CRC=OK)—a "ck" is sent from the receiving side to the sending side. The sending side, upon receiving the "Ack," sends the next data.

[0028] Other examples of retransmission methods are illustrated below. If a CRC error occurs at the receiving end, a retransmission request is sent from the receiving end to the sending end. The retransmission request is made via a switch to NDI (New Data Indicator). The sending end, upon receiving the retransmission request, retransmits the data. If no CRC error occurs at the receiving end, no retransmission request is sent. If the sending end does not receive a retransmission request within a specified time, it is assumed that no CRC error was transmitted at the receiving end.

[0029] The logical channel described in Non-Patent Document 1 (Chapter 6) will be explained. The Broadcast Control Channel (BCCH) is a downlink channel used to broadcast system control information. The BCCH, as a logical channel, is mapped to either the broadcast channel (BCH) as a transmission channel or the downlink shared channel (DL-SCH).

[0030] The Paging Control Channel (PCCH) is a downlink channel used to transmit paging information and system information updates. The PCCH, as a logical channel, is mapped to the Paging Channel (PCH), which is a transport channel.

[0031] The Common Control Channel (CCCH) is a channel used to transmit control information between a communication terminal and a base station. The CCCH is used when there is no RRC connection between the communication terminal and the network. In the downlink direction, the CCCH is mapped to the Downlink Shared Channel (DL-SCH) used as a transport channel. In the uplink direction, the CCCH is mapped to the Uplink Shared Channel (UL-SCH) used as a transport channel.

[0032] The Dedicated Control Channel (DCCH) is a channel used to transmit dedicated control information between a communication terminal and the network in a point-to-point manner. The DCCH is used when there is an RRC connection between the communication terminal and the network. In the uplink, the DCCH is mapped to the Uplink Shared Channel (UL-SCH), and in the downlink, it is mapped to the Downlink Shared Channel (DL-SCH).

[0033] A Dedicated Traffic Channel (DTCH) is a channel used for sending user information and conducting point-to-point communication with communication terminals. DTCH exists in both the uplink and downlink. In the uplink, DTCH is mapped to the Uplink Shared Channel (UL-SCH), and in the downlink, it is mapped to the Downlink Shared Channel (DL-SCH).

[0034] Location tracking of a communication terminal is performed on a unit consisting of one or more cells. Location tracking is used to locate the communication terminal even in standby mode, enabling calls to the terminal; in other words, it is performed to enable calls to the communication terminal. The area used for location tracking of this communication terminal is called the Tracking Area (TA).

[0035] In NR, calls from communication terminals within a range smaller than the tracking area are supported. This range is called the RAN Notification Area (RNA). Paging of communication terminals in the RRC_INACTIVE state, as described later, occurs within this range.

[0036] In NR, to support wider transmission bandwidths, carrier aggregation (CA) has been studied, which involves combining two or more component carriers (CCs). CA is described in Non-Patent Literature 1.

[0037] In the case of a CA (Communication Terminal), the UE, as a communication terminal, has a unique RRC (Remote Reference Cell) connection with the network (NW). Within the RRC connection, a serving cell provides NAS (Non-Access Stratum) mobility information and security input. This cell is called the Primary Cell (PCell). Based on the UE's capabilities, secondary serving cells (SCells) are formed to create a group of serving cells together with the PCell. For a single UE, a group of serving cells is formed consisting of one PCell and one or more SCells.

[0038] Furthermore, 3GPP includes dual connectivity (DC), where the UE communicates with two base stations to further increase communication capacity. DC is described in non-patent documents 1 and 22.

[0039] Sometimes, one of the base stations performing dual connectivity (DC) is called the "Master Node (MN)," and the other is called the "Secondary Node (SN)." The serving cells comprised of the Master Nodes are sometimes collectively referred to as the Master Cell Group (MCG), and the serving cells comprised of the Secondary Nodes are sometimes collectively referred to as the Secondary Cell Group (SCG). In DC, the Master Cell in the MCG or SCG is called a Special Cell (SpCell or SPCell). The Special Cell in the MCG is called a PCell, and the Special Cell in the SCG is called the Primary SCG Cell (PSCell).

[0040] In addition, in NR, the base station pre-defines a portion of the carrier frequency band for the UE (hereinafter sometimes called the Bandwidth Part (BWP)). The UE performs transmission and reception between itself and the base station in this BWP, thereby reducing the power consumption in the UE.

[0041] Furthermore, 3GPP has explored services (or applications) that support side-link (SL) communication (also known as PC5 communication) in both the EPS (Evolved Packet System) and 5G core systems (described later) (see Non-Patent Documents 1, 2, 26-28). SL communication involves communication between terminals. Examples of services using SL communication include V2X (Vehicle-to-everything) services and proxy services. In addition to direct communication between terminals, SL communication also proposes communication between the UE and the NW via a relay (see Non-Patent Documents 26, 28).

[0042] The physical channel used for SL (refer to Non-Patent Documents 2, 11) is described below. The Physical Sidelink Broadcast Channel (PSBCH) transmits information related to system synchronization and is sent from the UE.

[0043] The Physical Sidelink Control Channel (PSCCH) transmits control information from the UE for sidelink communication and V2X sidelink communication.

[0044] The Physical Sidelink Shared Channel (PSSCH) transmits data from the UE for sidelink communication and V2X sidelink communication.

[0045] The Physical Sidelink Feedback Channel (PSFCH) transmits HARQ feedback from the sidelink from the UE that received the PSSCH to the UE that sent the PSSCH.

[0046] The transmission channel used for SL (refer to Non-Patent Document 1) will be described. The sidelink broadcast channel (SL-BCH) has a predetermined transmission channel format and is mapped to the PSBCH as a physical channel.

[0047] The Sidelink Shared Channel (SL-SCH) supports broadcast transmission. SL-SCH supports both UE autonomous resource selection and resource allocation via base station scheduling. While UE autonomous resource selection carries a risk of conflict, there are no conflicts when the UE allocates dedicated resources via the base station. Furthermore, SL-SCH supports dynamic link adaptation by modifying transmit power, modulation, and coding. SL-SCH is mapped to the Physical Channel Separate Channel (PSSCH).

[0048] The logical channels used for SL (refer to Non-Patent Document 2) will be described. The Sidelink Broadcast Control Channel (SBCCH) is a sidelink channel used to broadcast sidelink system information from one UE to other UEs. The SBCCH is mapped to the SL-BCH, which serves as the transport channel.

[0049] The Sidelink Traffic Channel (STCH) is a one-to-many traffic channel used to send user information from one UE to other UEs. The STCH is used only by UEs with sidelink communication capabilities and UEs with V2X sidelink communication capabilities. One-to-one communication between two UEs with sidelink communication capabilities is also achieved through the STCH. The STCH is mapped to the SL-SCH, which serves as the transport channel.

[0050] The Sidelink Control Channel (SCCH) is a control channel used to send control information from one UE to other UEs. The SCCH is mapped to the SL-SCH, which serves as the transport channel.

[0051] In LTE, SL communication only involves broadcast. In NR, in addition to broadcast, support for unicast and groupcast has also been studied for SL communication (see Non-Patent Document 27 (3GPP TS23.287)).

[0052] In SL's unicast and multicast communications, it supports HARQ feedback (Ack / Nack), CSI reports, and more.

[0053] In addition, 3GPP is studying Integrated Access and Backhaul (IAB), which uses wireless methods to serve as both access links between UEs and base stations and backhaul links between base stations (see Non-Patent Literature 2, 20, 29).

[0054] For mobile communication systems, some new technologies are required. For example, to improve communication performance, the introduction of Artificial Intelligence (AI) / Machine Learning (ML) for the RAN is required. This new technology has begun to be discussed in 3GPP (non-patent documents 30, 31).

[0055] Existing technical documents

[0056] Non-patent literature

[0057] Non-patent document 1: 3GPP TS36.300 V17.3.0

[0058] Non-patent document 2: 3GPP TS38.300 V17.3.0

[0059] Non-patent literature 3: "Scenarios, requirements and KPIs for 5G mobile and wireless system", ICT-317669-METIS / D1.1

[0060] Non-patent literature 4: 3GPP TR23.799 V14.0.0

[0061] Non-patent literature 5: 3GPP TR38.801 V14.0.0

[0062] Non-patent document 6: 3GPP TR38.802 V14.2.0

[0063] Non-patent document 7: 3GPP TR38.804 V14.0.0

[0064] Non-patent document 8: 3GPP TR38.912 V16.0.0

[0065] Non-Patent Document 9: 3GPP RP-172115

[0066] Non-patent document 10: 3GPP TS23.501 V18.0.0

[0067] Non-patent document 11: 3GPP TS38.211 V17.4.0

[0068] Non-patent document 12: 3GPP TS38.212 V17.4.0

[0069] Non-patent document 13: 3GPP TS38.213 V17.4.0

[0070] Non-patent document 14: 3GPP TS38.214 V17.4.0

[0071] Non-patent document 15: 3GPP TS38.321 V17.3.0

[0072] Non-patent document 16: 3GPP TS38.322 V17.2.0

[0073] Non-patent document 17: 3GPP TS38.323 V17.3.0

[0074] Non-patent document 18: 3GPP TS37.324 V17.0.0

[0075] Non-patent document 19: 3GPP TS38.331 V17.3.0

[0076] Non-patent document 20: 3GPP TS38.401 V17.3.0

[0077] Non-patent document 21: 3GPP TS38.413 V17.3.0

[0078] Non-patent document 22: 3GPP TS37.340 V17.3.0

[0079] Non-patent document 23: 3GPP TS38.423 V17.3.0

[0080] Non-patent document 24: 3GPP TS38.305 V17.3.0

[0081] Non-patent document 25: 3GPP TS23.273 V18.0.0

[0082] Non-patent document 26: 3GPP TR23.703 V12.0.0

[0083] Non-patent document 27: 3GPP TS23.287 V17.5.0

[0084] Non-patent document 28: 3GPP TS23.303 V17.0.0

[0085] Non-patent document 29: 3GPP TS38.340 V17.3.0

[0086] Non-patent document 30: 3GPP TR 37.817 V17.0.0

[0087] Non-patent document 31: 3GPP RP-213602

[0088] Non-patent document 32: 3GPP TS38.314 V17.2.0 Summary of the Invention

[0089] The technical problem that the invention aims to solve

[0090] In mobile communication systems, the number of required KPIs (Key Performance Indicators) such as latency, reliability, connection density, user experience, and energy efficiency is increasing, necessitating continuous optimization. Therefore, research has begun on technologies that integrate AI / ML into mobile communication systems to improve multiple KPIs through data collection and analysis (Non-Patent Literature 30, 31). However, the use cases for integrating AI / ML are currently limited to network energy saving, load balancing, and mobility optimization, and have not yet been studied in other use cases. This leads to the problem that multiple KPIs cannot be optimized under various communication conditions.

[0091] In view of the above problems, one of the purposes of this disclosure is to implement a communication system that can import AI / ML under various communication conditions and implement effective processing corresponding to the communication conditions.

[0092] Technical means for solving technical problems

[0093] The communication system disclosed herein includes: a base station corresponding to a fifth-generation wireless access system; and a communication terminal connected to the base station. The base station uses a trained model to determine the settings used in the communication processing implemented by the communication terminal and notifies the communication terminal connected to the base station of the determined settings. The training uses information obtained from the communication terminal connected to the base station and other adjacent base stations. After the communication terminal receives the setting notification from the base station that determined the settings (i.e., the first base station), it uses the settings to implement communication processing for the first base station or for a base station different from the first base station (i.e., the second base station) after terminating its connection with the first base station.

[0094] Invention Effects

[0095] According to this disclosure, a communication system can be implemented that can perform effective processing corresponding to the communication status.

[0096] The purpose, features, aspects, and advantages of this disclosure will become more apparent from the following detailed description and accompanying drawings. Attached Figure Description

[0097] Figure 1 This is an explanatory diagram showing the structure of a wireless frame used in an NR communication system.

[0098] Figure 2 This is a block diagram showing the overall structure of a communication system 210 using the NR method discussed in 3GPP.

[0099] Figure 3 This is a structural diagram of a DC based on a base station connected to the NG core.

[0100] Figure 4 It is shown Figure 2 The diagram shows the structure of the mobile terminal 202.

[0101] Figure 5 It is shown Figure 2 The diagram shows the structure of base station 213.

[0102] Figure 6 This is a block diagram showing the structure of the 5GC section.

[0103] Figure 7 This is a flowchart illustrating the process from cell search to standby mode in a communication terminal (UE) in an NR-based communication system.

[0104] Figure 8 This is a diagram illustrating an example of cell structure in an NR system.

[0105] Figure 9 This is a connection structure diagram illustrating an example of the connection structure of a terminal in SL communication.

[0106] Figure 10 This is a connection structure diagram illustrating an example of the connection structure of a base station that supports access backhaul integration.

[0107] Figure 11 This is a diagram illustrating a structural example of a learning device related to a communication system in Embodiment 1.

[0108] Figure 12 This is a flowchart related to the learning process of the learning device in Implementation 1.

[0109] Figure 13 This is a diagram illustrating a structural example of a reasoning device related to a communication system in Embodiment 1.

[0110] Figure 14 This is a flowchart related to the reasoning process of the reasoning device in Implementation 1.

[0111] Figure 15 This is a diagram illustrating an example of a sequence of RA processing using AI / ML for RA processing in Implementation 1.

[0112] Figure 16This is a diagram illustrating a sequence example of RA processing using AI / ML for RA processing in a variation 1 of embodiment 1.

[0113] Figure 17 This is a diagram illustrating an example of a sequence processed using SDT processing AI / ML in Implementation 2.

[0114] Figure 18 This is a diagram illustrating an example sequence of IAB processing using AI / ML for IAB processing in Implementation 4.

[0115] Figure 19 This is a diagram showing the first half of a sequence example of BH RLF processing using AI / ML in Modification 1 of Embodiment 4.

[0116] Figure 20 This is a diagram showing the latter half of a sequence example of BH RLF processing using AI / ML in Modification 1 of Embodiment 4. Detailed Implementation

[0117] Implementation method 1.

[0118] Figure 2 This is a block diagram illustrating the overall structure of a communication system 210 using the NR method discussed in 3GPP. Figure 2 The following explanation is provided. The radio access network is referred to as NG-RAN (Next Generation Radio Access Network) 211. The communication terminal device, i.e., the mobile terminal device (hereinafter referred to as "user equipment" UE) 202, can wirelessly communicate with the base station device (hereinafter referred to as "NR base station (NG-RAN NodeB) gNB)" 213, and uses wireless communication to transmit and receive signals. NG-RAN 211 consists of one or more NR base stations 213.

[0119] Here, "communication terminal device" includes not only mobile terminal devices such as mobile phone terminals, but also stationary devices such as sensors. In the following description, "communication terminal device" will sometimes be abbreviated as "communication terminal".

[0120] Between UE202 and NG-RAN 211, the AS (Access Stratum) protocol is terminated. Protocols used as the AS include, for example, RRC (Radio Resource Control), SDAP (Service Data Adaptation Protocol), PDCP (Packet Data Convergence Protocol), RLC (Radio Link Control), MAC (Medium Access Control), and PHY (Physical Layer). RRC is used for the control layer (hereinafter sometimes referred to as C-Plane, C-Plane, or CP), SDAP is used for the user layer (hereinafter sometimes referred to as U-Plane, U-Plane, or UP), and PDCP, MAC, RLC, and PHY are used for both C-Plane and U-Plane.

[0121] The Radio Resource Control (RRC) protocol between UE202 and NR base station 213 performs broadcasting, paging, and RRC connection management. The states between NR base station 213 and UE202 in RRC include RRC_IDLE, RRC_CONNECTED, and RRC_INACTIVE.

[0122] During RRC_IDLE, PLMN (Public Land Mobile Network) selection, System Information (SI) broadcasting, paging, cell re-selection, and mobility operations are performed. During RRC_CONNECTED, the mobile terminal has an RRC connection and can send and receive data with the network. Additionally, during RRC_CONNECTED, handover (HO) and neighbor cell determination (measurement) are performed. During RRC_INACTIVE, the connection between the 5G core unit 214 and the NR base station 213 is maintained while simultaneously performing System Information (SI) broadcasting, paging, cell re-selection, and mobility operations.

[0123] The gNB213 connects to the 5G core (hereinafter sometimes referred to as the "5GC unit") 214 via the NG interface, which includes the Access and Mobility Management Function (AMF), Session Management Function (SMF), or User Plane Function (UPF). Control information and / or user data communication occurs between the gNB213 and the 5GC unit 214. The NG interface is a collective term for the N2 interface between the gNB213 and AMF220, the N3 interface between the gNB213 and UPF221, the N11 interface between AMF220 and SMF222, and the N4 interface between UPF221 and SMF222. One gNB213 can connect to multiple 5GC units 214. The gNBs213 are connected to each other via the Xn interface, enabling communication of control information and / or user data between them.

[0124] The 5GC unit 214 is a host device, specifically a host node, that controls the connection between the NR base station 213 and the mobile terminal (UE) 202, and allocates paging signals for one or more NR base stations (gNB) 213 and / or LTE base stations (E-UTRAN NodeB: eNB). Additionally, the 5GC unit 214 performs mobility control in the idle state. The 5GC unit 214 manages the tracking area list when the mobile terminal 202 is in the idle state, and in the inactive and active states. The 5GC unit 214 initiates the paging protocol by sending paging messages to cells belonging to the registered tracking area of ​​the mobile terminal 202.

[0125] gNB213 can form one or more cells. When one gNB213 forms multiple cells, each cell is configured to communicate with UE202.

[0126] The gNB213 can be divided into a Central Unit (CU) 215 and a Distributed Unit (DU) 216. A CU 215 constitutes one unit within the gNB213. One or more DUs 216 constitute one or more cells within the gNB213. A single DU 216 constitutes one or more cells. The CU 215 connects to the DU 216 via an F1 interface, facilitating communication of control information and / or user data between the CU 215 and DU 216. The F1 interface consists of an F1-C interface and an F1-U interface. The CU 215 handles the functions of various protocols including RRC, SDAP, and PDCP, while the DU 216 handles the functions of various protocols including RLC, MAC, and PHY. One or more Transmission Reception Points (TRPs) 219 are sometimes connected to the DU 216. The TRP 219 transmits and receives radio signals with the UE.

[0127] CU215 can be divided into CU (CU-C) 217 ​​for the C layer and CU (CU-U) 218 ​​for the U layer. CU-C 217 constitutes one unit within CU215. CU-U 218 constitutes one or more units within CU215. CU-C 217 connects to CU-U 218 via an E1 interface, facilitating control information communication between CU-C 217 and CU-U 218. CU-C 217 connects to DU216 via an F1-C interface, facilitating control information communication between CU-C 217 and DU216. CU-U 218 connects to DU216 via an F1-U interface, facilitating user data communication between CU-U 218 and DU216.

[0128] 5G communication systems may include the Unified Data Management (UDM) function and Policy Control Function (PCF) described in Non-Patent Document 10 (3GPP TS23.501). UDM and / or PCF may be included in... Figure 2 In section 5GC214.

[0129] In a 5G communication system, a Location Management Function (LMF) as described in Non-Patent Document 24 (3GPP TS38.305) can be configured. As disclosed in Non-Patent Document 25 (3GPP TS23.273), the LMF can be connected to the base station via the AMF.

[0130] In 5G communication systems, the non-3GPP interworking function (N3IWF) described in Non-Patent Document 10 (3GPP TS23.501) can also be included. The N3IWF can use the access network (AN) as a terminal between the user and the UE in non-3GPP access.

[0131] Figure 3 This is a diagram illustrating a structure based on a DC (dual-connection) linked to the NG core. Figure 3 In the diagram, solid lines represent U-Plane connections, and dashed lines represent C-Plane connections. Figure 3 In this configuration, the primary base station 240-1 can be either a gNB or an eNB. Similarly, the secondary base station 240-2 can also be either a gNB or an eNB. For example, in... Figure 3 In some contexts, the DC structure where the primary base station 240-1 is a gNB and the secondary base station 240-2 is an eNB is sometimes referred to as NG-EN-DC. Figure 3 The example shown illustrates a U-Plane connection between the 5GC unit 214 and the secondary base station 240-2 via the primary base station 240-1, but it can also be established directly between the 5GC unit 214 and the secondary base station 240-2. Additionally, Figure 3 In this configuration, the core network EPC (Evolved Packet Core) connected to the LTE and LTE-A systems can replace the 5GC unit 214 and connect to the main base station 240-1. The U-Plane connection between the EPC and the secondary base station 240-2 can be directly established.

[0132] Figure 4 It is shown Figure 2 The diagram shows the structure of the mobile terminal 202. Figure 4The transmission processing of the mobile terminal 202 shown will be described. First, control data from the control unit 310 and user data from the application unit 302 are sent to the protocol processing unit 301. Buffering of the control data and user data is possible. This buffering can be set in the control unit 310, the application unit 302, or the protocol processing unit 301. The protocol processing unit 301 performs protocol processing such as SDAP, PDCP, RLC, and MAC, for example, determining the transmission target base station in DC and assigning headers to various protocols. The protocol-processed data is transmitted to the encoding unit 304 for error correction and other encoding processing. Alternatively, data may be output directly from the protocol processing unit 301 to the modulation unit 305 without encoding processing. The data encoded by the encoding unit 304 is modulated in the modulation unit 305. Precoding for MIMO may also be performed in the modulation unit 305. After the modulated data is converted into a baseband signal, it is output to the frequency conversion unit 306 and converted into a wireless transmission frequency. Subsequently, the transmitted signal was sent from antennas 307-1 to 307-4 to base station 213. Figure 4 The example shown has four antennas, but the number of antennas is not limited to four.

[0133] Furthermore, the receiving process of the mobile terminal 202 is performed as follows: Wireless signals from the base station 213 are received via antennas 307-1 to 307-4. The received signal is converted from the wireless receiving frequency to a baseband signal by the frequency conversion unit 306, and demodulation processing is performed in the demodulation unit 308. Waiting calculations and multiplication processes can be performed in the demodulation unit 308. The demodulated data is transmitted to the decoding unit 309 for error correction and other decoding processing. The decoded data is transmitted to the protocol processing unit 301, where protocol processing such as MAC, RLC, PDCP, and SDAP is performed, including actions such as header removal in each protocol. Of the data after protocol processing, control data is transmitted to the control unit 310, and user data is transmitted to the application unit 302.

[0134] The series of processes of the mobile terminal 202 are controlled by the control unit 310. Therefore, although in Figure 4 The details have been omitted, but the control unit 310 is also connected to each of the units 302, 304 to 309.

[0135] Each part of the mobile terminal 202, such as the control unit 310, protocol processing unit 301, encoding unit 304, and decoding unit 309, is implemented, for example, by a processing circuit comprising a processor and a memory. For example, the control unit 310 is implemented by the processor executing a program describing a series of processes of the mobile terminal 202. The program describing the series of processes of the mobile terminal 202 is stored in a memory. Examples of memory are non-volatile or volatile semiconductor memories such as RAM (Random Access Memory), ROM (Read Only Memory), and flash memory. Each part of the mobile terminal 202, such as the control unit 310, protocol processing unit 301, encoding unit 304, and decoding unit 309, can be implemented by dedicated processing circuits such as FPGA (Field Programmable Gate Array), ASIC (Application Specific Integrated Circuit), and DSP (Digital Signal Processor). Figure 4 In this context, the number of antennas used for transmitting and the number of antennas used for receiving in the mobile terminal 202 may be the same or different.

[0136] Figure 5 It is shown Figure 2 The diagram shows the structure of base station 213. Figure 5 The transmission processing of the base station 213 shown will be described. The EPC communication unit 401 transmits and receives data between the base station 213 and the EPC. The 5GC communication unit 412 transmits and receives data between the base station 213 and the 5GC (5GC unit 214, etc.). The other base station communication units 402 transmit and receive data with other base stations. The EPC communication unit 401, the 5GC communication unit 412, and the other base station communication units 402 exchange information with the protocol processing unit 403. Control data from the control unit 411, and user data and control data from the EPC communication unit 401, the 5GC communication unit 412, and the other base station communication units 402 are sent to the protocol processing unit 403. Buffering of control data and user data can be performed. This buffering can be provided in the control unit 411, the EPC communication unit 401, the 5GC communication unit 412, or the other base station communication units 402.

[0137] The protocol processing unit 403 performs protocol processing for SDAP, PDCP, RLC, MAC, etc., such as routing transmitted data in DC and assigning headers to various protocols. The protocol-processed data is transmitted to the encoding unit 405 for error correction and other encoding processing. Alternatively, data may be output directly from the protocol processing unit 403 to the modulation unit 406 without encoding processing. Furthermore, data can be transmitted from the protocol processing unit 403 to other base station communication units 402. For example, in DC, data transmitted from the 5GC communication unit 412 or the EPC communication unit 401 can be transmitted to other base stations, such as auxiliary base stations, via other base station communication units 402. The encoded data undergoes modulation processing in the modulation unit 406. Precoding for MIMO can also be performed in the modulation unit 406. After the modulated data is converted into a baseband signal, it is output to the frequency conversion unit 407 and converted into a wireless transmission frequency. Then, using antennas 408-1 to 408-4, the transmission signal is transmitted to one or more mobile terminals 202. Figure 5 The example shown has four antennas, but the number of antennas is not limited to four.

[0138] Furthermore, the reception processing of base station 213 is performed as follows: Wireless signals from one or more mobile terminals 202 are received by antennas 408-1 to 408-4. The received signals are converted from the wireless receiving frequency to a baseband signal by frequency conversion unit 407, and demodulated in demodulation unit 409. The demodulated data is transmitted to decoding unit 410 for error correction and other decoding processing. The decoded data is transmitted to protocol processing unit 403, where protocol processing such as MAC, RLC, PDCP, and SDAP is performed, including actions such as header removal in each protocol. Of the data after protocol processing, control data is transmitted to control unit 411, 5GC communication unit 412, EPC communication unit 401, or other base station communication unit 402, and user data is transmitted to 5GC communication unit 412, EPC communication unit 401, or other base station communication unit 402. Data sent from other base station communication units 402 can be transmitted to 5GC communication unit 412 or EPC communication unit 401. This data could be, for example, uplink data transmitted from the DC to the 5GC communication unit 412 or the EPC communication unit 401 via other base stations.

[0139] The series of processes of base station 213 are controlled by control unit 411. Therefore, although in Figure 5 The details have been omitted, but the control unit 411 is also connected to each of the parts 401, 402, 405 to 410, 412.

[0140] The various parts of base station 213, such as control unit 411, protocol processing unit 403, 5GC communication unit 412, EPC communication unit 401, other base station communication unit 402, encoding unit 405, and decoding unit 410, are implemented similarly to the mobile terminal 202, by a processing circuit consisting of a processor and memory, or by a dedicated processing circuit such as FPGA, ASIC, or DSP. Figure 5 In this system, the number of antennas used for transmitting and the number of antennas used for receiving in base station 213 can be the same or different.

[0141] As Figure 2 The example of the structure of CU 215 shown, except Figure 5 In addition to the encoding unit 405, modulation unit 406, frequency conversion unit 407, antennas 408-1 to 408-4, demodulation unit 409, and decoding unit 410 shown, a structure with a DU communication unit is sometimes used. The DU communication unit is connected to the protocol processing unit 403. The protocol processing unit 403 in CU215 performs protocol processing for PDCP, SDAP, etc.

[0142] As Figure 2 The example of the structure of DU216 shown, except Figure 5 In addition to the EPC communication unit 401, other base station communication units 402, and 5GC communication unit 412 shown, a structure with a CU communication unit is sometimes used. The CU communication unit is connected to the protocol processing unit 403. The protocol processing unit 403 in DU216 performs protocol processing for PHY, MAC, RLC, etc.

[0143] Figure 6 This is a block diagram showing the structure of the 5GC section. Figure 6 The above is shown in the figure. Figure 2 The structure of the 5GC section 214 shown. Figure 6 It shows in Figure 2 The 5GC section 214 shown includes the structures of AMF, SMF, and UPF. Figure 6In the example shown, the AMF can have the functions of the control layer control unit 525, the SMF can have the functions of the session management unit 527, and the UPF can have the functions of the user layer communication unit 523 and the data network communication unit 521. The data network communication unit 521 performs data transmission and reception between the 5GC unit 214 and the data network. The base station communication unit 522 performs data transmission and reception between the 5GC unit 214 and the base station 21 via the NG interface. User data sent from the data network is transmitted from the data network communication unit 521 to the base station communication unit 522 via the user layer communication unit 523, and then sent to one or more base stations 213. User data sent from the base station 213 is transmitted from the base station communication unit 522 to the data network communication unit 521 via the user layer communication unit 523, and then sent to the data network.

[0144] Control data sent from base station 213 is transmitted from base station communication unit 522 to control layer control unit 525. Control layer control unit 525 can transmit control data to session management unit 527. Control data can be sent from the data network. Control data sent from the data network can be sent from data network communication unit 521 to session management unit 527 via user layer communication unit 523. Session management unit 527 can send control data to control layer control unit 525.

[0145] The user layer communication unit 523 includes a PDU processing unit 523-1, a mobility anchoring unit 523-2, etc., and performs overall processing for the user layer (hereinafter sometimes referred to as U-Plane). The PDU processing unit 523-1 processes data packets, such as sending and receiving packets with the data network communication unit 521 and with the base station communication unit 522. The mobility anchoring unit 523-2 is responsible for connecting the data path when the UE moves.

[0146] The session management unit 527 manages the PDU sessions set up between the UE and the UPF. The session management unit 527 includes a PDU session control unit 527-1 and a UE IP address allocation unit 527-2. The PDU session control unit 527-1 manages the PDU sessions between the mobile terminal 202 and the 5GC unit 214. The UE IP address allocation unit 527-2 allocates IP addresses for the mobile terminal 202.

[0147] The control layer control unit 525 includes the NAS security unit 525-1, the Idle State mobility management unit 525-2, etc., and performs overall processing for the control layer (hereinafter sometimes referred to as C-Plane). The NAS security unit 525-1 performs security protection for NAS (Non-Access Stratum) messages. The Idle State Mobility Management unit 525-2 performs mobility management in standby state (Idle State: RRC_IDLE state, or simply idle), generation and control of paging signals in standby state, addition, deletion, updating, retrieval of tracking areas for one or more mobile terminals 202 within the coverage area, and tracking area list management.

[0148] The series of processes in the 5GC unit 214 are controlled by the control unit 526. Therefore, although in Figure 6 The details are omitted, but the control unit 526 is connected to each of the units 521-523, 525, and 527. The various parts of the 5GC unit 214 are similar to the control unit 310 of the mobile terminal 202 described above, for example, implemented by a processing circuit comprising a processor and memory, or a dedicated processing circuit such as an FPGA, ASIC, or DSP.

[0149] Next, an example of a cell search method in a communication system is shown. Figure 7 This is a flowchart illustrating the process of a communication terminal (UE) in an NR-based communication system from cell search to standby operation. If the communication terminal starts cell search, in step ST601, the first synchronization signal (P-SS) and the second synchronization signal (S-SS) sent from the surrounding base stations are used to obtain the synchronization of time slot timing and frame timing.

[0150] P-SS and S-SS are collectively referred to as the Synchronization Signal (SS). The Synchronization Signal (SS) contains a synchronization code that corresponds one-to-one with the PCI (Physical Cell Identifier) ​​assigned to each cell. The number of PCIs is set to 1008. The communication terminal uses these 1008 PCIs to achieve synchronization and detects (determines) the PCIs of synchronized cells.

[0151] In step ST602, the communication terminal receives the PBCH for the next cell to be synchronized. The BCCH on the PBCH maps to the MIB (Master Information Block), which contains cell structure information. Therefore, by receiving the PBCH and obtaining the BCCH, the MIB can be obtained. Information in the MIB includes, for example, the SFN (System Frame Number), scheduling information of SIB (System Information Block) 1, subcarrier spacing of SIB1, and DM-RS location information.

[0152] Additionally, the communication terminal obtains the SS block identifier via the PBCH. A portion of the bit sequence of the SS block identifier is contained in the MIB. The remaining bit sequence is contained in the identifier used to generate the DM-RS sequence accompanying the PBCH. The communication terminal uses the MIB contained in the PBCH and the DM-RS sequence accompanying the PBCH to obtain the SS block identifier.

[0153] Next, in step ST603, the communication terminal measures the received power of the SS block.

[0154] Next, in step ST604, the communication terminal selects the cell with the best reception quality from the more than one cell detected up to step ST603, for example, selecting the cell with the highest reception power, i.e., the optimal cell. Additionally, the communication terminal selects the beam with the best reception quality, for example, selecting the beam with the highest reception power in the SS block, i.e., the optimal beam. The selection of the optimal beam is, for example, using the reception power of the SS block identified by each SS block.

[0155] Next, in step ST605, the communication terminal receives the DL-SCH based on the scheduling information of SIB1 contained in the MIB, and obtains SIB1 (System Information Block) from the broadcast information BCCH. SIB1 contains information related to access to the cell, cell structure information, and scheduling information of other SIBs (SIBk: an integer k ≥ 2). In addition, SIB1 contains the Tracking Area Code (TAC).

[0156] Next, in step ST606, the communication terminal compares the TAC of SIB1 received in step ST605 with the TAC portion of the Tracking Area Identity (TAI) in the tracking area list already stored by the communication terminal. The tracking area list is also called the TAI list. TAI is identification information used to identify the tracking area, consisting of the MCC (Mobile Country Code), MNC (Mobile Network Code), and TAC (Tracking Area Code). MCC is the country code. MNC is the network code. TAC is the tracking area code number.

[0157] If the comparison result obtained in step ST606 is the same as the TAC received in step ST605, and it is also included in the tracking area list, then the communication terminal enters standby mode in that cell. If the comparison result is the same as the TAC received in step ST605, and it is not included in the tracking area list, then the communication terminal requests a change of tracking area from the core network (EPC) containing the MME, etc., through that cell to perform a TAU (Tracking Area Update).

[0158] The apparatus constituting the core network (hereinafter sometimes referred to as "core network-side apparatus") updates the tracking area list based on the TAU request signal and the identification number (UE-ID, etc.) of the communication terminal sent from the communication terminal. The core network-side apparatus sends the updated tracking area list to the communication terminal. The communication terminal rewrites (updates) its own TAC list based on the received tracking area list. Afterward, the communication terminal enters standby mode in the cell.

[0159] Next, examples of random access methods in a communication system are shown. In random access, 4-step random access and 2-step random access are used. Furthermore, for 4-step and 2-step random access, there are contention-based random access, which may generate timing contention with other mobile terminals, and contention-free random access.

[0160] An example of a conflict-based four-step random access method is shown. As step 1, the mobile terminal sends a random access preamble to the base station. The random access preamble can be selected by the mobile terminal from a predefined range, or it can be assigned separately to the mobile terminal and notified by the base station.

[0161] As a second step, the base station sends a random access response to the mobile terminal. The random access response includes uplink scheduling information used in the third step, and the terminal identifier used in the uplink transmission in the third step.

[0162] As step 3, the mobile terminal sends an uplink transmission to the base station. The mobile terminal uses the information obtained in step 2 in this uplink transmission. As step 4, the base station notifies the mobile terminal whether there was contention resolution. Mobile terminals notified of no contention end the random access process. Mobile terminals notified of a contention restart the process from step 1.

[0163] The conflict-free 4-step random access method differs from the conflict-based 4-step random access method in the following ways: Before step 1, the base station pre-assigns a random access preamble and uplink scheduling to the mobile terminal. Furthermore, notification regarding conflict resolution in step 4 is not required.

[0164] An example of a collision-based two-step random access method is shown. In step 1, the mobile terminal sends a random access preamble and an uplink transmission to the base station. In step 2, the base station notifies the mobile terminal of whether a collision has occurred. Mobile terminals notified of no collision end the random access process. Mobile terminals notified of a collision restart the process from step 1.

[0165] The conflict-free two-step random access method differs from the conflict-based two-step random access method in the following way: Before step 1, the base station pre-assigns a random access preamble and uplink scheduling to the mobile terminal. Additionally, in step 2, the base station sends a random access response to the mobile terminal.

[0166] Figure 8 This illustrates an example of the structure of a cell in NR. In an NR cell, a narrow beam is formed and its direction is changed for transmission. Figure 8 In the example shown, base station 750 uses beam 751-1 to transmit and receive data with the mobile terminal at certain times. At other times, base station 750 uses beam 751-2 to transmit and receive data with the mobile terminal. Similarly, base station 750 uses one or more of beams 751-3 to 751-8 to transmit and receive data with the mobile terminal. Thus, base station 750 constitutes a wide-range cell 752.

[0167] exist Figure 8 The example shown depicts a base station 750 using 8 beams, but the number of beams can be different from 8. Additionally, in... Figure 8 In the example shown, the number of beams used simultaneously by base station 750 is set to 1, but it can also be multiple.

[0168] Beam identification uses the concept of QCL (Quasi-CoLocation) (refer to Non-Patent Document 14 (3GPP TS 38.214)). That is, it is identified by information indicating which reference signal (e.g., SS block, CSI-RS) the beam can be considered to be the same as. This information sometimes includes the type of information about the viewpoints that can be considered the same beam, such as information about Doppler shift, Doppler shift spread, average delay, average delay spread, and spatial Rx parameters (refer to Non-Patent Document 14 (3GPP TS 38.214)).

[0169] In 3GPP, due to D2D (Device to Device) communication and V2V (Vehicle to Vehicle) communication, sidelinks (SL) are supported (see Non-Patent Literature 1 and Non-Patent Literature 16). SL is specified through the PC5 interface.

[0170] In SL communication, in addition to broadcasting, support for PC5-S signaling was studied to support unicast and groupcast (see Non-Patent Document 27 (3GPP TS23.287)). For example, PC5-S signaling was implemented to establish SL, i.e., the link used to implement PC5 communication. This link is implemented in the V2X layer and is also known as a Layer 2 link.

[0171] In addition, support for RRC signaling is being researched in SL communication (see Non-Patent Document 27 (3GPP TS23.287)). RRC signaling in SL communication is also referred to as PC5 RRC signaling. For example, the ability to notify UEs of each other during PC5 communication, or to notify the AS layer settings used for V2X communication via PC5 communication, has been proposed.

[0172] Figure 9 The diagram shows an example of the connection structure of a mobile terminal in SL communication. Figure 9 In the example shown, UE805 and UE806 exist within the coverage area 803 of base station 801. UL / DL communication 805 occurs between base station 801 and UE806. UL / DL communication 808 occurs between base station 801 and UE806. SL communication 810 occurs between UE805 and UE806. UE811 and UE812 exist outside the coverage area 803. SL communication 814 occurs between UE805 and UE811. Additionally, SL communication 816 occurs between UE811 and UE812.

[0173] As an example of communication between the UE and NW via relay in SL communication, Figure 9 The UE805 shown relays the communication between UE811 and base station 801.

[0174] UEs that perform relays sometimes use with Figure 4 Same structure. Use Figure 4 The relay processing in the UE will be explained. The relay processing of UE805 in communication from UE811 to base station 801 will be explained. Radio signals from UE811 are received via antennas 307-1 to 307-4. The received signal is converted from the radio receiving frequency to a baseband signal by frequency conversion unit 306, and demodulation processing is performed in demodulation unit 308. In demodulation unit 308, waiting calculations and multiplication processes can be performed. The demodulated data is transmitted to decoding unit 309 for error correction and other decoding processing. The decoded data is transmitted to protocol processing unit 301, where protocol processing for communication with UE811, such as MAC, RLC, etc., is performed, including actions such as header removal in each protocol. Additionally, protocol processing for communication with base station 801, such as RLC, MAC, etc., is performed, including actions such as header assignment in each protocol. In the protocol processing unit 301 of UE811, PDCP and SDAP protocol processing are sometimes also performed. The data that has undergone protocol processing is transmitted to the encoding unit 304 for error correction and other encoding processing. Alternatively, data may be output directly from the protocol processing unit 301 to the modulation unit 305 without undergoing encoding processing. The data encoded by the encoding unit 304 is then modulated in the modulation unit 305. MIMO precoding may also be performed in the modulation unit 305. After the modulated data is converted into a baseband signal, it is output to the frequency conversion unit 306 and converted into a wireless transmission frequency. The transmission signal is then transmitted from antennas 307-1 to 307-4 to the base station 801.

[0175] The above content illustrates an example of UE805 relaying communication from UE811 to base station 801, but the same process is used in the relaying of communication from base station 801 to UE811.

[0176] 5G base stations can support Integrated Access and Backhaul (IAB) (see Non-Patent Documents 2, 20). An IAB-supporting base station (hereinafter sometimes referred to as an IAB base station) consists of a CU (IAB Host CU) acting as an IAB host, a DU (IAB Host DU) acting as an IAB host, and IAB nodes that connect to the IAB Host DU and the UE via radio interfaces. An F1 interface is provided between the IAB nodes and the IAB Host CU (see Non-Patent Document 2).

[0177] Figure 10 The diagram illustrates an example of IAB base station connections. IAB host CU901 is connected to IAB host DU902. IAB node 903 connects to IAB host DU902 using a radio interface. IAB node 903 connects to IAB node 904 using a radio interface. That is, multi-level connections of IAB nodes are sometimes performed. UE905 connects to IAB node 904 using a radio interface. UE906 sometimes connects to IAB node 903 using a radio interface, and UE907 sometimes connects to IAB host 902 using a radio interface. Multiple IAB host DU902s can connect to IAB host CU901, multiple IAB nodes 903 can connect to IAB host DU902, and multiple IAB nodes 904 can connect to IAB node 903.

[0178] In the connections between the IAB host DU and IAB nodes, and between IAB nodes, a BAP (Backhaul Adaptation Protocol) layer is set up (see Non-Patent Document 29). The BAP layer performs actions such as routing received data to the IAB host DU and / or IAB nodes, and mapping it to RLC channels (see Non-Patent Document 29).

[0179] As an example of the structure of the IAB host CU, the same structure as CU215 is used.

[0180] As an example of the structure of the IAB host DU, it uses the same structure as DU216. In the protocol processing section of the IAB host DU, BAP layer processing is performed, such as assigning BAP headers to downlink data, routing for IAB nodes, and removing BAP headers from uplink data.

[0181] As an example of the structure of IAB nodes, sometimes in addition to Figure 5 The structure shown is excluding the EPC communication unit 401, other base station communication units 402, and 5GC communication unit 412.

[0182] use Figure 5 , Figure 10The transmit / receive processing in the IAB node will be explained. The transmit / receive processing of IAB node 903 in communication between IAB host CU 901 and UE 905 will be described. In uplink communication from UE 905 to IAB host CU 901, the radio signal from IAB node 904 is received through antenna 408 (part or all of antennas 408-1 to 408-4). The received signal is converted from the radio receiving frequency to a baseband signal by frequency conversion unit 407, and demodulation processing is performed in demodulation unit 409. The demodulated data is transmitted to decoding unit 410 for error correction and other decoding processing. The decoded data is transmitted to protocol processing unit 403, where protocol processing for communication with IAB node 904, such as MAC, RLC, etc., and actions such as header removal in each protocol, are performed. In addition, routing to the IAB host DU 902 using the BAP header is performed, and protocol processing for communication with the IAB host DU 902, such as assigning headers to each protocol, is performed. The protocol-processed data is transmitted to the encoding unit 405 for error correction and other encoding processing. Alternatively, data may be output directly from the protocol processing unit 403 to the modulation unit 406 without encoding processing. The encoded data is modulated in the modulation unit 406. MIMO precoding may also be performed in the modulation unit 406. After the modulated data is converted into a baseband signal, it is output to the frequency conversion unit 407 and converted into a radio transmission frequency. Then, the transmission signal is transmitted to the IAB host DU 902 using antennas 408-1 to 408-4. The same processing is performed in downlink communication from the IAB host CU 901 to the UE 905.

[0183] In IAB node 904, the same send and receive processing is performed as in IAB node 903. In the protocol processing unit 403 of IAB node 903, as part of the BAP layer processing, such as assigning BAP headers in uplink communication and routing to IAB node 904, and removing BAP headers in downlink communication, etc.

[0184] In the 3GPP mobile communication system standard, the UE needs to perform RA (Random Access) processing during cell connection handling, such as cell selection or cell reselection. In RA processing, the UE sends a PRACH to the cell. If the UE cannot receive a PRACH response from the base station, it increases the transmit power by a specified amount and retransmits the PRACH. While significantly increasing the transmit power can establish a connection with fewer transmissions, it can also interfere with other UL communications. Therefore, the following problem arises: without deriving an appropriate PRACH transmit power, it is impossible to improve many of the KPIs required by the mobile communication system.

[0185] In this embodiment, a method is disclosed that enables more efficient RA processing corresponding to communication conditions.

[0186] In this embodiment, AI / ML is introduced into the RA process. Examples of RA processes include PRACH transmission processing, RA problem (Random Access Problem) processing, and backoff processing (see Non-Patent Document 15). For example, AI / ML is used in the RA process to optimize the information used in PRACH transmission processing, RA problem processing, or backoff processing.

[0187] The learning process of the model (sometimes referred to as training in this manual) is described. In this manual, the model (including the fully learned model) is sometimes referred to as an AI / ML model.

[0188] In the communication system involved in this embodiment, the following is adopted: Figure 11 The structure of the learning device 1100 shown. Figure 11 This is a diagram illustrating a structural example of a learning device 1100 related to a communication system. The learning device 1100 includes a data acquisition unit 1110 and a model generation unit 1120. The learned completed model generated by the model generation unit 1120 is stored by the learned completed model storage unit 1130. Figure 11 The diagram shows a structural example where the learning completion model storage unit 1130 is located outside the learning device 1100, but it is also possible for the learning device 1100 to include the learning completion model storage unit 1130. Figure 11 The learning device 1100 shown can be applied to a base station.

[0189] The data acquisition unit 1110 acquires input and feedback information for use as learning data.

[0190] The model generation unit 1120 learns the output information based on learning data containing input and feedback information. That is, it generates a learning-complete model that infers output information from the input information of the communication system.

[0191] The learning algorithm used by the model generation unit 1120 can be a known algorithm such as supervised learning, unsupervised learning, or reinforcement learning. As an example, we will illustrate the application of reinforcement learning. In reinforcement learning, an agent (acting entity) in an environment observes the current state (parameters of the environment) and decides on the action to be taken. Based on the agent's actions, the environment changes dynamically, and the agent receives rewards based on these changes. The agent repeats this process, learning the action strategy that yields the highest reward through a series of actions. Representative methods of reinforcement learning include Q-learning and TD-learning. For example, in the case of Q-learning, the general update formula for the action value function Q(s, a) is represented by equation (1).

[0192] [Mathematical Expression 1]

[0193]

[0194] In equation (1), s t a represents the state of the environment at time t. t This represents the action at time t. Through action a... t The state changes to s t+1 r t+1 Let γ represent the reward based on changes in state, α represent the discount rate, and α represent the learning coefficient. Additionally, γ is set to the range of 0 < γ ≤ 1, and α is set to the range of 0 < α ≤ 1. Use the input information as action a. t State s t And learn the state s at time t. t The best action under the following circumstances t .

[0195] If the action value Q of action a at time t+1, which has the highest Q value, is greater than the action value Q of action a performed at time t, then the update formula shown in equation (1) increases the action value Q; conversely, it decreases the action value Q. In other words, the action value function Q(s, a) is updated so that the action value Q of action a at time t is close to the optimal action value at time t+1. Thus, the optimal action value in a certain environment will propagate to the action values ​​in previous environments.

[0196] As described above, when a learning-completed model is generated through reinforcement learning, the model generation unit 1120 includes a reward calculation unit 1121 and a function update unit 1122.

[0197] The compensation calculation unit 1121 calculates compensation based on input information and feedback information. The compensation calculation unit 1121 calculates compensation r based on a compensation benchmark. For example, if the compensation increases benchmark, the compensation r is increased (e.g., a compensation of "1" is given); on the other hand, if the compensation decrease benchmark, the compensation r is decreased (e.g., a compensation of "-1" is given).

[0198] The function update unit 1122 updates the function used to determine the output according to the reward calculated by the reward calculation unit 1121, and outputs it to the learning completion model storage unit 1130. For example, in the case of Q-learning, the action value function Q(s) shown in equation (1) is used to calculate the action value function. t a t Use the function that outputs the information.

[0199] The learning device 1100 repeatedly performs the learning process as described above. Upon completion of the learning, the model storage unit 1130 stores the action value function Q(s) updated by the function update unit 1122. t a t That is, learning to complete the model.

[0200] Next, use Figure 12 This describes the learning process performed by the learning device 1100. Figure 12 This is a flowchart related to the learning process of the learning device.

[0201] In step b1, the data acquisition unit 1110 acquires input information and feedback information to be used as learning data.

[0202] In step b2, the model generation unit 1120 calculates the reward based on the input information and feedback information. Specifically, the reward calculation unit 1121 acquires the input information and feedback information, and determines whether to increase the reward (step b3) or decrease the reward (step b4) based on a predetermined reward benchmark.

[0203] If it is determined that the reward will increase, the reward calculation unit 1121 increases the reward in step b3. On the other hand, if it is determined that the reward will decrease, the reward calculation unit 1121 decreases the reward in step b4.

[0204] In step b5, the function update unit 1122 updates the action value function Q(s) shown in equation (1) stored in the learning completion model storage unit 1130 based on the reward calculated by the reward calculation unit 1121. t a t ).

[0205] The learning device 1100 repeats steps b1 to b5 above and generates the action value function Q(s). t a tThe learned model is stored in the learned model storage unit 1130 as a learned model.

[0206] The reasoning that uses the learned completion model is explained.

[0207] Figure 13 This is a structural diagram of an inference device 1200 related to a communication system. The inference device 1200 includes a data acquisition unit 1210 and an inference unit 1220.

[0208] The data acquisition unit 1210 acquires input information.

[0209] The reasoning unit 1220 uses the learned model to reason about output information. That is, by inputting the input information acquired by the data acquisition unit 1210 into the learned model, it can reason about output information appropriate to the input information.

[0210] In addition, this embodiment describes using a learning completion model learned by the model generation unit 1120 of the communication system to output the output information, but it is also possible to obtain a learning completion model from other communication systems and output the output information based on the learning completion model.

[0211] Figure 13 The diagram shows a structural example where the learning completion model storage unit 1130 is located outside the inference device 1200, but it is also possible for the inference device 1200 to include the learning completion model storage unit 1130. Figure 13 The inference device 1200 shown can be applied to a base station. The inference device 1200 can be applied to the same node as the learning device 1100, or it can be applied to a different node.

[0212] Next, use Figure 14 This describes the processing used to obtain output information using the inference device 1200.

[0213] In step c1, the data acquisition unit 1210 acquires the input information.

[0214] In step c2, the inference unit 1220 obtains the learned model from the learned model storage unit 1130, inputs the input information into the obtained learned model, and obtains the output information.

[0215] In step c3, the inference unit 1220 outputs the output information obtained from the learned model to the communication system.

[0216] In step c4, the communication system uses the output information to perform communication processing. This enables more efficient communication processing using the learned model.

[0217] Furthermore, this embodiment describes the application of reinforcement learning to the learning algorithm used by the inference unit 1220, but it is not limited to this. Regarding the learning algorithm, in addition to reinforcement learning, supervised learning, unsupervised learning, or semi-supervised learning can also be applied.

[0218] Furthermore, the learning algorithm used in the model generation unit 1120 can also use deep learning, which extracts the learning features themselves, and can perform machine learning based on other known methods, such as neural networks, genetic programming, inductive logic programming, support vector machines, etc.

[0219] Furthermore, the learning device 1100 and the inference device 1200 can be connected to a communication system, for example, via a network. Additionally, the learning device 1100 and the inference device 1200 can be built into the communication system. Furthermore, the learning device 1100 and the inference device 1200 can reside on a cloud server.

[0220] Furthermore, the model generation unit 1120 can learn output information using learning data acquired from multiple nodes. Additionally, the model generation unit 1120 can acquire learning data from multiple nodes used in the same area, or it can learn output information using learning data collected from multiple nodes operating independently in different areas. Furthermore, nodes that collect learning data can be added to or removed from the object along the way. Moreover, the learning device 1100, which has learned output information about a certain node, can be applied to nodes different from it, and the output information can be learned and updated again for those other nodes.

[0221] In this embodiment, the AI / ML model used for RA processing, i.e., the AI / ML model for RA processing, is set up at the base station. Learning (sometimes called model training) devices and inference devices can be set up in the base station. The AI / ML model for RA processing can be set up at the RAN node. For example, it can be set up at the CU or the DU. The AI / ML model for RA processing can be set up at the IAB host. For example, it can be set up at the IAB host CU or the IAB host DU. The AI / ML model for RA processing can be set up at the IAB node. The IAB nodes constitute an IAB-DU with DU functionality. For example, it can be set up at the IAB-DU. The base station and each node can use the AI / ML model to derive the configuration information required for RA processing, thus making configuration easy.

[0222] The base station can send information about whether it has an AI / ML model for RA processing (or information about having such a model) to the UE. This transmission can use RRC signaling. The base station can also send this information to neighboring base stations. This transmission can use Xn signaling. The base station can send this information to a CN (Core Network) node (or a function). This CN node could be, for example, an NWDAF (Network Data Analysis Function). The base station can also send this information to a management node. This management node could be, for example, an MnS (Management Service) node or an OAM (Operation, Administration and Management) node. Thus, the UE or node that received this information can identify whether the source base station has an AI / ML model for RA processing. It can then determine whether AI / ML can be imported into the RA processing using that base station.

[0223] Configure the input information for AI / ML models used in RA processing. The following are 30 examples of input information for AI / ML models used in RA processing.

[0224] (1) Information related to the state of the UE.

[0225] (2) Information related to the trajectory of the UE.

[0226] (3) Information related to the UE’s connection to the base station.

[0227] (4) Information related to the UE’s connection base station from neighboring base stations.

[0228] (5) Information related to UE cell selection and cell reselection.

[0229] (6) Information from neighboring base stations related to the UE’s cell selection and cell reselection.

[0230] (7) Information related to RA processing of the UE.

[0231] (8) Information related to RA processing of the UE from neighboring base stations.

[0232] (9) Information related to RA issues of UE.

[0233] (10) Information related to the RA problem of the UE from neighboring base stations.

[0234] (11) Information related to RLF (Radio Link Failure) caused by RA problems of UE.

[0235] (12) Information from neighboring base stations related to RLF caused by RA problems of the UE.

[0236] (13) Information related to the wireless measurement of the UE.

[0237] (14) Information related to the UE’s MDT (Minimization of Drive Tests).

[0238] (15) CLI (Cross-Link Interference Management) information in the UE.

[0239] (16) Information related to RACH settings.

[0240] (17) Information related to the base station’s resources.

[0241] (18) Information related to the UE’s traffic volume.

[0242] (19) Information related to RRC_Idle UE.

[0243] (20) Information related to RRC_Inactive UE.

[0244] (21) Information related to the load of the base station.

[0245] (22) Information related to UL interference.

[0246] (23) RIM (Remote Interference Management) information in the base station.

[0247] (24) RIM information in adjacent base stations.

[0248] (25) Information related to L2 (layer 2) measurement of UE.

[0249] (26) Information related to L2 measurement from the base station.

[0250] (27) Information related to the performance of the UE.

[0251] (28) Information related to the UE's QoE (Quality of Experience).

[0252] (29) Information about the propagation delay between the UE and the cell.

[0253] The combination of (30)(1) to (29).

[0254] (1) For example, it could be the UE's location, speed, and direction of movement. The UE's location could be coordinates, for example, a region, for example, information identifying the TA, RAN, base station, cell, CU, DU, or TRP to which the UE is located or connected, for example, information identifying the TA, RAN, base station, cell, CU, DU, or TRP with the best reception quality (or reception power). In addition, it could be time-related information. Time-related information could be, but is not limited to, hours, minutes, and seconds, but could also be years, months, days, days of the week, etc. In addition, it could be time-related information used in NW such as HFN (Hyper Frame Number) and SNF. This information could be associated with that time. In addition, it could be the UE's remaining battery power. In addition, it could be information related to the UE's power consumption. This information could be combined. The UE's state could be set as input information for training or inference.

[0255] (2) Information related to the UE's trajectory can, for example, appropriately apply information related to the UE's state disclosed in (1). For example, it can be information about connected base stations disclosed in (3). For example, it can be information related to cell selection and cell reselection disclosed in (5). It can be information related to the UE's past and current trajectories. Information related to the UE's past and current trajectories can be historical information. Trajectory-related information can be information related to the UE's future trajectories. Information related to the future trajectories can be predicted information. This information can be listed. In addition, it can include time-related information. Information about the UE's trajectory can be associated with information about time. Information about the UE's trajectory can be set as input information for training or inference.

[0256] (3) For example, it could be information about RAN nodes (e.g., base stations) that the UE has previously connected to. RAN nodes could be, for example, base stations, cells, cell groups, CUs, DUs, TRPs, etc. For example, base station information could be information identifying the base station. For example, it could be configuration information of base stations that the UE has previously connected to. For example, it could be historical information about base stations that the UE has previously connected to. For example, it could be a list of information about that base station. In addition, it could include information related to the UE's location. It could include time-related information. Information about previously connected base stations could be associated with information about location or time. Base stations that the UE has previously connected to could be limited to information about base stations connected through RA processing.

[0257] (4) For example, information about the base stations that the UE has previously connected to, which are located in adjacent base stations. Information related to the base stations to which the UE has connected may be the information disclosed in (3).

[0258] (5) For example, it could be information about base stations where the UE has previously performed cell selection and / or cell reselection. For example, it could be information identifying the base station. For example, it could be base station configuration information. It could be historical information about base stations where the UE has previously performed cell selection and / or cell reselection. For example, it could be a list of information about that base station. In addition, it could include information related to the UE's location. It could include time-related information. It could associate information about base stations that have previously performed cell selection and / or cell reselection with information about location or time. It could use information about the UE in RRC_Idle or RRC_Inactive states as input information for training or inference. Information about base stations where the UE has previously performed cell selection and / or cell reselection could be limited to information about base stations that have undergone RA processing.

[0259] (6) For example, information about neighboring base stations where the UE has previously performed cell selection and / or cell reselection. Information related to the UE's cell selection and / or cell reselection may be the information disclosed in (5).

[0260] (7) For example, it could be information related to past RA processing by the UE. For example, it could be information about whether the RA processing was successful. For example, it could be information about the base station to which the RA processing was performed. For example, it could be information identifying the base station. For example, it could be setting information used by the UE in RA processing. It could be historical information about past RA processing by the UE. For example, it could be information derived by the UE for performing RA processing (hereinafter referred to as UE derived information). For example, it could be the received power (RSRP) of the UE's SSB (Synchronization Signal Block) used to determine the PRACH transmit power, the path loss derived from it, the number of PRACH retransmissions, the PRACH transmit power determined using them, the selected RA resources, the time from PRACH transmission to RAR (Random Access Response) reception, the history of RA problems, etc. It could be historical information about past UE derived information. For example, it could be a list of such information. In addition, it could include information related to the UE's location. It could include time-related information. It could associate information about RA processing with information about location or time. For example, it could be that which base station the UE previously performed RA processing on as input information for training or inference.

[0261] (8) For example, information related to the RA processing of the UE that is possessed by a neighboring base station. The information related to the RA processing of the UE may be the information disclosed in (7).

[0262] (9) For example, information related to the UE's past RA problems. For example, it could be historical information about past RA problems of the UE. For example, it could be information related to the base stations where the UE previously experienced RA problems. For example, it could be information related to the settings used in the RA processing where the UE previously experienced RA problems. For example, it could be information related to the RACH settings at the time the RA problem occurred. For example, it could be the PRACH transmission power at the time the RA problem occurred. For example, it could be a list of this information. In addition, it could include information related to the UE's location. It could include time-related information. Information about RA problems could be associated with information about location or time. For example, it could be used as input information for training or inference, specifying which base stations and RA settings the UE previously experienced RA problems in.

[0263] (10) For example, information related to past RA problems of the UE that is held by neighboring base stations. Information related to the RA problems of the UE may be the information disclosed in (9).

[0264] (11) For example, information related to RLFs caused by past RA problems of the UE. For example, it could be historical information about RLFs caused by past RA problems of the UE. For example, it could be information related to the base stations where RLFs caused by past RA problems of the UE occurred. For example, it could be information related to the settings used in RA processing when RLFs occurred due to past RA problems of the UE. For example, it could be a list of such information. In addition, it could include information related to the location of the UE. It could include information related to time. It could associate information about RLFs caused by RA problems with information about location or time. For example, it could be used as input information for training or inference, such as which base stations and RA settings in which RLFs caused by RA problems of the UE occurred in the past.

[0265] (12) is information held by neighboring base stations related to RLF caused by past RA problems of the UE. The information related to RLF caused by past RA problems of the UE may be the information disclosed in (11).

[0266] (13) For example, it could be the UE's radio measurement results. For example, it could be RSRP, RSRQ, SINR (Signal to Interference Noise Ratio). It could contain information related to the UE's location. It could contain information related to time. It could associate the radio measurement results with information about location or time. It could set the reception quality based on the UE's location or state as input information for training or inference.

[0267] (14) For example, the measurement results of MDT. For example, it can be the measurement results of immediate MDT or the measurement results of logged MDT. It can be the measurement results of management-based MDT. It can be the measurement results of signaling-based MDT. It can be the measurement results of tracking function. The measurement results of MDT can be associated with information about the location or time of the UE. Information about the location or time of the UE can be obtained using the information measured by MDT. By setting up MDT for the UE, the measurement results of MDT can be obtained as input information for AI / ML models used in RA processing. More states or more information of the UE can be used as input information for training or inference.

[0268] (15) For example, it could be CLI measurement results in the UE (see Non-Patent Literature 2, Chapter 17.2). The CLI measurement results can be associated with information about the location or time of the UE. By setting up CLI measurements for the UE, CLI measurement results can be obtained as input information for training or inference. Information related to uplink interference from other UEs can be set as input information for training or inference.

[0269] (16) For example, information related to the RACH settings of the base station. This could include, for example, the base station's target receive power, the base station's RS transmit power, the damping factor, the maximum number of PRACH transmissions, the RAR window, or backoff indicator information. It could be information related to the current RACH settings or information related to past RACH settings. For example, it could be a list of this information. Furthermore, it could include information related to the location of the base station. It could include information related to the time when the RACH settings were performed. It could include information related to the load of the base station. Information about RACH settings could be associated with information about location, time, or load. For example, which RACH settings were performed in which base station could be used as input information for training or inference.

[0270] (17) For example, the resource usage of a base station. This can be the current resource usage or the predicted resource usage. For example, it can be the resource usage of each base station, each cell group, each cell, each DU, or each TRP. This resource can be frequency resource or time resource. Frequency resource usage includes the number of carriers (CC (Component Carrier), RB, subcarrier, etc. For example, it can be the usage of BWP. It can be a carrier unit, a PRB (Physical Resource Block) unit, or a subcarrier unit. Time resource usage includes the number of subframes, time slots, symbols, etc. It can be a subframe unit, a time slot unit, or a symbol unit. The amount of base station resources used can be used as input information for training or inference.

[0271] (18) For example, the traffic volume of UEs in a base station. This can be the current traffic volume or the predicted traffic volume. The traffic volume of a UE can be, for example, the number of UEs connected to the base station, or the number of UEs using the base station. For example, it can be the number of UEs connected or used per base station, per cell group, per cell, per DU, or per TRP. The number of UEs connected to the base station, etc., can be used as input information for training or inference.

[0272] (19) For example, the number of UEs in RRC_Idle. This can be the current number of UEs or the predicted number of UEs. The number of UEs in RRC_Idle can be limited to UEs that have completed location registration (sometimes called registration). For example, the number of UEs can be set by subtracting the number of UEs in RRC connection state and RRC_Inacive state from the number of UEs that have completed location registration. For example, it can be the number of UEs in RRC_Idle for each PLMN and each TA. For example, it can be the number of UEs in RRC_Idle within the paging notification area. For example, the number of UEs in RRC_Idle that can perform RA processing can be used as input information for training or inference.

[0273] (20) For example, the number of UEs with RRC_Inactive. This can be the current number of UEs or the predicted number of UEs. The number of UEs with RRC_Inactive can be, for example, the number of UEs in the base station that hold the UE context for RRC_Inactive. For example, it can be the number of UEs that the base station has notified RRC_Inactive of RRC_Inactive. For example, it can be the number of UEs with RRC_Inactive for each PLMN, each TA, or each RNA. For example, the number of RRC_Inactive UEs that can perform RA processing can be used as input information for training or inference.

[0274] (21) For example, base station load information. This can be current load information or predicted load information. Load information includes, for example, the number of DRBs (Data Radio Bearers) set by the base station, buffer usage, etc. It can be load information other than the resource information disclosed above, the traffic volume information of UEs, and the number of UEs with RRC_Idle or RRC_Inactive. The level of load of the base station can be set as input information for training or inference.

[0275] (22) For example, information related to UL interference in the base station. For example, it could be the UL interference power derived from the UL signal received by the base station from the UE. For example, it could be the UL interference power from a UE not connected to the base station. The UL interference power could be SINR. It could be the current UL interference power or the predicted UL interference power. The level of UL interference generated in the base station could be used as input information for training or inference.

[0276] (23) For example, it could be the RIM measurement results in the base station (see Chapter 17.1 of Non-Patent Literature 2). Information about interference from other base stations can be set as input information for training or inference.

[0277] (24) For example, it could be the RIM measurement results of adjacent base stations.

[0278] (25) For example, it could be the UE's measurement result of L2 measurement information. It could be the current L2 measurement result or the predicted L2 measurement result. The L2 measurement information could be, for example, the information measured by the UE as described in Non-Patent Document 32. For example, it could be the packet delay amount. The UE's L2 measurement result could be the L2 measurement result of a UE connected to a base station via RA processing. For example, it could be the UE's L2 measurement result for a PCell, PSCell, SPCell, or SCell connected via RA processing. In addition, it could include information related to the UE's location. It could include time-related information. It could associate information about the UE's L2 measurement with information about location or time. It could set information about the communication performance with the base station as input information for training or inference.

[0279] (26) For example, it can be the L2 measurement result of a base station. It can be the L2 measurement result of a neighboring base station. It can be the current L2 measurement result or the predicted L2 measurement result. The L2 measurement information of the base station can be, for example, the information measured by the base station as described in Non-Patent Document 32. For example, it can be the number of received RA preambles, the packet delay, the number of UEs in the connected state, the number of stored deactivated UE contexts, the packet loss rate, the PRB information for MIMO (Multiple Input Multiple Output), the number of PDCP packets in the forked DRB, the total RAN (Radio Access Network) delay in the forked DRB, etc. It can be the L2 measurement result of a neighboring base station. In addition, the L2 measurement result of the base station can be the L2 measurement result of the base station related to the UE. For the L2 measurement result of the base station related to the UE, it can include information related to the location of the UE. It can include information related to time. It can associate the L2 measurement information of the base station related to the UE with information about location or time. Information about the communication performance in the base station can be used as input for training or inference.

[0280] (27) For example, it could be the UE's QoS. It could be packet loss rate or latency. For example, it could be data communication quality. For example, it could be the UE's SSB reception quality. For example, it could be the UE's PDCCH and PDSCH reception quality. Reception quality could be RSRP, RSRQ, SINR, or BER. Information related to the UE's performance could be information after connection to the base station via RA processing. It could be information related to the performance required by the UE. In addition, it could include information related to the UE's location. It could include time-related information. Information about the UE's performance could be associated with information about location or time. Information about the UE's performance could be set as input information for training or inference.

[0281] (28) For example, it could be QoE measurement results (sometimes also called application layer measurement results). It could be information after connection to the base station via RA processing. It could be information related to the QoE requested by the UE. In addition, it could include information related to the UE's location. It could include time-related information. Information about QoE could be associated with information about the UE's location or time. Measurement results in the UE's application layer could be used as input information for training or inference.

[0282] (29) For example, it could be propagation delay information between the UE and the cell. Propagation delay information could be timing advance information, such as RTT (Round Trip Time) information. RTT could be a measured value of the transmit / receive time difference, for example, information measured by the UE or by the base station. It is not limited to the cell, but could also be the RAN node. For example, by using propagation delay information, a more accurate distance between the UE and the base station, and the location of the UE, can be obtained.

[0283] The aforementioned input information can be current information or predicted information. By setting the predicted information as input to the AI / ML model, more effective output information can be derived in the future.

[0284] The input information disclosed above can be associated with a base station. A base station can be, for example, a base station, a cell, a cell group, a CU, a DU, a TRP, etc. By associating it with a base station, information about which base station is being referred to can be taken into account in the input information.

[0285] The input information disclosed above can be associated with information about the UE's state. The disclosed information can be appropriately applied to the information about the UE's state. By associating it with the UE's state, it becomes possible to consider which UE states the input information pertains to.

[0286] During RA processing, there is information known only to the UE. For example, this includes the UE's SSB receive power (RSRP) used to determine the PRACH transmit power, as disclosed in example (7) above; the path loss derived from this; the number of PRACH retransmissions; the PRACH transmit power used to determine this; the selected RA resources; the time from PRACH transmission to RAR reception; and the history of RA issues. These are determined or selected by the UE, not set by the base station. By setting this information as input to the AI / ML model, the processing performed in the UE can be implemented more effectively.

[0287] RAN nodes send input information for AI / ML models used in RA processing to RAN nodes that have RA processing AI / ML models. This notification can include information indicating what the information is. For example, it can include information indicating that it is input information. Distinguishing between input information and feedback information becomes easier. The notification can include information indicating what the input information is. For example, it can include information indicating that it is training input information. Distinguishing it from other input information becomes easier.

[0288] A base station equipped with an AI / ML model for RA processing inputs the aforementioned disclosed AI / ML model for RA processing into the AI / ML model for training or inference. Thus, the AI / ML model for RA processing can be used to perform more efficient RA processing.

[0289] A base station sends input information for an AI / ML model used for RA processing to neighboring base stations. A base station can send this input information to a base station that has an AI / ML model for RA processing. A base station with an AI / ML model for RA processing can send a request for this input information to a neighboring base station. This request can include information related to the UE that is the object of RA processing. The information related to the UE can be the UE's identifier. The base station receiving the request sends the input information for the AI / ML model for RA processing to the base station that received the request. The transmission of the input information for the AI / ML model for RA processing between base stations, and the request for this information, can use inter-base station signaling. For example, Xn signaling can be used. For example, an Xn message can be set for this transmission. For example, current information and predicted information can be sent through different messages.

[0290] The DU sends input information for the AI / ML model used for RA processing to the CU. The DU can send this input information to a CU that has an AI / ML model for RA processing. The CU can be a CU-CP (Control Plane) or a CU-CP (User Plane). A CU with an AI / ML model for RA processing can send a request for input information to the DU. This request can contain information related to the UE that is the object of RA processing. The information related to the UE can be the UE's identifier. Upon receiving the request, the DU sends the input information for the AI / ML model used for RA processing to the CU that received the request. The transmission of the input information for the AI / ML model used for RA processing between CU and DU, and the request for this information, can use CU-DU signaling. For example, F1 signaling can be used. For example, an F1 message can be set for this transmission. For example, current information and predicted information can be sent through different messages.

[0291] The CU-UP can send input information for the AI / ML model used for RA processing to the CU-CP. The CU-UP can also send this input information to the CU-CP, which has an AI / ML model for RA processing. The CU-CP, which has an AI / ML model for RA processing, can send a request for this input information to the CU-UP. This request can contain information related to the UE, which is the target of RA processing. The information related to the UE can be the UE's identifier. Upon receiving the request, the CU-UP sends the input information for the AI / ML model used for RA processing to the CU-CP that is the source of the request. The transmission of the input information for the AI / ML model used for RA processing between the CU-UP and CU-CP, and the request for this information, can use signaling between the CU-UP and CU-CP. For example, E1 signaling can be used. For example, an E1 message can be configured for this transmission. For example, current information and predicted information can be sent through different messages.

[0292] The UE sends input information for the AI / ML model used for RA processing to the base station. The UE can send this input information to a base station that has an AI / ML model for RA processing. The base station with the AI / ML model for RA processing can send a request for this input information to the UE. This request can contain information related to the UE as the object of RA processing. The information related to the UE can be the UE's identifier. Upon receiving the request, the UE sends the input information for the AI / ML model used for RA processing to the base station that received the request. The transmission of the input information for the AI / ML model used for RA processing between the UE and the base station, and the request for this information, can be done using signaling between the UE and the base station. For example, RRC signaling can be used. Alternatively, MAC signaling can be used. L1 / L2 signaling can also be used. It can be included in the UCI. PUCCH can be used. The information and the request for this information should be sent as early as possible.

[0293] The transmission of this information and the request for this information between the UE and the base station can, for example, occur after the RA process is successful. For instance, it can be performed during the RRC connection process. This reduces the signaling required for this information and the request for this information.

[0294] Even without a request to receive this information, the UE-base station can still transmit it. For example, it can be transmitted after successful RA processing. Or, it can be transmitted after RRC connection processing is complete. This reduces the signaling required to request the information.

[0295] When a base station receives input information from a UE, it can send the input information to neighboring base stations. Even without receiving a request for input information from a neighboring base station, the base station can still send input information. This reduces the signaling required to request the information.

[0296] The above-described transmission methods can be combined for each request for this information between the UE and the base station. This allows for flexible configuration.

[0297] The output information of AI / ML models used for RA processing is disclosed. Information used in RA processing can be set as output information. Below are 14 examples of output information from AI / ML models used for RA processing.

[0298] (1) Information related to the trajectory of the UE.

[0299] (2) Information related to RACH settings.

[0300] (3) The priority order set by RACH.

[0301] (4) Information related to the base station candidates for RA processing.

[0302] (5) The arrival probability of the candidate base station to perform RA processing.

[0303] (6) Connection time for RA processing.

[0304] (7) Information related to the base station’s resources.

[0305] (8) Information related to the UE’s traffic volume.

[0306] (9) Information related to RRC_Idle UE.

[0307] (10) Information related to RRC_Inactive UE.

[0308] (11) Information related to the load of the base station.

[0309] (12) Information related to UL interference.

[0310] (13) Information about the propagation delay between the UE and the cell.

[0311] (14) Combinations of (1) to (13).

[0312] (1) For example, it could be the trajectory information of the predicted future UE. The information related to the UE's trajectory can be set as the information disclosed in example (2) of the input information for the AI / ML model used in RA processing. It can include information related to past and current trajectories. It can include historical information of past and current UE trajectories together with the trajectory information of the predicted future UE. This information can be listed. In addition, it can include time-related information. It can associate information about the UE's trajectory with information about time.

[0313] (2) Information related to the RACH setting predicted by the AI / ML model. The information related to the RACH setting can be the information disclosed in the example (16) of the input information for the AI / ML model used in RA processing. Furthermore, it can include information related to the predicted location of the UE. It can include information related to the predicted time. Information about the RACH setting can be associated with information about location or time. The information related to the RACH setting is not limited to one item; it can be multiple items. It can include information on the prediction accuracy of the information related to the RACH setting predicted by the AI / ML model. The UE can use the RACH setting derived from the AI / ML model. More efficient RA processing can be performed.

[0314] (3) is the priority order of information related to RACH settings as shown in (2). Priority can be set for RACH settings. Predicted RACH settings and their priorities can be correlated. Priority can be assigned to multiple output RACH settings. For example, if an RA problem occurs in the initial RACH setting, the next priority RACH setting can be used for RA processing. More flexible and efficient RA processing is possible.

[0315] (4) Information related to the base station candidates for predictive RA processing. Base station candidates for RA processing can be set. There are multiple base station candidates for RA processing, not just one. Information on the prediction accuracy of the base station candidates for predicted RA processing can be included. For example, the UE can identify which base station is more effective for RA processing.

[0316] (5) is the arrival probability information of the candidate base stations to which the predicted RA processing is performed. It may include information on the confidence interval of the arrival probability of the candidate base stations predicted by the AI / ML model. For example, the UE can identify which base station is more effective for RA processing.

[0317] (6) is the information on the predicted connection time of RA processing. The predicted connection time of RA processing can be set, for example, by setting candidates for each RACH. For example, it can be set for each base station candidate to perform RA processing. For example, the UE can identify which base station is more efficient to perform RA processing on.

[0318] (7) is the predicted resource status information of the base station. The resource status information of the base station can be set as the information disclosed in the input information example (17) of the AI / ML model for RA processing. The resource status of the base station can be the predicted resource status among the base station candidates. For example, neighboring base stations can identify the predicted resource status of the base station that has undergone RA processing.

[0319] (8) is the predicted traffic volume related to the UE. The traffic volume related to the UE can be the information disclosed in the input information example (18) of the AI / ML model for RA processing. The information about the predicted UE traffic volume can be the predicted UE traffic volume in the base station candidates. For example, neighboring base stations can identify the UE traffic volume predicted by the base station that has performed RA processing.

[0320] (9) For example, it could be information related to the predicted RRC_Idle UE. The information related to the RRC_Idle UE could be the information disclosed in the example (19) of the input information for the AI / ML model for RA processing. For example, neighboring base stations can identify the number of RRC_Idle UEs when RA processing has been performed.

[0321] (10) For example, it could be information related to predicted RRC_Inactive UEs. Information related to RRC_Inactive UEs could be the information disclosed in the example (20) of input information for the AI / ML model for RA processing. For example, neighboring base stations can identify the number of RRC_Inactive UEs when RA processing has been performed.

[0322] (11) For example, it could be information related to the load of the base station. The information related to the load of the base station could be the information disclosed in the example (21) of the input information of the AI / ML model for RA processing. For example, neighboring base stations can identify the load status of the base station when RA processing has been performed.

[0323] (12) For example, it could be information related to predicted UL interference. Information related to UL interference could be the information disclosed in example (22) of the input information for the AI / ML model used in RA processing. The information related to predicted UL interference could be information related to predicted UL interference in the base station candidates. For example, the UE and neighboring base stations can identify information related to UL interference predicted by the base station that has undergone RA processing.

[0324] (13) For example, it could be information related to the predicted propagation delay between the UE and the cell. The information related to the propagation delay between the UE and the cell can be the information disclosed in the example (29) of the input information for the AI / ML model used in RA processing. It can include information related to the propagation delay between the UE and the cell in the past and present. It can include information related to the predicted propagation delay between the UE and the cell in the future, and it can include historical information related to the propagation delay between the UE and the cell in the past and present. This information can be listed. In addition, it can include time-related information. It can associate the information about the propagation delay between the UE and the cell with the information about time. For example, the UE and the base station can identify information related to the propagation delay between the UE and the cell predicted by the base station that has performed RA processing.

[0325] Offsets can be set for the output information of the AI / ML model used in RA processing. Offsets can be set for some or all of the output information disclosed above. Offset values ​​are derived through the AI / ML model used in RA processing. This offset can be applied to each UE. By adding UE-specific offsets to the output information of each cell, individual settings can be configured for each UE. Output information optimization can be performed individually for each UE. This offset can be applied to each UE group. Offsets can be set for each specified group. For example, it can be set for UE groups providing the same service. For example, it can be set for UE groups existing in the same area. Output information can be optimized for each UE group. This offset can be applied to each base station, each cell group, each RAN, and each TA, for example. Predicted offsets can be set. Predicted offset values ​​can be derived through the AI / ML model used in RA processing. This allows for more flexible settings of output information.

[0326] For example, an offset for PRACH transmit power derivation can be set. For example, a UE-specific offset can be set for the target receive power of the base station. For example, a UE-specific offset can be set for the damping factor. Not only can positive values ​​be set as offset values, but also 0 or negative values ​​can be set. Therefore, each UE's offset value can be added to the RACH settings set for each cell, enabling RACH settings for each UE. By setting offsets, the RACH settings for each cell can be utilized as before.

[0327] The offset value can be set as the input information for the AI / ML model used in RA processing. This allows you to obtain AI / ML model and inference results that take the offset value into account.

[0328] As another method, the AI / ML model output information for RA processing can be configured for each UE. Some or all of the output information disclosed above can be set as the output information for each UE. As mentioned above, it is not possible to configure it for each individual UE, but rather for each group of UEs. For example, UE-specific output information can be configured via RRC, and the output information notified via SIB can be updated. Flexible configuration of the output information is possible.

[0329] Information relating to the relationship between the UE's state and the output information can be set. This information can be set as a list. The information relating to the UE's state can appropriately apply the information related to the UE's state disclosed in the above-described input information. For example, it could be a list indicating the relationship between the UE's location and RACH setting information. For example, it could be a list indicating the relationship between the RA processing start time and the RACH setting information. The information relating to the UE's state is not limited to one type; it can be a combination of multiple types of information disclosed above. Thus, the relationship between the UE's state and the output information can be shown.

[0330] Information relating to the relationship between the UE's state and the output information can be communicated in advance from the node with the AI / ML model to the node that needs the output information. The node needing the output information can be, for example, the UE, or, for example, a base station. Thus, for example, no notification from the base station to the UE is needed when the UE performs RA processing. Even if the UE is RRC_Idle or RRC_Inactive, the UE itself can determine which RACH setting to use based on its state.

[0331] The output information of the AI / ML model for RA processing disclosed above is appropriately notified to the RAN node or UE performing RA processing. A node with an AI / ML model can notify the UE of this information via the RAN node most recently connected to the UE. A node with an AI / ML model can broadcast this information to UEs within its coverage area. Thus, more efficient RA processing can be performed using the AI / ML model for RA processing. The notification can include information indicating what kind of information it is. For example, it can include information indicating that it is output information. Distinguishing between input information and feedback information becomes easier.

[0332] A base station sends AI / ML model output information for RA processing to neighboring base stations. A base station with an AI / ML model for RA processing can send this information to base stations related to RA processing. A base station related to RA processing can send a request for AI / ML model output information for RA processing to a base station with an AI / ML model for RA processing. The base station receiving the request sends the AI / ML model output information for RA processing to the requesting base station. The transmission of AI / ML model output information for RA processing between base stations, and the request for this information, can utilize inter-base station signaling. For example, Xn signaling can be used. For example, an Xn message can be configured for this transmission.

[0333] The CU sends AI / ML model output information for RA processing to the DU. A CU with an AI / ML model for RA processing can send this information to the DU. A DU can send a request for AI / ML model output information for RA processing to a CU with an AI / ML model for RA processing. The CU receiving the request sends the AI / ML model output information for RA processing to the requesting DU. The transmission of AI / ML model output information for RA processing between CU and DU, and the request for this information, can use CU-DU signaling. For example, F1 signaling can be used. For example, an F1 message can be configured for this transmission.

[0334] The base station sends AI / ML model output information for RA processing to the UE. A base station with an AI / ML model for RA processing can send this information to the UE. The UE can send a request for AI / ML model output information for RA processing to a base station with an AI / ML model for RA processing. The base station receiving the request sends the AI / ML model output information for RA processing to the UE that initiated the request. The transmission of the AI / ML model output information for RA processing and the request for this information between the UE and the base station can utilize signaling between the UE and the base station. For example, RRC signaling can be used. Alternatively, MAC signaling or L1 / L2 signaling can be used. The information and the request for this information should be sent as early as possible.

[0335] The base station can send the configuration information of each cell in the output information to the UE on a per-cell basis. For example, this information can be included in the SIB. The SIB can be broadcast. For example, this information can be sent to UEs within the cell coverage area via RRC signaling. The configuration information of each cell can be easily changed, and RA processing can be optimized on a per-cell basis.

[0336] The base station can send the configuration information for each cell in the output information to the UE individually. For example, this information can be sent using dedicated RRC signaling, MAC signaling, or L1 / L2 signaling. It can send the configuration information for each UE and optimize RA processing on a per-UE basis.

[0337] The methods for sending these output messages can be combined. The output messages can be flexibly configured, further optimizing RA processing.

[0338] Feedback information on RA processing using an AI / ML model for RA processing is disclosed (hereinafter, this feedback information is referred to as "RA processing feedback information using an AI / ML model for RA processing"). Information about each node in the RA processing, communication performance, etc., can be referred to as feedback information. Fifteen examples of RA processing feedback information using an AI / ML model for RA processing are disclosed below.

[0339] (1) Information related to the performance of the UE.

[0340] (2) Information related to the QoE of the UE.

[0341] (3) Information related to L2 measurement of UE.

[0342] (4) Information related to L2 measurement from the base station.

[0343] (5) Information related to RA processing of the UE.

[0344] (6) Information related to RA issues of UE.

[0345] (7) Information related to RLF caused by RA problems of UE.

[0346] (8) Information related to the base station’s resources.

[0347] (9) Information related to the UE’s traffic volume.

[0348] (10) Information related to RRC_Idle UE.

[0349] (11) Information related to RRC_Inactive UE.

[0350] (12) Information related to the load of the base station.

[0351] (13) Information related to UL interference.

[0352] (14) Information about the propagation delay between the UE and the cell.

[0353] (15) Combinations of (1) to (14).

[0354] (1) For example, it could be the input information example of the AI / ML model used in the above RA processing (27).

[0355] (2) For example, it could be the input information example of the AI / ML model used for RA processing mentioned above (28).

[0356] (3) For example, it could be the input information example of the AI / ML model used in the above RA processing (25).

[0357] (4) For example, it could be the input information example of the AI / ML model used in the above RA processing (26).

[0358] (5) For example, it could be (7) and (8) of the above RA processing AI / ML model using input information.

[0359] (6) For example, it could be (9) or (10) of the above RA processing AI / ML model using input information.

[0360] (7) For example, it could be (11) or (12) of the above RA processing AI / ML model using input information.

[0361] (8) For example, it could be (17) of the above RA processing using AI / ML model input information.

[0362] (9) For example, it could be (18) of the above RA processing using AI / ML model input information.

[0363] (10) For example, it could be (19) of the above RA processing using AI / ML model input information.

[0364] (11) For example, it could be (20) of the above RA processing using AI / ML model input information.

[0365] (12) For example, it could be (21) of the above RA processing using AI / ML model input information.

[0366] (13) For example, it could be (22) of the above RA processing using AI / ML model input information.

[0367] (14) For example, it could be the input information example of the AI / ML model used in the above RA processing (29).

[0368] The RA processing node appropriately notifies the nodes with the RA processing AI / ML model of the RA processing feedback information disclosed above. This allows for evaluation of the reward calculation of the RA processing AI / ML model and enables model updates. Consequently, more efficient RA processing can be performed. The notification may include information indicating what the information is. For example, it may include information indicating that it is feedback information. Distinguishing between input and output information becomes easier.

[0369] A base station sends RA processing feedback information using an AI / ML model for RA processing to neighboring base stations. A base station related to RA processing can send RA processing feedback information using an AI / ML model for RA processing to a base station with an AI / ML model for RA processing. A base station with an AI / ML model for RA processing can send a request for RA processing feedback information using an AI / ML model for RA processing to a base station related to RA processing. The base station receiving the request sends RA processing feedback information using an AI / ML model for RA processing to the requesting base station. The transmission of RA processing feedback information using an AI / ML model for RA processing between base stations, and the request for this information, can use inter-base station signaling. For example, Xn signaling can be used.

[0370] A DU sends RA processing feedback information using an AI / ML model for RA processing to a CU. A DU related to RA processing can also send RA processing feedback information using an AI / ML model for RA processing to a CU. A CU with an AI / ML model for RA processing can send a request for RA processing feedback information using an AI / ML model for RA processing to a DU. The DU receiving the request sends RA processing feedback information using an AI / ML model for RA processing to the requesting CU. The transmission of RA processing feedback information using an AI / ML model for RA processing and the request for this information between CUs and DUs can use CU-DU signaling. For example, F1 signaling can be used.

[0371] The UE sends RA processing feedback information using the AI / ML model for RA processing to the base station. The UE can send RA processing feedback information using the AI / ML model for RA processing to a base station that has an AI / ML model for RA processing. A base station with an AI / ML model for RA processing can send a request for RA processing feedback information using the AI / ML model for RA processing to the UE. Upon receiving the request, the UE sends RA processing feedback information using the AI / ML model for RA processing to the requesting base station. The transmission of RA processing feedback information using the AI / ML model for RA processing between the UE and the base station, and the request for this information, can be done using UE-base station signaling. For example, RRC signaling can be used. Alternatively, MAC signaling or L1 / L2 signaling can be used. The information and the request for this information should be sent as early as possible.

[0372] Figure 15 This is a diagram illustrating an example sequence of RA processing using AI / ML for RA processing in Implementation 1. An example of a RAN node having an AI / ML model for RA processing and performing training and inference is disclosed. The RAN node can be, for example, a base station.

[0373] In steps ST1501 and ST1502, RAN nodes #1 and #2 are equipped with AI / ML models for RA processing. AI / ML models for RA processing can be set for all RAN nodes capable of performing RA processing. The same AI / ML model can be set within a PLMN. The same model can be set within an RNA. The same model can be set across all regions. In step ST1503, the UE and RAN node #1 are in an RRC connection state. RAN node #1, equipped with the AI / ML model for RA processing, sends a request for model training input information to the UE in step ST1511, and sends the request to surrounding RAN nodes (e.g., RAN node #2) in step ST1512. Similarly, RAN node #2, equipped with the AI / ML model for RA processing, sends a request for model training input information to surrounding RAN nodes (e.g., RAN node #1) and to the UE (not shown) connected to RAN node #2 in step ST1513.

[0374] This request can be made based on receiving requests for input information from AI / ML models used for RA processing from surrounding RAN nodes. Alternatively, RAN node #1 can send a model training implementation request to surrounding RAN nodes. For example, if this RAN node decides to implement model training, it can send a model training implementation request to surrounding RAN nodes. Thus, not only RAN node #1, but also surrounding RAN nodes can implement model training using AI / ML models for RA processing.

[0375] The request may contain information indicating the required information. It may contain information indicating the purpose of the input information for the AI / ML model used in RA processing. For example, the request may contain information indicating it is for RA processing. For example, the request may contain information indicating it is for model training. This is effective when the input information for the AI / ML model used in RA processing includes information for model training. For example, if the information held by the UE includes information for model training, inference, and feedback, the UE can quickly grasp the information to be notified to the RAN node, resulting in rapid notification from the UE to the RAN node.

[0376] The transmission of this request between the RAN node and the UE can, for example, use RRC signaling. The RAN node's request for input information using the AI / ML model can be included in the measurement settings sent from the RAN node to the UE. This can reduce signaling volume.

[0377] The transmission of this request between RAN nodes can, for example, use Xn signaling.

[0378] In steps ST1514 and ST1516, the UE and RAN node #2 send model training input information to RAN node #1. RAN node #2 can send the model training input information obtained from the connected UE to RAN node #1. RAN node #1 can input the information of the UE connected to RAN node #2 into the AI / ML model. In step ST1515, RAN node #1 sends model training input information to RAN node #2. Furthermore, although not shown, the UE connected to RAN node #2 sends model training input information to RAN node #2. RAN node #1 can send the model training input information obtained from the UE to RAN node #2. RAN node #2 can input the information of the UE connected to RAN node #1 into the AI / ML model.

[0379] The UE's measurement results can be sent to the RAN node via a measurement result report. These results can be wireless measurements, L2 measurements, or MDT (Multi-Level Testing) results. MDT results can be recorded when the UE is in RRC_Inactive or RRC_Idle state. These results can be sent using the same signaling or different signaling. The UE's measurement results can be used as input information for model training.

[0380] The transmission of input information for model training between the RAN node and the UE can, for example, use RRC signaling. Alternatively, it can use UE Information Response messages. For instance, the transmission of radio measurement results and L2 measurement results can use Measurement Report messages. By using existing messages, processing complexity can be avoided.

[0381] The transmission of this input information between RAN nodes can, for example, use Xn signaling.

[0382] In steps ST1517 and ST1518, RAN nodes #1 and #2 use the input information of the AI / ML model for RA processing stored in their own nodes, as well as the input information of the AI / ML model for RA processing obtained from the UE and neighboring RAN nodes, to train the AI / ML model for RA processing. The AI / ML model for RA processing is updated through training. These processes can be performed appropriately. They can be performed periodically, or each time RA processing is performed in the local node or surrounding RAN nodes.

[0383] In step ST1521, RAN node #1 determines to release RRC for the UE. In steps ST1531 and ST1532, RAN node #1 sends an inference input information request to surrounding RAN nodes. RAN node #2, which has an AI / ML model for RA processing, also sends an inference input information request to surrounding RAN nodes in step ST1533, and sends the inference input information request to the UE (not shown) connected to RAN node #2. This request can be made based on receiving requests for input information for the AI / ML model for RA processing from surrounding RAN nodes. Alternatively, RAN node #1 can send an inference implementation request to surrounding RAN nodes. For example, RAN node #1, which has determined to release RRC for the UE, can send an inference implementation request to surrounding RAN nodes. Thus, not only RAN node #1, but also surrounding RAN nodes can implement inference using the AI / ML model for RA processing.

[0384] The request may contain information indicating the desired information. It may also contain information indicating the purpose of the input information for the AI / ML model used in RA processing. For example, the request may contain information indicating that it is for inference. This is effective, for example, when the input information for the AI / ML model used in RA processing includes information for inference.

[0385] In steps ST1534 and ST1536, the UE and RAN#2 node send inference input information to RAN#1. RAN#2 can send the inference input information obtained from the connected UE to RAN#1. RAN#1 can input the information of the UE connected to RAN#2 into the AI / ML model. In step ST1535, RAN#1 sends inference input information to RAN#2. Furthermore, although not illustrated, the UE connected to RAN#2 sends inference input information to RAN#2. RAN#1 can send the inference input information obtained from the UE to RAN#2. RAN#2 can input the information of the UE connected to RAN#1 into the AI / ML model.

[0386] RA processing uses AI / ML models to request input information (equivalent to...) Figure 15 The input information requests for model training and inference described in the document, and the input information for AI / ML models used in RA processing (equivalent to...) Figure 15 The method for sending the model training input information and inference input information described herein can be appropriately applied to the method disclosed above.

[0387] In steps ST1537 and ST1538, RAN nodes #1 and #2 use the input information of the AI / ML model for RA processing stored in their own nodes, as well as the input information of the AI / ML model for RA processing obtained from the UE and neighboring base stations, to perform inference using the AI / ML model for RA processing. The RAN nodes can then derive the output information of the AI / ML model for RA processing through this inference.

[0388] In step ST1541, RAN node #1 sends AI / ML model output information for RA processing to the UE. For example, it can send UE-specific output information, such as UE-specific RACH settings or UE-specific offsets added to the cell's RACH settings. Before the UE transitions to the RRC_Idle or RRC_Inactive state, such as before or during RRC release processing, the transmission and reception of AI / ML model output information for RA processing can be performed. Thus, the UE can obtain UE-specific output information.

[0389] AI / ML model output information for RA processing can be transmitted between RAN nodes. RAN node #1, which performed inference in step ST1537, can send output information (inference result) to RAN node #2. RAN node #2, which performed inference in step ST1538, can send output information (inference result) to RAN node #1. AI / ML model output information for RA processing can be shared between RAN nodes. RAN node #1 can send the output information received from RAN node #2 to the UE. For example, this can be done in step ST1541. The UE can obtain the AI / ML model output information for RA processing from each RAN node. Furthermore, this is effective, for example, when the AI / ML model output information for RA processing of RAN node #1 differs from that of RAN node #2.

[0390] In step ST1551, RAN node #1 performs RRC release processing with the UE. The UE then transitions to either the RRC_Idle or RRC_Inactive state.

[0391] In steps ST1561 and ST1562, RAN nodes #1 and #2 send output information derived from the AI / ML model used for RA processing. For example, they send RACH settings. This output information can be included in the SIB for broadcast, for example. These processes can be performed before the processing in step ST1551, or after steps ST1537 and ST1538. Thus, RAN node #1 and surrounding RAN nodes can send the AI / ML model output information for RA processing, and the UE can use this output information to perform RA processing.

[0392] The UE determines the RRC connection with the RAN node. The UE can move. In step ST1571, the UE uses the AI / ML model output information for RA processing received from RAN node #2 and / or the AI / ML model output information for RA processing received from RAN node #1 to perform RA processing with RAN node #2. Figure 15 The example shown illustrates the case where the UE selects RAN node #2. Therefore, RA processing can be performed between the UE and RAN node #2 using the output information derived from the AI / ML model of RA processing.

[0393] In step ST1572, the UE performs RRC connection processing with RAN node #2. This RRC connection processing can be part of the RA process. As a result, the UE can transition to an RRC connection state with RAN node #2. For example, data communication can be performed between the UE and the NW.

[0394] In step ST1581, the UE sends RA processing feedback information using the AI / ML model for RA processing to RAN node #2. The feedback information may contain information related to the RAN node with which the UE last made an RRC connection. For example, it may contain information identifying the RAN node. The RAN node receiving this information can identify the RAN node with which the UE last made an RRC connection and can then send the feedback information to that RAN node.

[0395] Alternatively, the UE can include this feedback information in the message sent to RAN node #2, or it can send a separate request for feedback information to RAN node #2. RAN node #2 can then send feedback information to neighboring RAN nodes based on the received information. RAN node #2 can use information received from the UE regarding the RAN node with which the last RRC connection was established to send feedback information to RAN node #1. The UE can appropriately request feedback information from RAN node #2 to RAN node #1. This allows for flexible control.

[0396] The feedback information sent in step ST1581 can be performed when the UE has performed RA processing. The UE can perform RA processing and communicate with the RAN node connected to the RRC. Therefore, the RAN node can obtain feedback information each time the UE performs RA processing.

[0397] As another method, the RAN node #2 connected to the UE during RA processing can send a request for feedback information to the UE. Upon receiving this request, the UE can send feedback information to RAN node #2. Thus, feedback information can be sent from the UE to RAN node #2 as needed, thereby reducing signaling load.

[0398] In step ST1581, RAN node #2, having received feedback information from the UE, sends the UE's feedback information to RAN node #1. In step ST1582, RAN node #2 sends the feedback information it possesses to RAN node #1. In step ST1583, RAN node #1 can send the feedback information it possesses to RAN node #2.

[0399] RA was used to process feedback information requests using AI / ML models. Figure 15 (The text is omitted here). It uses RA processing feedback information from AI / ML models used for RA processing (equivalent to...). Figure 15 The method of sending feedback information described herein may be appropriately applied to the methods disclosed above for input information.

[0400] Input and feedback requests for AI / ML models using RA processing can be eliminated. Each node can appropriately send input and feedback information for the AI / ML models used for RA processing to nodes with RA processing AI / ML models. For example, this can be done periodically. Alternatively, input and feedback information can be sent upon receiving new input or feedback information. This aims to reduce signaling load. When a node with an AI / ML model using RA processing receives this information from other nodes or its own node, it can send input requests to nodes other than that node. This reduces unnecessary signaling, such as sending identical input information.

[0401] Alternatively, any RAN node can train an AI / ML model. Updates to the trained AI / ML model can be sent from the RAN node to other RAN nodes. Other RAN nodes can then use the received update information to update their AI / ML models. The base station where the UE last established an RRC connection can perform AI / ML model inference and send the output information to the UE via RRCRelease processing. This simplifies the processing of AI / ML models used in the system.

[0402] Therefore, the RAN node can obtain information related to the performance of RA processing using the AI / ML model for RA processing. The RAN node can use the RA processing feedback information using the AI / ML model for RA processing stored in its own node, as well as the RA processing feedback information using the AI / ML model for RA processing obtained from the UE and surrounding RAN nodes, to evaluate the effect of RA processing when using the AI / ML model for RA processing.

[0403] Therefore, importing AI / ML into RA processing can lead to more efficient RA processing.

[0404] By employing the method disclosed in this embodiment, training and inference in RA processing can be performed using AI / ML models. By deriving the output information of the AI / ML model using various input information, RA processing can be optimized. Furthermore, by setting UE-specific RACH settings, RACH settings suitable for the UE state can be optimized. More appropriate RACH settings can be performed, thus shortening the time required for RA processing and cell connection via the UE. In addition, various KPIs can be optimized in RA processing, such as reducing UL interference, reducing UE transmit power, and reducing RA processing failures.

[0405] Implementation Method 1, Variation 1.

[0406] In order to solve the problems disclosed in Implementation 1, other methods that can perform more efficient RA processing corresponding to the communication status are disclosed.

[0407] This variation summarizes the setup of an AI / ML model for RA processing on the UE. A learning device and an inference device are configured on the UE. The UE can use the AI / ML model to derive the configuration information required for RA processing. The UE can be an IAB node. IAB nodes constitute an IAB-MT with terminal functionality. For example, the AI / ML model for RA processing can be configured on the IAB-MT. The IAB node can use the AI / ML model to derive the configuration information required for RA processing.

[0408] The input, output, and feedback information of the AI / ML model used in RA processing can be appropriately applied using the information disclosed in Implementation Method 1, achieving the same effect.

[0409] The UE can send information about whether it has an AI / ML model for RA processing (or information about having such a model) to the base station. This transmission can use RRC signaling. The base station receiving this information from the UE can then send the information about whether it has an AI / ML model for RA processing to neighboring base stations. This transmission can use Xn signaling. The UE can also send this information about whether it has an AI / ML model for RA processing to a CN (Core Network) node (or a function). The base station receiving this information from the UE can then send it to the CN node. The CN node could be, for example, an NWDAF (Network Data Analysis Function). The UE can also send this information about whether it has an AI / ML model for RA processing to a management node. The base station or CN node receiving this information from the UE can then send it to the management node. The management node could be, for example, an MnS (Management Service) node or an OAM (Operation, Administration and Management) node. Thus, the base station, CN node, or management node that notified the UE can identify whether the UE has an AI / ML model for RA processing. It can then determine whether AI / ML can be imported into the RA processing of the UE.

[0410] The UE receives AI / ML model input information for RA processing from the RAN node. This RAN node can be, for example, a base station or the serving cell. The UE can receive input information while in RRC connected state. To receive input information, the UE can transition to RRC connected state. The UE can perform RRC connected processing on the RAN node. Alternatively, the UE can use SDT (Small Data Transmission) processing to receive input information. This allows the UE to receive input information without transitioning to RRC connected state, reducing signaling load and avoiding processing complexity.

[0411] To enable RA processing for training AI / ML models, the UE can transition to RRC connected state. For example, it can receive input information for model training via a base station connected to RRC. Alternatively, the UE can continue training AI / ML models for RA without transitioning to RRC connected state. For example, it can use SDT processing to receive input information for model training.

[0412] Appropriate RA processing AI / ML model training can be performed. For example, it can be performed periodically. The model can be updated periodically. This period can be statically determined in advance through standards, etc., and can be set by the NW node. Alternatively, RA processing AI / ML model training can be implemented when the UE has performed RA processing. When a UE in RRC_Inactive or RRC_Idle makes an RRC connection to a cell that has undergone RA processing, it can receive input information for model training. Alternatively, RA processing AI / ML model training can be implemented when the UE has performed cell selection or cell reselection. For example, model training can be implemented when RAN update processing has been performed. For example, model training can be implemented when TA update processing has been performed. When a UE in RRC_Inactive or RRC_Idle makes an RRC connection to a new cell, it can receive input information for model training.

[0413] Inference using the AI / ML model for RA processing is implemented when the UE initiates RA processing. The UE can perform this in RRC_Idle or RRC_Inactive. The input information for inference can be limited to information related to the UE's RA processing and / or information related to the UE's state. Input information from other nodes can be omitted during inference within the UE.

[0414] The NW node can send a request to the UE to implement model training. The NW node can also send a request to the UE to implement inference. This request can be sent using paging processing. The request can be included in the paging information and sent from the base station. Alternatively, the request can be included in the SIB and broadcast by the base station. Thus, the NW node can send requests to the UE to implement model training and inference. The UE can implement model training and inference based on the requests from the NW node. The NW node can determine whether the UE should implement model training and inference.

[0415] The target users for model training and inference are not limited to one UE; they can be multiple UEs. They can be specific UEs. For example, they can be a defined group of UEs. For example, they can be a group of UEs providing the same service. For example, they can be UEs within the base station's coverage area. For example, they can be UEs within the RNA (Radio Area) or the TA (Target Area).

[0416] Figure 16 This is a diagram illustrating a sequence example of RA processing using AI / ML for RA processing in Variation 1 of Implementation 1. An example is disclosed where the UE has an AI / ML model for RA processing and is trained and inferred. Regarding... Figure 15 Common steps are labeled with the same step number, and common descriptions are omitted. In step ST1601, the UE has an AI / ML model for RA processing. In step ST1503, the UE and RAN node #1 are in RRC connection state.

[0417] In step ST1611, the UE with the AI / ML model for RA processing sends a request for model training input information to the connected RAN node (RAN node #1 in the example). In step ST1612, RAN node #1 sends a request for model training input information to the surrounding RAN node (RAN node #2 in the example). This request can be made based on the receipt of a request for AI / ML model input information for RA processing from the UE.

[0418] The methods disclosed in Implementation 1 can be appropriately applied to the information contained in the request. The request may contain information indicating that the UE has an AI / ML model or information identifying the UE. The RAN node (including surrounding RAN nodes) can quickly grasp the information to be notified to the UE, resulting in rapid notification from the RAN node to the UE.

[0419] The request for and transmission of this input information between the RAN node and the UE can, for example, use RRC signaling. The RAN node's request for input information using the AI / ML model can be included in the measurement settings sent from the RAN node to the UE. This can reduce signaling volume.

[0420] The request for and transmission of this input information between RAN nodes can, for example, use Xn signaling.

[0421] In step ST1615, the UE uses the input information of the AI / ML model for RA processing stored in the UE and the input information of the AI / ML model for RA processing obtained from the RAN node to train the AI / ML model for RA processing. The AI / ML model for RA processing is updated through training. These processes can be performed appropriately. They can be performed periodically. Through appropriate training, the AI / ML model is updated to be more suitable for more effective RA processing.

[0422] In step ST1551, RAN node #1 performs RRC release processing on the UE. Unlike implementation method 1, for example, RAN node #1 can initiate RRC release processing immediately after deciding to release the UE's RRC. The UE then transitions to either the RRC_Idle or RRC_Inactive state.

[0423] The transmission and reception of AI / ML input information for RA processing can be performed before the RRC release process in step 1551. For example, RAN node #1 can send AI / ML input information for RA processing to the UE. For example, RAN node #1 can send a request for AI / ML input information for RA processing to RAN node #2. Upon receiving the request, RAN node #2 can send the AI / ML input information for RA processing to RAN node #1. RAN node #1 can then send the input information received from RAN node #2 to the UE. RAN node #1 can also send the input information from both RAN node #1 and RAN node #2 to the UE. This input information can be used for training the AI / ML model or for inference. Thus, the UE can input the input information received from the RAN nodes into the AI / ML model before the RRC release. This shortens the time from receiving the input information from the RAN nodes until the UE performs RA processing. Therefore, especially in the case of inference, more efficient output information of the AI / ML model for RA processing can be obtained.

[0424] RAN nodes #1 and #2 send RACH settings in steps ST1561 and ST1562, respectively.

[0425] The UE determines the RRC connection with the RAN node. The UE can move. In step ST1631, the UE collects AI / ML model input information for RA processing within the UE. For example, it collects RA processing association information of the UE, information about the UE's state, etc.

[0426] In step ST1632, the UE uses the input information of the AI / ML model for RA processing in this UE and / or the RACH settings obtained from RAN node #1 to perform inference using the AI / ML model for RA processing. The UE can derive the output information of the AI / ML model for RA processing through this inference. As input information for this inference, the most recent model training input information obtained by the UE from RAN node #1 and RAN node #2 can be used. More types of input information can be input into this inference.

[0427] In step ST1571, the UE performs RA processing with RAN node #2. The UE uses the output information derived from the AI / ML model for RA processing to perform RA processing. Thus, RA processing using AI / ML can be performed.

[0428] In step ST1572, the UE performs RRC connection processing with RAN node #2. This RRC connection processing can be part of the RA process. As a result, the UE can transition to an RRC connection state with RAN node #2. For example, data communication can be performed between the UE and the NW.

[0429] In step ST1681, RAN node #2 sends RA processing feedback information to the UE using the AI / ML model for RA processing.

[0430] The feedback information in step ST1681 can be sent when the UE has performed RA processing. The UE can perform RA processing, which is carried out by the RAN node connected to the RRC. Thus, the UE can obtain feedback information each time RA processing is performed.

[0431] The UE can send a feedback request to RAN node #2. RAN node #2 can explicitly identify that the UE is requesting feedback. This request can include information related to the RAN node with which the UE last made an RRC connection. For example, it can include information identifying the RAN node. Upon receiving this information, RAN node #2 can identify the RAN node with which the UE last made an RRC connection. RAN node #2 can request feedback from the RAN node with which the UE last made an RRC connection (RAN node #1 in this example). This request can include information related to the UE that underwent RA processing. It can include information identifying the UE. Upon receiving the feedback request, RAN node #1 sends feedback to RAN node #2. RAN node #2 can then send the feedback information obtained from RAN node #1 to the UE. Thus, the UE can obtain feedback information not only from RAN node #2 but also from RAN node #1. RAN node #2 is not limited to RAN node #1 and can also request feedback from surrounding RAN nodes. Upon receiving this request, the RAN node sends feedback to RAN node #2. RAN node #2 can then send the feedback information obtained from surrounding RAN nodes to the UE. Therefore, the UE can obtain feedback information from the surrounding RAN nodes of RAN node #2.

[0432] Therefore, the UE can obtain information related to the performance of RA processing using the AI / ML model for RA processing. The UE can use the RA processing feedback information using the AI / ML model for RA processing stored in the UE and the RA processing feedback information using the AI / ML model for RA processing obtained from the RAN node to evaluate the effect of RA processing when using the AI / ML model for RA processing.

[0433] Input and feedback requests for AI / ML models using RA processing can be eliminated. Each node can appropriately send input and feedback information for AI / ML models using RA processing to the UEs that have such models. For example, this can be done periodically. Alternatively, input and feedback information can be sent upon receiving new input or feedback information. This aims to reduce signaling load. When a node with an AI / ML model using RA processing receives this information from other nodes or its own node, it can send input requests to nodes other than that node. This reduces unnecessary signaling, such as sending identical input information.

[0434] Therefore, importing AI / ML into RA processing can lead to more efficient RA processing.

[0435] By employing the method disclosed in this variation, the same effects as in Implementation 1 can be achieved. Furthermore, inference can be performed when the UE initiates RA processing, thus enabling more appropriate RA processing corresponding to the UE's state. Moreover, since the UE performs inference, the signaling volume used for inference and the inference processing can be reduced and simplified.

[0436] For nodes that set up AI / ML models for RA processing (in this specification, UE is sometimes also included in the node), the methods disclosed in Embodiment 1 and the methods disclosed in this variation can be combined. For example, a learning device for training an AI / ML model for RA processing can be set up in the RAN node, and an inference device can be set up in the UE. Model training can be performed using the method disclosed in Embodiment 1, and inference can be performed using the method disclosed in Variation 1 of Embodiment 1. The input information required for model training can be sent to the RAN node, and the information required for inference can be sent to the UE. For example, by training the model based on input information from multiple RAN nodes, the accuracy of the model can be improved, and by performing inference in the UE, the output information can be derived at a timing closer to RA processing, thereby enabling more efficient RA processing.

[0437] For example, RA processing can be configured using an AI / ML model at the RAN node and the UE. Input and feedback information for this model can be sent through the RAN node and the UE. For instance, by applying RA processing to a UE in the RRC_Connected state, more efficient RA processing can be performed.

[0438] The methods disclosed in Implementation 1 and this variation can be applied not only to RA processing when the UE is connected to the PCell of the MCG, but also to RA processing when connected to the SCell, PSCell, or SCell of the SCG. They can also be applied to RA processing when connected to a cell. Furthermore, they can be applied to RA processing performed by IAB nodes. Moreover, they can be applied not only to RA processing performed by UEs in the RRC_Idle and RRC_Inactive states, but also to RA processing performed by UEs in the RRC_Connected state. Furthermore, they can be applied to RA processing when the UE is connected to the RAN node of a HO target or the RAN node of a DC target. The same effects can be achieved.

[0439] Implementation method 2.

[0440] 3GPP supports SDT (Small Data Transmission) (see Non-Patent Literature Chapter 218). SDT enables data transmission and reception for RRC_INACTIVE UEs without transitioning to RRC_CONNECTED. Therefore, the UE can transmit and receive data as quickly as possible. For example, SDT is effective for intermittent data communication due to its low latency. However, SDT can only be activated when the UL data volume is below a specified threshold and the received power is higher than the specified threshold. The UL data volume threshold and received power threshold are set by the base station. Optimization is needed to ensure that RRC_INACTIVE UEs utilize the UL data volume threshold and received power threshold of SDT as much as possible.

[0441] In this embodiment, a method is disclosed that enables more efficient SDT processing corresponding to communication conditions.

[0442] In this embodiment, AI / ML is incorporated into the SDT processing. Examples of SDT processing include UL data volume thresholds, receive power thresholds, and settings for SDT timer values. AI / ML is used in the SDT processing to optimize this information.

[0443] The AI / ML models, learning devices, and inference devices can appropriately apply the methods disclosed in Implementation 1.

[0444] In this embodiment, the AI / ML model used for SDT processing, i.e., the AI / ML model for SDT processing, is set at the base station. The learning device and inference device can be set at the base station. The AI / ML model for SDT processing can be set at the RAN node. For example, it can be set at the CU or the DU. The base station and each node can use the AI / ML model to derive the configuration information required for SDT processing, thus simplifying the setup process.

[0445] Regarding whether an AI / ML model for SDT processing is available, the method disclosed in Implementation 1 can be appropriately applied.

[0446] Configure the input information for AI / ML models used in SDT processing. The following are 25 examples of input information for AI / ML models used in RA processing.

[0447] (1) Information related to the state of the UE.

[0448] (2) Information related to the trajectory of the UE.

[0449] (3) Information related to the UE’s connection to the base station.

[0450] (4) Information related to the UE’s connection base station from neighboring base stations.

[0451] (5) Information related to UE cell selection and cell reselection.

[0452] (6) Information from neighboring base stations related to the UE’s cell selection and cell reselection.

[0453] (7) Information related to the UE’s SDT processing.

[0454] (8) Information related to the UE’s SDT processing from neighboring base stations.

[0455] (9) Information related to the failure of the SDT timer for the UE.

[0456] (10) Information from neighboring base stations related to the failure of the UE's SDT timer.

[0457] (11) Information related to the wireless measurement of the UE.

[0458] (12) Information related to the UE’s MDT.

[0459] (13) Information related to SDT settings.

[0460] (14) Information related to the base station’s resources.

[0461] (15) Information related to the UE’s traffic volume.

[0462] (16) Information related to RRC_Idle UE.

[0463] (17) Information related to RRC_Inactive UE.

[0464] (18) Information related to the load of the base station.

[0465] (19) Information related to L2 (layer 2) measurement of UE.

[0466] (20) Information related to L2 measurement from the base station.

[0467] (21) Information related to the performance of the UE.

[0468] (22) Information related to the UE's QoE (Quality of Experience).

[0469] (23) Information related to the performance of the base station.

[0470] (24) Information about the propagation delay between the UE and the cell.

[0471] (25) (1) ~ (24) combinations.

[0472] (1) to (6) can respectively apply the input information example of the AI / ML model for RA processing disclosed in Embodiment 1. Regarding the information example (3) and (5) disclosed in Embodiment 1, the description of RA processing can be replaced with SDT processing.

[0473] (7) For example, it may be information related to SDT processing performed by the UE in the past. For example, it may be information on whether the SDT processing was successful. For example, it may be information about the base station that performed the SDT processing. For example, it may be information identifying the base station. For example, it may be setting information used by the UE in the SDT processing. It may be historical information about the SDT processing performed by the UE in the past. For example, it may be information exported by the UE for performing SDT processing (hereinafter referred to as UE exported information). For example, UL data volume, received power, etc. It may be historical information about past UE exported information. For example, it may be a list of such information. In addition, it may include information related to the UE's location. It may include information related to time. It may associate information about SDT processing with information about location or time. For example, it may set which base stations the UE has performed SDT processing on in the past as input information.

[0474] (8) For example, information related to the UE's SDT processing held by neighboring base stations. The information related to the UE's SDT processing may be the information disclosed in (7).

[0475] (9) For example, information related to past SDT timer failures of the UE. For example, it could be historical information about past SDT timer failures of the UE. For example, it could be information related to the base stations where past SDT timer failures of the UE occurred. For example, it could be information related to the settings used in the SDT processing where past SDT timer failures of the UE occurred. For example, it could be a list of such information. In addition, it could include information related to the UE's location. It could include time-related information. It could associate information about SDT timer failures with information about location or time. For example, it could set input information such as which base stations and SDT settings in which past SDT timer failures of the UE occurred.

[0476] (10) For example, information held by neighboring base stations related to past SDT timer failures of the UE. Information related to SDT timer failures of the UE may be the information disclosed in (9).

[0477] (11) and (12) can respectively apply (13) and (14) of the example of input information for the AI / ML model for RA processing disclosed in Implementation 1.

[0478] (13) Information related to the SDT settings of the base station, such as UL data volume threshold, receive power threshold, SDT timer value, etc. This information can be related to the current SDT settings or to past SDT settings. For example, it can be a list of such information. For example, it can be SDT settings sent in the SIB or SDT settings sent in the RRRCRelease message. For example, it can be SDT settings set for each cell or SDT settings set for each UE. Furthermore, it can include information related to the location of the base station. It can include information related to the time when the SDT settings were performed. It can include information related to the load of the base station. Information about SDT settings can be associated with information about location, time, or load. For example, it can be input information about which SDT settings were performed in which base station.

[0479] (14) to (18) can respectively apply (17) to (21) of the example of input information for the AI / ML model for RA processing disclosed in Implementation 1.

[0480] (19) to (22) can respectively apply (25) to (28) of the example input information for the AI / ML model for RA processing disclosed in Embodiment 1. Regarding (19), the phrase "can be the L2 measurement result of a UE connected to a base station through RA processing. For example, it can be the L2 measurement result of a UE connected to a PCell, PSCell, SPCell, or SCell through RA processing" in the input information (25) for the AI / ML model for RA processing disclosed in Embodiment 1 can be set to "can be the L2 measurement result of a UE communicating with a base station through SDT processing. For example, it can be the L2 measurement result of a UE communicating with a PCell, PSCell, SPCell, or SCell through SDT processing". Regarding (21), the phrase "can be the information after connecting to a base station through RA processing" in the input information (27) for the AI / ML model for RA processing disclosed in Embodiment 1 can be set to "can be the information after communicating with a base station through SDT processing". Regarding (22), the "It can be information after connecting to the base station through RA processing" in the input information (28) of the AI / ML model for RA processing disclosed in Implementation 1 can be set to "It can be information after communicating with the base station through SDT processing".

[0481] (23) For example, it could be information related to the throughput of the base station. For example, it could be information related to the latency of the base station. For example, it could be information related to the packet loss rate of the base station. It could be the total amount in the base station, it could be the average, or it could be the maximum or minimum value. For example, it could be information limited to connected UEs, it could be information limited to UEs that have undergone SDT processing, or it could be information combining them. In addition, it could include information related to the location of the base station. It could include time-related information. Information about the performance of the base station could be associated with information about location or time. Information about the performance of the base station could be set as input information for training or inference.

[0482] (24) The input information example of the AI / ML model for RA processing disclosed in Implementation 1 can be appropriately applied (29).

[0483] The input information for the AI / ML model used in the aforementioned RA processing can be either current information or predicted information. By setting the predicted information as the input information for the AI / ML model, more effective output information can be derived in the future.

[0484] The RA processing described above can be associated with a base station using an AI / ML model and the input information. The base station can be, for example, a base station, a cell, a cell group, a CU, a DU, a TRP, etc. By associating with a base station, information about which base station is being considered can be incorporated into the input information.

[0485] The RA processing described above can be correlated with input information using an AI / ML model, along with information about the UE's state. The disclosed information about the UE's state can be appropriately applied. By correlating this information with the UE's state, it becomes possible to consider which UE states the input information pertains to.

[0486] During SDT processing, there is information known only to the UE. For example, the amount of UL data and the received power disclosed in example (7) are input information. The UE uses this information to determine whether to perform SDT processing. By setting this information as input information to the AI / ML model, the processing performed in the UE can be implemented more effectively.

[0487] A base station equipped with an AI / ML model for SDT processing inputs the aforementioned disclosed AI / ML model for SDT processing into the model for training or inference. Thus, the AI / ML model for SDT processing can be used to perform more efficient SDT processing.

[0488] The input information transmission method for SDT processing between base stations, between CU-DU, and between UE and base station can appropriately apply the input information transmission method disclosed in Implementation 1.

[0489] The output information of AI / ML models used for SDT processing is disclosed. Information used for SDT processing can be set as output information. The following are 13 examples of output information from AI / ML models used for SDT processing.

[0490] (1) Information related to the trajectory of the UE.

[0491] (2) Information related to SDT settings.

[0492] (3) Priority order set by SDT.

[0493] (4) Information related to the base station candidates for SDT processing.

[0494] (5) The arrival probability of the candidate base station to perform SDT processing.

[0495] (6) Communication time processed by SDT.

[0496] (7) Information related to the base station’s resources.

[0497] (8) Information related to the UE’s traffic volume.

[0498] (9) Information related to RRC_Idle UE.

[0499] (10) Information related to RRC_Inactive UE.

[0500] (11) Information related to the load of the base station.

[0501] (12) Information about the propagation delay between the UE and the cell.

[0502] (13) Combinations of (1) to (12).

[0503] (1) The output information example of the AI / ML model for RA processing disclosed in Implementation 1 can be appropriately applied.

[0504] (2) Information related to the SDT settings predicted by the AI / ML model. The information related to the SDT settings can be the information disclosed in the example (13) of the input information for the AI / ML model used for SDT processing. Furthermore, it can include information related to the predicted location of the UE. It can include information related to the predicted time. Information about the SDT settings can be associated with information about location or time. The information related to the SDT settings is not limited to one, but can be multiple. It can include information on the prediction accuracy of the information related to the SDT settings predicted by the AI / ML model. The UE can use the SDT settings derived from the AI / ML model. More efficient SDT processing can be performed.

[0505] (3) is the priority order of information related to SDT settings as shown in (2). Priority can be set for SDT settings. Predicted SDT settings and their priorities can be correlated. Priority can be assigned to multiple output SDT settings. For example, if an SDT timer failure occurs in the initial SDT setting, the next priority SDT setting can be used for SDT processing. More flexible and efficient SDT processing is possible.

[0506] (4) and (5) can appropriately apply the output information examples disclosed in Embodiment 1. The RA processing described in the output information examples of AI / ML models disclosed in Embodiment 1 (4) and (5) can be replaced with SDT processing.

[0507] (6) is information about the predicted communication time for SDT processing. The predicted communication time for SDT processing can be set, for example, by setting candidates for each SDT. For example, it can be set for each base station candidate to perform SDT processing. For example, the UE can identify which base station is more effective for SDT processing.

[0508] (7) is the predicted resource status information of the base station. The resource status information of the base station can be set as the information disclosed in the input information example (14) of the AI / ML model for SDT processing. The resource status of the base station can be the predicted resource status among the base station candidates. For example, neighboring base stations can identify the predicted resource status of the base station that has undergone SDT processing.

[0509] (8) is the predicted traffic volume related to the UE. The traffic volume related to the UE can be the information disclosed in the example (15) of the input information for the AI / ML model used in SDT processing. The information about the predicted UE traffic volume can be the traffic volume of the UE predicted in the base station candidates. For example, neighboring base stations can identify the traffic volume of the UE predicted by the base station that has undergone RA processing.

[0510] (9) For example, it could be information related to the predicted RRC_Idle UE. The information related to the RRC_Idle UE could be the information disclosed in the example (16) of the input information for the AI / ML model for SDT processing. For example, neighboring base stations can identify the number of RRC_Idle UEs when SDT processing has been performed.

[0511] (10) For example, it could be the predicted information related to RRC_Inactive UEs. The information related to RRC_Inactive UEs could be the information disclosed in the example (17) of the input information for the AI / ML model for SDT processing. For example, neighboring base stations can identify the number of RRC_Inactive UEs when SDT processing has been performed.

[0512] (11) For example, it could be information related to the load of the base station. The information related to the load of the base station could be the information disclosed in the example (18) of the input information of the AI / ML model for SDT processing. For example, neighboring base stations can identify the load status of the base station when SDT processing has been performed.

[0513] (12) The output information example disclosed in Embodiment 1 (13) can be appropriately applied. The RA processing described in the output information example (13) of the AI / ML model disclosed in Embodiment 1 can be replaced with SDT processing.

[0514] An offset can be set for the output information of the AI / ML model used for SDT processing. The method for setting an offset for the output information of the AI / ML model used for RA processing disclosed in Implementation Method 1 can be appropriately applied.

[0515] For example, an offset for each UE can be set for the UL data volume threshold. Similarly, an offset for each UE can be set for the received power threshold. Not only can positive values ​​be set as offset values, but also 0 or negative values. Therefore, an offset value for each UE can be added to the SDT settings set for each cell, enabling SDT settings for each UE. By setting the offset, the SDT settings for each cell can be utilized as before.

[0516] The offset value can be set as the input information for the AI / ML model processed by SDT. This allows you to obtain AI / ML model and inference results that take the offset value into account.

[0517] As another method, the output information of the AI / ML model for SDT processing can be set for each UE. Some or all of the output information disclosed above can be set as the output information for each UE. As mentioned above, it can be set not for each UE, but for each UE group, etc. For example, UE-specific output information can be set via RRC, and the output information notified via SIB can be updated. Flexible setting of output information is possible.

[0518] Information relating to the relationship between the UE's state and the output information can be set. The method for setting this information and the method for notifying this information can appropriately apply the method disclosed in Embodiment 1.

[0519] The output information of the AI / ML model for SDT processing disclosed above is appropriately notified to the RAN node or UE performing SDT processing. A node with an AI / ML model can notify the UE of this information via the RAN node most recently connected to the UE. A node with an AI / ML model can broadcast this information to UEs within its coverage area. Thus, more efficient SDT processing can be performed using the AI / ML model for SDT processing. The notification can include information indicating what kind of information it is. For example, it can include information indicating that it is output information. Distinguishing between input information and feedback information becomes easier.

[0520] The method for transmitting output information of the AI / ML model for SDT processing between base stations, between CU-DU, and between UE and base station can appropriately apply the method for transmitting output information disclosed in Implementation 1.

[0521] Feedback information on SDT processing using an AI / ML model for SDT processing is disclosed (hereinafter, this feedback information is referred to as "SDT processing feedback information using an AI / ML model for SDT processing"). Information about each node in SDT processing, communication performance, etc., can be referred to as feedback information. The following are 14 examples of SDT processing feedback information using an AI / ML model for SDT processing.

[0522] (1) Information related to the performance of the UE.

[0523] (2) Information related to the QoE of the UE.

[0524] (3) Information related to L2 measurement of UE.

[0525] (4) Information related to L2 measurement from the base station.

[0526] (5) Information related to the UE’s SDT processing.

[0527] (6) Information related to the failure of the SDT timer for the UE.

[0528] (7) Information related to the base station’s resources.

[0529] (8) Information related to the UE’s traffic volume.

[0530] (9) Information related to RRC_Idle UE.

[0531] (10) Information related to RRC_Inactive UE.

[0532] (11) Information related to the load of the base station.

[0533] (12) Information related to the performance of the base station.

[0534] (13) Information about the propagation delay between the UE and the cell.

[0535] (14) Combinations of (1) to (13).

[0536] (1) For example, it could be the input information of the AI / ML model used in the above SDT processing example (1).

[0537] (2) For example, it could be the input information of the AI / ML model used in the above SDT processing example (22).

[0538] (3) For example, it could be the input information of the AI / ML model used in the above SDT processing example (19).

[0539] (4) For example, it could be the input information of the AI / ML model used in the above SDT processing example (20).

[0540] (5) For example, it could be (7) or (8) of the above-mentioned SDT processing AI / ML model using input information.

[0541] (6) For example, it could be (9) or (10) of the above-mentioned SDT processing AI / ML model using input information.

[0542] (7) For example, it could be the input information of the AI / ML model used in the above SDT processing example (14).

[0543] (8) For example, it could be the input information of the AI / ML model used in the above SDT processing example (15).

[0544] (9) For example, it could be the input information of the AI / ML model used in the above SDT processing example (16).

[0545] (10) For example, it could be (17) of the above SDT processing using AI / ML model input information.

[0546] (11) For example, it could be the input information of the AI / ML model used in the above SDT processing example (18).

[0547] (12) For example, it could be the input information of the AI / ML model used in the above SDT processing example (23).

[0548] (13) For example, it could be the input information of the AI / ML model used in the above SDT processing example (24).

[0549] The feedback information disclosed above can be an increase or decrease. For example, it could be an increase or decrease in the number of UEs' service volume. For example, it could be an increase or decrease in the number of RRC_Inactive UEs.

[0550] The SDT processing node appropriately notifies the nodes with the SDT processing AI / ML model of the SDT processing feedback information disclosed above. This allows for evaluation of the reward calculation of the SDT processing AI / ML model and enables model updates. Consequently, more efficient SDT processing can be performed.

[0551] The method for transmitting SDT processing feedback information using the AI / ML model for SDT processing between base stations, between CU-DU, and between UE and base stations can appropriately apply the method for transmitting RA processing feedback information using the AI / ML model for RA processing disclosed in Implementation 1.

[0552] Figure 17 This is a diagram illustrating an example sequence of SDT processing using AI / ML for SDT processing in Implementation 2. An example of a RAN node having an AI / ML model for SDT processing and performing training and inference is disclosed. The RAN node can be, for example, a base station. Regarding... Figure 15 Common steps are labeled with the same step number, and common descriptions are omitted.

[0553] In steps ST1701 and ST1702, RAN nodes #1 and #2 are equipped with AI / ML models for SDT processing. AI / ML models for SDT processing can be set for all RAN nodes capable of executing AI / ML for SDT processing. In step ST1503, the UE and RAN node #1 are in an RRC connection state. RAN node #1, equipped with an AI / ML model for SDT processing, sends a request for model training input information to the UE in step ST1711, and sends the request to surrounding RAN nodes (e.g., RAN node #2) in step ST1712. Similarly, RAN node #2, equipped with an AI / ML model for SDT processing, sends a request for model training input information to surrounding RAN nodes (e.g., RAN node #1) and to the UE (not shown) connected to RAN node #2 in step ST1713.

[0554] This request can be made based on receiving requests for input information from SDT processing AI / ML models from surrounding RAN nodes. Alternatively, RAN node #1 can send a model training implementation request to surrounding RAN nodes. For example, if this RAN node decides to implement model training, it can send a model training implementation request to surrounding RAN nodes. Thus, not only RAN node #1, but also surrounding RAN nodes can implement model training using SDT processing AI / ML models.

[0555] The request may contain information indicating the required information. It may contain information indicating the purpose of the input information for the AI / ML model used in SDT processing. For example, the request may contain information indicating it is for SDT processing. For example, the request may contain information indicating it is for model training. This is effective when the input information for the AI / ML model used in SDT processing includes information for model training. For example, if the information held by the UE includes information for model training, inference, and feedback, the UE can quickly grasp the information to be notified to the RAN node, resulting in rapid notification from the UE to the RAN node.

[0556] The transmission of this request between the RAN node and the UE can, for example, use RRC signaling. The RAN node's request for input information using the AI / ML model can be included in the measurement settings sent from the RAN node to the UE. This can reduce signaling volume.

[0557] The transmission of this request between RAN nodes can, for example, use Xn signaling.

[0558] In steps ST11714 and ST1716, the UE and RAN#2 node send model training input information to RAN#1. RAN#2 can send the model training input information obtained from the connected UE to RAN#1. RAN#1 can input the information of the UE connected to RAN#2 into the AI / ML model. In step ST1715, RAN#1 sends model training input information to RAN#2. Furthermore, although not shown, the UE connected to RAN#2 sends model training input information to RAN#2. RAN#1 can send the model training input information obtained from the UE to RAN#2. RAN#2 can input the information of the UE connected to RAN#1 into the AI / ML model.

[0559] The UE's measurement results can be sent to the RAN node via a measurement result report. These results can be wireless measurements, L2 measurements, or MDT (Multi-Level Testing) results. MDT results can be recorded when the UE is in RRC_Inactive or RRC_Idle state. These results can be sent using the same signaling or different signaling. The UE's measurement results can be used as input information for model training.

[0560] The transmission of input information for model training between the RAN node and the UE can, for example, use RRC signaling. Alternatively, it can use UE Information Response messages. For instance, the transmission of radio measurement results and L2 measurement results can use Measurement Report messages. By using existing messages, processing complexity can be avoided.

[0561] The transmission of this input information between RAN nodes can, for example, use Xn signaling.

[0562] In steps ST1717 and ST1718, RAN nodes #1 and #2 use the input information of the SDT processing AI / ML model stored in their own nodes, as well as the input information of the SDT processing AI / ML model obtained from the UE and neighboring RAN nodes, to train the SDT processing AI / ML model. The training is used to update the SDT processing AI / ML model. These processes can be performed appropriately. They can be performed periodically, or each time SDT processing is performed in the local node or surrounding RAN nodes.

[0563] In step ST1521, RAN node #1 determines to release RRC for the UE. In steps ST1731 and ST1732, RAN node #1 sends an inference input information request to surrounding RAN nodes. RAN node #2, which has an AI / ML model for SDT processing, also sends an inference input information request to surrounding RAN nodes in step ST1733, and sends the inference input information request to the UE (not shown) connected to RAN node #2. This request can be made based on receiving requests for input information for the AI / ML model for SDT processing from surrounding RAN nodes. Alternatively, RAN node #1 can send an inference implementation request to surrounding RAN nodes. For example, RAN node #1, which has determined to release RRC for the UE, can send an inference implementation request to surrounding RAN nodes. Thus, not only RAN node #1, but also surrounding RAN nodes can implement inference using the AI / ML model for SDT processing.

[0564] The request may contain information indicating the required information. The request may contain information indicating the purpose of the input information for the AI / ML model used in SDT processing. For example, the request may contain information indicating that it is for inference. This is effective, for example, when the input information for the AI / ML model used in SDT processing includes information for inference.

[0565] In steps ST1734 and ST1736, the UE and RAN#2 node send inference input information to RAN#1. RAN#2 can send the inference input information obtained from the connected UE to RAN#1. RAN#1 can input the information of the UE connected to RAN#2 into the AI / ML model. In step ST1735, RAN#1 sends inference input information to RAN#2. Furthermore, although not illustrated, the UE connected to RAN#2 sends inference input information to RAN#2. RAN#1 can send the inference input information obtained from the UE to RAN#2. RAN#2 can input the information of the UE connected to RAN#1 into the AI / ML model.

[0566] SDT processing uses AI / ML models to request input information (equivalent to...) Figure 17 The input information requests for model training and inference described herein, and the input information for AI / ML models used in SDT processing (equivalent to...) Figure 17 The method for sending the model training input information and inference input information described herein can be appropriately applied to the method disclosed above.

[0567] In steps ST1737 and ST1738, RAN nodes #1 and #2 use the input information of the SDT processing AI / ML model stored in their own nodes, as well as the input information of the SDT processing AI / ML model obtained from the UE and neighboring base stations, to perform inference using the SDT processing AI / ML model. The RAN nodes can then derive the output information of the SDT processing AI / ML model through this inference.

[0568] In step ST1741, RAN node #1 sends SDT processing AI / ML model output information to the UE. For example, UE-specific output information can be sent. For example, UE-specific SDT settings can be sent. For example, UE-specific offsets added to the cell's SDT settings can be sent. The transmission and reception of SDT processing AI / ML model output information can be performed before the UE transitions to the RRC_Idle or RRC_Inactive state, for example, before or during RRC release processing. Thus, the UE can obtain UE-specific output information.

[0569] SDT processing AI / ML model output information can be transmitted between RAN nodes. RAN node #1, which performed inference in step ST1737, can send output information (inference result) to RAN node #2. RAN node #2, which performed inference in step ST1738, can send output information (inference result) to RAN node #1. SDT processing AI / ML model output information can be shared between RAN nodes. RAN node #1 can send the output information received from RAN node #2 to the UE. For example, this can be done in step ST1741. The UE can obtain the SDT processing AI / ML model output information of each RAN node. Furthermore, this is effective, for example, when the SDT processing AI / ML model output information of RAN node #1 differs from that of RAN node #2.

[0570] In step ST1551, RAN node #1 performs RRC release processing with the UE. The UE then transitions to either the RRC_Idle or RRC_Inactive state.

[0571] In steps ST1761 and ST1762, RAN nodes #1 and #2 send output information derived from the AI / ML model using SDT processing. For example, they send SDT settings. This output information can be included in the SIB for broadcast, for example. These processes can be performed before the processing in step ST1551, or after steps ST1737 and ST1738. Thus, RAN node #1 and surrounding RAN nodes can send SDT processing AI / ML model output information, and the UE can use this output information to perform SDT processing.

[0572] The UE determines its RRC connection with the RAN node. The UE can move. In step ST1771, the UE uses the SDT processing AI / ML model output information received from RAN node #2 and / or the SDT processing AI / ML model output information received from RAN node #1 to evaluate whether SDT can be performed. The RAN node to perform SDT processing can be determined. The figure illustrates the case where the UE selects RAN node #2. Thus, SDT processing can be performed between the UE and RAN node #2 using the output information derived from the SDT processing AI / ML model.

[0573] In step ST1772, the UE performs SDT processing with RAN node #2. This allows the UE to communicate without transitioning to an RRC connection state with RAN node #2. For example, data communication can be performed between the UE and the NW.

[0574] The UE sends SDT processing feedback information to RAN node #2 using the AI / ML model for SDT processing. The transmission of feedback information can occur during SDT processing. The feedback information may contain information related to the RAN node with which the UE last made an RRC connection. For example, it may include information identifying the RAN node. The RAN node receiving this information can identify the RAN node with which the UE last made an RRC connection and can then send the feedback information to that RAN node.

[0575] Alternatively, the UE can include the feedback information in the message and send it to RAN node #2, or it can send a separate request for feedback information to RAN node #2. This transmission can be performed during SDT processing. RAN node #2 can then send feedback information to neighboring RAN nodes based on the received information. RAN node #2 can use information received from the UE regarding the RAN node with which the last RRC connection was established to send feedback information to RAN node #1. The UE can appropriately request feedback information from RAN node #2 to RAN node #1. This allows for flexible control.

[0576] Feedback information can be sent even after the UE has performed SDT (Software-Defined Technology) processing. SDT processing can be performed and communicated to the RAN node connected to the RRC (Remote Radio Frequency) system. Therefore, the RAN node can obtain feedback information each time the UE performs SDT processing.

[0577] Alternatively, the RAN node #2 connected to the UE via SDT processing can send a request for feedback information to the UE. This request can be sent within the SDT process. Upon receiving the request, the UE can send feedback information to RAN node #2. Thus, feedback information can be sent from the UE to RAN node #2 as needed, potentially reducing signaling load.

[0578] In the SDT processing of step ST1772, RAN node #2, which receives feedback information from the UE, sends the UE's feedback information to RAN node #1. In step ST1782, RAN node #2 sends the feedback information it possesses to RAN node #1. In step ST1783, RAN node #1 can send the feedback information it possesses to RAN node #2.

[0579] SDT was used to process the RA (Research and Analysis) requests for feedback information using AI / ML models. Figure 17 (The text is omitted here). It uses SDT processing feedback information from AI / ML models used for SDT processing (equivalent to...). Figure 17 The method of sending feedback information described herein may be appropriately applied to the methods disclosed above for input information.

[0580] Therefore, the RAN node can obtain information related to the performance of RA processing using the AI / ML model for SDT processing. The RAN node can use the SDT processing feedback information using the AI / ML model for SDT processing stored in its own node, as well as the SDT processing feedback information using the AI / ML model for SDT processing obtained from the UE and surrounding RAN nodes, to evaluate the effect of SDT processing when using the AI / ML model for SDT processing.

[0581] Input and feedback requests for AI / ML models using SDT processing can be eliminated. Each node can appropriately send input and feedback information for AI / ML models using RA processing to nodes with SDT processing. For example, this can be done periodically. Alternatively, input and feedback information can be sent upon receiving new input or feedback information. This aims to reduce signaling load. When a node with an AI / ML model using SDT processing receives this information from other nodes or itself, it can send input requests to nodes other than that node. This reduces unnecessary signaling, such as sending identical input information.

[0582] Alternatively, any RAN node can perform AI / ML model training. The method disclosed in Implementation Method 1 can be applied appropriately. This simplifies the processing of AI / ML models used as the system.

[0583] Therefore, importing AI / ML into SDT processing enables more efficient SDT processing.

[0584] By employing the method disclosed in this embodiment, AI / ML models can be used for training and inference in SDT processing. By deriving the output information of the AI / ML model using various input information, SDT processing can be optimized. Furthermore, by setting UE-specific SDT settings, SDT settings suitable for the UE's state, such as UL data volume thresholds, received power thresholds, and optimization of SDT timers, can be performed. More appropriate SDT settings can be achieved. Thus, for example, UEs with RRC_INACTIVE can utilize SDT as much as possible, and the UE can transmit and receive data as early as possible. In SDT processing, various KPIs can be optimized, such as reduced latency, improved reliability, and enhanced user experience.

[0585] The above discloses that SDT processing can be configured using an AI / ML model on the RAN node. Alternatively, the SDT processing can be configured using an AI / ML model on the UE. Learning and inference devices can be configured on the UE. The UE can use the AI / ML model to derive the configuration information required for SDT processing.

[0586] The method for setting up an AI / ML model for SDT processing for a UE can be appropriately applied to the method for setting up an AI / ML model for RA processing for a UE disclosed in Variation 1 of Implementation 1.

[0587] Thus, the same effect as described above can be achieved. Furthermore, inference can be performed when the UE initiates SDT processing, allowing for more appropriate SDT processing corresponding to the UE's state. Moreover, since the UE performs inference, the signaling volume used for inference and the inference processing can be reduced and simplified.

[0588] Regarding the configuration of nodes using the AI / ML model for SDT processing, the methods for configuring RAN nodes and UE nodes can be combined to achieve more efficient SDT processing.

[0589] Implementation method 3.

[0590] UEs in RRC_Idle and RRC_Inactive states do not connect to the base station. During communication processing in these states, such as cell selection and cell reselection, the UE uses information configured for each cell. Therefore, the UE cannot perform communication processing tailored to its specific state, resulting in suboptimal communication performance.

[0591] In this embodiment, as a method to solve the above problems, a method for improving the communication performance of the UE in communication processing under RRC_Idle and RRC_Idle is disclosed.

[0592] In this embodiment, information provided to the UE for communication processing of the UE in RRC_Idle and RRC_Inactive states, or information obtained by the UE, is associated with information about the UE's state. Sometimes, this associated information is called state-related information. The information provided to the UE or the information obtained by the UE is not limited to one type; it can be multiple types. The information related to the UE's state is not limited to one type; it can be multiple types. State-related information can be set up as a list.

[0593] Information about the UE's state can be appropriately applied using the input information example (1) of the AI / ML model for RA processing disclosed in Implementation 1. For example, the UE's location, time, moving speed, moving direction, etc.

[0594] Communication processing for UEs under RRC_Idle and RRC_Inactive includes cell selection processing, cell reselection processing, DRX processing, RA processing, SDT processing, EDT (Early Data Transmission) processing, Control Plane CIoT 5GS Optimization processing, and User Plane CIoT 5GS Optimization processing.

[0595] The base station sends state association information to the UE. This UE can be a UE within the base station's coverage area or a standalone UE. The UE uses the state association information in communication processing under RRC_Idle and RRC_Inactive. The UE can use the state association information to derive information used in communication processing under RRC_Idle and RRC_Inactive. The UE uses the derived information to perform communication processing under RRC_Idle and RRC_Inactive. Therefore, when initiating communication processing under RRC_Idle and RRC_Inactive, the UE can use information that takes into account its current state to implement communication processing.

[0596] AI / ML can be imported into the communication processing of UEs with RRC_Idle and RRC_Inactive. AI / ML can be used to export configuration information for this communication processing. The AI / ML model used for this communication processing is called the AI / ML model for communication processing. AI / ML can be used to export state association information. The exported information can be set as the output information of the AI / ML model.

[0597] In this embodiment, a communication processing AI / ML model is set for the RAN node. The RAN node can be, for example, a base station, CU, DU, IAB node, IAB host, etc. By setting it in the RAN, for example, the communication processing configuration information set in the RAN can be directly sent to the UE. This avoids processing complexity. The method of setting the communication processing AI / ML model for UEs with RRC_Idle or RRC_Inactive in the RAN node can appropriately be the method of setting the RA processing AI / ML model in the RAN node disclosed in Embodiment 1, or the method of setting the SDT processing AI / ML model in the RAN node disclosed in Embodiment 2. Information specific to each process, such as information related to each process of the AI / ML model input information, can be set as information related to the individual communication processes (e.g., the cell selection process, cell reselection process, etc. disclosed above). Alternatively, the communication processing AI / ML model can be set in the CN node. The CN node can be, for example, an AMF, NWDAF, etc. By setting it in the CN node, for example, setting the communication processing configuration information for UEs moving across multiple RAN nodes becomes easier. As another method, the communication processing AI / ML model can be set up on a management node. This management node could be, for example, an MnS node or an OAM node. By setting it up on a management node, for example, the management and transmission of communication processing configuration information can be performed together with the management of the base station and UE. Control combined with other management functions becomes possible. As a control combined with other management functions, mobility robustness optimization, etc., are possible. The learning device and inference device can also be set up on the RAN node, the CN node, or the management node, achieving the same effect. Furthermore, the above methods can be combined, enabling flexible application of AI / ML.

[0598] The validity period can be set for the configuration information used in communication processing of UEs in RRC_Idle and RRC_Inactive states. The validity period can be a time limit. It can also be a timer. The validity period is not limited to hours, minutes, or seconds; it can also be years, months, days, days of the week, etc. Furthermore, it can be information related to time used in NWs such as HFN and SNF. For example, the validity period is activated when the UE receives the configuration information, and is not used when the validity period expires. The validity period is activated again when new configuration information is notified, and communication processing is performed using this new configuration information. This validity period can be notified to the UE from the base station. For example, the base station can include it in the SIB and send it to UEs within its coverage area, or it can send it to the UE separately via RRC signaling. In the configuration information used in communication processing of UEs in RRC_Idle and RRC_Inactive states, factors such as the radio wave propagation environment and the base station load that change over time can be considered.

[0599] The validity period of the configuration information used for communication processing of RRC_Idle and RRC_Inactive UEs can be set as the input, output, and feedback information of the AI / ML model used for communication processing of RRC_Idle and RRC_Inactive UEs. AI / ML models that take into account time-varying radio wave propagation environments and base station load can be used.

[0600] By employing the method disclosed in this embodiment, the UE can consider its state during communication processing under RRC_Idle and RRC_Inactive. Since the state of each UE can be considered, communication processing suitable for each UE can be performed. Furthermore, by incorporating AI / ML into the communication processing under RRC_Idle and RRC_Inactive, the UE can use AI / ML models for training and inference during communication processing under these conditions, enabling more efficient communication processing.

[0601] The above content discloses the method of configuring the communication processing of RRC-Idle and RRC-Inactive UEs using an AI / ML model at the RAN node. Alternatively, the communication processing of RRC-Idle and RRC-Inactive UEs can be configured using an AI / ML model at the UE itself. Learning and inference devices can be configured on the UE. The UE can use the AI / ML model to derive the configuration information required for the communication processing of RRC-Idle and RRC-Inactive UEs.

[0602] The method of setting the communication processing of RRC_Idle and RRC_Inactive UEs using an AI / ML model can be appropriately applied to the method of setting the RA processing of UEs using an AI / ML model as disclosed in Variation 1 of Implementation 1.

[0603] Thus, the same effect as described above can be achieved. Furthermore, inference can be performed during the communication processing of UEs initiating RRC_Idle or RRC_Inactive states, allowing for more appropriate communication processing for RRC_Idle or RRC_Inactive states that corresponds to the UE's condition. Moreover, since the UE performs inference, the signaling volume used for inference and the inference processing can be reduced, thus simplifying the process.

[0604] Regarding the communication processing of UEs with RRC_Idle and RRC_Inactive settings using the AI / ML model, the methods used for setting these settings at the RAN node and the UE can be combined. This allows for more efficient communication processing of UEs with RRC_Idle and RRC_Inactive settings.

[0605] The above content discloses the AI / ML model for communication processing of UEs with RRC_Idle and RRC_Inactive settings. However, the AI / ML model for communication processing can also be set for each individual communication process of UEs with RRC_Idle and RRC_Inactive settings (e.g., the cell selection process, cell reselection process, etc. disclosed above).

[0606] Implementation method 4.

[0607] 3GPP supports IABs (see non-patent documents 2, 20, 29). An IAB node consists of an IAB-MT (Mobile Termination) with terminal functionality and an IAB-DU with DU functionality. The parent IAB-DU of the IAB-MT is selected by the IAB host CU based on the measurement results of the IAB-MT. This method only considers the radio wave propagation environment between the IAB-MT and the parent IAB-DU. Depending on the status of the IAB nodes on the backhaul (BH) path and the radio wave propagation environment between IAB nodes, the selected parent IAB-DU may not be optimal. For example, even if the radio wave propagation environment between the IAB-MT and the selected parent IAB-DU is optimal, if the radio wave propagation environment between the IAB node containing the parent IAB-DU and its parent IAB node is poor, or if the load of any IAB node on the BH path is high, communication performance deteriorates, resulting in decreased throughput, etc., in communication using the BH path.

[0608] In this embodiment, a method is disclosed that can solve the above problems and achieve more efficient IAB processing corresponding to the communication status.

[0609] AI / ML is imported into IAB processing. IAB processing includes, for example, BAP configuration processing and IAB node resource configuration processing. Examples include IAB node integration processing, IAB-MT configuration processing, IAB-DU configuration processing, BH RCL channel establishment processing, and path update processing. AI / ML is used to optimize the information required for IAB processing.

[0610] The AI / ML model, learning device, and inference device can be appropriately applied using the methods disclosed in Implementation 1. Hereinafter, the AI / ML model used for IAB processing will sometimes be referred to as the AI / ML model for IAB processing.

[0611] In this embodiment, the AI / ML model used for IAB processing, i.e., the AI / ML model for IAB processing, is set at the base station. The learning device and inference device can be set at the base station. The AI / ML model for IAB processing can be set at the RAN node. For example, it can be set at the CU or the DU. The AI / ML model for IAB processing can be set at the IAB host. For example, it can be set at the IAB host CU or the IAB host DU. The base station and each node can use the AI / ML model to derive the configuration information required for IAB processing, thus simplifying the configuration process.

[0612] Regarding whether information is available for AI / ML models used for IAB processing, the method disclosed in Implementation 1 can be appropriately applied.

[0613] Configure the input information for AI / ML models used in IAB processing. The following are 22 examples of input information for AI / ML models used in IAB processing.

[0614] (1) Information related to the state of the IAB node.

[0615] (2) Information related to the trajectory of the IAB node.

[0616] (3) Information related to the IAB nodes connected to the IAB node.

[0617] (4) Information related to the IAB host to which the IAB node is connected.

[0618] (5) Information related to wireless measurements of IAB nodes.

[0619] (6) Information related to the MDT of the IAB node.

[0620] (7) Information related to the BAP settings of the IAB node.

[0621] (8) Information related to the BH link.

[0622] (9) Information related to the resource settings of the IAB node.

[0623] (10) Information related to the resource usage of IAB nodes.

[0624] (11) Information related to UE traffic volume of IAB nodes.

[0625] (12) Information related to the load of the IAB node.

[0626] (13) Information related to UE-NW communication using IAB nodes.

[0627] (14) Information related to resource conflicts in IAB nodes.

[0628] (15) Information related to the communication quality of IAB nodes.

[0629] (16) Information related to the L2 (layer 2) determination of the IAB node.

[0630] (17) Information related to L2 (layer 2) determination of IAB host.

[0631] (18) Information related to L2 (Layer 2) determination of UE via IAB node.

[0632] (19) Information related to the performance of the UE via the IAB node.

[0633] (20) Information related to the QoE of the UE via the IAB node.

[0634] (21) Information related to the propagation delay of IAB nodes.

[0635] (22) Combinations of (1) to (21).

[0636] (1) For example, it could be the location, speed, and direction of movement of the IAB node. The location of the IAB node could be, for example, coordinates. For example, it could be a region. For example, it could be information identifying the TA, RAN, base station, cell, CU, DU, and TRP where the UE is located. For example, it could be information identifying the TA, RAN, base station, cell, CU, DU, and TRP with the best reception quality (or reception power). In addition, it could be time-related information. This information can be associated with that time. In addition, it could be the remaining battery power of the IAB node. In addition, it could be information related to the power consumption of the IAB node. This information can be combined. The state of the IAB node can be set as input information for training or inference.

[0637] In (2), information related to the trajectory of an IAB node can, for example, appropriately apply information related to the state of the IAB node disclosed in (1). For example, it can be information about the connected IAB nodes disclosed in (3). For example, it can be information about the connected IAB hosts disclosed in (4). It can be information related to the trajectories of past and current IAB nodes. Information related to the trajectories of past and current IAB nodes can be historical information. Trajectory-related information can be information related to the trajectories of future IAB nodes. Information related to future trajectories can be predicted information. This information can be listed. In addition, time-related information can be included. Information about the trajectory of an IAB node can be associated with time-related information. Information about the trajectory of an IAB node can be set as input information for training or inference.

[0638] (3) For example, it could be information about IAB nodes that the IAB node has previously connected to. An IAB node can be a parent IAB node or a child IAB node. For example, the information about an IAB node could be information identifying the IAB node. For example, it could be configuration information about the IAB nodes that the IAB node has previously connected to. For example, it could be historical information about the IAB nodes that the IAB node has previously connected to. For example, it could be a list of information about the IAB node. Furthermore, it could include information related to the location of the IAB node. It could include time-related information. Information about previously connected IAB nodes could be associated with information about location or time. The IAB nodes that the IAB node has previously connected to could be limited to information about IAB nodes connected through IAB processing.

[0639] (4) For example, it could be information about the IAB hosts that the IAB node is connected to. The connection is not limited to direct connections, but can also be indirect connections. For example, it could be information about the IAB hosts that the IAB node is connected to via other IAB nodes. The IAB host could be IAB host CU or IAB host DU. The information about the IAB host could be information that identifies the IAB host. For example, it could be configuration information of the IAB hosts that the IAB node has previously connected to. For example, it could be historical information about the IAB hosts that the IAB node has previously connected to. For example, it could be a list of information about the IAB host. In addition, it could include information related to the location of the IAB node. It could include time-related information. Information about previously connected IAB hosts could be associated with information about the location or time of the IAB node. The IAB hosts that the IAB node has previously connected to could be limited to information about IAB hosts connected through IAB processing.

[0640] (5) For example, wireless measurement results from an IAB node. The IAB node can be an IAB-MT. Wireless measurement results can be, for example, RSRP, RSRQ, SINR. Information related to the location of the IAB node can be included. Information related to time can be included. Wireless measurement results can be associated with information about location or time. The reception quality based on the location or state of the IAB node can be set as input information for training or inference.

[0641] (6) For example, the measurement results of MDT for an IAB node. An IAB node can be an IAB-MT. An IAB node can perform MDT-based measurements. For example, it can be an immediate MDT measurement or a logged MDT measurement. It can perform management-based MDT measurements. It can perform signaling-based MDT measurements. An IAB node can support tracking functionality. It can perform tracking-based measurements. An IAB node can associate the MDT measurement results with information about the location or time of the IAB node. Information about the location or time of the IAB node can be obtained using information measured by MDT. By setting up an MDT for the IAB node, the MDT measurement results can be obtained as input information for AI / ML models used in IAB processing. More states or more information about the IAB node can be used as input information for training or inference.

[0642] (7) For example, BAP configuration information set at IAB nodes. For example, BAP mapping settings, BH RLC channel settings, etc. BAP mapping settings include, for example, information related to BH routing, information related to traffic mapping, etc. Information about BH routing includes, for example, BAP routing information, such as BAP address, path identifier, etc. Information related to traffic mapping includes, for example, information related to the mapping between BAP routing information and BH RLC channels. Information related to the number of IAB nodes on the BH path can be set. Information can be set for each BH path. This information can be included in information related to BAP settings. This information can be included in information related to BH routing. Information about BAP settings can be set by the IAB host CU. BAP configuration information set as an IAB node can be set as input information for training or inference.

[0643] (8) For example, the number of BH links set in the IAB node. For example, it could be information identifying the BH links. For example, it could be information indicating the number of BH links in use, the number of unused BH links, which BH link is being used, which BH link is not being used, etc. Information about the BH links in the IAB node can be set as input information for training or inference.

[0644] (9) is the resource configuration information set at the IAB node. This includes cell settings, information about activating cells, information about deactivating cells, information about child IAB nodes, information about neighboring node cells, and information about serving cells. The resource configuration information can be set by the IAB host CU. The resource configuration information set at the IAB node can be used as input information for training or inference.

[0645] (10) For example, the resource usage of IAB nodes. This can be the current resource usage or the predicted resource usage. For example, it can be the resource usage of each IAB node, each cell group, each cell, each DU, and each TRP. This resource can be frequency resource or time resource. Frequency resource usage includes the number of carriers, the number of RBs, the number of subcarriers, etc. For example, it can be the usage of BWPs. It can be a carrier unit, a PRB unit, or a subcarrier unit. Time resource usage includes the number of subframes, the number of time slots, the number of symbols, etc. It can be a subframe unit, a time slot unit, or a symbol unit. The amount of base station resources used can be used as input information for training or inference.

[0646] (11) For example, the traffic volume of UEs at IAB nodes. This can be the current traffic volume or the predicted traffic volume. The traffic volume of a UE can be, for example, the number of UEs connected to the IAB node, or the number of UEs using the IAB node. For example, it can be the number of UEs connected or used per IAB node, per cell group, per cell, per DU, or per TRP. The number of UEs connected to the IAB node, etc., can be used as input information for training or inference.

[0647] (12) This is the load information of the IAB node. It can be the current load information or the predicted load information. Load information can be, for example, the number of DRBs set by the IAB node, the amount of buffer usage, etc. For example, it can be the number of child IAB nodes, the number of parent IAB nodes, etc. For example, it can be the number of BH paths including this IAB node. It can be load information other than the information on resource usage and the information on UE traffic disclosed above. The level of load of the IAB node can be set as input information for training or inference.

[0648] (13) For example, information related to UE-NW communication using an IAB node. For example, it could be the number of UE-NW communication sessions. For example, it could be UE-NW communication configuration information. For example, it could be information about PDU sessions, information about bearers, information about slices, etc. For example, it could be the number of PDU sessions, PDU session configuration information, information identifying PDU sessions, the number of bearers, bearer configuration information, information identifying bearers, the number of slices, slice configuration information, information identifying slices, etc. Communication configurations when the UE communicates with the NW via the IAB node can be set as input information for training or inference.

[0649] (14) For example, information related to conflicts between resources to child IAB nodes and resources to parent IAB nodes. This includes conflict information between resources used in communication between IAB-DU and child IAB nodes, and between resources used in communication between IAB-MT and parent IAB nodes (or IAB hosts). Information about resource conflicts can be, for example, information related to interference. Resources can be the information disclosed in (9) and (10). For example, information related to conflicts between resources that are scheduled or scheduled by IAB nodes can be used as input information for training or inference.

[0650] (15) For example, information related to the communication quality between an IAB node and its parent IAB node. For example, information related to the communication quality between an IAB node and its child IAB node. Information related to the communication quality between an IA node and its IAB host. For example, the communication quality between IAB-MT and its IAB host CU. For example, the communication quality between IAB-MT and its parent IAB node's IAB-DU. For example, the communication quality between IAB-DU and its IAB host CU. For example, the communication quality between IAB-DU and its child IAB node's IAB-MT. Communication quality can be, for example, packet loss rate, latency. For example, data communication quality. For example, the reception quality of SSBs sent by IAB-MT from IAB-DU. For example, the reception quality of PDCCH and PDSCH sent by IAB-MT from IAB-DU. Reception quality can be RSRP, RSRQ, SINR, BER. In addition, information related to the location of the IAB node can be included. Information related to time can be included. Information about the communication quality of an IAB node can be correlated with information about its location or time. Information related to the communication quality between an IAB node and other IAB nodes or the IAB host can be used as input for training or inference.

[0651] (16) For example, it could be the measurement result of the IAB node on the L2 measurement information. It could be the current L2 measurement result or the predicted L2 measurement result. The L2 measurement information could be the information measured by the UE as disclosed in the example (25) and (26) of the AI / ML model input information for RA processing in Implementation 1, or it could be the information measured by the base station. The L2 measurement result of the IAB node could be the L2 measurement result when connected to the IAB host, parent IAB node or child IAB node through IAB processing. In addition, it could include information related to the location of the IAB node. It could include time-related information. It could associate the L2 measurement information of the IAB node with information about location or time. It could set the information about the communication performance in the IAB node as input information for training or inference.

[0652] (17) For example, it could be the measurement result of the IAB host on the L2 measurement information. It could be the current L2 measurement result or the predicted L2 measurement result. The L2 measurement information could be the information measured by the base station as disclosed in the example of input information for the AI / ML model for RA processing in Implementation 1 (26). In addition, it could include information related to the location of the IAB host. It could include time-related information. It could associate the L2 measurement information about the IAB host with information about location or time. It could set the information about the communication performance in the IAB host as input information for training or inference.

[0653] (18) For example, it can be the measurement result of L2 measurement information measured by the UE via the IAB node. It can be the current L2 measurement result or the predicted L2 measurement result. The L2 measurement information can be, for example, the information measured by the UE as disclosed in the example of input information for the AI / ML model for RA processing in Implementation 1 (25). For example, it can be the packet delay amount. The L2 measurement result of the UE can be the L2 measurement result of the UE connected to the NW via the IAB node through IAB processing. In addition, it can include information related to the location of the UE. It can include information related to time. It can associate the information about the L2 measurement of the UE with the information about location or time. Information related to the communication performance when the UE communicates with the NW via the IAB node can be set as input information for training or inference.

[0654] (19) For example, it could be the UE's QoS. It could be packet loss rate or latency. For example, it could be data communication quality. For example, it could be the UE's SSB reception quality. For example, it could be the UE's PDCCH and PDSCH reception quality. Reception quality could be RSRP, RSRQ, SINR, or BER. Information related to the UE's performance could be information processed by the IAB and connected to the NW via the IAB node. It could be information related to the performance required by the UE. In addition, it could include information related to the UE's location. It could include time-related information. Information about the UE's performance could be associated with information about location or time. Information about the UE's performance could be set as input information for training or inference.

[0655] (20) For example, it could be a QoE measurement result. It could be information processed by the IAB and connected to the NW via the IAB node. It could be information related to the QoE requested by the UE. In addition, it could include information related to the UE's location. It could include time-related information. Information about QoE could be associated with information about the UE's location or time. Measurement information in the UE's application layer could be set as input information for training or inference.

[0656] (21) For example, it could be propagation delay information between an IAB node and its parent or child IAB node. It could be information about the propagation delay between an IAB node and its host DU, or between an IAB node and its UE, but not limited to propagation delay between IAB nodes. Information related to propagation delay could be the propagation delay information disclosed in the example (29) of the AI / ML model input information for RA processing. For example, it could be information measured by the IAB-MT, or by the IAB-DU. For example, it could be information measured by the UE, or by the IAB host DU. Information related to propagation delay between IAB nodes could be information derived from the IAB host CU. Higher precision distances between IAB nodes and the positions of IAB nodes can be obtained.

[0657] The IAB nodes disclosed above can be IAB nodes that are the objects of IAB processing. In IAB processing using AI / ML, information about IAB nodes can be set as input information. IAB nodes can be some or all of the IAB nodes within a BH path. Information about the IAB nodes constituting the BH path can be set as input information for training or inference. IAB nodes can be IAB nodes adjacent to the IAB node that is the object of IAB processing. In IAB processing, not only the object IAB node, but also information related to adjacent IAB nodes can be set as input information for training or inference. IAB nodes can be some or all of the IAB nodes within a PLMN. Information related to IAB nodes located within a PLMN that supports IAB can be set as input information for training or inference.

[0658] The aforementioned input information can be current information or predicted information. By setting the predicted information as input to the AI / ML model, more effective output information can be derived in the future.

[0659] The input information disclosed above can be associated with information about the state of IAB nodes. The disclosed information can be appropriately applied to the information about the state of IAB nodes. By associating it with the state of IAB nodes, it is possible to take into account which IAB nodes' states the input information belongs to.

[0660] The input information disclosed above can be historical information. Historical information about IAB nodes can be set. Historical information about IAB-MT, IAB-DU, IAB host, IAB host DU, and IAB host CU can be set. This historical information contains associated input information. This historical information can be included in the UE's historical information. It can include information on whether it is an IAB node and information identifying the IAB node. It can identify IAB nodes and UEs. Therefore, it is possible to aggregate and manage information related to IAB nodes, simplifying processing. Furthermore, the method of including it in the UE's historical information can use processing related to the history of previous UEs, thus avoiding processing complexity.

[0661] IAB processing allows input information for AI / ML models to be sent before model training and inference.

[0662] IAB nodes send input information for AI / ML models used for IAB processing to other IAB nodes. IAB nodes can send this input information to IAB nodes that have AI / ML models for IAB processing. IAB nodes with AI / ML models for IAB processing can send requests for input information to other IAB nodes. This request can contain information related to the IAB node being processed. This information can be the IAB node's identifier. The IAB node receiving the request sends the input information for its AI / ML model to the requesting IAB node. The transmission of input information for AI / ML models between IAB nodes, and the request for this information, can use RRC signaling. For example, RRC messages can be configured for this transmission. For example, current information and predicted information can be sent using different messages.

[0663] An IAB host can send input information for an AI / ML model used for IAB processing to other IAB hosts. An IAB host can be an IAB host CU. Input information for an AI / ML model used for IAB processing can be sent to an IAB host that has an AI / ML model for IAB processing. An IAB host that has an AI / ML model used for IAB processing can send a request for input information for an AI / ML model used for IAB processing to other IAB hosts. This request can contain information related to the IAB host that is the object of IAB processing. The information related to the IAB host can be the IAB host's identifier. The IAB host receiving the request sends the input information for an AI / ML model used for IAB processing to the IAB host that received the request. The transmission of input information for an AI / ML model used for IAB processing between IAB hosts, and the request for this information, can use inter-base station signaling. For example, Xn signaling can be used. For example, an Xn message can be set for this transmission. For example, current information and predicted information can be sent through different messages.

[0664] An IAB-DU (or IAB host DU) can send input information for an AI / ML model used for IAB processing to an IAB host CU. An IAB host CU with an AI / ML model for IAB processing can send a request for input information for an AI / ML model for IAB processing to an IAB-DU. This request can contain information related to the IAB-DU, which is the object of IAB processing. This information can be the identifier of the IAB node or the identifier of the IAB-DU. Upon receiving the request, the IAB-DU sends the input information for an AI / ML model for IAB processing to the IAB host CU that received the request. The transmission of the input information for an AI / ML model for IAB processing between the IAB-DU and the IAB host CU, and the request for this information, can use DU-CU signaling. For example, F1 signaling can be used. For example, an F1 message can be set for this transmission. For example, current information and predicted information can be sent using different messages.

[0665] The IAB-MT can send input information for the AI / ML model used for IAB processing to the IAB host CU. It can send this input information to the IAB host CU, which has an AI / ML model for IAB processing. The IAB host CU, which has an AI / ML model for IAB processing, can send a request for this input information to the IAB-MT. This request can contain information related to the IAB-MT, which is the object of IAB processing. This information can be the identifier of the IAB node or the identifier of the IAB-MT itself. Upon receiving the request, the IAB-MT sends the input information for the AI / ML model used for IAB processing to the IAB host CU that received the request. The transmission of this request for the input information for the AI / ML model used for IAB processing between the IAB-MT and the IAB host CU can use RRC signaling. For example, an RRC message can be configured for this transmission.

[0666] This document discloses the output information of AI / ML models used for IAB processing. Information used in IAB processing can be set as output information. The following are 13 examples of output information from AI / ML models used for IAB processing.

[0667] (1) Information related to the trajectory of the IAB node.

[0668] (2) Information related to the BAP settings of the IAB node.

[0669] (3) Information related to the BH link.

[0670] (4) Information related to the resource settings of the IAB node.

[0671] (5) The priority order is set.

[0672] (6) Information related to the connected IAB node candidates.

[0673] (7) The probability of reaching the candidate IAB node connected to.

[0674] (8) Information related to the resource usage of IAB nodes.

[0675] (9) Information related to UE traffic volume of IAB nodes.

[0676] (10) Information related to the load of the IAB node.

[0677] (11) Information related to resource conflicts in IAB nodes.

[0678] (12) Information related to the propagation delay of IAB nodes.

[0679] (13) Combinations of (1) to (12).

[0680] (1) For example, it could be the trajectory information of predicted future IAB nodes. Information related to the trajectory of IAB nodes can be set as the information disclosed in example (2) of the input information for AI / ML models used for IAB processing. It can include information related to past and current trajectories. It can include historical information of past and current IAB node trajectories together with the trajectory information of predicted future IAB nodes. This information can be listed. In addition, it can include time-related information. It can associate information about the trajectory of IAB nodes with information about time.

[0681] (2) Information related to the BAP setting predicted by the AI / ML model. The information related to the BAP setting can be the information disclosed in Example (7) of the input information for the AI / ML model used in IAB processing. Furthermore, it can include information related to the predicted location of the IAB node. It can also include information related to the predicted time. Information about the BAP setting can be associated with information about location or time. The information related to the BAP setting is not limited to one item; it can be multiple items. It can include information on the prediction accuracy of the information related to the BAP setting predicted by the AI / ML model. IAB nodes can use the BAP setting derived from the AI / ML model. More efficient IAB processing can be performed.

[0682] (3) Information related to the BH link predicted by the AI / ML model. The information related to the BH link can be the information disclosed in example (8) of the input information for the AI / ML model used in IAB processing. Furthermore, it can include information related to the predicted location of the IAB node. It can also include information related to the predicted time. Information about the BH link can be associated with information about location or time. The information related to the BH link is not limited to one item; it can be multiple items. It can include information about the prediction accuracy of the information related to the BH link predicted by the AI / ML model. IAB nodes can use the BH link derived from the AI / ML model. More efficient IAB processing can be performed.

[0683] (4) Information related to the resource settings predicted by the AI / ML model. This information can be the information disclosed in Example (9) of the input information for the AI / ML model used in IAB processing. Additionally, it can include information related to the predicted location of the IAB node. It can also include information related to the predicted time. Information about the resource settings can be correlated with information about location or time. The information related to the resource settings is not limited to one item; it can include multiple items. It can include information about the prediction accuracy of the information related to the resource settings predicted by the AI / ML model. IAB nodes can use the resource settings derived from the AI / ML model. This enables more efficient IAB processing.

[0684] (5) is the priority order of information related to the settings as shown in (2) to (4). Priority can be set for some or all of this information. Predicted settings and their priorities can be linked. Priority can be assigned to multiple output settings separately. For example, if BH causes a communication interruption in the initial settings, the next priority setting can be used for IAB processing. More flexible and efficient IAB processing is possible.

[0685] (6) is the predicted information related to the connected IAB node candidates. Connected IAB node candidates can be set. There can be multiple IAB node candidates, not just one. Information on the prediction accuracy of the connected IAB node candidates can be included. For example, it can identify which IAB node is more effectively connected to.

[0686] (7) is the predicted arrival probability information to the candidate IAB nodes to which the IAB node is connected. It may include information on the confidence interval of the arrival probability of the candidate IAB node predicted by the AI / ML model. For example, it can identify which IAB node is more effective to connect to.

[0687] (8) is the information on the predicted resource usage of the connected IAB node candidates. The resource usage information of the IAB node can be set as the information disclosed in the input information example (10) of the AI / ML model for IAB processing. The resource usage of the IAB node can be the predicted resource usage among the IAB node candidates. For example, the surrounding IAB nodes can identify the predicted resource usage of the IAB node that has undergone IAB processing.

[0688] (9) is the predicted traffic volume related to the UE. The information related to the UE's traffic volume can be the information disclosed in example (11) of the input information of the AI / ML model for IAB processing. The information about the predicted UE's traffic volume can be the UE's traffic volume predicted in the IAB node candidates. For example, the surrounding IAB nodes can identify the UE's traffic volume predicted by the IAB node that has undergone IAB processing.

[0689] (10) For example, it could be information related to the predicted load of the connected IAB nodes. The information related to the load of the IAB nodes could be the information disclosed in the example (12) of the input information for the AI / ML model for IAB processing. For example, neighboring IAB nodes can identify the load status of the IAB nodes when IAB processing has been performed.

[0690] (11) is the predicted information related to resource conflicts of the connected IAB nodes. The information on resource conflicts of IAB nodes can be set as the information disclosed in the input information example (14) of the AI / ML model for IAB processing. The information related to resource conflicts of IAB nodes can be the information related to resource conflicts predicted from the IAB node candidates. For example, neighboring IAB nodes can identify the information related to resource conflicts predicted by the IAB nodes that have undergone IAB processing.

[0691] (12) For example, it could be information related to the predicted propagation delay between future IAB nodes. Information related to the propagation delay between IAB nodes can be the information disclosed in example (21) of the input information for the AI / ML model for IAB processing. It can include information related to the propagation delay between past and current IAB nodes. It can include predicted information related to the propagation delay between future IAB nodes, and it can also include historical information related to the propagation delay between past and current IAB nodes. This information can be listed. In addition, it can include time-related information. Information about the propagation delay between IAB nodes can be associated with time-related information. For example, neighboring IAB nodes can identify information related to the predicted propagation delay of IAB nodes that have undergone IAB processing.

[0692] The output information of the AI / ML model for IAB processing disclosed above is appropriately communicated to the IAB node or IAB host DU performing the IAB processing. This allows for more efficient IAB processing using the AI / ML model for IAB processing.

[0693] The method for sending output information from AI / ML models in IAB processing can be appropriately applied to the method for sending configuration information in IAB processing.

[0694] The IAB host CU can send output information of the AI / ML model used for IAB processing to IAB nodes (or IAB host DUs). An IAB host CU with an AI / ML model for IAB processing can send this output information to IAB host DUs and IAB nodes related to IAB processing. IAB host DUs and IAB nodes related to IAB processing can send a request for the output information of the AI / ML model to the IAB host CU with the AI / ML model. Upon receiving the request, the IAB host CU sends the AI / ML model output information to the requesting IAB host DU or IAB node. This request and output information may contain information related to the IAB host DU or IAB node that is the object of IAB processing. This information may be an identifier. The transmission of the AI / ML model output information and the request for this information between the IAB host CU and the IAB host DU or IAB node's IAB-DU can use DU-CU inter-signaling. For example, F1 signaling can be used. For example, an F1 message can be configured for this transmission. The IAB host CU and the IAB-MT in the IAB node process the output information of the AI / ML model, and the request to send this information can use RRC signaling. For example, an RRC message can be configured for this transmission.

[0695] Feedback information on IAB processing using the AI / ML model for IAB processing is disclosed (hereinafter referred to as "IAB processing feedback information using the AI / ML model for IAB processing"). This includes information related to the performance of UE-NW inter-communication, the performance of IAB nodes, and the performance of the IAB host. The performance of the IAB node and the performance of the IAB host can refer to the performance of the IAB node and the IAB host used for UE-NW inter-communication. Fifteen examples of IAB processing feedback information using the AI / ML model for IAB processing are disclosed below.

[0696] (1) Information related to L2 (Layer 2) determination of UE via IAB node.

[0697] (2) Information related to the performance of the UE via the IAB node.

[0698] (3) Information related to the QoE of the UE via the IAB node.

[0699] (4) Information related to the resource usage of IAB nodes.

[0700] (5) Information related to the UE traffic volume of the IAB node.

[0701] (6) Information related to the load of the IAB node.

[0702] (7) Information related to UE-NW communication using IAB nodes.

[0703] (8) Information related to resource conflicts in IAB nodes.

[0704] (9) Information related to the communication quality of IAB nodes.

[0705] (10) Information related to the L2 (layer 2) determination of the IAB node.

[0706] (11) Information related to L2 (layer 2) determination of IAB host.

[0707] (12) Information related to the IAB nodes to which the IAB nodes are connected.

[0708] (13) Information related to the IAB host to which the IAB node is connected.

[0709] (14) Information related to propagation delay between IAB nodes.

[0710] (15) Combinations of (1) to (14).

[0711] (1) to (3) could be (18) to (20) of the above IAB processing AI / ML model using input information example.

[0712] (4) to (11) could be (10) to (17) of the above IAB processing AI / ML model using input information.

[0713] (12), (13) For example, it could be (3), (4) of the above IAB processing AI / ML model using input information.

[0714] (14) For example, it could be (21) of the above IAB processing AI / ML model using input information.

[0715] The IAB processing node and the UE appropriately notify the node with the IAB processing AI / ML model of the IAB processing feedback information disclosed above. This allows for evaluation of the reward calculation of the IAB processing AI / ML model and enables model updates. Consequently, more efficient IAB processing can be performed.

[0716] The method for sending IAB processing feedback information using the AI / ML model for IAB processing between base stations, between CU-DU, and between UE and base stations can appropriately apply the input information sending method disclosed above.

[0717] Figure 18 This is a diagram illustrating an example sequence of IAB processing using AI / ML for IAB processing in Implementation 4. An example is disclosed where the IAB host CU has an AI / ML model for IAB processing and performs training and inference.

[0718] In step ST1801, the IAB host CU has an AI / ML model for IAB processing. In steps ST1811 to ST1813, the IAB host CU with the AI / ML model for IAB processing sends input information requests for model training to the IAB host DU, IAB node #2, and IAB node #1. The IAB host DU, IAB node #1, and IAB node #2 can be limited to the coverage area of ​​the IAB host CU.

[0719] This request can be processed based on IAB processing. For example, it can be processed based on IAB node integration processing from IAB nodes. For instance, it can be processed based on BAP configuration update processing, IAB node resource configuration update processing, and BH route update processing for IAB nodes. Model training can be performed based on the status of newly added IAB nodes and the input information under updated configuration statuses.

[0720] The request may contain information indicating the desired information. It may include information indicating the purpose of the input information for the AI / ML model used in IAB processing. For example, the request may contain information indicating it is for IAB processing. For example, the request may contain information indicating it is for model training. This is effective when the input information for the AI / ML model used in IAB processing includes information for model training. For example, if the information held by the IAB node includes information for model training, inference, and feedback, the IAB node can quickly grasp the information to be notified to the IAB host CU, resulting in rapid notification from the IAB node to the IAB host CU.

[0721] The transmission of this request between the IAB host CU and the IAB host DU or IAB node can, for example, use F1 signaling. The transmission of this request between the IAB host CU and the IAB node can, for example, use RRC signaling. For example, the measurement settings sent from the IAB host CU to the IAB node can include a request for input information. This can reduce the signaling load.

[0722] In steps ST1814 to ST1816, IAB node #1, IAB node #2, and IAB host DU send model training input information to IAB host CU. IAB nodes and IAB host DU can send model training input information obtained from connected UEs to the IAB host CU. The IAB host CU can input information from UEs connected to IAB nodes and IAB host DU into the AI / ML model.

[0723] Input requests can be sent between the IAB host DU or IAB node and the IAB host CU, for example, using F1 signaling. Input requests can also be sent between the IAB node and the IAB host CU, for example, using RRC signaling.

[0724] Measurement results from IAB nodes can be sent to the IAB host CU via measurement result reports. These results can be wireless or L2 measurements. They can be sent using the same signaling or different signaling methods. The measurement results from IAB nodes can be used as input for model training.

[0725] In step ST1817, the IAB host CU uses the input information of the IAB processing AI / ML model stored in its own node, and the input information of the IAB processing AI / ML model obtained from the IAB nodes and the IAB host DU, to train the IAB processing AI / ML model. The IAB processing AI / ML model is updated through training. These processes can be performed appropriately. They can be performed periodically or each time IAB processing is performed. By performing them periodically, for example, the time-varying load status of the IAB nodes can be appropriately input as input information. By performing them each time IAB processing is performed, for example, the status of newly added IAB nodes can be appropriately input, and the input information under the set status can be updated. More appropriate training of the IAB processing AI / ML model can be implemented.

[0726] In steps ST1821 to ST1823, the IAB host CU sends inference input information requests to the IAB host DU, IAB node #2, and IAB node #1. The IAB host DU, IAB node #1, and IAB node #2 can be limited to the coverage area of ​​the IAB host CU. This request can be processed according to IAB processing. Inference input information can be obtained through the inference input information request and input into the AI / ML model for IAB processing, and output information for IAB processing can be exported.

[0727] The request may contain information indicating the desired information. It may also contain information indicating the purpose of the input information for the AI / ML model used in IAB processing. For example, the request may contain information indicating that it is for inference. This is effective, for instance, when the input information for the AI / ML model used in IAB processing includes information for inference.

[0728] In steps ST1824 to ST1826, IAB node #1, IAB node #2, and IAB host DU send inference input information to IAB host CU. IAB nodes and IAB host DU can send inference input information obtained from connected UEs to the IAB host CU. The IAB host CU can input information from UEs connected to IAB nodes and IAB host DU into the AI / ML model.

[0729] The methods for requesting and sending input information for inference can appropriately apply the methods for requesting and sending input information for model training disclosed above.

[0730] In step ST1827, the IAB host CU uses the input information of the IAB processing AI / ML model stored in this node, and the input information of the IAB processing AI / ML model obtained from the IAB node and the IAB host DU, to perform inference using the IAB processing AI / ML model. The IAB host CU then derives the output information of the IAB processing AI / ML model through this inference.

[0731] In step ST1831, IAB processing is performed between the IAB host CU, IAB host DU, IAB node #2, and IAB node #1. For example, IAB node integration processing is performed on IAB node #1. During IAB processing, the IAB host CU sends AI / ML model output information for IAB processing to the IAB host DU, IAB node #2, and IAB node #1. For example, BAP setting information and IAB node resource setting information are sent to IAB node #1. To update IAB node #2 and IAB host DU, output information can be sent to IAB node #2 and IAB host DU. Thus, the IAB nodes and IAB host DU used in IAB processing can obtain the AI / ML model output information for IAB processing. This output information can be used to perform IAB processing before IAB host CU, IAB host DU, IAB node #2, and IAB node #1.

[0732] In the IAB host CU, the transmission of AI / ML model output information for IAB processing of IAB host DU, IAB node #2, and IAB node #1 can be performed separately from IAB processing. For example, it can be performed before IAB processing. Alternatively, it can be performed after the AI / ML model output information for IAB processing is exported and transmitted. Finally, IAB processing can be performed after the AI / ML model output information for IAB processing is transmitted.

[0733] The method for sending output information can appropriately apply the method for sending setting information in IAB processing, and the method for sending input information requests for AI / ML models in IAB processing disclosed above.

[0734] In steps ST1851 to ST1853, IAB node #1, IAB node #2, and IAB host DU send IAB processing feedback information to IAB host CU, indicating that the AI / ML model used for IAB processing has been implemented. The feedback information may include information identifying the IAB node and the IAB host DU. Upon receiving this information, the IAB host CU can identify the node that sent the feedback information.

[0735] Therefore, the IAB host CU can obtain information related to the performance of IAB processing using the AI / ML model for IAB processing. The IAB host CU can use the IAB processing feedback information using the AI / ML model for IAB processing stored in this node, as well as the IAB processing feedback information using the AI / ML model for IAB processing obtained from IAB node #1, IAB node #2, and the IAB host DU, to evaluate the effect of IAB processing when using the AI / ML model for IAB processing.

[0736] The method for sending feedback information can appropriately apply the method for sending input information for model training disclosed above.

[0737] The IAB node configured in step ST1831 can be used for communication between the UE and the NW. In step ST1861, the UE performs RRC connection processing via IAB node #1, IAB node #2, IAB host DU, and IAB host CU. This RRC connection is used for communication between the UE and the NW.

[0738] Information related to the performance of communication between the UE and NW can be set as feedback information. In steps ST1871 to ST1874, the UE, IAB node #1, IAB node #2, and IAB host DU send IAB processing feedback information using the AI / ML model for IAB processing to the IAB host CU. The feedback information may include information identifying the UE, the current IAB node, and the IAB host DU. The IAB host CU that receives this information can identify the node that sent the feedback information.

[0739] Therefore, the IAB host CU can obtain information related to the communication performance between the UE and NW via the IAB node and the IAB host, where the IAB node and IAB host use output information derived from the AI / ML model for IAB processing. The IAB host CU can use the IAB processing feedback information using the AI / ML model held by this node, as well as the IAB processing feedback information using the AI / ML model obtained from the UE, IAB node #1, IAB node #2, and IAB host DU, to evaluate the effectiveness of IAB processing when using the AI / ML model for IAB processing.

[0740] The method for sending feedback information can appropriately apply the method for sending input information for the AI / ML model used in IAB processing disclosed above. For example, RRC signaling can be used to send feedback information from the UE to the IAB host CU.

[0741] The transmission of feedback information on the performance of IAB processing alone, and the transmission of feedback information on the communication performance between the UE and NW via the IAB node, are disclosed, but are not limited to these; the transmission of these feedback information can also be combined. For example, the transmission of this feedback information can be performed in steps ST1871 to ST1874 after the RRC processing connection in step ST1861. Alternatively, a portion of the feedback information on the performance of IAB processing can be transmitted in steps ST1851 to ST1853, and other feedback information can be transmitted in steps ST1871 to ST1874 in combination with feedback information on the communication performance between the UE and NW. Multiple types of feedback information can be flexibly transmitted according to the communication processing. The IAB host CU uses this feedback information to evaluate the effectiveness of IAB processing, thereby enabling more effective IAB processing.

[0742] Input and feedback requests for AI / ML models using IAB processing can be eliminated. Each node can appropriately send input and feedback information for the AI / ML models using IAB processing to the nodes that have them. For example, this can be done periodically. Alternatively, input and feedback information can be sent upon receiving new input or feedback information. This aims to reduce signaling load. When a node with an AI / ML model using IAB processing receives this information from other nodes or itself, it can send input requests to nodes other than that node. This reduces unnecessary signaling, such as sending the same input information.

[0743] Therefore, importing AI / ML into IAB processing enables more efficient IAB processing.

[0744] Other IAB host CUs can have AI / ML models for IAB processing. For example, an IAB host CU that supports IAB can have an AI / ML model for IAB processing. Partial or complete transmission and reception of input information, requests for such input information, output information, and feedback information can be performed between the CUs and other IAB host CUs. This transmission and reception can use Xn signaling.

[0745] Therefore, for example, even when IAB processing is implemented due to changes in the IAB host CU connected via the movement of an IAB node, input or feedback information from other IAB host CUs, IAB nodes within the coverage area of ​​that IAB host CU, IAB host DUs, etc., can be input into the AI / ML model for IAB processing. By acquiring more node information and inputting it into the AI / ML model for IAB processing, output information suitable for various states of the IAB node and the UE communicating using that IAB node can be derived. By performing IAB processing using this output information, more efficient IAB processing can be implemented.

[0746] By employing the method disclosed in this embodiment, training and inference in IAB processing can be performed using AI / ML models. By deriving the output information of the AI / ML model using various input information, IAB processing can be optimized. Furthermore, information beyond the IAB-MT measurement results can be considered in the BAP settings of IAB nodes and the settings of BH paths. For example, optimization of IAB processing suitable for the state of the IAB nodes becomes possible, including which IAB nodes are included in the BH path, how to configure those IAB nodes, and resource settings. More efficient IAB processing is possible, thus improving the communication performance between UEs and NWs using IAB.

[0747] Figure 18 This document discloses an example of setting up an AI / ML model for IAB processing on the IAB host CU. Alternatively, an AI / ML model for IAB processing can be set up on an IAB node. Learning and inference mechanisms can be configured on the IAB node. The IAB node can then use the AI / ML model to derive the configuration information required for IAB processing.

[0748] An IAB node equipped with an AI / ML model for IAB processing can send requests for input and feedback information for the AI / ML model to other IAB nodes, the IAB host DU, and the IAB host CU. Other IAB nodes, the IAB host DU, and the IAB host CU can also send input and feedback information for the AI / ML model to an IAB node equipped with an AI / ML model for IAB processing.

[0749] IAB nodes that use AI / ML models for inference and derive the configuration information required for IAB processing can send output information containing this configuration information to the IAB host CU. For example, this can be sent during IAB processing or before IAB processing. Thus, the IAB host CU can use the output information to configure the IAB nodes and the IAB host DU for IAB processing.

[0750] Alternatively, IAB nodes can send the AI / ML model output information from IAB processing to other IAB nodes, the IAB host DU, and the IAB host CU. For example, this can be sent during IAB processing or before IAB processing. This allows each node to use the output information to implement IAB processing. For example, it can reduce the amount of configuration information for IAB processing sent from the IAB host CU to each node.

[0751] An IAB node equipped with an AI / ML model for IAB processing can perform inference using that model once the IAB processing method has been determined. The IAB node can send a request for inference input information to each node after deciding on the IAB processing method. Each node can then send its own inference input information to the IAB node based on this request. The IAB node can use the inference input information to derive the AI / ML model output information. Thus, for example, when an IAB node determines IAB processing for IAB integration, the AI / ML model can be used to derive the configuration information for the IAB processing. This allows for the use of input information that is closer to the timing of the IAB processing, enabling more efficient IAB processing.

[0752] Regarding the configuration of nodes for AI / ML models used in IAB processing, the methods for configuring the IAB host CU and the methods for configuring IAB nodes can be combined to achieve more efficient IAB processing.

[0753] Implementation method 4, variation 1.

[0754] In the IAB, RLF (Rapid Link Default) sometimes occurs in the BH (Browser Height). When an IAB node experiences a BH RLF, it uses a new parent IAB node to perform RRC (Re-establishment) processing. The IAB node uses the new parent IAB node for BH path setup and BAP (Browser Access Point) setup between itself and the IAB host CU. However, if the IAB node moves, or due to changes in the surrounding environment over time, the radio wave propagation environment between the IAB node and its parent IAB node, or in the BH path using the new parent IAB node, is constantly changing. Therefore, RRC reconstruction sometimes fails. Furthermore, even if RRC reconstruction succeeds, if the communication condition of the BH link immediately deteriorates, the following problem can occur: a BH RLF will recur, making communication between the UE and NW (Network Wire) impossible.

[0755] In this variation, a method for solving the above problem is disclosed.

[0756] AI / ML is imported into BH RLF processing. For example, BH RLF processing includes parent IAB node selection, BAP configuration processing, and IAB node resource configuration processing when BH becomes RLF. For instance, AI / ML is used in BH RLF processing to optimize information such as the selection of parent IAB nodes.

[0757] The AI / ML models, learning devices, and inference devices can appropriately apply the methods disclosed in Implementation 1.

[0758] In this variation, the AI / ML model used for BH RLF processing, i.e., the AI / ML model for BH RLF processing, is set at the base station. The learning device and inference device can be set at the base station. The AI / ML model for BH RLF processing can be set at the RAN node. For example, it can be set at the CU or the DU. The AI / ML model for BH RLF processing can be set at the IAB host. For example, it can be set at the IAB host CU or the IAB host DU. The base station and each node can use the AI / ML model to derive the configuration information required for BH RLF processing, thus simplifying the setup process.

[0759] Regarding whether an AI / ML model for BH RLF processing is available, the method disclosed in Implementation 1 can be appropriately applied.

[0760] Configure the input information used by the AI / ML model for BH RLF processing. For example, it could be the input information for the AI / ML model for IAB processing, or it could be RLF association information. Seven examples of input information for the AI / ML model for BH RLF processing are provided below.

[0761] (1) IAB processing uses AI / ML models to input information.

[0762] (2) Information related to RLF detection.

[0763] (3) Information related to RLF detection indication.

[0764] (4) Information related to the BH link in the IAB node.

[0765] (5) Information related to RLF recovery.

[0766] (6) Information related to RLF recovery instructions.

[0767] (7) Combinations of (1) to (6).

[0768] (1) For example, it could be the input information of the AI / ML model for IAB processing disclosed in Implementation 4.

[0769] (2) For example, it could be information indicating which BH link the RLF occurred on. For example, it could include information identifying the BH link. For example, it could be information related to the IAB node that generated the BH RLF. For example, it could include information identifying the IAB node. This IAB node could be a parent IAB node, a child IAB node, an IAB host DU, etc. For example, it could be information indicating which parent IAB node generated the BH RLF. For example, it could be information indicating in which path the RLF occurred on. For example, it could include information identifying the path. For example, it could be a BAP path identifier. For example, it could be a BAP route identifier. Information on where the RLF was detected can be set as input information for training or inference. In addition, it can include time-related information. Information about RLF detection can be associated with information about time.

[0770] (3) For example, it could be the transmission information of the BH RLF detection indication. It could be information indicating which sub-IAB node the RLF detection indication was sent to. For example, it could be the reception information of the RLF detection indication. It could be information indicating which parent IAB node the RLF detection indication was received from. It could include information identifying the IAB node. Information on sending and receiving RLF detection indications could be set as input information for training or inference. In addition, it could include time-related information. Information about the RLF detection indication could be associated with information about time.

[0771] (4) For example, it could be information related to available BH links in the IAB node. For example, it could be information related to BH links other than those that generated BH RLFs. For example, it could be information related to dis-available BH links in the IAB node. For example, it could be information about BH links that received BH RLF detections. The status of BH links in the IAB node can be set as input information for training or inference. In addition, it can include time-related information. Information about BH links in the IAB node can be associated with time-related information.

[0772] (5) For example, it could be information related to BH RLF recovery. For example, it could be information about successful BH RLF recovery. For example, it could be information about the IAB node with which BH RLF recovery was successful. It could include information identifying the IAB node. For example, it could be information indicating which parent IAB node the BH RLF recovery was successful with, or information indicating which child IAB node the BH RLF recovery was successful with. The status of BH RLF recovery can be set as input information for training or inference. In addition, it can include time-related information. Information about BH RLF recovery can be associated with information about time.

[0773] (6) For example, it could be the transmission information of the BH RLF recovery indication. It could be information indicating which sub-IAB node the BH RLF recovery indication was sent to. For example, it could be the reception information of the BH RLF recovery indication. It could be information indicating from which parent IAB node the BH RLF recovery indication was received. It could include information identifying the IAB node. Information on sending and receiving BH RLF recovery indications could be used as input information for training or inference. In addition, it could include time-related information. Information about the BH RLF recovery indication could be associated with time-related information.

[0774] The aforementioned input information can be current information or predicted information. By setting the predicted information as input to the AI / ML model, more effective output information can be derived in the future.

[0775] The input information disclosed above can be associated with information about the state of IAB nodes. The disclosed information can be appropriately applied to the information about the state of IAB nodes. By associating it with the state of IAB nodes, it is possible to take into account which IAB nodes' states the input information belongs to.

[0776] The input information disclosed above can be historical information. BH RLF historical information can be set. It can be included in IAB node historical information, IAB-MT historical information, IAB-DU historical information, IAB host historical information, IAB host DU historical information, IAB host CU historical information, UE historical information, etc. Therefore, information about BH RLF can be associated with and managed with IAB nodes, simplifying processing. Furthermore, the method of including it in the UE's historical information allows for processing related to the history of previous UEs, thus avoiding processing complexity.

[0777] BH RLF processing allows input information for AI / ML models to be sent before model training and inference.

[0778] The input information sending method disclosed in Implementation 4 can be appropriately applied to the input information sending method for BH RLF processing AI / ML model.

[0779] The output information of the AI / ML model used for BH RLF processing is disclosed. Information used for BH RLF processing can be set as output information. Below are eight examples of output information from the AI / ML model used for BH RLF processing.

[0780] (1) IAB processes the output information of AI / ML models.

[0781] (2) Information related to community selection and community reselection.

[0782] (3) Information related to the cells that can be selected and reselected.

[0783] (4) Information related to the frequency of cell selection and cell reselection.

[0784] (5) Information related to the BH path.

[0785] (6) Information related to the BH link.

[0786] (7) Information related to the setting of BH RLF.

[0787] (8) Combinations of (1) to (7).

[0788] (1) For example, it could be the output information of the AI / ML model for IAB processing disclosed in Implementation 4.

[0789] (2) For example, it could be setting information used in cell selection processing. For example, it could be setting information used in cell reselection processing. For example, it could be a threshold for received power used in cell selection processing. For example, it could be a threshold for received power used in cell reselection processing. For example, an offset could be set for the setting information used in cell selection processing. For example, an offset could be set for the setting information used in cell reselection processing. An offset used by the IAB node could be set. For example, an offset associated with the state of the IAB node could be set. Cell selection processing and cell reselection processing suitable for the IAB node could be implemented.

[0790] This offset can be broadcast via SIB or notified via RRC signaling. It can be included in cell selection and cell reselection configuration information. It can also include information indicating that the offset is used by the IAB node. In cell selection and cell reselection processing, it can differentiate the actions of a typical UE and an IAB node. It can implement cell selection and cell reselection processing suitable for IAB nodes.

[0791] (3) For example, it can be a cell of an IAB node capable of cell selection and cell reselection. For example, it can be a candidate cell of an IAB node capable of cell selection and cell reselection. The cell is not limited to one, but can be multiple. The IAB node is not limited to one, but can be multiple. It can contain information identifying the IAB node. It can contain information identifying the cell. It can set a priority order for the IAB nodes. For example, it can be information indicating the priority order of IAB nodes for cell selection and cell reselection, or information indicating the priority order of cells. For example, it can be information that prioritizes the last connected IAB node, etc. In the cell selection or cell reselection process of BHRLF, IAB nodes, cells, or their candidates can be specified, and therefore, more efficient processing can be implemented.

[0792] (4) For example, it can be a carrier frequency capable of cell selection and cell reselection. For example, it can be a frequency band capable of cell selection and cell reselection. It can be a candidate carrier frequency or frequency band capable of cell selection and cell reselection. The carrier frequency or frequency band is not limited to one and can be multiple. It can contain information identifying the carrier frequency or frequency band. For example, it can be information indicating the priority order of the carrier frequency or frequency band for cell selection and cell reselection. For example, it can be information that prioritizes the last connected carrier frequency, etc. In the cell selection or cell reselection process of BH RLF, the carrier frequency, frequency band or their candidates can be specified, and therefore, more efficient processing can be implemented.

[0793] (5) For example, it could be a BH path set in the IAB node during BH RLF recovery. For example, it could be a candidate BH path. The BH path could be, for example, BAP path information. It could contain information identifying the BAP path. For example, it could be BAP routing information. It could contain information identifying the BAP route. The BH path is not limited to one, but can also be multiple. For example, it could be information indicating the priority order of the BH paths. For example, it could be information that prioritizes the last connected BH path, etc. Since BH paths or their candidates can be set during BH RLF recovery, more efficient processing can be implemented.

[0794] (6) For example, it could be a BH link selected by the IAB node during BH RLF recovery. For example, it could be a candidate BH link. It could contain information identifying the BH link. The BH link is not limited to one, but can be multiple. For example, it could be information indicating the priority order of the BH links. For example, it could be information that prioritizes the last connected BH link, etc. In BH RLF recovery, BH links or their candidates can be set, thus enabling more efficient processing.

[0795] (7) For example, it could be a timer used in BH RLF processing. For example, it could be a timer from the detection of out-of-sync until the inability to synchronize and the initiation of RRC reconstruction. For example, it could be a timer from the initiation of RRC reconstruction until the inability to perform cell selection and the transition to RRC_Idle state. For example, it could be a counter for physical problem detection. Alternatively, it could be an out-of-sync counter or an in-sync counter.

[0796] Information relating to the relationship between the IAB node's state and the output information can be set. This information can be a list. The information relating to the IAB node's state can appropriately apply the information disclosed in the above-described input information. For example, it could be a list representing the relationship between the IAB node's location and information regarding cell selection. The information relating to the IAB node's state is not limited to one piece; it can be a combination of multiple pieces of information disclosed above. Thus, the relationship between the IAB node's state and the output information can be shown.

[0797] The method for sending output information using the AI / ML model in BH RLF processing can appropriately apply the method for sending output information disclosed in Implementation 4.

[0798] The output information of the AI / ML model for BH RLF processing disclosed above is appropriately notified to the IAB node or IAB host DU performing BH RLF processing. Therefore, the AI / ML model for BH RLF processing can be used to perform more efficient BH RLF processing.

[0799] Feedback information for BH RLF processing using the AI / ML model for BH RLF processing is disclosed. This includes information related to the performance of communication between the UE and NW, information related to the performance of the IAB node, and information related to the performance of the IAB host. The performance of the IAB node and the IAB host can refer to the performance of the IAB node and IAB host used for communication between the UE and NW. This information can be set to be related to the performance after BH RLF. The input information of the AI / ML model for BH RLF processing disclosed above can be appropriately applied. The feedback information of the IAB processing using the AI / ML model for IAB processing disclosed in Embodiment 4 can be appropriately applied.

[0800] The BH RLF processing node and the UE appropriately notify the node with the BH RLF processing AI / ML model of the BH RLF processing feedback information disclosed above. This allows for evaluation of the reward calculation of the BH RLF processing AI / ML model and enables model updates. Consequently, more efficient BH RLF processing can be performed.

[0801] The method for sending feedback information in BH RLF processing using the AI / ML model between base stations, between CU-DU, and between UE and base stations, which uses BH RLF processing, can appropriately apply the input information sending method disclosed above.

[0802] Figure 19 and Figure 20 This is a diagram illustrating a sequence example of BH RLF processing using AI / ML for BH RLF processing in Variation 1 of Embodiment 4. Figure 19 The first half of the sequence example is shown. Figure 20 The latter half of the sequence example is shown. Figure 19 and Figure 20 The IAB host CU is publicly available, featuring BH RLF processing of AI / ML models, and examples of training and inference.

[0803] In step ST1901, the IAB host CU is equipped with an AI / ML model for BH RLF processing. In step ST1905, data communication occurs between the UE, the recovery IAB node, the parent IAB node of the initial path, the intermediate IAB nodes, the IAB host DU, and the IAB host CU.

[0804] In step ST1911, the IAB host CU with the BH RLF processing AI / ML model sends model training input information requests to each IAB node and each IAB host DU. The IAB nodes and IAB host DUs are not limited to the initial path and recovery path shown in the example in the figure. The IAB nodes and IAB host DUs can be limited to the coverage area of ​​the IAB host CU.

[0805] The request can be made based on BH RLF processing. For example, it can be made based on BH RLF detection-based BH RLF processing in any IAB node within the coverage area of ​​the IAB host CU. BH RLF information can be used as input information for model training. Furthermore, for example, as disclosed in Embodiment 4, the request can be made based on IAB processing from IAB nodes. IAB nodes can be processed based on BAP setting update processing, IAB node resource setting update processing, and BH route update processing. Model training can be performed based on the state of newly added IAB nodes and the input information in the updated setting state.

[0806] The request may contain information indicating the required information. It may include information indicating the purpose of the input information for the AI / ML model used in BH RLF processing. For example, the request may contain information indicating it is for BH RLF processing. For example, the request may contain information indicating it is for model training. This is effective when the input information for the AI / ML model used in BH RLF processing includes information for model training. For example, if the information held by the IAB node includes information for model training, inference, and feedback, the IAB node can quickly grasp the information to be notified to the IAB host CU, resulting in rapid notification from the IAB node to the IAB host CU.

[0807] The transmission of this request between the IAB host CU and the IAB host DU or IAB node can, for example, use F1 signaling. The transmission of this request between the IAB host CU and the IAB node can, for example, use RRC signaling. For example, the measurement settings sent from the IAB host CU to the IAB node can include a request for input information. This can reduce the signaling load.

[0808] In step ST1912, each IAB node and each IAB host DU sends model training input information to the IAB host CU. IAB nodes and IAB host DUs can send model training input information obtained from connected UEs to the IAB host CU. The IAB host CU can input information from UEs connected to the IAB nodes and IAB host DUs into the AI / ML model.

[0809] Input requests can be sent between the IAB host DU or IAB node and the IAB host CU, for example, using F1 signaling. Input requests can also be sent between the IAB node and the IAB host CU, for example, using RRC signaling.

[0810] Measurement results from IAB nodes can be sent to the IAB host CU via measurement result reports. These results can be wireless or L2 measurements. They can be sent using the same signaling or different signaling methods. The measurement results from IAB nodes can be used as input for model training.

[0811] In step ST1913, the IAB host CU uses the input information for the BH RLF processing AI / ML model stored in its own node, and the input information for the BH RLF processing AI / ML model obtained from the IAB node and the IAB host DU, to train the BH RLF processing AI / ML model. The BH RLF processing AI / ML model is updated through training. These processes can be performed appropriately. They can be performed periodically, either each time BH RLF processing is performed or each time IAB processing is performed. By performing them periodically or each time IAB processing is performed, the same effect as disclosed in Implementation Method 4 can be obtained. Furthermore, by performing them each time BH RLF processing is performed, for example, the input information of the state where a new path is set and the state of the IAB node whose path has been updated can be appropriately input. More appropriate training of the BH RLF processing AI / ML model can be implemented.

[0812] In step ST1921, the IAB host CU sends inference input information requests to each IAB node and each IAB host DU. The IAB nodes and IAB host DUs are not limited to the initial path and recovery path shown in the example in the diagram. The IAB host DUs and IAB nodes can be limited to the coverage area of ​​the IAB host CU. This request can be processed according to BH RLF processing. The request for inference input information allows the acquisition of inference input information and its input into the AI / ML model used for BH RLF processing, and it also allows the export of output information for BH RLF processing.

[0813] The request may contain information indicating the required information. The request may contain information indicating the purpose of the input information for the AI / ML model used in BHRLF processing. For example, the request may contain information indicating that it is for inference. This is effective, for example, when the input information for the AI / ML model used in BHRLF processing includes information for inference.

[0814] In step ST1922, each IAB node and each IAB host DU sends inference input information to the IAB host CU. IAB nodes and IAB host DUs can send inference input information obtained from connected UEs to the IAB host CU. The IAB host CU can input information from UEs connected to IAB nodes and IAB host DUs into the AI / ML model.

[0815] The sending of inference input information requests and inference input information can be performed periodically. Output information can be updated periodically. The sending of inference input information requests and inference input information can also occur when a physical problem is detected. If the physical problem remains unresolved, communication is transferred to the BH RLF; therefore, communication can sometimes continue during the physical problem detection phase. Alternatively, the sending of inference input information requests and inference input information can occur when a predetermined number of synchronization deviations are detected. This predetermined number can be set to a value smaller than the predetermined number required for physical problem detection. The predetermined number can be preset. For example, it can be statically set through standards or sent from the IAB host CU. By obtaining inference input information at these stages, input information that better reflects the status of the BH RLF can be obtained. More appropriate output information can be obtained.

[0816] The methods for requesting and sending input information for inference can appropriately apply the methods for requesting and sending input information for model training disclosed above.

[0817] In step ST1923, the IAB host CU uses the input information of the BH RLF processing AI / ML model stored in this node, and the input information of the BH RLF processing AI / ML model obtained from the IAB node and the IAB host DU, to perform inference using the BH RLF processing AI / ML model. The IAB host CU then derives the output information of the BH RLF processing AI / ML model through this inference.

[0818] In step ST1931, the IAB host CU sends the AI / ML model output information for BH RLF processing to each IAB host DU and each IAB node. For example, it sends cell selection and cell reselection processing configuration information to each IAB host DU and each IAB node. For example, it sends BAP configuration information and IAB node resource configuration information to each IAB node. The sent output information can be the configuration information used during BH RLF recovery. Thus, the IAB nodes and IAB host DUs used in BH RLF processing can obtain the AI / ML model output information for BH RLF processing. When an IAB node detects BH RLF, it can use this output information to perform BH RLF processing among the IAB host CU, each IAB host DU, and each IAB node.

[0819] The method for sending output information can be appropriately applied to the method for sending setting information for BH RLF processing, and the method for sending input information requests for the AI / ML model for BH RLF processing disclosed above.

[0820] In step ST1941, the IAB node detection BH RLF is restored.

[0821] In step ST1951, the recovery IAB node performs cell selection processing and selects a parent IAB node for the recovery path. The recovery IAB node then performs RRC reconstruction processing via the selected parent IAB node, intermediate IAB nodes, IAB host DU, and IAB host CU. This processing can utilize BH RLF processing with AI / ML model output information. Thus, a parent node that becomes a more efficient recovery path is selected.

[0822] In step ST1952, IAB processing is performed between the IAB host CU and the parent IAB node, intermediate IAB nodes, IAB host DU, and recovery IAB node of the recovery path. This includes setting the BAP for each IAB node and configuring its resources. Furthermore, the BH path from the recovery IAB node to the IAB host CU is defined. This processing can utilize BH RLF processing to output information from the AI / ML model. This allows for more efficient configuration of the recovery path and IAB nodes.

[0823] In step ST1953, BAP configuration release processing is performed between the IAB host CU and the parent IAB node, intermediate IAB nodes, and IAB host DU of the initial path. Resource configuration release processing can be performed on IAB nodes. Path configurations from the recovery IAB node can be released. Through this release processing, the configurations of the IAB nodes and IAB host DU of the initial path are released, thus reducing the load.

[0824] The release process in step ST1953 can be limited to restoring the path used by the IAB node. Other IAB nodes may not require release processing. Alternatively, other IAB nodes may be updated. For example, when initial paths are set for multiple IAB nodes, it avoids complicating control in cases where a BH RLF is detected between an IAB node.

[0825] Therefore, the restored IAB node connects to the IAB host CU via the recovery path. In step ST1961, data communication occurs between the UE and the IAB host CU via the restored IAB node, the parent IAB node of the recovery path, the intermediate IAB node, and the IAB host DU.

[0826] In steps ST1971 to ST1975, the restored IAB node, the parent IAB node of the restored path, the intermediate IAB nodes, the IAB host DU, and the UE send BH RLF processing feedback information using the AI / ML model for BH RLF processing to the IAB host CU. The feedback information may include information identifying the IAB node, the IAB host DU, and the UE. Upon receiving this information, the IAB host CU can identify the node that sent the feedback information.

[0827] Therefore, the IAB host CU can obtain information related to the performance of BH RLF processing using the AI / ML model for BH RLF processing. The IAB host CU can use the BH RLF processing feedback information using the AI / ML model for BH RLF processing stored in this node, as well as the BH RLF processing feedback information using the AI / ML model for BH RLF processing obtained from the recovery IAB node, the parent IAB node of the recovery path, intermediate IAB nodes, the IAB host DU, and the UE, to evaluate the effect of BH RLF processing when using the AI / ML model for BH RLF processing.

[0828] The method for sending feedback information can appropriately apply the method for sending input information for the AI / ML model used in IAB processing disclosed above. For example, RRC signaling can be used to send feedback information from the UE to the IAB host CU.

[0829] Input and feedback requests for the BH RLF processing AI / ML model can be eliminated. Each node can appropriately send input and feedback information for the BH RLF processing AI / ML model to nodes with the BH RLF processing AI / ML model. For example, this can be done periodically. Alternatively, input and feedback information can be sent upon receiving new input or feedback information. This aims to reduce signaling load. When a node with the BH RLF processing AI / ML model receives this information from other nodes or itself, it can send input requests to nodes other than that node. This reduces unnecessary signaling, such as sending identical input information.

[0830] Therefore, importing AI / ML into BH RLF processing enables more efficient BH RLF processing.

[0831] Other IAB host CUs can have AI / ML models for BH RLF processing. For example, an IAB host CU that supports IABs can have an AI / ML model for BH RLF processing. Partial or complete transmission and reception of input information, requests for such input information, output information, and feedback information can be performed between IAB host CUs and other IAB host CUs. This transmission and reception can use Xn signaling.

[0832] Therefore, for example, even when BH RLF processing is implemented due to changes in the IAB host CU connected via the movement of an IAB node, input or feedback information from other IAB host CUs, IAB nodes within the coverage area of ​​that IAB host CU, IAB host DUs, etc., can be input into the AI / ML model for BH RLF processing. By acquiring more node information and inputting it into the AI / ML model for BH RLF processing, output information suitable for various states of the IAB node and the UE communicating using that IAB node can be derived. By performing BH RLF processing using this output information, more efficient BH RLF processing can be implemented.

[0833] By employing the method disclosed in this variation, AI / ML models can be used for training and inference in BH RLF processing. By deriving the output information of the AI / ML model using various input information, BH RLF processing can be optimized. In BH RLF processing, more suitable parent IAB node selection, BH path settings, IAB node BAP settings, and resource settings can be performed. Even if IAB node movement or changes in the radio wave propagation environment in the BH path occur, the failure of reconstruction processing caused by BHRLF and the recurrence of BH RLF can be reduced. This reduces communication interruptions between UE and NW, improving communication performance between UE and NW.

[0834] Figure 19 , Figure 20 This document discloses an example of setting up an AI / ML model for BH RLF processing on the IAB host CU. Alternatively, an AI / ML model for BH RLF processing can be set up on an IAB node. Learning and inference mechanisms can be configured on the IAB node. The IAB node can use the AI / ML model to derive the configuration information required for BH RLF processing.

[0835] An IAB node equipped with an AI / ML model for BH RLF processing can send requests for input and feedback information for the BH RLF processing AI / ML model to other IAB nodes, the IAB host DU, and the IAB host CU. Other IAB nodes, the IAB host DU, and the IAB host CU can also send input and feedback inform...

Claims

1. A communication system, characterized in that, include: The base station corresponds to the fifth-generation wireless access system. as well as A communication terminal, which is connected to the base station. The base station uses a trained model to determine the settings used in the communication processing implemented by the communication terminal, and notifies the communication terminal connected to the base station of the determined settings. The training process utilizes information obtained from the communication terminal already connected to the base station and other neighboring base stations. After receiving the notification of the settings from the base station that determined the settings, i.e., the first base station, the communication terminal uses the settings to perform communication processing for the first base station or for a base station different from the first base station, i.e., the second base station, after ending the connection with the first base station.

2. The communication system as described in claim 1, characterized in that, The first base station uses the model to determine the settings used in the random access process implemented when the communication terminal connects to the first base station or the second base station, and notifies the communication terminal connected to this base station of the determined settings. The communication terminal that receives the setting notification from the first base station uses the setting to implement random access processing for reconnecting to the first base station after the connection with the first base station has ended, or for connecting to a base station different from the first base station, namely a second base station.

3. The communication system as described in claim 1, characterized in that, The first base station uses the model to determine the settings used in the data transmission process when the communication terminal is not connected to the base station, and notifies the communication terminal connected to the base station of the determined settings. After receiving the configured notification from the first base station, the communication terminal, without being connected to either the first or the second base station, uses the configured settings to perform data transmission processing to either the first or the second base station.

4. A base station that operates as an integrated access and backhaul host in a communication system corresponding to a fifth-generation wireless access system, characterized in that: The trained model is used to determine the settings for the processes performed when constituting the access backhaul integration, and the determined settings are communicated to each node of the access backhaul integration, wherein... The training used information obtained from each node of the integrated access and backhaul system.

5. The base station as described in claim 4, characterized in that, It consists of a central unit and one or more distributed units. The central unit uses a trained model to determine the settings of the processing to be performed when the access backhaul integration is constructed, and notifies the distributed units and each node of the access backhaul integration of the determined settings, wherein the training uses information obtained from the distributed units and information obtained from each node of the access backhaul integration.

6. A base station that operates as an integrated access and backhaul host in a communication system corresponding to a fifth-generation wireless access system, characterized in that, The trained model is used to determine the settings used in the reconnection process between the nodes of the access backhaul integration, and the determined settings are communicated to each node of the access backhaul integration. The training used information obtained from each node of the integrated access and backhaul system.

7. The base station as described in claim 6, characterized in that, It consists of a central unit and one or more distributed units. The central unit uses a trained model to determine the settings used in the reconnection process between the distributed units and the nodes of the access backhaul integration, as well as in the reconnection process between the nodes of the access backhaul integration, and notifies the distributed units and the nodes of the access backhaul integration of the determined settings. The training uses information obtained from the distributed units and information obtained from the nodes of the access backhaul integration.

8. A communication terminal connected to a base station corresponding to a fifth-generation wireless access system, characterized in that, Using a trained model, the settings used in the random access process implemented by this communication terminal to connect to a base station are determined. These determined settings are then used to implement random access processes for connecting to the first base station again after the connection with the first base station has ended, or for connecting to a second base station, which is different from the first base station. The training used information obtained from the connected base station, namely the first base station, and a second base station that is different from the first base station.

Images