Communication method, user device, mobile communication system, chipset, and program
By controlling availability notifications based on buffer and power conditions with timers, the system addresses inefficiencies in AI/ML data collection, reducing resource waste and power consumption in mobile communication systems.
Patent Information
- Application Number
- PCT/JP2026/017352
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2025-05-06
- Filing Date
- 2026-05-01
- Publication Date
- 2026-11-12
AI Technical Summary
Existing mobile communication systems face inefficiencies in data collection for AI/ML model training due to frequent availability notifications from user equipment, leading to resource waste, increased processing load, and power consumption.
Implementing a mechanism where user equipment suppresses availability notifications based on settings or requests from the network, using trigger conditions such as buffer and power states, and employing timers to control notification frequency.
This approach reduces unnecessary resource usage and power consumption by minimizing frequent availability notifications, optimizing data collection for AI/ML model training.
Smart Images

Figure JP2026017352_12112026_PF_FP_ABST
Abstract
Description
Communication method, user apparatus, mobile communication system, chipset, and program
[0001] The present disclosure relates to a communication method, a user apparatus, and a network node used in a mobile communication system.
[0002] In the 3rd Generation Partnership Project (3GPP (registered trademark; the same applies hereinafter)), which is a standardization project for mobile communication systems, formulation of technical specifications for applying artificial intelligence or machine learning (AI / ML) technology to wireless communication is in progress.
[0003] In AI / ML technology, a learning model called an AI / ML model is used, for example, for mobility optimization, channel state information (CSI) feedback enhancement, beam management improvement, or positioning improvement.
[0004] 3GPP Technical Specification "3GPP TS 38.300 V18.5.0 (2025-03)"
[0005] A communication method according to a first aspect of the present disclosure is a communication method executed by a user apparatus in a mobile communication system, the method comprising: transmitting, by the user apparatus, a notification including information indicating a low power state to a network; receiving, by the user apparatus, in response to transmitting the notification, a data collection stop instruction that instructs stopping of collection of training data used for training an artificial intelligence or machine learning (AI / ML) model from the network; and stopping, by the user apparatus, the collection of the training data in response to receiving the data collection stop instruction.
[0006] A user apparatus according to a second aspect of the present disclosure is a user apparatus in a mobile communication system, comprising: a transmitter that transmits a notification including information indicating a low power state to a network; a receiver that receives, in response to transmission of the notification, a data collection stop instruction that instructs stopping of collection of training data used for training an artificial intelligence or machine learning (AI / ML) model from the network; and a controller that stops the collection of the training data in response to receiving the data collection stop instruction.
[0007] A network node according to a third aspect of the present disclosure is a network node used in a mobile communication system, and includes a receiving unit that receives a notification transmitted from a user device that collects training data used for training artificial intelligence or machine learning (AI / ML) models on the network side, in response to the fulfillment of a trigger condition based on the amount of training data collected or the power state of the user device, and a transmitting unit that transmits a setting or request to the user device to suppress the transmission of the notification.
[0008] A mobile communication system according to a fourth aspect of the present disclosure is a mobile communication system having a user device, wherein the user device transmits a notification to a network including information indicating a low power state, the user device receives a data collection stop instruction from the network instructing it to stop collecting training data used for training an artificial intelligence or machine learning (AI / ML) model in response to transmitting the notification, and the user device stops collecting the training data in response to receiving the data collection stop instruction.
[0009] A fifth aspect of the present disclosure is a chipset for a user device in a mobile communication system, which includes transmitting a notification to a network that includes information indicating a low power state, receiving a data collection stop instruction from the network in response to transmitting the notification, which instructs the cessation of collecting training data used for training an artificial intelligence or machine learning (AI / ML) model, and stopping the collection of the training data in response to receiving the data collection stop instruction.
[0010] A program according to a sixth aspect of this disclosure includes the following processes: sending a notification to a network containing information indicating a low power state to a user device in a mobile communication system; receiving a data collection stop instruction from the network in response to the transmission of the notification, instructing the cessation of the collection of training data used for training an artificial intelligence or machine learning (AI / ML) model; and stopping the collection of the training data in response to the receipt of the data collection stop instruction.
[0011] This figure shows an example configuration of a mobile communication system according to the embodiment. This figure shows an example configuration of the protocol stack of the U-plane (UP) radio interface that handles data. This figure shows an example configuration of the protocol stack of the C-plane radio interface that handles signaling (control signals). This figure shows an example configuration of a UE (User Equipment) according to the embodiment. This figure shows an example configuration of a base station (an example of a network node) according to the embodiment. This figure is for explaining the functions of AI / ML. This figure shows an operation scenario related to the suppression of availability notifications according to the embodiment. This figure shows a basic operation example when a buffer condition is used as a trigger condition in the mobile communication system according to the embodiment. This figure shows a basic operation example when a power condition is used as a trigger condition in the mobile communication system according to the embodiment. This figure shows an overview of the operation of the UE related to the suppression of availability notifications according to the embodiment. This figure shows a first operation example of the suppression of availability notifications according to the embodiment. This figure shows a second operation example of the suppression of availability notifications according to the embodiment. This figure shows a fourth operation example of the suppression of availability notifications according to the embodiment. This figure shows a fifth operation example of the suppression of availability notifications according to the embodiment. This figure shows an operation scenario related to the data collection operation of multiple AI / ML according to the embodiment. This figure shows an overview of the operation of the UE related to the data collection operation of multiple AI / ML according to the embodiment. This is a diagram illustrating a first management method for a memory area according to the embodiment. This is a diagram illustrating a second management method for a memory area according to the embodiment. This is a diagram illustrating an example of the operation of data collection operations for multiple AI / MLs when a buffer condition is used as a trigger condition in the mobile communication system according to the embodiment. This is a diagram illustrating an example of the operation of data collection operations for multiple AI / MLs when a power condition is used as a trigger condition in the mobile communication system according to the embodiment.
[0012] AI / ML models are sometimes implemented on the network side, and these cases are also referred to as network-side models. In network-side models, the network may need to collect training data using user equipment. The training data is used to train the AI / ML model (also referred to as "model training").
[0013] This disclosure aims to enable appropriate data collection in the case of a network-side model.
[0014] In the following, a mobile communication system according to an embodiment will be described with reference to the drawings. In the drawings, the same or similar parts are denoted by the same or similar reference numerals.
[0015] (1) System Configuration First, the system configuration according to this embodiment will be described. Figure 1 is a diagram showing an example of the configuration of the mobile communication system according to this embodiment. The mobile communication system according to this embodiment is a system that conforms to the 3GPP standard. For example, the mobile communication system according to this embodiment may be a fifth-generation (5G) system or a sixth-generation (6G) system.
[0016] The mobile communication system according to this embodiment comprises a network (NW) 1 and a user device (UE) 100. The UE 100 is a mobile communication device that performs wireless communication with the NW 1. The UE 100 may be any device used by a user, such as a mobile phone terminal (including a smartphone), a tablet terminal, a notebook PC (Personal Computer), a communication module (including a communication card or chipset), a sensor or a device installed on a sensor, a vehicle or a device installed on a vehicle (Vehicle UE), or an aircraft or a device installed on an aircraft (Aerial UE).
[0017] NW1 includes a Radio Access Network (RAN) 2, which is the part for wireless communication, and a Core Network (CN) 3, which is the part for mobility management, session management, and service control. When the mobile communication system is a fifth-generation system (5GS), RAN2 is called NG-RAN (Next Generation Radio Access Network) and CN3 is called 5GC (5G Core Network). However, in 6G, different names may be used.
[0018] RAN2 includes multiple base stations 200 (base stations 200a to 200c in the illustrated example). The base stations 200 are interconnected via inter-base station interfaces. The term "connection" refers to either a physical or logical connection. Base station 200 is an example of a network node. A network node is an element that constitutes NW1 and is configured physically or logically.
[0019] The base station 200 is composed of a CU (Central Unit) and a DU (Distributed Unit) (i.e., functionally divided), and the two units may be connected by a front-haul interface. When the mobile communication system is 5GS, the base station 200 is referred to as a gNB, the inter-base station interface is referred to as an Xn interface, and the front-haul interface is referred to as an F1 interface. However, in 6G, different names may be used.
[0020] Each base station 200 manages one or more cells. The base station 200 performs wireless communication with UE 100 that has established a connection with its own cell. Each base station 200 has a radio resource management (RRM) function, a user data routing function (also simply referred to as "data"), and a measurement and control function for mobility control and scheduling.
[0021] The term "cell" is used to indicate the smallest unit of a wireless communication area. It is also used to indicate a function or resource for wireless communication with the UE100. One cell belongs to one carrier frequency. One cell may be associated with one downlink component carrier and one uplink component carrier. The bandwidth corresponding to one cell (system bandwidth) may be divided into multiple bandwidth parts (BWPs).
[0022] Furthermore, the terms "downlink" and "DL" may be used interchangeably, and the terms "uplink" and "UL" may be used interchangeably. In addition, the terms "base station," "network node," "RAN node," "cell," and "network" may be used interchangeably.
[0023] CN3 includes a CN device 300, which is another example of a network node. The CN device 300 may include a CP-CN device 300a (see Figure 3) corresponding to the control plane (CP) and a UP-CN device corresponding to the user plane (UP). The CP-CN device 300a performs various mobility controls and paging for the UE100. The CP-CN device 300a may also perform session management for the UE100. The CP-CN device 300a communicates with the UE100 using Non-Access Stratum (NAS) signaling. The UP-CN device, on the other hand, performs data transfer control, etc.
[0024] In the case of a 5GS mobile communication system, the CP-CN device 300a is referred to as the AMF (Access and Mobility Management Function) and / or SMF (Session Management Function), the UP-CN device is referred to as the UPF (User Plane Function), and the interface between the base station 200 and the CN device 300 is referred to as the NG interface. However, in 6G, different names may be used.
[0025] Figure 2 shows an example of the protocol stack configuration for a user plane (UP) wireless interface that handles data.
[0026] A user-plane wireless interface protocol includes, for example, a physical (PHY) layer, a MAC (Medium Access Control) layer, an RLC (Radio Link Control) layer, a PDCP (Packet Data Convergence Protocol) layer, and an SDAP (Service Data Adaptation Protocol) layer. The term "layer" may be used interchangeably with the term "protocol."
[0027] The PHY layer performs encoding / decoding, modulation / demodulation, antenna mapping / demapping, and resource mapping / demapping. Data and control information are transmitted between the PHY layer of UE100 and the PHY layer of base station 200 via a physical channel. The PHY layer of UE100 receives downlink control information (DCI) transmitted from base station 200 on the physical downlink control channel (PDCCH). Specifically, UE100 performs blind decoding of the PDCCH using a Radio Network Temporary Identifier (RNTI) and acquires the successfully decoded DCI as the DCI addressed to its own UE. The DCI transmitted from base station 200 has a CRC (Cyclic Redundancy Check) parity bit added, which is scrambled by the RNTI.
[0028] The MAC layer performs data priority control and retransmission processing using Hybrid ARQ (HARQ: Hybrid Automatic Repeat reQuest). Data and control information are transmitted between the MAC layer of UE100 and the MAC layer of base station 200 via the transport channel. The MAC layer of base station 200 includes a scheduler. The scheduler determines the transport format for the uplink and downlink (transport block size, modulation scheme (MCS)) and the resources to be allocated to UE100.
[0029] The RLC layer transmits data to the receiving RLC layer using the functions of the MAC layer and PHY layer. Data and control information are transmitted between the RLC layer of UE100 and the RLC layer of base station 200 via a logical channel.
[0030] The PDCP layer performs header compression / decompression, encryption / decryption, etc.
[0031] The SDAP layer maps IP flows, which are the units in which CN3 performs QoS (Quality of Service) control, to wireless bearers, which are the units in which AS (Access Stratum) performs QoS control. However, if CN3 performs QoS control at the same unit as RAN2 (i.e., QoS control is performed at the bearer level rather than the IP flow level, and a one-to-one mapping with wireless bearers is possible), the SDAP layer is not necessary.
[0032] Figure 3 shows an example of the protocol stack configuration for a wireless interface of a control plane (CP) that handles signaling (control signals).
[0033] The protocol stack of the control plane's wireless interface includes, for example, an RRC (Radio Resource Control) layer and a NAS layer instead of the SDAP layer shown in Figure 2.
[0034] RRC signaling for various settings is transmitted between the RRC layer of UE100 and the RRC layer of base station 200. The RRC layer controls the logical channel, transport channel, and physical channel in response to the establishment, re-establishment, and release of the radio bearer. If there is a connection (RRC connection) between the RRC of UE100 and the RRC of base station 200, UE100 is in the RRC connected state. If there is no connection (RRC connection) between the RRC of UE100 and the RRC of base station 200, UE100 is in the RRC idle state. If the connection between the RRC of UE100 and the RRC of base station 200 is suspended, UE100 is in the RRC inactive state.
[0035] The NAS (also simply referred to as "NAS"), located above the RRC layer, performs session management and mobility management, etc. NAS signaling is transmitted between the NAS layer of UE100 and the NAS layer of CP-CN device 300a. In addition to the wireless interface protocol, UE100 also has an application layer, etc. Furthermore, the layer below the NAS is called the access layer (AS).
[0036] Furthermore, if the base station 200 is functionally divided into a CU and a DU, the DU may be responsible for processing the PHY, MAC, and RLC layers, while the CU may be responsible for processing the PDCP, SDAP, and RRC layers.
[0037] Figure 4 shows an example configuration of UE100 (user device) according to this embodiment.
[0038] The UE100 includes a receiving unit 110, a transmitting unit 120, and a control unit 130. The receiving unit 110 and the transmitting unit 120 constitute a wireless communication unit 140 (transceiver unit) that performs wireless communication with the base station 200.
[0039] The receiving unit 110 performs various types of reception under the control of the control unit 130. The receiving unit 110 includes an antenna and a receiver. The receiver converts the radio signal received by the antenna into a baseband signal (received signal) and outputs it to the control unit 130. The transmitting unit 120 performs various types of transmission under the control of the control unit 130. The transmitting unit 120 includes an antenna and a transmitter. The transmitter converts the baseband signal (transmitted signal) output by the control unit 130 into a radio signal and transmits it from the antenna.
[0040] The control unit 130 performs various controls and processes in the UE 100. The operation of the UE 100 described above and below may be controlled by the control unit 130. The control unit 130 includes at least one processor and at least one memory. The memory stores programs executed by the processor and information used for processing by the processor. The processor may include a baseband processor and a CPU (Central Processing Unit). The baseband processor performs modulation, demodulation, encoding, and decoding of baseband signals. The CPU executes programs stored in memory and performs various processes.
[0041] Figure 5 shows an example configuration of a base station 200 (an example of a network node) according to this embodiment.
[0042] The base station 200 includes a transmitting unit 210, a receiving unit 220, a control unit 230, and an NW communication unit 240. The transmitting unit 210 and the receiving unit 220 constitute a wireless communication unit 250 (transceiver unit) that performs wireless communication with the UE 100.
[0043] The transmitting unit 210 performs various types of transmissions under the control of the control unit 230. The transmitting unit 210 includes an antenna and a transmitter. The transmitter converts the baseband signal (transmission signal) output by the control unit 230 into a radio signal and transmits it from the antenna. The receiving unit 220 performs various types of receptions under the control of the control unit 230. The receiving unit 220 includes an antenna and a receiver. The receiver converts the radio signal received by the antenna into a baseband signal (received signal) and outputs it to the control unit 230.
[0044] The control unit 230 performs various types of control and processing in the base station 200. The operations of the base station 200 described above and below may be operations under the control of the control unit 230. The control unit 230 includes at least one processor and at least one memory. The memory stores programs to be executed by the processor and information used for processing by the processor. The processor may include a baseband processor and a CPU. The baseband processor performs modulation / demodulation, encoding / decoding, and the like of baseband signals. The CPU executes programs stored in the memory to perform various types of processing.
[0045] The NW communication unit 240 is connected to adjacent base stations via an inter-base station interface. Further, the NW communication unit 240 is connected to the CN apparatus 300 via an interface between the base station and the CN.
[0046] (2) Outline of AI / ML Technology Next, an outline of the AI / ML technology according to the present embodiment will be described. AI / ML may be applied to the RAN 2 and / or the CN 3 in the mobile communication system according to the present embodiment. For example, a learning model (referred to herein as an AI / ML model) may be used to optimize the mobility of the UE 100. Further, the AI / ML model may be used for purposes such as improvement of channel state information (CSI) feedback (e.g., overhead reduction, accuracy improvement, prediction), improvement of beam management (e.g., accuracy improvement, prediction in time / spatial domain), improvement of position measurement (e.g., improvement of position estimation / prediction), and the like.
[0047] The AI / ML model may be a data-driven algorithm that applies AI / ML technology to generate a series of outputs based on a series of inputs. The AI / ML model may output at least one piece of information such as an estimated value, a predicted value, a selected operation, or classification, based on input information. The UE 100 or the base station 200 may input channel state information and / or reference signal measurement values or the like to the AI / ML model, and output highly accurate channel state information, measurement values, beam selection, and / or position, or may output future channel state information and / or radio link quality or the like. The term "inference" may be read as "estimation" or "prediction".
[0048] The AI / ML model may be provided on the network side (for example, the base station 200). Such a case may be referred to as a NW-side model. The AI / ML model may be provided on the UE 100 side. Such a case may be referred to as a UE-side model. In the present embodiment, the NW-side model will be mainly described.
[0049] FIG. 6 is a diagram for describing functions of AI / ML. AI / ML may include the following functions.
[0050] - Data collection: Providing input data for model training and model inference functions. Data collection may be a process of collecting data by the base station 200, the AI / ML management function 4 (see FIG. 7), or the UE 100 for the purpose of AI / ML model training, data analysis, and inference.
[0051] - Model training: Executing training, verification, and testing of the AI / ML model. As part of the model testing procedure, model performance indicators may be generated. Model training may be a process of training the AI / ML model by learning the relationship between inputs and outputs, and obtaining a trained AI / ML model for inference.
[0052] - Model inference: Providing inference outputs (such as predictions or decisions). The model inference function may provide control details of model inference to the model management or performance monitoring function. Model inference may mean a process for deriving a set of outputs from a set of inputs using a trained AI / ML model.
[0053] - Management (model management or performance monitoring): Managing the AI / ML model and monitoring the performance of the model.
[0054] Furthermore, an AI / ML model may be provided for each AI / ML function. For example, an AI / ML model may be provided for each use case (or each sub-use case). An AI / ML model may also be provided for each function within the same use case (or the same sub-use case). The terms "use case," "sub-use case," "function," "AI / ML function," "(AI / ML) dataset," and "(AI / ML) resource set" may be interchangeable.
[0055] For example, the use case (or sub-use case) may be "beam management," and the sub-use cases (or functions) may be "spatial beam prediction" and "temporal beam prediction." The use case (or sub-use case) may be "positioning accuracy," and the sub-use cases (or functions) may be "direct AI / ML positioning" and "AI / ML-assisted positioning." The use case (or sub-use case) may be "CSI feedback," and the sub-use cases (or functions) may be "spatial and frequency domain CSI compression" and "temporal domain CSI prediction."
[0056] Furthermore, the use case (or sub-use case) may be "measurement prediction," and the sub-use cases (or functions) may be "spatial domain measurement prediction," "interfrequency measurement prediction," and "beam level prediction." The use case (or sub-use case) may be "handover optimization," and the sub-use cases (or functions) may be "radio link failure (RLF) prediction," "conditional handover triggering," and "early timing advance (TA) update."
[0057] (3) Suppression of Availability Notifications Next, the suppression of availability notifications according to this embodiment will be described.
[0058] The availability notification according to this embodiment is a notification transmitted by the UE 100, which performs data collection to collect training data used for model training in the NW-side model, and is a notification regarding the availability of training data. The term "availability notification" may be read as "notification regarding the availability of training data," "availability indication," or "availability indicator." The UE 100 may include the availability notification in the RRC message transmitted to the base station 200.
[0059] (3.1) Operation Scenario Figure 7 is a diagram showing the operation scenario related to the suppression of availability notifications according to this embodiment.
[0060] In the illustrated example, NW1 has RAN2 and AI / ML model 5. The term "RAN" may be read as "base station". AI / ML model 5 may be provided in RAN2. NW1 may have an AI / ML management function 4. AI / ML management function 4 may be a network function (NF) provided in RAN2 or CN3. AI / ML management function 4 may be a server provided outside of RAN2 and CN3. When AI / ML management function 4 manages an AI / ML model, AI / ML management function 4 may transfer the AI / ML model 5 used for model training and / or model inference to RAN2.
[0061] The term "AI / ML management function" may be replaced with "server," "OTT (Over-The-Top)," "OTT server," "application," or "application server," etc.
[0062] In step S1, NW1 (RAN2) sends data collection settings, which are settings for data collection, to UE100. UE100 receives the data collection settings. The data collection settings may be included in an RRC message sent from NW1 to UE100. For example, the data collection settings may be included in dedicated signaling (e.g., an RRC Reconfiguration message). The data collection settings may also be included in broadcast signaling (e.g., a System Information Block (SIB)).
[0063] In step S2, UE100 collects training data to be used for model training based on the received data collection settings. The collected training data may differ depending on the AI / ML function being targeted.
[0064] In step S3, UE100 and NW1 (RAN2) perform a data acquisition procedure for NW1 (RAN2) to acquire training data. In this procedure, UE100 sends an availability notification to NW1 (RAN2). UE100 sends an availability notification to NW1 (RAN2) when the trigger condition for sending the availability notification is met. The trigger condition may be set by NW1 (RAN2) in UE100.
[0065] The trigger conditions may include two types: buffer conditions (when the amount of collected training data reaches a predetermined amount) and power conditions (when the UE100 enters a low-power state). Here, a low-power state refers to a state in which the remaining power (or remaining time) of the UE100's power source (e.g., battery) is low.
[0066] Furthermore, there may be two types of buffer conditions: • Full buffer condition: The buffer has reached its full capacity, i.e., the amount of collected data has reached the upper limit of the storage area. • Buffer threshold condition: The buffer threshold has been reached, i.e., the amount of collected data has reached the threshold.
[0067] Therefore, there may be three types of trigger conditions for availability notifications: full buffer conditions, buffer threshold conditions, and power conditions.
[0068] UE100 may, upon meeting the trigger conditions, send an RRC message to NW1 (RAN2) as an availability notification, containing at least one of the following pieces of information. The RRC message may also be a UAI (UE Assistance Information) message.
[0069] Information indicating that data is available; information indicating the trigger cause (e.g., full buffer or threshold); and information indicating a low-power state.
[0070] The method for determining when the buffer threshold has been reached and for determining the low-power state may be left to the implementation of UE100.
[0071] In step S4, NW1 performs model training using the training data acquired in step S3. For example, RAN2 and / or AI / ML management function 4 may train the AI / ML model 5 using the training data. If the AI / ML management function 4 performs model training, RAN2 may transfer the acquired training data to the AI / ML management function 4.
[0072] Figure 8 shows a basic example of operation when a buffer condition is used as a trigger condition in the mobile communication system according to this embodiment. In the following embodiments, NW1 may be read as a base station 200 or an AI / ML management function 4.
[0073] In step S11, NW1 transmits the data acquisition settings to UE100. UE100 receives the data acquisition settings. In this example, the data acquisition settings may include information to set buffer conditions (e.g., full buffer conditions and / or buffer threshold conditions) as trigger conditions.
[0074] In step S12, UE100 collects training data to be used for model training based on the received data collection settings. The collected training data may differ depending on the AI / ML function being targeted. UE100 stores the collected training data in its own memory area.
[0075] In step S13, UE100 determines whether the collected training data has reached a predetermined amount (e.g., a full buffer or buffer threshold) (i.e., performs a trigger evaluation). If the collected training data has not reached a predetermined amount (step S13: NO), UE100 returns to step S12.
[0076] On the other hand, if the collected training data reaches a predetermined amount (step S13: YES), in step S14, UE100 sends an availability notification to NW1. NW1 receives the availability notification.
[0077] In step S15, NW1 sends a transmission request for the collected training data to UE100. UE100 receives the transmission request. The transmission request may be an RRC message. The RRC message may be a UE Information Request message.
[0078] In step S16, UE100 sends a message containing the collected training data to NW1 in response to a transmission request. NW1 acquires the training data by receiving the message. The message may be an RRC message. The RRC message may be a UE Information Response message.
[0079] In step S17, NW1 performs model training using the acquired training data.
[0080] Figure 9 shows a basic example of operation when a power condition is used as a trigger condition in the mobile communication system according to this embodiment. However, redundant explanations of operations similar to those in Figure 8 are omitted.
[0081] In step S21, NW1 transmits the data acquisition settings to UE100. UE100 receives the data acquisition settings. In this example, the data acquisition settings may include information that sets the power conditions as trigger conditions.
[0082] In step S22, UE100 collects training data to be used for model training based on the received data collection settings. The collected training data may differ depending on the AI / ML function being targeted. UE100 stores the collected training data in its own memory area.
[0083] In step S23, UE100 determines whether its power state has entered a low-power state (i.e., performs a trigger evaluation). If it has not entered a low-power state (step S23: NO), UE100 returns to step S22.
[0084] On the other hand, if a low-power state is entered (step S23: YES), in step S24, UE100 sends an availability notification to NW1 that includes information indicating that a low-power state is in place. NW1 receives the availability notification.
[0085] In step S25, NW1 may send a transmission request for the collected training data to UE100.
[0086] In step S26, UE100 may, in response to a transmission request, send a message containing the collected training data to NW1.
[0087] In step S27, NW1 may perform model training using the acquired training data.
[0088] In step S28, NW1 sends a data collection stop instruction to UE100, which instructs the UE100 to stop (pause) data collection. UE100 receives the data collection stop instruction. NW1 may also send an instruction to UE100 to end or cancel data collection. In this case, the operations from step S30 onward may be omitted. These instructions may be included in the RRC message (for example, the RRC Reconfiguration message) sent from NW1 to UE100.
[0089] In step S29, UE100 stops data collection in response to receiving the instruction in step S28.
[0090] In step S30, UE100 may determine whether or not it has emerged from the low-power state. For example, UE100 may emerge from the low-power state by being charged. If it has not emerged from the low-power state (step S30: NO), UE100 returns to step S29.
[0091] On the other hand, if the system exits the low-power state (step S30: YES), in step S31, UE100 may send a notification to NW1 indicating that the system has exited the low-power state. This notification may be a type of availability notification. This notification may be an RRC message containing information indicating that the system has exited the low-power state. This RRC message may be a UAI message.
[0092] In step S32, NW1 may send an instruction to UE100 to resume data collection in response to receiving the notification in step S31. This instruction may be included in an RRC message sent from NW1 to UE100. For example, this instruction may be included in dedicated signaling (e.g., an RRC Reconfiguration message).
[0093] In step S33, UE100 may resume data collection in response to receiving the instruction in step S32. In this case, UE100 may return to processing in step S22.
[0094] (3.2) Operation Overview Based on the above-described configuration and operation, an overview of the suppression operation of availability notifications according to this embodiment will be explained.
[0095] As described above, the determination of the trigger conditions for availability notifications, such as determining when the buffer threshold has been reached and determining when a low-power state is present, may be left to the implementation of UE100. In this case, UE100 can send availability notifications to NW1 any number of times at any time, which may result in frequent transmission of availability notifications. This can lead to problems such as wasted resources and increased processing load and power consumption. The following describes the operation to solve these problems.
[0096] Figure 10 is a diagram showing an overview of the operation of UE100 regarding the suppression of availability notifications according to this embodiment.
[0097] In step S101, UE100 collects training data to be used for training the AI / ML model 5 on the NW1 side.
[0098] In step S102, UE 100 evaluates whether the trigger conditions, which are based on the amount of training data collected or the power status of UE 100, have been met, which are conditions for sending an availability notification regarding the availability of training data to the base station 200.
[0099] In step S103, UE100 suppresses the transmission of availability notifications based on settings or requests from NW1 (e.g., base station 200).
[0100] Thus, in this embodiment, UE 100 suppresses the transmission of availability notifications based on settings or requests from NW1 (e.g., base station 200). This prevents UE 100 from sending availability notifications to NW1 multiple times at arbitrary times, thereby suppressing frequent transmission of availability notifications. Consequently, it is possible to suppress wasted resources and increases in processing load and power consumption.
[0101] The UE 100 performing this operation has a control unit 130 that collects training data used for training the AI / ML model 5 on the NW1 side (see Figure 4). The control unit 130 evaluates whether the conditions for sending an availability notification regarding the availability of training data to the base station 200, which are trigger conditions based on the amount of training data collected or the power status of the UE 100, have been met. The control unit 130 also suppresses the transmission of availability notifications based on settings or requests from NW1.
[0102] On the other hand, NW1 (for example, base station 200) has a receiving unit 220 that receives availability notifications transmitted from UE 100, which collects training data used for training the AI / ML model 5 on the NW1 side, in response to the fulfillment of trigger conditions based on the amount of training data collected or the power status of UE 100, and a transmitting unit 210 that transmits settings or requests to UE 100 to suppress the transmission of availability notifications (see Figure 5).
[0103] (3.3) Examples of Operation Below, the first to fifth examples of operation of the availability notification suppression operation according to this embodiment will be described. The first to fifth examples of operation may be implemented separately and independently, or two or more examples may be combined and implemented.
[0104] In the following operations, in addition to or instead of including availability notice in the UAI message, UE100 may include availability notice in at least one of the following signalings that it transmits to base station 200.
[0105] Connection-related messages Connection-related messages are RRC messages, and may include, for example, RRC Setup Request messages, RRC Reestablishment Request messages, RRC Reconfiguration Complete messages, RRC Resume Complete messages, and RRC Reestablishment Complete messages.
[0106] • Capability notification message: The capability notification message is an RRC message, which may be, for example, a UE Capability Information message.
[0107] • Measurement Report Message The measurement report message is an RRC message, which may be, for example, a Measurement Report message. The Measurement Report message is a message that contains the measurement results for the serving cell and / or adjacent cells.
[0108] AI / ML messages AI / ML messages may be RRC messages or NAS messages. AI / ML messages may be messages handled by a new layer (new protocol) for AI / ML. AI / ML messages may be model monitoring messages or inference result reporting messages, etc.
[0109] Layer 1 or Layer 2 signaling: Layer 1 signaling may be UCI (Uplink Control Information). For example, availability notifications may be included in the CSI Report transmitted via UCI. Layer 2 signaling may be MAC CE (Control Element).
[0110] (3.3.1) First Operation Example In the first operation example of the suppression operation of availability notification according to this embodiment, when UE100 sends an availability notification regarding the availability of training data, it starts a timer set by NW1 (for example, base station 200). While the timer is operating, UE100 is prohibited from sending availability notifications.
[0111] As a result, UE100 cannot send another availability notification until the timer has elapsed after sending the previous one. Therefore, frequent sending of availability notifications by UE100 can be suppressed.
[0112] Figure 11 shows a first example of the suppression operation of availability notifications according to this embodiment. In the following description of operations, redundant explanations of operations that overlap with the operations described above will be omitted.
[0113] In step S111, NW1 transmits the data acquisition settings to UE100. UE100 receives the data acquisition settings. In this example, the data acquisition settings may include information to set buffer conditions (e.g., full buffer conditions and / or buffer threshold conditions) and / or power conditions as trigger conditions.
[0114] Furthermore, in this example, the data acquisition settings may include a timer setting value indicating the timer's set time. The data acquisition settings may include the AI / ML function identifier described later, and the timer setting value may be associated with the AI / ML function identifier. In other words, UE100 may manage an independent timer for each AI / ML function. Alternatively, UE100 may manage a timer common to multiple AI / ML functions.
[0115] In step S112, UE100 performs data collection to collect training data to be used for model training, based on the received data collection settings.
[0116] In step S113, UE100 determines whether the trigger conditions for availability notification set by NW1 have been met (i.e., performs a trigger evaluation). If the trigger conditions for availability notification have not been met (step S113: NO), UE100 returns to step S112.
[0117] On the other hand, if the trigger condition for availability notification is met (step S113: YES), in step S114, UE100 sends an availability notification to NW1. NW1 receives the availability notification.
[0118] In step S115, UE100 starts the timer to which the above-mentioned timer setting value has been set.
[0119] In step S116, NW1 may send a transmission request for the collected training data to UE100.
[0120] In step S117, UE100 may, in response to a transmission request, send a message containing the collected training data to NW1.
[0121] In step S118, NW1 may perform model training using the acquired training data.
[0122] In step S119, UE100 performs data collection to collect training data to be used for model training, based on the data collection settings.
[0123] In step S120, UE100 determines whether the trigger conditions for availability notification set by NW1 have been met (i.e., performs a trigger evaluation). If the trigger conditions for availability notification have not been met (step S120: NO), UE100 returns to step S119.
[0124] On the other hand, if the trigger condition for availability notification is met (step S120: YES), in step S121, UE100 checks whether the timer is running (i.e., whether the timer has expired). If the timer is running (i.e., the timer has not expired) (step S121: YES), UE100 returns to step S119 without sending an availability notification.
[0125] If the timer is not running (i.e., the timer has expired) (step S121: NO), UE100 returns to step S114 and sends an availability notification to NW1.
[0126] In this example, the order of steps S120 and S121 may be reversed. In this case, UE100 may perform the trigger evaluation in step S120 if the timer is not operating (i.e., the timer has expired).
[0127] Furthermore, if the trigger condition for availability notification is "full buffer," UE100 cannot acquire any further logs. Therefore, if UE100 determines "full buffer" in step S120, it may wait for the timer to expire without returning to step S119, or it may send the availability notification to NW1 without waiting for the timer to expire. In addition, the behavior of these actions, namely whether or not to send an availability notification to NW1 if UE100 detects a full buffer before the timer expires, may be set in advance from NW1 to UE100.
[0128] (3.3.2) Second Operation Example In a second operation example of the suppression operation of availability notifications according to this embodiment, the availability notification may be a notification indicating that UE 100 is in a low-power state. NW1 (for example, base station 200) sets at least one condition (also called an "additional condition") from among a power condition that prevents the transmission of availability notifications (also called "power condition A"), a power condition that enables the transmission of another notification indicating that the low-power state has been discontinued (also called "power condition B"), and a time condition that indicates the duration for which the trigger condition should be continuously met in order to transmit an availability notification.
[0129] By setting these additional conditions for UE100, the frequent sending of availability notifications by UE100 can be suppressed.
[0130] Figure 12 shows a second example of the suppression operation of availability notifications according to this embodiment. Here, it is mainly assumed that the availability notification is a notification of a low power state, but the availability notification may also be a notification indicating that the buffer condition has been met.
[0131] In step S131, NW1 transmits the data acquisition settings to UE100. UE100 receives the data acquisition settings. In this example, the data acquisition settings may include information for setting power conditions.
[0132] Furthermore, in this example, the data collection settings may include at least one of the following additional conditions.
[0133] - Power condition A that prevents the transmission of availability notifications Power condition A may be a power condition that causes UE100 to stop or prohibit such transmission, on the premise that UE100 can repeatedly or periodically transmit availability notifications after the trigger condition is met.
[0134] For example, UE100 may recognize a low-power state when its power (voltage) or remaining power is 40% (threshold 1) and repeatedly or periodically send availability notifications to NW1. Threshold 1 may be set by NW1 to UE100. Under such assumptions, 70% (threshold 2) may be set as power condition A. When UE100's power (voltage) or remaining power reaches 70% (threshold 2), it recognizes that it is not possible to send availability notifications and stops sending them.
[0135] Alternatively, power condition A may represent an offset value (hysteresis value) relative to threshold 1. For example, UE100 recognizes a low-power state when its power (voltage) or remaining time is 40% (threshold 1) and sends an availability notification to NW1. Threshold 1 may be set from NW1 to UE100. Under such assumptions, power condition A may be set to "-10%", i.e., 30% (threshold 2). UE100 recognizes that it cannot send availability notifications until its power (voltage) or remaining time reaches 30% (threshold 2) and stops sending availability notifications. When UE100's power (voltage) or remaining time reaches 30% (threshold 2), it sends an availability notification to NW1.
[0136] - Power condition B that sends another notification indicating that the low-power state is over Power condition B may be a power condition that causes UE100 to send a notification to NW1 indicating that it has emerged from a low-power state. Power condition B may be an offset value (hysteresis value) with respect to threshold 1. For example, UE100 recognizes a low-power state when the power (voltage) or remaining time is 40% (threshold 1) and sends an availability notification to NW1. Threshold 1 may be set from NW1 to UE100. Under such assumptions, power condition B may be set to "+20%", i.e., 60% (threshold 2). UE100 recognizes that it cannot send notifications until its power (voltage) or remaining time reaches 60% (threshold 2) and stops sending notifications. When UE100's power (voltage) or remaining time reaches 60% (threshold 2), it sends a notification to NW1.
[0137] - Time condition indicating the duration for which the trigger condition must be continuously met in order to send an availability notification The time condition indicates the duration for which the trigger condition (e.g., low power state) must be continuously met, and a time threshold such as 5 minutes is set. For example, if UE100 recognizes a low power state when the power (voltage) or remaining time is 40% (threshold 1), and this state continues for 5 minutes, it will send an availability notification to NW1. In other words, even if UE100 reaches 40% (threshold 1) of power (voltage) or remaining time, it will not send an availability notification to NW1 until the set time condition (duration) is met.
[0138] Furthermore, the data collection settings may include the AI / ML function identifier described later, and additional conditions may be associated with the AI / ML function identifier. That is, UE100 may apply independent additional conditions for each AI / ML function. Alternatively, UE100 may apply additional conditions common to multiple AI / ML functions.
[0139] In step S132, UE100 performs data collection to collect training data to be used for model training, based on the received data collection settings.
[0140] In step S133, UE100 determines whether the trigger condition for availability notification set by NW1 (e.g., low power state) has been met (i.e., trigger evaluation). If the trigger condition for availability notification has not been met (step S133: NO), UE100 returns to step S132.
[0141] On the other hand, if the trigger condition for availability notification is met (step S133: YES) and power condition A or power condition B is set, in step S134, UE100 may send an availability notification to NW1.
[0142] In step S135, UE100 determines whether the set additional conditions have been met.
[0143] If power condition A or power condition B is set as an additional condition, UE100 will not send a notification until power condition A or power condition B is met. If power condition B is set, UE100 may send a notification to NW1 indicating that it has exited the low-power state when power condition B is met (step S136).
[0144] If a time condition is set as an additional condition, UE100 will not send an availability notification until the time condition is met. If a time condition is set, UE100 may send an availability notification to NW1 when the time condition is met (step S136).
[0145] (3.3.3) Third Operation Example The third operation example of the availability notification suppression operation according to this embodiment may be performed in combination with the first operation example or the second operation example described above, or in combination with the fourth operation example or the fifth operation example described later.
[0146] In the third operational example, if NW1 (e.g., base station 200) has configured UE100 not to suppress the transmission of availability notifications, UE100 will send an availability notification to NW1 (e.g., base station 200) when the trigger condition is met. For example, if NW1 is not congested and has ample resources, NW1 may configure UE100 not to suppress the transmission of availability notifications. In this case, UE100 may recognize that the trigger condition has been determined to be UE implementation-dependent and send an availability notification to NW1 at any time.
[0147] A setting indicating that the transmission of availability notifications should not be suppressed may be made in the data collection settings transmitted from NW1 to UE100. On the other hand, if no setting indicating that the transmission of availability notifications should not be suppressed is made, UE100 will recognize that the transmission of availability notifications should be suppressed and may perform the operation of the first or second operation example described above, or the operation of the fourth or fifth operation example described later. The setting of which operation example is applied may be made in the data collection settings transmitted from NW1 to UE100.
[0148] (3.3.4) Fourth Operation Example In the fourth operation example of the availability notification suppression operation according to this embodiment, the UE 100 has a plurality of trigger conditions set by the NW1 (e.g., base station 200) to cause it to send availability notifications in stages. When any of the plurality of trigger conditions is met, the UE 100 sends an availability notification to the NW1 (e.g., base station 200). This prevents the UE 100 from sending availability notifications multiple times at arbitrary timings, and suppresses frequent sending of availability notifications by the UE 100.
[0149] UE100 may send an availability notification to NW1 (e.g., base station 200) containing information indicating which trigger condition was met if any of the configured trigger conditions are met. This allows NW1 to know which trigger condition has been met.
[0150] Figure 13 shows a fourth example of the suppression operation of availability notifications according to this embodiment.
[0151] In step S141, NW1 transmits the data acquisition settings to UE100. UE100 receives the data acquisition settings. In this example, the data acquisition settings may include information to set buffer conditions (e.g., full buffer conditions and / or buffer threshold conditions) and / or power conditions as trigger conditions.
[0152] Furthermore, in this example, the data collection settings include information that configures multiple trigger conditions for sending availability notifications in stages.
[0153] Multiple trigger conditions may also be multiple buffer conditions. For example, the following settings may be made: - Execute a low power notification with priority 4 when the buffer is 50% occupied - Execute a low power notification with priority 3 when the buffer is 70% occupied - Execute a low power notification with priority 2 when the buffer is 90% occupied - Execute a low power notification with priority 1 when the buffer is full (100%) These buffer thresholds and priority mappings may be set from NW1 to UE100.
[0154] Multiple trigger conditions may also be multiple power conditions. For example, the following settings may be made: - Execute low power notification with priority 3 at 50% power - Execute low power notification with priority 2 at 40% power - Execute low power notification with priority 1 at 30% power These power thresholds and priority mappings may be set from NW1 to UE100.
[0155] Note that "priority" may be interpreted as "level" or "stage."
[0156] The data collection settings include the AI / ML function identifier described later, and multiple trigger conditions may be associated with the AI / ML function identifier. That is, UE100 may apply multiple trigger conditions for each AI / ML function. Alternatively, UE100 may apply multiple trigger conditions common to multiple AI / ML functions.
[0157] In step S142, UE100 performs data collection to collect training data to be used for model training, based on the received data collection settings.
[0158] In step S143, UE100 determines whether any of the multiple trigger conditions for availability notification set by NW1 have been met (i.e., trigger evaluation). If the trigger conditions for availability notification have not been met (step S143: NO), UE100 returns to step S142.
[0159] On the other hand, if any trigger condition is met (step S143: YES), in step S144, UE100 sends an availability notification to NW1. The availability notification may include information indicating the priority corresponding to the met trigger condition. NW1 receives the availability notification. Subsequently, UE100 returns to step S142.
[0160] (3.3.5) Fifth Operation Example In the fifth operation example of the suppression operation of availability notification according to this embodiment, UE100 receives a request for confirmation (so-called polling) from NW1 (e.g., base station 200) regarding the amount of training data collected or the power status of UE100 after sending an availability notification. In response to receiving the request for confirmation, UE100 sends a notification to NW1 (e.g., base station 200) regarding the amount of training data collected or the power status of UE100.
[0161] In other words, in the fifth operational example, after sending an availability notification once (for example, the initial transmission), UE100 will send notifications in response to polling from NW1, and will not send the notification to NW1 if there is no polling. This prevents UE100 from sending availability notifications multiple times at arbitrary times, thereby suppressing frequent sending of availability notifications by UE100.
[0162] Figure 14 shows a fifth example of the suppression operation of availability notifications according to this embodiment.
[0163] In step S151, NW1 transmits the data acquisition settings to UE100. UE100 receives the data acquisition settings. In this example, the data acquisition settings may include information to set buffer conditions (e.g., full buffer conditions and / or buffer threshold conditions) and / or power conditions as trigger conditions.
[0164] In this example, the data collection settings may include polling settings that configure the system to send an availability notification once (for example, the initial transmission) and then send notifications in response to polls from NW1.
[0165] The data collection settings include the AI / ML function identifier described later, and the polling settings may be associated with the AI / ML function identifier. That is, UE100 may apply polling to each AI / ML function. Alternatively, UE100 may apply polling to multiple AI / ML functions in common.
[0166] In step S152, UE100 performs data collection to collect training data to be used for model training, based on the received data collection settings.
[0167] In step S153, UE100 determines whether the trigger conditions for availability notification set by NW1 have been met (i.e., performs a trigger evaluation). If the trigger conditions for availability notification have not been met (step S153: NO), UE100 returns to step S152.
[0168] On the other hand, if the trigger condition is met (step S153: YES), in step S154, UE100 sends an availability notification to NW1. NW1 receives the availability notification.
[0169] In step S155, NW1 sends an acknowledgment request to UE100. UE100 receives the acknowledgment request.
[0170] If NW1 receives an availability notification regarding buffer conditions in step S154, it may send a confirmation request to UE100 regarding the amount of training data collected (i.e., buffer state). On the other hand, if NW1 receives an availability notification regarding power conditions in step S154, it may send a confirmation request to UE100 regarding the power state of UE100. The confirmation request may include information indicating whether the item to be confirmed is in a buffer state, a power state, or both a buffer state and a power state.
[0171] In step S156, upon receiving the acknowledgment request, UE100 sends a notification to NW1 indicating the latest buffer state and / or latest power state. NW1 receives the notification. The notification may be a type of availability notification. The latest buffer state and / or latest power state may be expressed by priority as described in the fourth example of operation above.
[0172] In step S157, NW1 may send a transmission request for the collected training data to UE100.
[0173] In step S158, UE100 may, in response to a transmission request, send a message containing the collected training data to NW1.
[0174] In step S159, NW1 may perform model training using the acquired training data.
[0175] In step S160, NW1 may send a data collection stop instruction to UE100 to stop (pause) data collection. NW1 may also send a data collection end or cancellation instruction to UE100. These instructions may be included in an RRC message (for example, an RRC Reconfiguration message) sent from NW1 to UE100.
[0176] (4) Data collection operation of multiple AI / ML Next, the data collection operation of multiple AI / ML according to this embodiment will be explained, mainly in terms of the differences from the configuration and operation described above.
[0177] In the following embodiment, we will describe the operation assuming the use of multiple AI / ML models 5 (multiple AI / ML functions) in the case of the NW side model. The data collection operation of the multiple AI / ML according to this embodiment may be performed in combination with or without the availability notification suppression operation described above.
[0178] (4.1) Operation Scenario and Operation Overview Diagram 15 is a diagram showing the operation scenario related to the data collection operation of multiple AI / ML according to this embodiment.
[0179] In the illustrated example, NW1 has multiple AI / ML models 5 (AI / ML model 5 #1, #2, #3, etc.). Each AI / ML model 5 is associated with one of the AI / ML functions. Note that #1, #2, #3, etc. may each represent an identifier for an AI / ML model 5 or an AI / ML function.
[0180] The flow of operations in Step S1 (data acquisition setup), Step S2 (data acquisition), Step S3 (data acquisition procedure), and Step S4 (model training) is the same as described above.
[0181] Since the content (type) of training data required differs for each AI / ML function, it is desirable to be able to identify which AI / ML function is collecting or has collected training data for in the operation scenario shown in Figure 15. Furthermore, if UE100 does not collect data for only one of the multiple AI / ML functions, there is a problem in that efficient and sufficient data collection cannot be performed. The operation to solve these problems will be described below.
[0182] Figure 16 is a diagram showing an overview of the operation of UE100 regarding the data collection operation of multiple AI / ML according to this embodiment.
[0183] In step S201, UE100 collects training data to be used for training the AI / ML model 5 on the NW1 side (data collection). In this example, UE100 may collect data for each of the multiple AI / ML functions. The collected training data may differ depending on the target AI / ML function. UE100 stores the collected training data in its own memory area (area within memory).
[0184] In step S202, UE100 sends an availability notification regarding the availability of the training data to NW1 (for example, base station 200). The trigger conditions for sending the availability notification are the same as those described above.
[0185] In this example, the availability notification regarding the availability of training data may include an AI / ML function identifier indicating the AI / ML function corresponding to the training data. This allows NW1 (for example, base station 200) to determine which AI / ML function the training data held by UE100 belongs to, based on the AI / ML function identifier included in the availability notification.
[0186] The term "AI / ML function identifier" may be interpreted as "AI / ML model identifier," "model identifier," or "AI / ML identifier." The term "identifier" may be interpreted as "ID," "number," or "index."
[0187] In step S203, UE100 receives a transmission request for training data from NW1 (for example, base station 200).
[0188] In this example, a transmission request for training data may include an AI / ML function identifier indicating the AI / ML function corresponding to that training data. This allows UE100 to determine which AI / ML function's training data is being requested based on the AI / ML function identifier included in the transmission request.
[0189] In step S204, UE100 sends a message containing training data to NW1 (e.g., base station 200) in response to a transmission request. Here, if the transmission request includes an AI / ML function identifier, UE100 sends a message to NW1 containing training data corresponding to that AI / ML function identifier.
[0190] In this example, a message containing training data may also include an AI / ML function identifier indicating the AI / ML function corresponding to that training data. This allows NW1 (for example, base station 200) to determine which AI / ML function the training data contained in the message belongs to, based on the AI / ML function identifier included in the message.
[0191] Thus, according to the data collection operation of multiple AI / MLs in this embodiment, at least one of the availability notification, transmission request, and message includes an AI / ML function identifier indicating the AI / ML function corresponding to the training data. This makes it possible to identify which AI / ML function is collecting or has collected training data. Furthermore, the UE 100 can perform data collection for multiple AI / ML functions in parallel, making it easier to perform efficient and sufficient data collection.
[0192] The UE 100 performing this operation includes a control unit 130 that collects training data used for training the AI / ML model 5 on the NW1 side, a transmission unit 120 that sends availability notifications regarding the availability of training data to NW1 (e.g., base station 200), and a reception unit 110 that receives transmission requests for training data from NW1 (e.g., base station 200) (see Figure 4). In response to a transmission request, the transmission unit 120 sends a message containing the training data to NW1 (e.g., base station 200). At least one of the availability notification, transmission request, and message includes an AI / ML function identifier indicating the AI / ML function corresponding to the training data.
[0193] On the other hand, NW1 (for example, base station 200) has a receiving unit 220 that receives availability notifications regarding the availability of training data from UE 100, which collects training data used for training the AI / ML model 5 on the NW1 side, and a transmitting unit 210 that sends a transmission request for training data to UE 100 (see Figure 5). The receiving unit 220 receives a message sent from UE 100 in response to the transmission request, which includes the training data. At least one of the availability notification, the transmission request, and the message includes an AI / ML function identifier indicating the AI / ML function corresponding to the training data.
[0194] The AI / ML function identifier may be set (notified) to the UE 100 by the data collection settings from NW1. In this case, the UE 100 receives a setting message (data collection setting) regarding the settings for collecting training data from NW1 (for example, base station 200), and this setting message includes the AI / ML function identifier. In this way, the AI / ML function identifier may be set (notified) to the UE 100. Such an AI / ML function identifier may be an Associated ID, which is an ID indicating the training target.
[0195] The AI / ML function identifier may be an identifier assigned to uniquely identify and manage the AI / ML model 5, related data, and inference processing, etc., within the NW1. The AI / ML function identifier may also be managed by the AI / ML management function 4.
[0196] The AI / ML function identifier may also be the CSI-Report Config ID in beam management. The CSI-Report Config ID may also be an identifier that uniquely identifies which report config the UE100 will use to report channel status information (CSI) to the NW1 (base station 200).
[0197] The correspondence between AI / ML function identifiers and AI / ML functions may be predefined in the technical specifications of the mobile communication system.
[0198] The AI / ML function identifier may be transmitted from the AI / ML management function 4 to RAN2 (base station 200). For example, when the AI / ML model 5 is transferred from the AI / ML management function 4 to RAN2 (base station 200), the AI / ML function identifier may be assigned to the transferred AI / ML model 5.
[0199] The AI / ML function identifier may be transmitted from RAN2 (base station 200) to the AI / ML management function 4. For example, when RAN2 (base station 200) transfers training data acquired to the AI / ML management function 4, the AI / ML function identifier may be assigned to the AI / ML model 5 being transferred.
[0200] (4.2) Method for managing memory area Next, a method for managing memory area related to the data collection operation of multiple AI / ML according to this embodiment will be described.
[0201] The UE100 collects training data for each of its multiple AI / ML functions. The UE100 stores the collected training data in its memory, associating it with the corresponding AI / ML function identifier. This allows the training data to be managed separately for each AI / ML function (AI / ML function identifier).
[0202] The number of memory areas for storing training data may be one. This case will be described later as the first method of managing memory areas. Alternatively, the number of memory areas for storing training data may be multiple. This case will be described later as the second method of managing memory areas.
[0203] UE100 may receive configuration information from base station 200 that specifies at least one of the number of memory areas for storing training data and the size of the memory areas. This configuration information may be included in the data acquisition settings. That is, NW1 (for example, base station 200) may transmit configuration information to UE100 that specifies at least one of the number of memory areas for storing training data and the size of the memory areas. This allows at least one of the number of memory areas for storing training data and the size of the memory areas to be set variably.
[0204] At least one of the number of memory areas for storing training data and the size of the memory areas may be predefined in the technical specifications of the mobile communication system. In this case, at least one of the number of memory areas for storing training data and the size of the memory areas is fixed.
[0205] (4.2.1) First Management Method Figure 17 is a diagram illustrating the first management method of the storage area according to this embodiment.
[0206] In the first management method, the UE100 has a single memory area common to multiple AI / ML functions as a memory area for storing collected training data. The UE100 stores each of the collected training data and its corresponding AI / ML function identifier (AI / ML function ID) in this single memory area. This allows the UE100 to manage training data for multiple AI / ML functions in a single memory area.
[0207] In the illustrated example, UE100 stores combinations of AI / ML function IDs and training data (Data1, Data2, ...) in a single memory area. As a result, for each of the multiple AI / ML function IDs, the corresponding training data (Data1, Data2, ...) is stored in a single memory area.
[0208] Regarding the buffer conditions described above, "full buffer" may mean a state in which the single memory area is completely full. Also, regarding the buffer conditions described above, the buffer threshold may be a data volume threshold set for the single memory area.
[0209] Alternatively, if UE100 counts the amount of data stored in the memory area for each of the multiple AI / ML functions, the buffer threshold may be set individually for each AI / ML function. In this case, one or more sets of AI / ML function identifiers and buffer thresholds may be included in the data acquisition settings. Based on these sets, UE100 may set the buffer threshold individually for each AI / ML function.
[0210] (4.2.2) Second Management Method Figure 18 is a diagram illustrating the second management method of the storage area according to this embodiment.
[0211] In the second management method, the memory area for storing training data consists of multiple memory areas (memory area #1, #2, ...) corresponding to multiple AI / ML functions. Each of the multiple memory areas is associated with a corresponding AI / ML function identifier (AI / ML function ID). The UE100 stores each of the collected training data in the corresponding memory area among the multiple memory areas. This eliminates the need to store the combination of collected training data and AI / ML function identifier in the memory area, thus streamlining data management.
[0212] In the illustrated example, UE100 has memory area #1 corresponding to AI / ML function ID = 1, memory area #2 corresponding to AI / ML function ID = 2, and so on. UE100 stores the training data collected for the AI / ML function with AI / ML function ID = 1 in memory area #1. Similarly, UE100 stores the training data collected for the AI / ML function with AI / ML function ID = 2 in memory area #2.
[0213] Regarding the buffer conditions described above, "full buffer" may mean a state where any one storage area is full, or a state where all storage areas are full. Also, regarding the buffer conditions described above, the buffer threshold may be a data volume threshold set for any one storage area, or a data volume threshold set for all storage areas.
[0214] The buffer threshold may be set individually for each storage area. In this case, one or more sets of AI / ML function identifiers and buffer thresholds may be included in the data acquisition settings. Based on these sets, the UE100 may set the buffer threshold individually for each storage area.
[0215] (4.3) Operation Example Next, an example of the operation of data collection operation of multiple AI / ML according to this embodiment will be described.
[0216] (4.3.1) Figure 19, the first example of operation, is a diagram showing an example of the operation of data collection operation of multiple AI / ML when a buffer condition is used as a trigger condition in the mobile communication system according to this embodiment. For operations similar to those shown in Figure 8, redundant explanations will be omitted.
[0217] In step S211, NW1 transmits the data collection settings to UE100. UE100 receives the data collection settings. In this example, the data collection settings may include information to set buffer conditions as trigger conditions. Also in this example, the data collection settings include the AI / ML function identifier of the target AI / ML function. The data collection settings may include multiple AI / ML function identifiers corresponding to multiple AI / ML functions. Based on the AI / ML function identifiers included in the data collection settings, UE100 can determine which AI / ML function should collect training data for.
[0218] In step S212, UE100 stores and applies the received data collection settings and sends a setting completion message to NW1 indicating the completion of the data collection settings. NW1 receives the setting completion message. In this example, the setting completion message includes the AI / ML function identifier for each AI / ML function for which the data collection settings have been completed. This allows NW1 to identify the AI / ML functions for which the data collection settings have been completed.
[0219] In step S213, UE100 performs data collection to collect training data to be used for model training, based on the data collection settings. In this example, UE100 performs data collection for the AI / ML functions indicated by the AI / ML function identifiers included in the data collection settings. If multiple AI / ML function identifiers are included in the data collection settings, UE100 performs data collection for each AI / ML function. UE100 stores the collected training data in its own memory area.
[0220] In step S214, UE100 determines whether the collected training data has reached a predetermined amount (e.g., a full buffer or buffer threshold) (i.e., performs a trigger evaluation). If the collected training data has not reached a predetermined amount (step S214: NO), UE100 returns to step S213.
[0221] On the other hand, if the collected training data reaches a predetermined amount (step S214: YES), in step S215, UE100 sends an availability notification to NW1 that includes an AI / ML function identifier. NW1 receives the availability notification. For example, UE100 may identify one or more AI / ML functions whose buffer conditions (full buffer or buffer threshold) have been met, and include the AI / ML function identifier of each identified AI / ML function in the availability notification. This allows NW1 to understand which AI / ML functions have met the buffer conditions.
[0222] In step S216, NW1 sends a transmission request for the collected training data to UE100. UE100 receives the transmission request. In this example, the transmission request includes the AI / ML function identifier for each AI / ML function for which the transmission of training data is requested. For example, NW1 may include the AI / ML function identifier for each AI / ML function whose buffer condition is met in the transmission request.
[0223] In step S217, UE100 sends a message containing the collected training data to NW1 in response to a transmission request. NW1 acquires the training data by receiving the message. In this example, the message includes an AI / ML function identifier associated with the training data contained in the message. If the message contains training data for multiple AI / ML functions, the message includes the AI / ML function identifier for each of those multiple AI / ML functions. This allows NW1 to acquire the training data separately for each AI / ML function, even if training data for multiple AI / ML functions is transmitted in a single message.
[0224] In step S218, NW1 performs model training using the acquired training data. In this example, if training data for multiple AI / ML functions is included in the message in step S217, NW1 uses the training data for each AI / ML function to train the corresponding AI / ML model 5.
[0225] (4.3.2) Figure 20, the second example of operation, is a diagram showing an example of the operation of data acquisition operation of multiple AI / ML when a power condition is used as a trigger condition in the mobile communication system according to this embodiment. For operations similar to those shown in Figure 9 and operations similar to those shown in Figure 19, redundant explanations will be omitted.
[0226] In step S221, NW1 transmits the data acquisition settings to UE100. UE100 receives the data acquisition settings. In this example, the data acquisition settings may include information that sets the power conditions as trigger conditions. Also in this example, the data acquisition settings include the AI / ML function identifier of the target AI / ML function. The data acquisition settings may include multiple AI / ML function identifiers corresponding to multiple AI / ML functions.
[0227] In step S222, UE100 stores and applies the received data collection settings and sends a setting completion message to NW1 indicating the completion of the data collection settings. NW1 receives the setting completion message. In this example, the setting completion message includes the AI / ML function identifier for each AI / ML function for which the data collection settings have been completed.
[0228] In step S223, UE100 performs data collection to collect training data to be used for model training, based on the received data collection settings. In this example, UE100 performs data collection for the AI / ML functions indicated by the AI / ML function identifiers included in the data collection settings. If multiple AI / ML function identifiers are included in the data collection settings, UE100 performs data collection for each AI / ML function. UE100 stores the collected training data in its own memory area.
[0229] In step S224, UE100 determines whether its power state has entered a low-power state (i.e., performs a trigger evaluation). If it has not entered a low-power state (step S224: NO), UE100 returns to step S223.
[0230] On the other hand, if a low-power state is entered (step S224: YES), in step S225, UE100 sends an availability notification to NW1 that includes information indicating that a low-power state is in place. NW1 receives the availability notification. For example, UE100 may include the AI / ML function identifier of each AI / ML function from which training data has been collected in the availability notification.
[0231] In step S226, NW1 may send a transmission request for the collected training data to UE100. In this example, the transmission request includes the AI / ML function identifier for each AI / ML function for which the transmission of training data is requested.
[0232] In step S227, UE100 may send a message containing the collected training data to NW1 in response to a transmission request. In this example, the message includes an AI / ML function identifier associated with the training data contained in the message. If the message contains training data for multiple AI / ML functions, the message includes the AI / ML function identifier for each of those multiple AI / ML functions.
[0233] In step S228, NW1 may perform model training using the acquired training data. In this example, if training data for multiple AI / ML functions is included in the message in step S227, NW1 will train the corresponding AI / ML model 5 using the training data for each AI / ML function.
[0234] In step S229, NW1 sends a data collection stop instruction to UE100, which instructs the UE100 to stop (pause) data collection. UE100 receives the data collection stop instruction. NW1 may also send a data collection end or cancellation instruction to UE100. In this case, the operations from step S231 onward may be omitted. In this example, the instruction in step S229 includes the AI / ML function identifier of each AI / ML function for which data collection stop, end, or cancellation is requested.
[0235] In step S230, UE100 stops data collection in response to receiving the instruction in step S229. UE100 may also stop data collection for the AI / ML function indicated by the AI / ML function identifier included in the instruction.
[0236] In step S231, UE100 may determine whether or not it has exited the low-power state. If it has not exited the low-power state (step S231: NO), UE100 returns to step S230.
[0237] On the other hand, if the system exits the low-power state (step S231: YES), in step S232, UE100 may send a notification to NW1 indicating that the system has exited the low-power state. This notification may be a type of availability notification. This notification may include an AI / ML function identifier.
[0238] In step S233, NW1 may send an instruction to UE100 to resume data collection in response to receiving the notification in step S232. This instruction may include the AI / ML function identifier of the AI / ML function for which data collection is requested to be resumed.
[0239] In step S234, UE100 may resume data collection in response to receiving the instruction in step S233. In this case, UE100 may return to step S223.
[0240] This example illustrates the inclusion of AI / ML function identifiers in availability notifications, transmission requests, messages, etc. However, in low-power conditions, it may be necessary for NW1 to quickly acquire all training data from UE100. Therefore, in low-power conditions, the AI / ML function identifier may be omitted.
[0241] (5) Other Embodiments In the embodiments described above, an example was given in which the signals (signaling) transmitted and received between UE100 and NW1 are mainly RRC messages. However, at least one signal (signaling) transmitted and received between UE100 and NW1 may be MAC-CE, UCI, DCI, AI / ML layer messages, LPP (LTE Positioning Protocol) messages, NRPPa (NR Positioning Protocol A) messages, NAS messages, TCP / IP, etc.
[0242] Capability information indicating that it supports the operation of the UE 100 according to the above embodiment may be transmitted from the UE 100 to the base station 200. This capability information may be transmitted in an RRC message (for example, a UE Capability message).
[0243] Furthermore, the base station 200 may transmit configuration information for setting the operation of the UE 100 according to the above embodiment to the UE 100. This configuration information may be transmitted as an RRC message. This RRC message may be a dedicated RRC signaling message for each UE (e.g., an RRC Reconfiguration message) or a broadcast signaling message (e.g., SIB).
[0244] The above-described operation flows can be performed not only independently, but also in combination of two or more operation flows. For example, some steps of one operation flow may be added to another operation flow, or some steps of one operation flow may be replaced with some steps of another operation flow. It is not necessary to execute all steps in each flow; only some steps may be executed. Furthermore, the order of steps in each flow may be changed as appropriate.
[0245] In the embodiments and examples described above, the base station may be an NR base station (gNB) or a 6G base station. Alternatively, the base station may be a relay node such as an IAB (Integrated Access and Backhaul) node. The base station may also be a DU of an IAB node. Furthermore, UE100 may be an MT (Mobile Termination) of an IAB node. That is, UE100 may be a terminal function unit (a type of communication module) for the base station to control a repeater that performs signal relay. Such a terminal function unit is called an MT. Examples of MTs other than IAB-MT include, for example, NCR (Network Controlled Repeater)-MT and RIS (Reconfigurable Intelligent Surface)-MT.
[0246] Furthermore, the term "network node" primarily refers to a base station, but may also refer to a core network device or a part of a base station (CU, DU, or RU). Additionally, a network node may consist of a combination of at least a part of the core network device and at least a part of a base station.
[0247] A program may be provided that causes a computer to execute each process performed by the UE 100 or base station 200. The program may be recorded on a computer-readable medium. Using a computer-readable medium, it is possible to install the program on a computer. Here, the computer-readable medium on which the program is recorded may be a non-transient recording medium. The non-transient recording medium is not particularly limited, but may be a recording medium such as a CD-ROM or DVD-ROM. Alternatively, the circuits that execute each process performed by the UE 100 or base station 200 may be integrated, and at least a part of the UE 100 or base station 200 may be configured as a semiconductor integrated circuit (chipset, SoC: System on a chip).
[0248] The functions realized by UE100 or base station 200 may be implemented in circuitry or processing circuitry, including general-purpose processors, application-specific processors, integrated circuits, ASICs (Application Specific Integrated Circuits), CPUs (Central Processing Units), conventional circuits, and / or combinations thereof, programmed to realize the described functions. A processor, including transistors and other circuits, is considered circuitry or processing circuitry. A processor may be a programmed processor that executes a program stored in memory. In this specification, circuitry, unit, and means are hardware programmed to realize or perform the described functions. Such hardware may be any hardware disclosed herein, or any hardware known to be programmed to realize or perform the described functions. If such hardware is a processor that is considered to be of the type of circuitry, then such circuitry, means, or unit is a combination of hardware and software used to constitute such hardware and / or processor.
[0249] The phrases “based on” and “depending on / in response to” used in this disclosure do not mean “based solely on” or “depending solely on” unless otherwise specified. The phrase “based on” means both “based solely on” and “at least partially on.” Similarly, the phrase “depending on” means both “at least partially on” and “at least partially on.” The terms “include,” “comprise,” and variations thereof do not mean that only the listed items are included, but that they may include only the listed items or that they may include additional items in addition to the listed items. Furthermore, the term “or” used in this disclosure is not intended to mean exclusive OR. In addition, any reference to elements using designations such as “first,” “second,” etc., used in this disclosure does not generally limit the quantity or order of those elements. These designations may be used herein as a convenient way to distinguish between two or more elements. Thus, references to first and second elements do not mean that only two elements may be employed therein, or that the first element must precede the second element in any way. In this disclosure, if articles are added by translation, such as a, an, and the in English, these articles shall be plural unless it is clearly indicated otherwise from the context.
[0250] Although the embodiments have been described in detail above with reference to the drawings, the specific configuration is not limited to those described above, and various design changes can be made without departing from the gist of the invention.
[0251] This application claims priority to U.S. Provisional Application No. 63 / 800841 (filed May 6, 2025), the entirety of which is incorporated into the specification of this application.
[0252] (6) Additional notes: Features of the above-described embodiments are noted below.
[0253] - Appendix 1 A communication method performed by a user device in a mobile communication system, comprising: collecting training data used for training an artificial intelligence or machine learning (AI / ML) model on the network side; conditions for sending a notification regarding the availability of the training data to a network node, which include evaluating whether a trigger condition based on the amount of the collected training data or the power status of the user device has been met; and suppressing the transmission of the notification based on a setting or request from the network node.
[0254] - Appendix 2 The communication method described in Appendix 1, wherein when the user device sends a notification regarding the availability of the training data, the user device starts a timer set by the network node, and the user device is prohibited from sending the notification while the timer is operating.
[0255] - Appendix 3 The notification is a notification indicating that the user device is in a low-power state, and the communication method according to Appendix 1 or 2, wherein the network node sets at least one of the following conditions: a power condition that prevents the transmission of the notification, a power condition that enables the transmission of another notification indicating that the low-power state has been discontinued, and a time condition that indicates the duration for which the trigger condition should be continuously met in order to transmit the notification.
[0256] - Appendix 4: If the network node has configured the user device not to suppress the transmission of the notification, the communication method according to any one of Appendix 1 to 3, wherein the user device transmits the notification to the network node in response to the trigger condition being met.
[0257] - Appendix 5 The communication method according to any one of Appendix 1 to 4, wherein the user device has a plurality of trigger conditions set by the network node for sending the notification in stages, and the user device sends the notification to the network node when any of the plurality of trigger conditions is met.
[0258] - Appendix 6 The communication method according to any one of Appendix 1 to 5, wherein the user device receives a request for confirmation from the network node regarding the amount of training data collected or the power status of the user device after sending the notification, and in response to receiving the request for confirmation, transmits a notification to the network node regarding the amount of training data collected or the power status of the user device.
[0259] - Appendix 7 The communication method according to any one of the appendices 1 to 6, wherein the user device includes the notification in any of the connection-related messages, capability notification messages, measurement report messages, AI / ML messages, or Layer 1 or Layer 2 signaling messages transmitted to the network node.
[0260] - Appendix 8 A user device used in a mobile communication system, having a control unit that collects training data used for training artificial intelligence or machine learning (AI / ML) models on the network side, wherein the control unit evaluates whether a trigger condition based on the amount of collected training data or the power status of the user device has been met, which is a condition for sending a notification regarding the availability of the training data to a network node, and suppresses the transmission of the notification based on a setting or request from the network node.
[0261] - Appendix 9 A network node used in a mobile communication system, comprising: a receiving unit that receives a notification transmitted from a user device that collects training data used for training artificial intelligence or machine learning (AI / ML) models on the network side, in response to the fulfillment of a trigger condition based on the amount of training data collected or the power status of the user device; and a transmitting unit that transmits a setting or request to the user device to suppress the transmission of the notification.
[0262] 1: NW 2: RAN 3: CN 4: AI / ML management function 5: AI / ML model 100: UE 110: Receiving unit 120: Transmitting unit 130: Control unit 140: Wireless communication unit 200: Base station 210: Transmitting unit 220: Receiving unit 230: Control unit 240: NW communication unit 250: Wireless communication unit 300: CN device
Claims
1. A communication method performed by a user device in a mobile communication system, comprising: the user device transmitting a notification to a network containing information indicating a low power state; the user device receiving a data collection stop instruction from the network instructing the user device to stop collecting training data used for training an artificial intelligence or machine learning (AI / ML) model in response to transmitting the notification; and the user device stopping the collection of the training data in response to receiving the data collection stop instruction.
2. The communication method according to claim 1, further comprising: the user device receiving a data collection resumption notification from the network; and the user device resuming the collection of the training data in response to receiving the data collection resumption notification.
3. The communication method according to claim 1, further comprising the user device receiving configuration information from the network for transmitting the notification to the network, wherein the transmission includes the user device transmitting the notification to the network based on the configuration information.
4. User device in a mobile communication system, comprising: a transmitting unit that transmits a notification to a network containing information indicating a low power state; a receiving unit that receives a data collection stop instruction from the network in response to transmitting the notification, instructing the cessation of collection of training data used for training an artificial intelligence or machine learning (AI / ML) model; and a control unit that stops the collection of the training data in response to receiving the data collection stop instruction.
5. A mobile communication system having a user device, wherein the user device transmits a notification to a network containing information indicating a low power state, the user device receives a data collection stop instruction from the network in response to transmitting the notification, instructing the user device to stop collecting training data used for training an artificial intelligence or machine learning (AI / ML) model, and the user device stops collecting the training data in response to receiving the data collection stop instruction.
6. A chipset for a user device in a mobile communication system, comprising: transmitting a notification to a network that includes information indicating a low power state; receiving a data collection stop instruction from the network in response to the transmission of the notification, which instructs the stopping of the collection of training data used for training an artificial intelligence or machine learning (AI / ML) model; and stopping the collection of the training data in response to the receipt of the data collection stop instruction.
7. A program that causes a user device in a mobile communication system to perform the following steps: send a notification to the network containing information indicating a low power state to the user device; receive a data collection stop instruction from the network in response to the transmission of the notification, instructing the user to stop collecting training data used for training an artificial intelligence or machine learning (AI / ML) model; and stop collecting the training data in response to the receipt of the data collection stop instruction.
