Method for performing operation of managing performance of artificial intelligence model in wireless communication system, and device therefor

Abstract

The present disclosure relates to a technology for monitoring the performance of an artificial intelligence model configured in a terminal, and to a method and a device. The method comprises the steps of: generating a beam prediction result by using an artificial intelligence model configured in a terminal; receiving at least one reference signal including a first reference signal associated with a first beam selected according to the beam prediction result; determining whether an event occurs, by using at least one of a measurement result of the first reference signal, reliability information regarding the first beam, and measurement results of reference signals except for the first reference signal; and transmitting report information to a base station according to the determination of whether the event has occurred.

Description

Method and device for performing performance management operation of an artificial intelligence model in a wireless communication system

[0001] The present disclosure relates to a communication technology using an AI / ML model, and relates to a method and device for performing performance management operations of an artificial intelligence model in a wireless communication system.

[0002] The International Telecommunication Union (ITU) is working on standardizing next-generation 5th generation (5G) and 6th generation (6G) mobile communications through the International Mobile Telecommunication (IMT)-2020 and IMT-2030 programs.

[0003] The IMT-2020 and IMT-2030 projects are proposing technologies such as expanded numerology, large-scale antennas, and mmWave to meet these technical requirements. In particular, they are discussing the development of new technologies for enhanced 5G mobile communication technologies and the next-generation 6G mobile communication system. For 6G systems, technologies are being discussed that will advance wireless communication technologies and expand usability, including THz bands, AI-based networks, and ultra-precision positioning / sensing.

[0004] Specifically, 5G mobile communication technology defines a wide frequency band to enable fast transmission speeds and new services, and can be implemented not only in the sub-6GHz frequency band ('Sub 6GHz') such as 3.5 gigahertz (3.5GHz), but also in the ultra-high frequency band ('Above 6GHz') called millimeter wave (mmWave) such as 28GHz and 39GHz. In addition, in the case of 6G mobile communication technology, which is called the system after 5G communication (Beyond 5G), implementation in the terahertz band (for example, the 3 terahertz (3THz) band at 95GHz) is being considered to achieve a transmission speed that is 50 times faster than 5G mobile communication technology and an ultra-low latency time that is reduced to one-tenth.

[0005] In addition, the development of these 5G mobile communication systems includes new waveforms to ensure coverage in the terahertz band of 6G mobile communication technology, multi-antenna transmission technologies such as Full Dimensional MIMO (FD-MIMO), Array Antenna, and Large Scale Antenna, metamaterial-based lenses and antennas to improve the coverage of terahertz band signals, high-dimensional spatial multiplexing technology using Orbital Angular Momentum (OAM), Reconfigurable Intelligent Surface (RIS) technology, as well as full duplex technology to improve the frequency efficiency and system network of 6G mobile communication technology, satellite, AI (Artificial Intelligence) from the design stage and AI-based communication technology that realizes system optimization by internalizing end-to-end AI support functions, and ultra-high-performance communication and computing resources to provide services with complexity that exceeds the limits of terminal computing capabilities. It can serve as a basis for the development of next-generation distributed computing technologies that can be realized by utilizing them.

[0006] In this way, AI-based networking is expected to become more active in next-generation mobile communication systems, but a specific framework for this has not yet been presented.

[0007] Therefore, various communication technologies and methods are urgently needed in these AI-based enhanced network systems.

[0008] The present embodiments aim to provide a method and device for performing performance management operations of an artificial intelligence model in a wireless communication system.

[0009] In one aspect, the present embodiments may include a method for a terminal to perform a performance management operation of an artificial intelligence model, the method including: generating a beam prediction result using an artificial intelligence model configured in the terminal; receiving at least one reference signal including a first reference signal associated with a first beam selected according to the beam prediction result; determining whether an event has occurred using at least one of a measurement result of the first reference signal, reliability information for the first beam, and measurement results for reference signals excluding the first reference signal; and transmitting report information to a base station according to the determination of whether an event has occurred.

[0010] In another aspect, the present embodiments may provide a method in which a base station performs a performance management operation for an artificial intelligence model of a terminal, the method including: transmitting to the terminal at least one reference signal including a first reference signal associated with a first beam predicted using an artificial intelligence model configured in the terminal; receiving report information from the terminal based on a result of determining whether an event has occurred in the terminal using at least one of a measurement result of the first reference signal, reliability information for the first beam, and measurement results for reference signals excluding the first reference signal; and transmitting performance management instruction information for the artificial intelligence model based on the report information.

[0011] In another aspect, the present embodiments provide a terminal device that performs a performance management operation of an artificial intelligence model, the terminal device including a control unit that generates a beam prediction result using an artificial intelligence model configured in the terminal, and a receiving unit that receives at least one reference signal including a first reference signal associated with a first beam selected according to the beam prediction result, wherein the control unit determines whether an event has occurred using at least one of a measurement result of the first reference signal, reliability information for the first beam, and measurement results for reference signals excluding the first reference signal, and further includes a transmitting unit that transmits report information to a base station according to the determination of whether an event has occurred.

[0012] In another aspect, the present embodiments provide a base station that performs a performance management operation for an artificial intelligence model of a terminal, comprising: a transmitter that transmits to the terminal at least one reference signal including a first reference signal associated with a first beam predicted using an artificial intelligence model configured in the terminal; and a receiver that receives report information from the terminal based on a result of determining whether an event has occurred in the terminal using at least one of a measurement result of the first reference signal, reliability information for the first beam, and measurement results for reference signals excluding the first reference signal, wherein the transmitter may provide a base station device that transmits performance management instruction information for the artificial intelligence model based on the report information.

[0013] According to the present embodiments, it is possible to provide AI-based networking operations that can monitor the performance of AI / ML models for beam management and prevent performance degradation.

[0014] FIG. 1 is a diagram schematically illustrating the structure of an NR wireless communication system to which the present embodiment can be applied.

[0015] FIG. 2 is a drawing for explaining a frame structure in an NR system to which the present embodiment can be applied.

[0016] FIG. 3 is a diagram for explaining a resource grid supported by a wireless access technology to which the present embodiment can be applied.

[0017] FIG. 4 is a diagram for explaining a bandwidth part supported by a wireless access technology to which the present embodiment can be applied.

[0018] FIG. 5 is a diagram illustrating an example of a synchronization signal block in a wireless access technology to which the present embodiment can be applied.

[0019] FIG. 6 is a diagram for explaining a random access procedure in a wireless access technology to which the present embodiment can be applied.

[0020] Figure 7 is a drawing for explaining CORESET.

[0021] FIG. 8 is a diagram illustrating the structure of a functional framework for RAN Intelligence according to one embodiment.

[0022] FIGS. 9 and 10 are diagrams for explaining a beam prediction operation using a UE-side AI / ML model according to one embodiment.

[0023] FIG. 11 is a diagram for explaining a beam management operation according to one embodiment.

[0024] Fig. 12 is a drawing for explaining terminal operation according to one embodiment.

[0025] Fig. 13 is a diagram for explaining base station operation according to one embodiment.

[0026] FIG. 14 is a diagram for explaining an operation for triggering beam prediction performance monitoring of a terminal according to one embodiment.

[0027] FIG. 15 is a diagram for explaining the performance monitoring operation of an artificial intelligence model according to one embodiment.

[0028] Fig. 16 is a diagram for explaining the performance monitoring operation of an artificial intelligence model according to another embodiment.

[0029] Fig. 17 is a diagram for explaining the performance monitoring operation of an artificial intelligence model according to another embodiment.

[0030] Fig. 18 is a diagram for explaining the performance monitoring operation of an artificial intelligence model according to another embodiment.

[0031] Fig. 19 is a diagram showing a terminal configuration according to one embodiment.

[0032] Figure 20 is a diagram showing a base station configuration according to one embodiment.

[0033] Hereinafter, some embodiments of the present disclosure will be described in detail with reference to exemplary drawings. When adding reference numerals to components in each drawing, identical components may have the same numerals as much as possible even if they are shown in different drawings. In addition, when describing the present embodiments, if it is determined that a detailed description of a related known configuration or function may obscure the gist of the technical idea of ​​the present invention, the detailed description may be omitted. When "includes," "has," "consists of," etc. are used in this specification, other parts may be added unless "only" is used. When a component is expressed in the singular, it may include a case in which the plural is included unless specifically stated otherwise.

[0034] Additionally, terms such as first, second, A, B, (a), (b), etc. may be used to describe components of the present disclosure. These terms are only intended to distinguish the components from other components, and the nature, order, sequence, or number of the components are not limited by the terms.

[0035] In a description of the positional relationship of components, when it is described that two or more components are "connected," "combined," or "connected," it should be understood that the two or more components may be directly "connected," "combined," or "connected," but that the two or more components may also be further "interposed" with another component to be "connected," "combined," or "connected." Here, the other component may be included in one or more of the two or more components that are "connected," "combined," or "connected" to each other.

[0036] In the description of the temporal flow relationship related to components, operation methods, or manufacturing methods, for example, when the temporal or flow relationship is described as “after”, “following”, “next to”, “before”, etc., it may also include cases where it is not continuous, unless “immediately” or “directly” is used.

[0037] Meanwhile, when numerical values ​​or corresponding information (e.g., levels, etc.) for components are mentioned, even without separate explicit description, the numerical values ​​or corresponding information may be interpreted as including an error range that may occur due to various factors (e.g., process factors, internal or external impact, noise, etc.).

[0038] The wireless communication system in this specification refers to a system for providing various communication services such as voice, data packets, etc. using wireless resources, and may include a terminal, a base station, or a core network.

[0039] The embodiments disclosed below can be applied to wireless communication systems using various wireless access technologies. For example, the embodiments can be applied to various wireless access technologies such as CDMA (code division multiple access), FDMA (frequency division multiple access), TDMA (time division multiple access), OFDMA (orthogonal frequency division multiple access), SC-FDMA (single carrier frequency division multiple access), or NOMA (non-orthogonal multiple access). In addition, the wireless access technology may not only refer to a specific access technology, but also to each generation of communication technologies established by various communication agreement organizations such as 3GPP, 3GPP2, WiFi, Bluetooth, IEEE, and ITU. For example, CDMA can be implemented with wireless technologies such as UTRA (universal terrestrial radio access) or CDMA2000. TDMA can be implemented with wireless technologies such as GSM (global system for mobile communications) / GPRS (general packet radio service) / EDGE (enhanced data rates for GSM evolution). OFDMA can be implemented in wireless technologies such as IEEE (Institute of Electrical and Electronics Engineers) 802.11 (Wi-Fi), IEEE 802.16 (WiMAX), IEEE 802-20, and E-UTRA (evolved UTRA). IEEE 802.16m is an evolution of IEEE 802.16e and provides backward compatibility with systems based on IEEE 802.16e.UTRA is part of UMTS (universal mobile telecommunications system). 3GPP (3rd generation partnership project) LTE (long term evolution) is part of E-UMTS (evolved UMTS) that uses E-UTRA (evolved-UMTS terrestrial radio access), employing OFDMA in the downlink and SC-FDMA in the uplink. Thus, the present embodiments can be applied to currently disclosed or commercialized wireless access technologies, as well as wireless access technologies currently under development or to be developed in the future.

[0040] Meanwhile, the term "terminal" in this specification is a comprehensive concept meaning a device including a wireless communication module that performs communication with a base station in a wireless communication system, and should be interpreted as a concept that includes not only UE (User Equipment) in WCDMA, LTE, NR, HSPA, and IMT-2020 / IMT-2030 (5G, New Radio, and 6G), but also MS (Mobile Station), UT (User Terminal), SS (Subscriber Station), and wireless device in GSM. In addition, the terminal may be a user portable device such as a smartphone depending on the usage type, and in a V2X communication system, it may mean a vehicle, a device including a wireless communication module in the vehicle, etc. In addition, in the case of a Machine Type Communication system, it may include an MTC terminal, an M2M terminal, a URLLC terminal, etc. equipped with a communication module to perform machine type communication.

[0041] The base station or cell in this specification refers to an end that communicates with a terminal in terms of a network, and includes various coverage areas such as Node-B, eNB (evolved Node-B), gNB (gNode-B), LPN (Low Power Node), Sector, Site, various types of antennas, BTS (Base Transceiver System), Access Point, Point (e.g., Transmission Point, Reception Point, Transmission / Reception Point), Relay Node, Mega Cell, Macro Cell, Micro Cell, Pico Cell, Femto Cell, RRH (Remote Radio Head), RU (Radio Unit), and Small Cell. In addition, a cell may mean including a BWP (Bandwidth Part) in the frequency domain. For example, a serving cell may mean an Activation BWP of a terminal.

[0042] Since the various cells listed above have a base station that controls one or more cells, the base station can be interpreted in two meanings. 1) It can be a device itself that provides a mega cell, macro cell, micro cell, pico cell, femto cell, or small cell in relation to a wireless area, or 2) it can indicate the wireless area itself. In 1), all devices that provide a given wireless area are controlled by the same entity or that interact to cooperatively configure the wireless area are all indicated as a base station. Depending on how the wireless area is configured, a point, a transceiver point, a transmission point, a reception point, etc. can be an embodiment of a base station. In 2), the wireless area itself that receives or transmits a signal from the perspective of a user terminal or a neighboring base station can also be indicated as a base station.

[0043] In this specification, a cell may mean a component carrier having coverage of a signal transmitted from a transmission / reception point or a transmission / reception point itself.

[0044] Uplink (UL, or uplink) refers to a method of transmitting and receiving data from a terminal to a base station, and downlink (DL, or downlink) refers to a method of transmitting and receiving data from a base station to a terminal. Downlink may refer to communication or a communication path from multiple transmission / reception points to a terminal, and uplink may refer to communication or a communication path from a terminal to multiple transmission / reception points. In this case, in the downlink, the transmitter may be part of the multiple transmission / reception points, and the receiver may be part of the terminal. In addition, in the uplink, the transmitter may be part of the terminal, and the receiver may be part of the multiple transmission / reception points.

[0045] Uplink and downlink transmit and receive control information through control channels such as PDCCH (Physical Downlink Control CHannel) and PUCCH (Physical Uplink Control CHannel), and transmit and receive data by configuring data channels such as PDSCH (Physical Downlink Shared CHannel) and PUSCH (Physical Uplink Shared CHannel).

[0046] Hereinafter, the present disclosure describes an NR system as an example, but it will be apparent that the present disclosure can be applied and used to the corresponding functions of a 6G mobile communication network, and the limitations of the present disclosure are not limited by the NR system. This is because, although NR (New Radio) is used as a specific embodiment, it is stated that the technical idea of ​​the present disclosure can be generalized to future systems such as 6G, and is characterized by not being limited to a specific system (NR). Accordingly, the technical scope of the invention is characterized by not being limited to the NR system, but can be implemented in various wireless systems including 6G.

[0047] 6G communication, to which this disclosure applies, is the next-generation wireless communication technology after 5G, and its standardization is underway with the goal of commercialization around 2030. 6G goes beyond simple speed improvement and aims for artificial intelligence, hyper-spatial scalability, and real-digital convergence, and is being prepared with the goal of becoming the core infrastructure of the future digital society. 6G communication organically combines core technologies such as THz communication capable of ultra-wideband transmission, AI-Native network that integrates AI / ML into all layers of the network, RIS (Reconfigurable Intelligent Surface) technology that actively reflects and manipulates radio waves, NTN (Non-Terrestrial Network) that provides 3D coverage such as satellites, Joint Communication & Sensing (JCAS) that integrates communication and sensing functions, ultra-precision positioning, and quantum key cryptography (QKD) for security to support the next-generation digital society.

[0048] This 6G communication network architecture is being discussed with a focus on the following directions and characteristics. First, as an end-to-end intelligent autonomous network (AI-Native Architecture) architecture, AI / ML is fundamentally embedded in all layers of the network to perform data-based policy optimization, network autonomous operation, anomaly detection, and quality of service (QoS / QoE) assurance, and it is expected that AI functions will be distributed and operated in RAN, Core, and service management. Second, it will adopt a 3D extended architecture to support a Non-Terrestrial Network (NTN) that integrates the terrestrial network (RAN), satellite, and high-altitude aerial platforms (HAPS). For this, an integrated cell structure and IAB-based backhaul / fronthaul design will be supported. Third, the Service-Based Architecture (SBA) introduced in 5G will be enhanced to strengthen the modularization of network functions (NF) and enable flexible API-based service creation through NEF, NWDAF, and PCF. Fourth, through a distributed intelligence structure, we plan to move away from a centralized structure and perform real-time AI-based judgment at edge nodes such as RAN, MEC, and UE, and support local decision-making in smart transportation, factories, and medical sites. Fifth, we plan to introduce a structure that integrates communication and sensing (Joint Communication & Sensing) so that RAN can perform environmental sensing functions beyond simple communication functions, and a sensing feature processing and asynchronous collection structure for this will also be designed. Finally, through a digital twin-based virtualization structure, we plan to support the intelligence and efficiency of the entire network by replicating the physical network status in real time and enabling predictive autonomous operation and network simulation.The next-generation wireless system to which this disclosure applies includes support for both sub-6 GHz frequency bands (FR1, Frequency Range 1) and 6 GHz or higher frequency bands (FR2, Frequency Range 2), and in particular, newly defines support for FR3 (Frequency Range 3) between FR1 and FR2, and proposes various support technologies related to end-to-end interoperability and data connectivity targeting the FR3 band (7.125-24.25 GHz).

[0049] For example, NR defines various operating scenarios by adding considerations for satellites, automobiles, and new verticals, and supports the eMBB (Enhanced Mobile Broadband) scenario in terms of services, the mMTC (Massive Machine Communication) scenario that has high terminal density but is deployed over a wide area and requires low data rates and asynchronous access, and the URLLC (Ultra Reliability and Low Latency) scenario that requires high responsiveness and reliability and can support high-speed mobility.

[0050] To meet these scenarios, NR introduces a wireless communication system that incorporates new waveform and frame structure technologies, low latency technologies, support for ultra-high frequency bands (mmWave), and forward compatibility technologies. In particular, NR systems offer various technological changes in terms of flexibility to ensure forward compatibility. The key technical features of NR are described below with reference to the drawings.

[0051] Figure 1 is a schematic diagram illustrating the structure of an NR system to which the present embodiment can be applied.

[0052] Referring to Fig. 1, the NR system is divided into 5GC (5G Core Network) and NR-RAN parts, and the NG-RAN is composed of gNBs and ng-eNBs that provide user plane (SDAP / PDCP / RLC / MAC / PHY) and control plane (RRC) protocol termination for UE (User Equipment). gNBs or gNBs and ng-eNBs are interconnected via the Xn interface. gNBs and ng-eNBs are each connected to the 5GC via the NG interface. The 5GC can be configured to include an AMF (Access and Mobility Management Function) that is responsible for the control plane such as terminal access and mobility control functions, and an UPF (User Plane Function) that is responsible for the control function for user data. NR includes support for both frequency bands below 6 GHz (FR1, Frequency Range 1) and frequency bands above 6 GHz (FR2, Frequency Range 2). In addition, the next-generation wireless communication system after 5G according to the present disclosure can support various technologies related to end-to-end interoperability and data connection targeting the FR3 band (7.125-24.25 GHz).

[0053] gNB refers to a base station that provides NR user plane and control plane protocol termination to terminals, and ng-eNB refers to a base station that provides E-UTRA user plane and control plane protocol termination to terminals. The base station described in this specification should be understood to encompass both gNB and ng-eNB, and may also be used to refer to gNB or ng-eNB separately as needed.

[0054] NR uses the CP-OFDM waveform with a cyclic prefix for downlink transmission, and CP-OFDM or DFT-s-OFDM for uplink transmission. OFDM technology is easily combined with MIMO (Multiple Input Multiple Output) and offers the advantages of high spectral efficiency and low-complexity receivers.

[0055] Meanwhile, in NR, requirements for data rates, latency, and coverage vary across scenarios. Therefore, it is necessary to efficiently satisfy these requirements across the frequency bands that comprise a given NR system. To achieve this, technologies are applied to efficiently multiplex radio resources based on multiple different numerologies.

[0056] Specifically, the NR transmission numerator is determined based on the sub-carrier spacing and CP (Cyclic prefix), and is changed exponentially using the μ value as an exponent value of 2 based on 15 kHz, as shown in Table 1 below.

[0057] μsubcarrier intervalCyclic prefixSupported for dataSupported for synch015NormalYesYes130NormalYesYes260Normal, ExtendedYesNo3120NormalYesYes4240NormalNoYes

[0058] As shown in Table 1 above, the numerology of NR can be divided into five types according to the subcarrier spacing. Specifically, the subcarrier spacing used for data transmission in NR is 15, 30, 60, and 120 kHz, and the subcarrier spacing used for synchronization signal transmission is 15, 30, 12, and 240 kHz. In addition, the extended CP is applied only to the 60 kHz subcarrier spacing. Meanwhile, the frame structure in NR is defined as a frame with a length of 10 ms, which is composed of 10 subframes with the same length of 1 ms. One frame can be divided into half frames of 5 ms, and each half frame contains 5 subframes. In the case of a 15 kHz subcarrier spacing, one subframe consists of 1 slot, and each slot consists of 14 OFDM symbols.

[0059] FIG. 2 is a drawing for explaining a frame structure in an NR system to which the present embodiment can be applied.

[0060] Referring to FIG. 2, a slot is fixedly composed of 14 OFDM symbols in the case of a normal CP, but the length of the slot in the time domain may vary depending on the subcarrier spacing. For example, in the case of a numerology with a 15 kHz subcarrier spacing, a slot is composed of 1 ms in length, which is the same length as a subframe. In contrast, in the case of a numerology with a 30 kHz subcarrier spacing, a slot is composed of 14 OFDM symbols, but two slots may be included in one subframe with a length of 0.5 ms. In other words, a subframe and a frame are defined with a fixed time length, and a slot is defined by the number of symbols, so the time length may vary depending on the subcarrier spacing.

[0061] Furthermore, the basic unit of scheduling is defined as a slot, and mini-slots (or sub-slots, or non-slot-based scheduling) are also introduced to reduce transmission delay in the wireless section. This offers the advantage of reducing transmission delay in the wireless section, as the length of a single slot is inversely shortened by using a wide subcarrier spacing. These mini-slots (or sub-slots) are designed to efficiently support URLLC scenarios and allow scheduling in units of 2, 4, or 7 symbols.

[0062] Additionally, NR defines uplink and downlink resource allocation at the symbol level within a single slot. To reduce HARQ delay, a slot structure is defined that allows HARQ ACK / NACK transmission within a transmission slot. This slot structure is defined as a self-contained structure.

[0063] NR is designed to support a total of 256 slot formats and supports a common frame structure that configures FDD or TDD frames through various combinations of slots. For example, it can support a slot structure in which all symbols in a slot are set to downlink, a slot structure in which all symbols are set to uplink, and a slot structure in which downlink and uplink symbols are combined. In addition, NR supports data transmission being distributed and scheduled across one or more slots. Therefore, a base station can use a slot format indicator (SFI) to inform a UE whether a slot is a downlink slot, an uplink slot, or a flexible slot. The base station can indicate the slot format by indicating an index of a table configured through UE-specific RRC signaling using the SFI, and can also indicate it dynamically through DCI (Downlink Control Information) or statically or semi-statically through RRC.

[0064] In relation to physical resources in NR, antenna ports, resource grids, resource elements, resource blocks, and bandwidth parts are considered.

[0065] Antenna ports are defined such that the channel through which a symbol on an antenna port is carried can be inferred from the channel through which another symbol on the same antenna port is carried. Two antenna ports are said to be quasi co-located (QC / QCL) if the large-scale properties of the channel through which a symbol on one antenna port is carried can be inferred from the channel through which a symbol on the other antenna port is carried. Here, the large-scale properties include one or more of delay spread, Doppler spread, frequency shift, average received power, and received timing.

[0066] FIG. 3 is a diagram for explaining a resource grid supported by a wireless access technology to which the present embodiment can be applied.

[0067] Referring to Figure 3, since the Resource Grid supports multiple numerals on the same carrier, a resource grid may exist for each numeral. Furthermore, resource grids may exist based on antenna ports, subcarrier spacing, and transmission direction.

[0068] For example, a resource block consists of 12 subcarriers and is defined only in the frequency domain. Furthermore, a resource element consists of one OFDM symbol and one subcarrier. Therefore, as shown in Figure 3, the size of a resource block can vary depending on the subcarrier spacing. NR also defines "Point A," which serves as a common reference point for the resource block grid, as well as common resource blocks and virtual resource blocks.

[0069] FIG. 4 is a diagram for explaining a bandwidth part supported by a wireless access technology to which the present embodiment can be applied.

[0070] In NR, the carrier bandwidth is fixed at 20 MHz, and the maximum carrier bandwidth can be set from 50 MHz to 400 MHz for each subcarrier interval. Here, it is not assumed that all terminals use the entire carrier bandwidth. Therefore, in NR, as illustrated in Figure 4, a bandwidth part (BWP) can be designated within the carrier bandwidth and instructed to be used by the terminal. In addition, a bandwidth part is associated with a single numerology, consists of a subset of consecutive common resource blocks, and can be dynamically activated over time. For example, a terminal is configured with up to four bandwidth parts each for uplink and downlink, and supports data transmission and reception using the bandwidth part activated at a given time.

[0071] Meanwhile, in the case of a paired spectrum, the uplink and downlink bandwidth parts are set independently, and in the case of an unpaired spectrum, the downlink and uplink bandwidth parts are set in pairs so that they can share a center frequency to prevent unnecessary frequency re-tuning between downlink and uplink operations.

[0072] Additionally, in NR, a terminal performs cell search and random access procedures to connect to a base station and perform communication.

[0073] Cell search is a procedure in which a terminal synchronizes to the cell of a corresponding base station, obtains a physical layer cell ID, and obtains system information using a synchronization signal block (SSB) transmitted by the base station.

[0074] FIG. 5 is a diagram illustrating an example of a synchronization signal block in a wireless access technology to which the present embodiment can be applied.

[0075] Referring to FIG. 5, SSB is composed of a primary synchronization signal (PSS) and a secondary synchronization signal (SSS), each occupying one symbol and 127 subcarriers, and a PBCH spanning three OFDM symbols and 240 subcarriers.

[0076] The terminal receives SSB by monitoring SSB in the time and frequency domain.

[0077] SSB can be transmitted up to 64 times in 5ms. Multiple SSBs are transmitted in different transmission beams within 5ms, and the terminal performs detection assuming that SSBs are transmitted every 20ms based on a specific beam used for transmission. The number of beams that can be used for SSB transmission within 5ms can increase as the frequency band increases. For example, up to 4 SSB beams can be transmitted below 3GHz, up to 8 in the frequency band between 3GHz and 6GHz, and up to 64 different beams can be used for SSB transmission in the frequency band above 6GHz.

[0078] SSB contains two symbols in one slot, and the starting symbol and number of repetitions within the slot are determined as follows depending on the subcarrier spacing.

[0079] Meanwhile, SSB can be transmitted at the center frequency of the carrier bandwidth, or can be transmitted at a location other than the center of the system bandwidth, and multiple SSBs can be transmitted in the frequency domain when supporting wideband operation. Accordingly, the terminal monitors SSB using the synchronization raster, which is a candidate frequency location for monitoring SSB. The carrier raster, which is the center frequency location information of the channel for initial access, and the synchronization raster are newly defined in NR. The synchronization raster is set with a wider frequency interval than the carrier raster, which can support the terminal's fast SSB search.

[0080] A UE can obtain the MIB through the PBCH of the SSB. The MIB (Master Information Block) includes the minimum information required for the UE to receive the remaining system information (RMSI, Remaining Minimum System Information) broadcast by the network. In addition, the PBCH may include information on the position of the first DM-RS symbol in the time domain, information for the UE to monitor SIB1 (e.g., SIB1 numerology information, information related to SIB1 CORESET, search space information, PDCCH-related parameter information, etc.), offset information between the common resource block and the SSB (the absolute position of the SSB within the carrier is transmitted through SIB1), etc. Here, the SIB1 numerology information is also applied equally to some messages used in the random access procedure for the UE to access the base station after completing the cell search procedure. For example, the numerology information of SIB1 may be applied to at least one of messages 1 to 4 for the random access procedure.

[0081] The aforementioned RMSI may refer to SIB1 (System Information Block 1), and SIB1 is broadcast periodically (e.g., every 160 ms) in the cell. SIB1 contains information necessary for the UE to perform the initial random access procedure and is periodically transmitted via PDSCH. In order for the UE to receive SIB1, it must receive numerology information used for SIB1 transmission and CORESET (Control Resource Set) information used for SIB1 scheduling via PBCH. The UE checks scheduling information for SIB1 using SI-RNTI in CORESET and acquires SIB1 on PDSCH according to the scheduling information. The remaining SIBs, excluding SIB1, may be transmitted periodically or upon request of the UE.

[0082] FIG. 6 is a diagram for explaining a random access procedure in a wireless access technology to which the present embodiment can be applied.

[0083] Referring to FIG. 6, once cell search is complete, the terminal transmits a random access preamble for random access to the base station. The random access preamble is transmitted via the PRACH. Specifically, the random access preamble is transmitted to the base station via the PRACH, which consists of consecutive radio resources in a specific slot that is periodically repeated. Generally, when a terminal initially accesses a cell, a contention-based random access procedure is performed, and when performing random access for beam failure recovery (BFR), a non-contention-based random access procedure is performed.

[0084] The terminal receives a random access response for the transmitted random access preamble. The random access response may include a random access preamble identifier (ID), an UL Grant (uplink radio resource), a temporary C-RNTI (Temporary Cell - Radio Network Temporary Identifier), and a TAC (Time Alignment Command). Since one random access response may include random access response information for one or more terminals, the random access preamble identifier may be included to indicate to which terminal the included UL Grant, temporary C-RNTI, and TAC are valid. The random access preamble identifier may be an identifier for the random access preamble received by the base station. The TAC may be included as information for the terminal to adjust uplink synchronization. The random access response may be indicated by a random access identifier on the PDCCH, i.e., a RA-RNTI (Random Access - Radio Network Temporary Identifier).

[0085] Upon receiving a valid random access response, the terminal processes the information contained in the random access response and performs scheduled transmission to the base station. For example, the terminal applies TAC and stores a temporary C-RNTI. Furthermore, using the UL Grant, the terminal transmits data stored in its buffer or newly generated data to the base station. In this case, information that identifies the terminal must be included.

[0086] Finally, the terminal receives a downlink message for contention resolution.

[0087] The downlink control channel in NR is transmitted in a CORESET (Control Resource Set) with a length of 1 to 3 symbols, and transmits uplink / downlink scheduling information, SFI (Slot format Index), and TPC (Transmit Power Control) information.

[0088] Additionally, to ensure system flexibility, the CORESET concept is introduced. CORESET (Control Resource Set) represents time-frequency resources for downlink control signals. A terminal can decode control channel candidates using one or more search spaces within the CORESET time-frequency resources. Quasi-CoLocation (QCL) assumptions are established for each CORESET, which are used to inform characteristics of analog beam direction in addition to delay spread, Doppler spread, Doppler shift, and average delay.

[0089] Figure 7 is a drawing for explaining CORESET.

[0090] Referring to Figure 7, a CORESET can exist in various forms within the carrier bandwidth within a single slot, and in the time domain, a CORESET can consist of up to three OFDM symbols. In addition, a CORESET is defined as a multiple of six resource blocks up to the carrier bandwidth in the frequency domain.

[0091] The first CORESET is indicated via the MIB as part of the initial bandwidth part configuration, allowing the terminal to receive additional configuration and system information from the network. After establishing a connection with the base station, the terminal can receive and configure one or more CORESET information via RRC signaling.

[0092] In this specification, the terms frequency, frame, subframe, resource, resource block, region, band, subband, control channel, data channel, synchronization signal, various reference signals, various signals or various messages related to NR include the meanings used in existing communication systems, and may be applied as concepts that are expanded, changed or added in consideration of newly defined terms in next-generation systems, thereby including the possibility of being interpreted in various meanings.

[0093] As described, communication systems are evolving and developing diverse technologies to support 6G beyond 5G through diverse research and application of next-generation wireless access technologies. To this end, research is being conducted on technologies to support large-scale access, communication in high-frequency bands, and faster, more seamless services. Technologies such as MIMO, which perform beamforming to target specific regions, directions, or terminals, are also being supported to provide more efficient data transmission and reception.

[0094] Furthermore, with the recent increase in terminal mobility, beamforming technology requires various measures to prevent the reduction of terminal service areas and provide uninterrupted services. In other words, there is a need to predict the quality of beams provided to terminals. To this end, AI technology can be applied and utilized in various forms for beam quality prediction. For example, beam quality prediction can include spatial prediction, temporal prediction, etc., and communication services can be provided by proactively providing better-quality beams to terminals based on the beam quality prediction results.

[0095] To this end, the present disclosure proposes a method for providing communication services using various artificial intelligence models, and in particular, proposes a method and device for managing artificial intelligence models to improve service efficiency. Considering that the artificial intelligence model utilizes training data, i.e., considering the potential deterioration of prediction performance due to terminal movement or changes in channel conditions, the present disclosure performs appropriate monitoring and control operations for prediction performance, thereby supporting more effective use of the artificial intelligence model.

[0096] FIG. 8 is a diagram illustrating the structure of a functional framework for Intelligence according to one embodiment of the present disclosure.

[0097] Referring to Figure 8, to apply an AI model to a communication service, data collection is first performed. The collected data serves as training data during the model training process. For AI models, training is performed using the training data, and the trained model is deployed / updated on the network or terminal.

[0098] Meanwhile, the collected data can also be used for inference in a trained model. When inference data is input, the AI ​​model generates output based on pre-trained information. For example, the output can be configured in various ways based on the training data, such as quality information for spatially predicted beams, quality information for temporally predicted beams, or information for predicting terminal locations.

[0099] AI models can degrade over time or due to environmental changes. Therefore, retraining may be performed based on performance feedback from the AI ​​model.

[0100] The output information of the AI ​​model can be fed back to the actor and collected as data. This data can then be reused as model training or inference data, allowing the aforementioned lifecycle management operations to operate cyclically. This overall operation can be described as a beam management operation.

[0101] Meanwhile, AI / ML capabilities vary depending on specific use cases and sub-use cases, and may focus on the interaction between the network and UE.

[0102] For example, collaboration levels between the network and the UE can be defined as follows.

[0103] 1. Level 0 (Collaboration Level 0): No collaboration

[0104] 2. Level 1 (Collaboration Level 1): Signaling-based collaboration without model transfer

[0105] 3. Level 2 (Collaboration Level 2): ​​Signaling-based collaboration with model transfer

[0106] More specifically, Collaboration Level 0 means that the UE operates through its own AI / ML model or only performs RRC / RRM procedures without AI functions, and the network does not separately configure or participate in the AI ​​model for the UE. In other words, it is characterized by the absence of AI / ML-related interactions between the network and the UE. On the other hand, Collaboration Level 1 is characterized by the fact that the model itself is not transmitted, but configuration / signaling information for assisting AI / ML functions is exchanged between the network and the UE. For example, the UE receives configuration information for using the AI / ML model from the network (e.g., reference signal configuration, evaluation conditions, reporting cycle, etc.), and considering this, the UE only reports the AI ​​inference result or measurement result, and the model itself exists independently within the UE. In other words, the UE can report the results of executing a local AI model used for CSI enhancement, beam management, mobility enhancement, etc. to the network. Additionally, Collaboration Level 2 allows the network to transmit the AI / ML model itself or some parameter / configuration information to the UE, and the UE performs AI functions based on the model received from the network (e.g., inference module, weight, etc.), or to provide feedback on the model results learned by the UE to the network, or to support bidirectional model updates. This can be referred to as collaborative inference and federated learning-like operations. To this end, the UE and the network can transmit and receive model-related signaling, such as the AI ​​model's configuration, parameters, and learning cycle. The technology described herein can be applied to at least one of the three collaboration levels described above. Examples of utilizing AI models for communication will be described in more detail below.

[0107] A beam management method according to an example of the present disclosure can be divided into an initial access phase and a post-cell connection establishment phase. A terminal performing initial access sets its initial transmit / receive (Tx / Rx) beam through a RACH procedure. In order to provide base station transmit beam settings to a terminal without cell connection, the base station periodically and repeatedly transmits SSBs to which beams in different directions are mapped. The terminal can select an optimal SSB through signal measurement for the periodically transmitted SSBs and transmit a PRACH preamble mapped to the corresponding SSB, thereby informing the base station of information about the selected transmit beam.

[0108] A terminal performing an initial access procedure receives cell-related parameter information, such as PRACH information corresponding to each SSB required in the initial access phase, via a system information message. The terminal measures RSRP for periodically transmitted SSBs. The terminal inputs beam information for the measured SSBs as input values ​​for an AI / ML model, and calculates predicted beam strengths or RSRP values ​​for all beams of the SSB as output values ​​for the input values. The terminal selects an SSB with the highest value among the calculated values, and can notify the base station of the selected predicted beam information by transmitting a preamble belonging to the PRACH resource corresponding to the selected SSB to the base station.

[0109] A base station that does not know the beam information of the first terminal entering the network can set up to 64 beams for a terminal that has not yet connected. The terminal sequentially measures all beams to find the optimal beam for its location. This not only causes delays in beam selection and cell connection as the number of beams within the cell increases, but can also increase the terminal's power consumption by requiring the terminal to measure a large number of beams.

[0110] To address the above issues, the base station can apply a wide beam for SSB to determine the approximate location / beam of the initially connected terminal, and then configure a narrow beam after the terminal accesses the cell. However, while narrow beams provide high data rates to the terminal, they are sensitive to the terminal's movement or environmental changes, which can easily lead to disconnection. Therefore, the base station assigns the terminal a CSI resource with a candidate beam mapped to it, allowing the terminal to continuously measure the surrounding beam strength and report the measurement results to the base station.

[0111] A UE configured for beam reporting performs CSI reporting by measuring its assigned reference signal. However, this method can lead to a problem in that as the number of UEs in a cell increases, the number of reference signal resources allocated to each UE also increases rapidly. To mitigate this resource overhead, the base station can allocate the same candidate beam (CSI resource) to UEs in similar locations. However, when UEs with different mobility share the same resource, the issue of assigning new candidate beam resources to UEs that leave the resource area arises. If a minimum number of candidate beams is allocated to a UE with high mobility to reduce resource overhead, the UE will experience frequent RRC reconfigurations. This reconfiguration of candidate beams through RRC incurs relatively large delays, which can lead to beam dropouts. The base station can mitigate this by appropriately increasing the number of beams, but this increased number of beams can increase the measurement burden on the UE.

[0112] AI / ML models can be applied to improve delay and terminal power consumption in beam search. For AI / ML-based beam management, BM-Case1 and BM-Case2 are supported for characterization and basic performance evaluation. Here, BM-Case1 performs spatial domain downlink beam prediction for beam set A based on measurement results of beam set B. BM-Case2 performs temporal DL beam prediction for beam set A based on historical measurement results of beam set B. In one example, for BM-Case1 and BM-Case2, the beams of sets A and B may be in the same frequency range. The present disclosure can be applied to both BM-Case 1 and 2.

[0113] FIGS. 9 and 10 are diagrams for explaining a beam prediction operation using a UE-side AI / ML model according to one embodiment.

[0114] Referring to FIG. 9, a terminal (901) can receive a reference signal from a base station (902) (S910). To receive the reference signal, the terminal (901) may also receive reference signal transmission resources and a measurement configuration for reference signal measurement from the terminal in advance via an RRC message.

[0115] The terminal (901) measures the received reference signal. Here, the reference signal may be associated with at least one beam set to Set B. That is, the base station (902) transmits the reference signal through a beam sweeping operation, and the terminal (901) may receive the reference signal.

[0116] Terminal (901) can measure the reception strength of the reference signal of Set B (S920). For example, terminal (901) can measure the quality of at least one reference signal comprised of Set B. Here, the quality may be information set to be measured through measurement configurations such as RSRP, RSRQ, and SINR.

[0117] The terminal (901) can predict the reference signal for SetA by inputting the reference signal measurement result of SetB into an artificial intelligence model configured in the terminal (S930).

[0118] SetA is composed of beams different from SetB, and predictions for beams set as SetA in the spatial domain can be performed by an artificial intelligence model based on the measurement results of the beams of SetB. Here, the artificial intelligence model can be trained so that when the quality measurement results of SetB, which are spatially distinct, are input, the quality measurement results of beams included in SetA, which is set in a different spatial domain, are output.

[0119] Referring to FIG. 10, a terminal (901) can receive a reference signal from a base station (902) (S910). The reference signal can be transmitted to the terminal (901) at multiple time instances. That is, measurements can be performed on multiple reference signals over time.

[0120] The terminal (901) derives historical measurement results of reference signals measured at multiple time instances (S925).

[0121] The terminal (901) can input historical measurement results as inputs to an artificial intelligence model configured in the terminal (901) and generate prediction measurement results of future reference signals as outputs (S935). Here, the historical measurement results may be described as SetB, and the future reference signal prediction results may be described as SetA.

[0122] For example, in the case of BM-Case2, the measurement results of K (K>=1) latest measurement instances are used as inputs to the AI / ML model. Regarding BM-Case2, the output of the AI / ML model should be F predictions for F future time instances, and each prediction can be composed for each time instance.

[0123] Set B is a subset of Set A, and among the beams in Set A, Set B can be determined according to a predetermined fixed pattern or a random pattern. Alternatively, Set A and Set B can be different from each other, for example, Set A can be composed of narrow beams and Set B can be composed of wide beams. Here, Set A can be used for DL ​​beam prediction and Set B can be used for DL ​​beam measurement.

[0124] With respect to BM-Case1 of the aforementioned FIG. 9, the AI / ML input may, for example, only use L1-RSRP measurements based on Set B. Alternatively, L1-RSRP measurements based on Set B and assistance information may be used. The assistance information may include transmit and / or receive beam shape information, such as transmit and / or receive beam patterns, transmit and / or receive beam directions (azimuth and elevation), 3dB beamwidths, etc., or beam angles of expected transmit and / or receive beams for prediction, transmit and / or receive beam IDs for prediction, terminal location information, terminal direction information, etc. Alternatively, CIR based on Set B or L1-RSRP measurements based on Set B and the corresponding DL transmit and / or receive beam ID may be used as input.

[0125] Thus, the goal of AI / ML-based BM (beam management) technology is to reduce measurement overhead and latency by optimizing the beam prediction process in the spatial / temporal domains for data transmission. Here, the AI / ML model needs to be trained before the beam prediction process. Two types of beam sets (Set A, Set B) are utilized in the training process. Set A is a set of beams for which prediction is targeted, and Set B is a set of beams used for measurement. BM-case 1 performs downlink TX beam prediction for Set A in the spatial domain based on the measurement results of the Set B beam. BM-case 2 performs downlink TX beam prediction for Set A in the temporal domain based on the past measurement results of the Set B beam.

[0126] Meanwhile, terminals can possess high mobility. Furthermore, channel conditions can change as they move. In response to these fluctuations, terminals and base stations must continuously manage their beams to ensure communication is seamless. This can be explained through beam management operations.

[0127] FIG. 11 is a diagram for explaining a beam management operation according to one embodiment.

[0128] Referring to FIG. 11, the base station (1002) transmits an SSB to the terminal (1001) (S1010). For example, the base station (1002) transmits a synchronization signal (SS) block to the terminal (1001) for optimal beam search and beam measurement when the terminal (1001) first connects. For beam pairing, the base station (1002) can sweep the SSB and transmit it by applying various spatial filters.

[0129] The terminal (1001) measures the RSRP value of each beam based on the received SSB (S1020). For example, the terminal (1001) measures the L1-RSRP for each beam to which a spatial filter is applied, thereby calculating a measurement value for each beam.

[0130] The terminal (1001) temporarily determines the optimal beam using the beam measurement results (S1030). In addition, the terminal (1001) reports the selected beam and related quality information to the base station (S1040). For example, the terminal (1001) can compare the measurement values ​​for each beam based on the beam measurement results and select the beam with the best quality. Alternatively, the terminal (1001) can identify K upper-quality beams configured by the base station (1002). The terminal (1001) reports the beam index and the measurement result values ​​for each beam to the base station (1002) for the identified K beams with the best quality (K is a natural number greater than or equal to 1).

[0131] The base station (1002) selects the optimal beam based on the beam report transmitted by the terminal (S1050). The base station (1002) performs beam pairing with the terminal (1001) using the selected optimal beam, and can communicate with the terminal (1001) using the corresponding beam.

[0132] Meanwhile, the base station (1002) transmits channel state information - reference signal and additional SSB to more accurately check the beam measurement performance of the terminal (1001) for the purpose of maintaining and updating the beam after connection (S1060). For example, as described above, the optimal beam may change depending on the movement of the terminal (1001) or changes in channel conditions. Accordingly, the base station (1002) periodically or aperiodically transmits a reference signal (e.g., CSI-RS or SSB) to the terminal (1001).

[0133] The terminal (1001) performs beam measurement and quality check operations repeatedly, periodically or aperiodically, using a reference signal transmitted by the base station (1002) (S1070). The terminal (1001) can report beam measurement results and quality information to the base station (1002) through periodic or aperiodical transmission according to an event occurrence, depending on the measurement report configuration (S1080).

[0134] The base station (1002) can perform a beam adjustment operation with the terminal (1001) based on the received reporting (S1090).

[0135] These actions ensure uninterrupted service provision between terminals and base stations through beamforming. This means that even when beam pairing is performed, beam management operations must be continuously performed to ensure beam coordination.

[0136] In the case of AI / ML, beam pairing is performed using beam prediction results from an AI model, rather than actual beam measurements. Therefore, even in this case, continuous beam management operations are required. Furthermore, the performance of the AI ​​model can change or deteriorate for various reasons. Therefore, monitoring the performance of the AI ​​model during beam management is crucial.

[0137] While AI / ML-based beam prediction technology holds the potential to dramatically improve beam management efficiency, ensuring prediction accuracy and real-time applicability in real-world wireless environments requires a robust design for an event-based reporting mechanism that can monitor the suitability of prediction results and effectively report any performance degradation. Therefore, a technology is needed that can monitor the performance of AI / ML-based beam prediction models embedded in terminals in real time, detect prediction failures or performance degradation early, and automatically report the results to an event-driven network.

[0138] Below, embodiments of monitoring the performance of an artificial intelligence model in beam management operations are described in more detail with reference to drawings.

[0139] In a beam prediction architecture, a performance monitoring system is required to verify the reliability and validity of prediction results and, if necessary, report them to the network. Accordingly, a performance monitoring framework can be defined. Performance monitoring frameworks are broadly categorized into Type 1 and Type 2. Type 1 performance monitoring is network-driven, where user terminals collect measurements or calculate performance indicators and then report the results to the network. Type 1 performance monitoring has two substructures.

[0140] First, Option 1 involves transmitting the measurements collected by the user terminal to the network in raw data form, and the network calculates the performance indicators independently. Option 2 involves the terminal calculating the performance indicators independently and reporting them directly to the network only when predefined threshold conditions are met.

[0141] Type 1 performance monitoring can be applied to both BM-Case1 and BM-Case2 described above. Furthermore, both Type 1 Option 1 and Option 2 can be applied.

[0142] Fig. 12 is a drawing for explaining terminal operation according to one embodiment.

[0143] Referring to FIG. 12, a method for a terminal to perform a performance management operation of an artificial intelligence model may include a step of generating a beam prediction result using an artificial intelligence model configured in the terminal (S1110).

[0144] For example, a terminal can predict a beam using an artificial intelligence model as described herein. For example, the terminal can predict the optimal beam in space, as in BM-Case 1. Alternatively, the terminal can predict the optimal beam in time, as in BM-Case 2.

[0145] To this end, the terminal can input the measurement results for the actual beam into an artificial intelligence model and obtain the beam prediction results as output. For example, the beam prediction results may be the Top-K beams predicted to have the best quality and the predicted predicted measurement values. K can be configured by the base station, and while the case where K is 1 is described as an example here, the present embodiments can also be applied when K is a natural number exceeding 1.

[0146] The terminal can transmit the predicted beam prediction results to the base station. This is to enable the terminal and base station to perform pairing using the beam prediction results in beam management.

[0147] A method for a terminal to perform a performance management operation of an artificial intelligence model may include a step of receiving at least one reference signal including a first reference signal associated with a first beam selected according to a beam prediction result (S1120).

[0148] For example, a terminal may periodically or aperiodically receive a first reference signal for which resources are set in association with a first beam selected based on a beam prediction result from a base station. Since the base station knows about the first beam predicted as the optimal beam by the terminal based on the beam prediction result, when performance management for the terminal's artificial intelligence model is required, the base station may transmit a reference signal for the first beam to the terminal to perform a performance management operation for the artificial intelligence model.

[0149] The terminal can measure the actual channel quality for the received first reference signal. For example, the terminal can measure the RSRP for the first reference signal. Alternatively, the terminal can derive various channel quality measurements, such as RSRQ and SINR. The terminal can determine which measurement scheme to use to derive the measurement results based on the configuration instructions of the base station.

[0150] As another example, a terminal may receive multiple reference signals, including a first reference signal, from a base station. For example, when transmitting a first reference signal predicted to be the optimal beam, the base station may also transmit candidate reference signals for other candidate beams to the terminal.

[0151] The terminal can perform measurements on the received first reference signal and candidate reference signals to generate measurement results.

[0152] A method for a terminal to perform a performance management operation of an artificial intelligence model may include a step of determining whether an event has occurred using at least one of a measurement result of a first reference signal, reliability information for a first beam, and measurement results for reference signals other than the first reference signal (S1130).

[0153] For example, the terminal can determine whether a preset event has occurred to manage the performance of an artificial intelligence model.

[0154] For example, the terminal may determine that an event has occurred if the measurement result of the first reference signal is less than a first threshold value. For example, the terminal receives the first reference signal associated with the optimal beam, compares the measured result with a preset first threshold value, and determines whether an event has occurred based on the comparison result. That is, if the actual measurement result for the first reference signal is lower than the first threshold value, the terminal may determine that there is a problem with the prediction of the first beam, which is the optimal beam predicted by the artificial intelligence model. Accordingly, the terminal may determine that the event condition configured in relation to artificial intelligence model management is satisfied in this case.

[0155] As another example, the terminal may determine that an event has occurred if the measurement result of the first reference signal is less than the first threshold value, and the highest measurement result among the measurement results for reference signals excluding the first reference signal exceeds the second threshold value. For example, the terminal determines whether the actual measurement result of the first reference signal is less than the preset first threshold value. In addition, when receiving multiple reference signals including the first reference signal, the terminal also generates measurement results for the multiple received reference signals. The terminal selects one reference signal having the highest measurement result among the measurement results of reference signals excluding the first reference signal. The terminal determines whether the measurement result of the selected one reference signal exceeds the second threshold value. That is, the terminal determines that the event is satisfied if the actual measurement value of the first reference signal associated with the existing predicted optimal beam deteriorates below the first threshold value, and the measurement results of other reference signals are determined to be greater than or equal to the second threshold value.

[0156] As another example, the terminal may determine that an event has occurred if the reliability information for the first beam is less than a third threshold. For example, the reliability information for the first beam selected as the optimal beam according to the beam prediction result may be calculated. The reliability information may be set as the prediction accuracy, probability value, etc. for the first beam through an artificial intelligence model. For example, the artificial intelligence model applies a softmax function in the output layer of the artificial intelligence model to calculate a probability distribution for N candidate beams. Here, the candidate beam calculated with the highest probability may be selected as the optimal first beam. Therefore, when performing a prediction for the first beam, a probability value for the first beam may be calculated, and the probability value may be used as the reliability information. The terminal may determine that the event condition is satisfied if the reliability information falls below a preset third threshold.

[0157] Here, the first threshold value, the second threshold value, the third threshold value, etc. may be included in the measurement configuration information received from the base station and stored in the terminal. The first threshold value, the second threshold value, and the third threshold value may be included in the threshold1 / 2 / 3 fields in measConfig -> ReportConfig and transmitted to the UE, and may be configured as a logic for event triggering operation by comparing the RSRP value of the AI-predicted beam with the threshold. In the present disclosure, as an example, an RSRP-based threshold value is applied to the AI-ML framework to describe the application of thresholds for beam performance and prediction, but as another example of the present disclosure, an AI-only threshold may be set and measurement and reporting operations may be performed accordingly.

[0158] The method of performing performance management operations of an artificial intelligence model by a terminal may include a step of transmitting report information to a base station based on a determination of whether an event has occurred (S1140).

[0159] For example, a terminal may transmit report information to a base station when a situation occurs that satisfies the aforementioned event conditions. In other words, the satisfaction of an event condition may serve as a trigger for the transmission of report information.

[0160] For example, the report information may include at least one of identification information for the first beam, measurement result information for the first reference signal, and performance monitoring result information for the artificial intelligence model. For example, the report information may include identification information for the first beam predicted by the terminal as the optimal beam and measurement result information for actual measurement performed using the first reference signal. Alternatively, the report information may include performance monitoring result information for the artificial intelligence model, which indicates whether the artificial intelligence model has deteriorated based on the result of determining whether the aforementioned event has occurred.

[0161] As another example, the report information may include identification information for the first beam, measurement result information for the first reference signal, identification information for the second beam indicating the highest measurement result, and the highest measurement result. When determining whether the above-mentioned event has occurred, the first reference signal and other additional reference signals may be used. In this case, the report information may include identification information for the first beam and an actual measurement result value for the first reference signal. In addition, the report information may further include identification information for the second beam selected as having the highest quality among the candidate reference signals and an actual measurement result value of the reference signal associated with the second beam. Through this information, the base station can determine whether a situation requires a beam change of the terminal, etc. In addition, the second beam may be identified as the target beam.

[0162] As another example, the report information may include indications indicating performance degradation of the AI ​​model. The indications may include reliability information for the first beam. Alternatively, the indications may simply indicate whether the AI ​​model is degraded.

[0163] By receiving the aforementioned report information, the base station can verify the performance of the UE-side AI / ML model configured in the terminal. If the terminal determines that an event has occurred, the terminal can perform at least one of the following operations: a fallback operation for the AI ​​model, an AI model switching operation, and an AI model deactivation operation. Alternatively, the terminal can perform at least one of the following operations: a fallback operation for the AI ​​model, an AI model switching operation, and an AI model deactivation operation, according to the instruction information indicated by the base station based on the report information.

[0164] For example, the terminal may perform a fallback operation to revert from the operation of performing beam management or beam prediction using an AI model to the existing legacy beam management operation. There is no limitation on the specific fallback operation procedure. Alternatively, the terminal may perform a switching operation to change the AI ​​model determined to be degraded to another AI model. Alternatively, the terminal may deactivate the AI ​​model determined to be degraded. These operations may be performed before the terminal transmits the aforementioned report information or after transmitting the report information. Alternatively, these operations may be performed when an instruction from the base station is received.

[0165] Through the above terminal operations, the terminal can monitor the performance of the artificial intelligence model and assist the base station in managing the performance of the artificial intelligence model.

[0166] Fig. 13 is a diagram for explaining base station operation according to one embodiment.

[0167] Referring to FIG. 13, a method for a base station to perform a performance management operation for an artificial intelligence model of a terminal may include a step of transmitting to the terminal at least one reference signal including a first reference signal associated with a first beam predicted using an artificial intelligence model configured in the terminal (S1210).

[0168] For example, the base station can receive information about the first beam predicted by the terminal in advance. The terminal can predict the optimal beam in space or predict the optimal beam at a future time instance in time using reference signals transmitted by the base station. For example, the beam prediction result can be a Top-K beam predicted to have the best quality and a predicted prediction measurement value. K can be configured by the base station, and although the case where K is 1 is described as an example here, the present embodiments can also be applied when K is a natural number exceeding 1.

[0169] The base station can receive the predicted beam prediction results from the terminal. This is to perform pairing between the terminal and the base station using the beam prediction results in beam management.

[0170] For example, the base station can periodically or aperiodically transmit a first reference signal for which resources are set in association with the first beam selected based on the beam prediction results. Since the base station knows the first beam predicted as the optimal beam by the terminal using the beam prediction results, when performance management for the terminal's artificial intelligence model is required, the base station can transmit a reference signal for the first beam to the terminal to perform a performance management operation for the artificial intelligence model.

[0171] The terminal can measure the actual channel quality for the received first reference signal. For example, the terminal can measure the RSRP for the first reference signal. Alternatively, the terminal can derive various channel quality measurements, such as RSRQ and SINR.

[0172] As another example, a terminal may receive multiple reference signals, including a first reference signal, from a base station. For example, when transmitting a first reference signal predicted to be the optimal beam, the base station may also transmit candidate reference signals for other candidate beams to the terminal.

[0173] The terminal can perform measurements on the received first reference signal and candidate reference signals to generate measurement results.

[0174] A method for a base station to perform a performance management operation for an artificial intelligence model of a terminal may include a step of receiving report information from a terminal based on a result of determining whether an event has occurred in the terminal using at least one of a measurement result of a first reference signal, reliability information for a first beam, and a measurement result for reference signals other than the first reference signal (S1220).

[0175] For example, the terminal can determine whether a preset event has occurred to manage the performance of an artificial intelligence model.

[0176] For example, the terminal may determine that an event has occurred if the measurement result of the first reference signal is less than a first threshold value. For example, the terminal receives the first reference signal associated with the optimal beam, compares the measured result with a preset first threshold value, and determines whether an event has occurred based on the comparison result. That is, if the actual measurement result for the first reference signal is lower than the first threshold value, the terminal may determine that there is a problem with the prediction of the first beam, which is the optimal beam predicted by the artificial intelligence model. Accordingly, the terminal may determine that the event condition configured in relation to artificial intelligence model management is satisfied in this case.

[0177] As another example, the terminal may determine that an event has occurred if the measurement result of the first reference signal is less than the first threshold value, and the highest measurement result among the measurement results for reference signals excluding the first reference signal exceeds the second threshold value. For example, the terminal determines whether the actual measurement result of the first reference signal is less than the preset first threshold value. In addition, when receiving multiple reference signals including the first reference signal, the terminal also generates measurement results for the multiple received reference signals. The terminal selects one reference signal having the highest measurement result among the measurement results of reference signals excluding the first reference signal. The terminal determines whether the measurement result of the selected one reference signal exceeds the second threshold value. That is, the terminal determines that the event is satisfied if the actual measurement value of the first reference signal associated with the existing predicted optimal beam deteriorates below the first threshold value, and the measurement results of other reference signals are determined to be greater than or equal to the second threshold value.

[0178] As another example, the terminal may determine that an event has occurred if the reliability information for the first beam is less than a third threshold. For example, the reliability information for the first beam selected as the optimal beam according to the beam prediction result may be calculated. The reliability information may be set as the prediction accuracy, probability value, etc. for the first beam through an artificial intelligence model. For example, the artificial intelligence model applies a softmax function in the output layer of the artificial intelligence model to calculate a probability distribution for N candidate beams. Here, the candidate beam calculated with the highest probability may be selected as the optimal first beam. Therefore, when performing a prediction for the first beam, a probability value for the first beam may be calculated, and the probability value may be used as the reliability information. The terminal may determine that the event condition is satisfied if the reliability information falls below a preset third threshold.

[0179] Here, the first threshold value, the second threshold value, the third threshold value, etc. may be included in the measurement configuration information received from the base station and stored in the terminal.

[0180] Meanwhile, the base station can receive report information from the terminal when a situation occurs that satisfies the aforementioned event conditions. In other words, the satisfaction of the event conditions can act as a trigger for reporting information transmission.

[0181] For example, the report information may include at least one of identification information for the first beam, measurement result information for the first reference signal, and performance monitoring result information for the artificial intelligence model. For example, the report information may include identification information for the first beam predicted by the terminal as the optimal beam and measurement result information for actual measurement performed using the first reference signal. Alternatively, the report information may include performance monitoring result information for the artificial intelligence model, which indicates whether the artificial intelligence model has deteriorated based on the result of determining whether the aforementioned event has occurred.

[0182] As another example, the report information may include identification information for the first beam, measurement result information for the first reference signal, identification information for the second beam indicating the highest measurement result, and the highest measurement result. When determining whether the above-mentioned event has occurred, the first reference signal and other additional reference signals may be used. In this case, the report information may include identification information for the first beam and an actual measurement result value for the first reference signal. In addition, the report information may further include identification information for the second beam selected as having the highest quality among the candidate reference signals and an actual measurement result value of the reference signal associated with the second beam. Through this information, the base station can determine whether a situation requires a beam change of the terminal, etc. In addition, the base station can also determine the suitability of the second beam as the target beam.

[0183] As another example, the report information may include indications indicating performance degradation of the AI ​​model. The indications may include reliability information for the first beam. Alternatively, the indications may simply indicate whether the AI ​​model is degraded.

[0184] A method in which a base station performs a performance management operation for an artificial intelligence model of a terminal may include a step of transmitting performance management instruction information for the artificial intelligence model based on report information (S1230).

[0185] By receiving the aforementioned report information, the base station can verify the performance of the UE-side AI / ML model configured in the terminal. If, based on the report information, the base station determines that the performance of the AI ​​model configured in the terminal has deteriorated, the base station can transmit instruction information to instruct at least one of the following actions: a fallback operation for the AI ​​model, an AI model switching operation, and an AI model deactivation operation.

[0186] For example, the base station may transmit instructions to instruct the base station to perform a fallback operation that reverts the operation performing beam management or beam prediction using an AI model to the existing legacy beam management operation. There are no restrictions on the specific fallback operation procedure. Alternatively, the base station may transmit instructions to instruct the terminal to perform a switching operation that changes the terminal's AI model to a different AI model. Alternatively, the base station may transmit deactivation instructions to deactivate the terminal's AI model.

[0187] Through the above base station operation, the terminal can monitor the performance of the artificial intelligence model and assist the base station in managing the performance of the artificial intelligence model.

[0188]

[0189] Below, each embodiment that can be performed by the aforementioned terminal and base station is described in more detail with reference to the drawings. Each of the embodiments below can be performed by the terminal and base station individually or in any combination.

[0190] We define various trigger conditions to monitor the performance of an AI / ML-based beam prediction system and to detect prediction failures or inaccurate beam selection early. The three conditions presented below cover various prediction error scenarios, including absolute quality of the predicted DL beam, mismatch between the predicted and actual beams, and insufficient prediction reliability. Each condition can be implemented using either Option 1 (NW-side monitoring) or Option 2 (UE-assisted monitoring), as described below.

[0191] FIG. 14 is a diagram for explaining an operation for triggering beam prediction performance monitoring of a terminal according to one embodiment.

[0192] Referring to FIG. 14, the terminal (1301) can transmit beam measurement results to the base station (1302) (S1310). The beam measurement results include quality measurement result information for the reference signal transmitted by the base station (1302).

[0193] The base station (1302) can evaluate the performance of the beam predicted by the terminal (1301) based on the beam measurement results (S1320). If the difference between the beam predicted by the terminal (1301) and the measured value received based on the beam measurement results exceeds a threshold value, the base station (1302) can determine that artificial intelligence model performance management is necessary.

[0194] In this case, the base station (1302) instructs the terminal (1301) to perform a performance evaluation (S1330). The terminal (1301) monitors or confirms the beam prediction performance of the artificial intelligence model configured in the terminal using the beam measurement results and beam prediction results described above (S1340). To monitor the beam prediction performance, the terminal (1301) may also receive additional reference signals from the base station (1302).

[0195] The terminal (1301) can transmit the performance evaluation results derived from the beam prediction performance monitoring results to the base station (1302) (S1350).

[0196] For example, in a situation where there is no trigger for separate beam prediction performance monitoring, the terminal (1301) periodically reports beam measurement results to the base station (1302). In this case, the performance of the artificial intelligence model is calculated by the base station (1302) as described above. In other words, the base station monitors performance according to the aforementioned option 1 method.

[0197] As above, the operation can be classified as follows based on the performance monitoring results of the base station.

[0198] For example, if a base station monitors performance and finds that performance falls below a threshold, it determines that an event has occurred. In this case, the base station can transmit instructions to the terminal to manage the AI ​​model.

[0199] Alternatively, if the performance monitoring result indicates that the performance is below a threshold, the base station may instruct the terminal to perform beam prediction performance monitoring for additional monitoring. That is, the operation of option 2 (UE auxiliary performance monitoring) described above may be instructed to be performed.

[0200] As another example, even if the base station monitors performance and finds that the performance is higher than the threshold, if continuous performance monitoring determines that the performance value is declining, instruction information for managing the artificial intelligence model can be transmitted as described above.

[0201] Alternatively, the base station may instruct the terminal to monitor beam prediction performance for additional monitoring, i.e., may instruct the operation of option 2 (UE auxiliary performance monitoring) described above to be performed.

[0202] This procedure allows the base station to periodically manage the AI ​​model performance using the terminal's beam measurement results and, when certain criteria are met, to request beam prediction performance management for the terminal. This procedure can reduce system overhead while continuously monitoring the AI ​​model. The operation of the terminal monitoring beam prediction performance can be applied as described with reference to FIGS. 12 and 13.

[0203] Below, the conditions for the occurrence of the aforementioned event and the actions for judging it are explained in detail.

[0204] 1) When the L1-RSRP of the predicted Top-1 beam is lower than the absolute threshold.

[0205] Ms + Hys <Thresh_RSRP (Ms: 예측 Top-1 빔의 측정 L1-RSRP)

[0206] The terminal (UE) predicts the optimal DL beam to be applied in the future through an AI / ML model based on the CSI-RS of the resource set (Set B) measured at a past time or location. At this time, the beam with the highest confidence score among the predicted results is defined as the Top-1 beam. The terminal measures the power of the DL reference signal (e.g., CSI-RS) received based on this Top-1 beam at the physical layer (Layer 1). The corresponding received power is defined as L1-RSRP (Reference Signal Received Power). In this embodiment, the L1-RSRP value is named Ms, and if Ms is lower than the absolute threshold Thresh_RSRP, it can be determined that the performance of the beam prediction is degraded.

[0207] That is, the terminal can configure the Event condition for comparison. The event condition is defined in the form of Ms + Hys < Thresh_RSRP, where Hys is a hysteresis margin to prevent malfunction due to temporary measurement fluctuations and is set to 2 dB by example. Thresh_RSRP is an absolute received power reference value and can be set to -110 dBm in mmWave environments. This value can be adjusted according to the cell coverage conditions in the system environment.

[0208] For example, if the above conditions are satisfied, the terminal reports the corresponding measurement value to the base station, and the base station determines whether the predicted beam performance has deteriorated based on the measurement value and performs the necessary control action.

[0209] As another example, if the above conditions are satisfied, the terminal may independently perform a performance degradation judgment and perform subsequent actions such as reporting to the base station.

[0210] Referring to FIGS. 15 and 16, an operational example is described when the L1-RSRP of the predicted Top-1 beam is lower than the absolute threshold.

[0211] FIG. 15 is a diagram for explaining the performance monitoring operation of an artificial intelligence model according to one embodiment.

[0212] Referring to FIG. 15, the terminal (1401) predicts the Top-1 DL beam through the AI / ML model (S1410).

[0213] The terminal continuously receives L1-RSRP Ms of signals (CSI-RS, etc.) received from the predicted corresponding beam (S1420).

[0214] The terminal (1401) measures the RSRP of the received Top-1 beam and compares it with a threshold value (S1430). For example, the terminal (1401) determines whether the predicted Top-1 beam's RSRP Ms is lower than the absolute threshold value, thereby satisfying the Low RSRP condition (i.e., Ms + Hys < Thresh_RSRP, thus satisfying the Event criterion).

[0215] If the event criteria are met, measurement result transmission is triggered (S1435), and the terminal (1401) reports the measurement results to the base station (1402) (S1440). This report may include the identifier of the current Top-1 beam and the measured RSRP value.

[0216] The base station (1402) confirms that the RSRP value reported from the terminal (1401) is below the threshold and recognizes it as a performance degradation event of the predicted beam. Based on this, the base station (1402) can perform the following response procedure (S1450).

[0217] i) Additional measurement instruction: The base station (1402) can issue an additional beam measurement instruction to the terminal (1401). For example, it requests that the terminal measure and report the RSRP of other candidate beams within the current cell. This allows for identifying alternative beam candidates.

[0218] ii) Beam reselection: If there is a beam with higher signal quality among the beams additionally reported by the terminal (1401), the base station (1402) instructs the terminal (1401) to switch to the optimal beam. The base station (1402) changes the downlink beam by transmitting a new beam ID to the terminal (1401) through an RRC reconfiguration or beam control message.

[0219] iii) Fallback execution: The base station (1402) may determine that the reliability of the current AI prediction-based procedure is low and may command a fallback to the existing beam management method. For example, it may revert to the standard beam sweeping / beam measurement procedure for a certain period of time to maintain a stable link.

[0220] iv) Model Performance Management: The base station (1402) collects information on the occurrence of the event and records that it is due to a performance degradation of the AI ​​model. If necessary, the base station (1402) considers model retraining or parameter adjustment and plans measures to improve future prediction accuracy.

[0221] The terminal (1401) performs the requested additional measurements under the control of the base station (1402), reports the results, and switches to a new beam indicated by the base station (1402). This ensures link quality in case of prediction failure due to absolute RSRP deficiency.

[0222] Fig. 16 is a diagram for explaining the performance monitoring operation of an artificial intelligence model according to another embodiment.

[0223] Referring to FIG. 16, the terminal (1501) predicts a Top-1 DL beam using an AI / ML model (S1510). The terminal continuously receives L1-RSRP Ms of signals (CSI-RS, etc.) received from the predicted beam (S1520).

[0224] The terminal (1501) measures the RSRP of the received Top-1 beam and compares it with a threshold (S1530). That is, the terminal (1501) monitors the L1-RSRP Ms of the beam (measured internally by the UE without reporting to the NW).

[0225] If the predicted beam RSRP Ms falls below the threshold value and satisfies the Low RSRP condition (Ms < Thresh_RSRP), the terminal (1501) immediately determines that the predicted beam performance has deteriorated.

[0226] Upon detecting a Low RSRP event, the terminal (1501) initiates a fallback procedure on its own (S1535). The fallback procedure may include at least one of the following actions.

[0227] i) Auxiliary beam search: The terminal (1501) additionally measures the RSRP of other available DL beams within the current cell. Through this, it finds an alternative beam to use instead of the current Top-1 beam.

[0228] ii) Immediate beam switching: If the terminal (1501) discovers a beam with better quality as a result of additional measurements, it switches the receiving beam at its own discretion. This is an operation in which the terminal (1501) temporarily performs autonomous beam reselection to maintain communication quality.

[0229] The terminal (1501) can report the occurrence of the Low RSRP event to the base station (1502) according to the established policy (S1540). In Option 2, the event is selectively reported to the base station (1502) only when a threshold condition is met, so the terminal (1501) transmits a "predicted beam RSRP degradation event" to the base station (1502) in the form of an RRC measurement report, etc. The report may include the RSRP value of the current Top-1 beam and (optionally) information about an alternative beam discovered by the UE.

[0230] Upon receiving an event report from the UE, the base station (1502) recognizes that the UE has replaced the predicted beam and, if necessary, takes additional action (S1545). For example, the base station (1502) verifies the replacement beam reported by the terminal (1501) and synchronizes the downlink beam settings to that beam (beam synchronization between the UE and the network). Furthermore, the base station (1502) can recognize a decrease in AI model reliability through the event and consider future model updates or parameter changes.

[0231] The terminal (1501) internally records occurrences of Low RSRP events and determines whether model retraining or model switching is necessary. For example, if this event occurs repeatedly, the terminal (1501) may determine that the prediction accuracy of the current AI model is inadequate for the environment and send a new model request or auxiliary data to the base station (1502).

[0232] 2) When there is a quality discrepancy between the predicted Top-1 beam and the actual optimal beam.

[0233] Mp + Hys <Thresh1 (Mp: 예측된 Top-1빔에서 측정된 실제 RSRP)

[0234] Mn - Hys >Thresh2 (Mn: highest RSRP among the actually measured Top-K candidate beams)

[0235] The terminal directly measures the L1-RSRP of each predicted Top-1 beam and actual beam candidates, and autonomously determines the prediction accuracy based on the quality difference between them. In this embodiment, if the L1-RSRP (Mp) of the predicted beam is lower than the threshold Thresh_1, and at the same time, the highest RSRP (Mn) among the actual candidate beams is higher than Thresh_2, the predicted beam is determined to be mismatched with the actual optimal beam.

[0236] Here, the actual candidate beam is a Top-K beam selected by sorting the top K beams with the highest L1-RSRP among the N total DL beam candidates that can be measured at the same time by the terminal, and the RSRP of the beam with the highest received power is defined as Mn. The total number of beam candidates N can be dynamically set according to the system configuration and frequency band. For example, in Sub-6GHz, it can be set to N=8 or 16, and in mmWave, it can be set to N=32 or 64. The value of K can vary depending on the system environment, UE capability, and AI prediction model structure, and is typically set to K=4 or K=8 as an example. The value of K can also be determined by the instruction of the base station.

[0237] That is, the terminal can make a judgment based on the corresponding Event. The corresponding Event is composed of dual comparison conditions in the form of Mp + Hys < Thresh1 and Mn - Hys > Thresh2, where Hys can be set to a margin of 2 dB. The Thresh1 and Thresh2 values ​​can be set differently depending on the Sub-6 GHz or mmWave environment, and can be configured to levels of -85 dBm and -80 dBm, respectively, for example. The terminal reports a mismatch event to the base station only when all of the above conditions are met, which allows selective reporting only in situations where there is a high possibility of error, without reporting all prediction results.

[0238] Referring to Fig. 17, a case where the quality between the predicted Top-1 beam and the actual optimal beam is inconsistent is described.

[0239] Fig. 17 is a diagram for explaining the performance monitoring operation of an artificial intelligence model according to another embodiment.

[0240] Referring to Figure 17, if there is a significant difference between the predicted Top-1 beam quality and the optimal beam quality in the actual environment, the prediction result can be considered incorrect. To this end, two threshold conditions are utilized simultaneously to determine whether an event is satisfied.

[0241] Mp: Predicted beam's measured power (RSRP) of the predicted Top-1 beam

[0242] Mn: The highest RSRP value among the actual candidate beams measured by the UE at the same time (signal strength of the optimal actual beam)

[0243] Condition: A mismatch event is considered when Mp + Hys < Thresh1 and Mn - Hys > Thresh2 are simultaneously satisfied. Here, Hys is the hysteresis margin (e.g., 2 dB), and Thresh1 and Thresh2 are thresholds set by the NW (e.g., -85 dBm and -80 dBm in sub-6 GHz). This dual condition reliably detects a prediction error by checking both the absolute performance degradation of the predicted beam and the good performance of the other beam.

[0244] The terminal (1601) predicts the Top-1 beam using its own AI model (S1610). After the prediction, the terminal (1601) receives multiple candidate beams in a real environment and measures the RSRP in parallel (S1620). (In Option 2, since the terminal determines the prediction accuracy on its own, additional measurements are performed under the leadership of the UE to verify the prediction results.) At this time, the terminal (1601) obtains the RSRP Mp of the current Top-1 beam and the maximum value Mn among the RSRPs of the top K candidate beams defined in advance.

[0245] The terminal compares the candidate beam measurements and thresholds (S1630). For example, the terminal (1601) internally determines whether the mismatch conditions (Mp + Hys < Thresh1 and Mn - Hys > Thresh2) are met. If both conditions are met, the terminal (1601) concludes that the predicted Top-1 beam is significantly mismatched with the actual optimal beam (S1635). If the conditions are not met, the terminal (1601) considers the prediction not to be significantly off, and continues to use the predicted beam without further reporting. If the conditions are met, the terminal (1601) initiates a response by triggering a prediction failure event.

[0246] When a mismatch event occurs, the terminal (1601) can take measures to maintain communication quality. For example, the measures may include the following actions.

[0247] i) Optimal beam utilization: Since the UE knows the beam with the highest RSRP among the candidates already measured, it immediately considers switching the reception beam to that beam.

[0248] ii) Stop model validation: Since the current prediction is wrong, the terminal (1601) temporarily distrusts the output of the AI ​​model and falls back to traditional measurement-based operation and prepares to operate.

[0249] The terminal (1601) optionally reports to the base station (1602) only when a mismatch condition is met. The report may include the predicted beam identifier and its RSRP (Mp), the identifier of the actual optimal beam and RSRP (Mn), etc. This allows the base station (1602) to immediately identify the current prediction failure situation and alternative beam information.

[0250] When the base station (1602) receives a mismatch event report from the terminal (1601), it can immediately recognize that it is a prediction failure situation and execute at least one of the following actions (S1645).

[0251] i) DL Beam Correction: Based on the actual optimal beam information reported by the terminal (1601), the base station (1602) changes the downlink beam to the corresponding beam. In Option 2, the terminal (1601) has already evaluated and reported the candidate beam, so the base station (1602) can quickly switch to the optimal beam without an additional command. The base station (1602) restores the link quality by transmitting data toward this beam.

[0252] ii) Model Management: The base station (1602) records the occurrence of a prediction error in the corresponding terminal (1601) and, if necessary, takes AI model management action. The base station (1602) comprehensively analyzes reports from multiple terminals (1601) and, if a common model performance issue is identified, considers overall model retraining.

[0253] iii) Additional instructions: The base station (1602) confirms that the terminal (1601) has already performed its own beam switching, and confirms / synchronizes the beam switching with an RRC signal. In addition, settings such as increasing the additional measurement frequency or adjusting the threshold value can be changed to respond to similar situations.

[0254] The terminal (1601) continues communication using the optimal beam according to the response instructions of the base station (1602) and maintains a beam state synchronized with the base station (1602). The terminal (1601) internally stores this mismatch case as learning data and considers model retraining or model replacement.

[0255] 3) If the reliability of the predicted Top-1 beam is below the threshold value

[0256] confidence < Thresh_confi

[0257] For example, a terminal-side AI / ML prediction module predicts the optimal downlink (DL) beam to be applied in the future based on measurement information, such as CSI-RS, from a resource set collected at a past time or location. The terminal outputs the top-1 beam identifier and its corresponding confidence score as a prediction result.

[0258] Here, the confidence score is a value indicating the probability that a specific predicted beam is actually the optimal beam, and is generally expressed as a probability distribution for N candidate beams by applying a softmax function to the output layer of the AI ​​model. The N candidate beams are determined according to the system configuration and AI model structure, and can be set to 8, 16, 32, etc. This probability distribution is based on the sum of 1, and the probability value corresponding to the top-1 beam can directly mean the prediction reliability.

[0259] The confidence score value is compared with a threshold probability value called Thresh_confi to determine whether the trigger condition is satisfied. The UE determines that a prediction failure or insufficient confidence condition exists if the confidence of the top-1 beam is less than Thresh_confi and uses this condition as a trigger for a prediction uncertainty event. The threshold Thresh_confi is configured according to the base station or terminal settings and can be set to, for example, 0.7 or 0.5. For example, in a Sub-6GHz environment, the channel stability is high, so the confidence threshold can be set to 0.7 or higher to strictly require prediction accuracy. In a mmWave environment, the channel variability is high and the number of candidate beams is large, so a looser standard of 0.5 can be set. In this way, by differentially configuring Thresh_confi for each environment, prediction failures can be effectively detected while controlling trigger sensitivity. If the above condition is satisfied, i.e., confidence < Thresh_confi, the terminal determines it as a prediction uncertainty event and performs subsequent beam management actions on its own, such as initiating an auxiliary beam quality verification procedure, requesting retraining of the existing model, or switching models.

[0260] Referring to Fig. 18, the operation when the reliability of the predicted Top-1 beam is less than the threshold value is described.

[0261] Fig. 18 is a diagram for explaining the performance monitoring operation of an artificial intelligence model according to another embodiment.

[0262] Referring to Figure 18, if the confidence score of the Top-1 beam is lower than a preset threshold (Thresh_confi) during the AI ​​model prediction stage, the prediction is judged to be unreliable. The threshold probability value Thresh_confi varies depending on the NW or UE settings (e.g., it can be set to 0.7 in Sub-6GHz with low channel fluctuations, or 0.5 in mmWave with high channel fluctuations) and is adjusted accordingly.

[0263] The terminal (1701) predicts the Top-1 beam using the AI ​​prediction module (S1710). The terminal (1701) calculates a confidence score for the Top-1 beam (S1720). The terminal (1701) compares this value with the threshold probability value Thresh_confi to independently evaluate whether the Low Confidence condition is satisfied (S1730). If confidence ≥ Thresh_confi, the terminal (1701) considers the prediction to be reliable and proceeds with normal beam setting. If confidence < Thresh_confi, the terminal (1701) determines that there is a lack of confidence in the prediction result and generates a Low Confidence trigger event.

[0264] When a Low Confidence event is triggered, the terminal (1701) independently performs a verification procedure for the prediction. The terminal (1701) performs at least one of the following actions.

[0265] i) Auxiliary beam quality verification: Instead of applying the currently predicted Top-1 beam as is, the terminal (1701) initiates an additional beam measurement procedure.

[0266] ii) Temporary beam setting: If additional measurements reveal a beam with a much stronger signal than the predicted beam, the terminal (1701) can ignore the prediction and adopt that beam. In other words, when confidence is low, decisions based on actual data are prioritized over those based on the AI ​​model, thereby protecting link quality.

[0267] The terminal (1701) may handle the Low Confidence situation on its own without reporting it, or may report related information to the base station (1702) or request support as needed (S1740).

[0268] When a new beam report or model-related request is received from a terminal (1701) that is different from the usual, the base station (1704) infers that the terminal (1701) is experiencing a prediction reliability issue. The base station (1702) adjusts communication to the new beam reported by the terminal (1701) and, if necessary, provides additional support to the terminal (1701) (S1745).

[0269] i) Model retraining: If the terminal (1701) has an on-device learning function, the model is partially retrained or local correction is performed using the data accumulated to date.

[0270] ii) Model switching: If the terminal (1701) has stored multiple AI models in advance, it can switch to an alternative model that is more suitable for the current environment.

[0271] iii) Model Deactivation: If a low confidence situation occurs continuously, the terminal (1701) can temporarily deactivate the AI ​​prediction function and return to a completely traditional beam management procedure.

[0272] This decision can be made by the terminal itself or by instruction from the base station and reported to the base station via an RRC reset.

[0273] According to the present disclosure, event-based triggering conditions can be defined and optimized to reduce unnecessary reporting and conserve network resources. Furthermore, signs of performance degradation can be detected early and reported in a timely manner based on internal terminal measurements or AI / ML model output.

[0274] Below, the aforementioned terminals and base stations are briefly described from a configuration perspective. To avoid unnecessary duplication, redundant portions may be omitted.

[0275] Fig. 19 is a diagram showing a terminal configuration according to one embodiment.

[0276] Referring to FIG. 19, a terminal (1800) that performs a performance management operation of an artificial intelligence model may include a control unit (1810) that generates a beam prediction result using an artificial intelligence model configured in the terminal, and a receiving unit (1830) that receives at least one reference signal including a first reference signal associated with a first beam selected according to the beam prediction result. In addition, the terminal (1800) may include a transmitting unit (1820) that transmits report information to a base station based on a determination of whether an event has occurred.

[0277] The control unit (1810) can determine whether an event has occurred by using at least one of the measurement results of the first reference signal, the reliability information for the first beam, and the measurement results for reference signals other than the first reference signal.

[0278] For example, the control unit (1810) can predict a beam using an artificial intelligence model as described herein. As an example, the control unit (1810) can predict an optimal beam in space, as in BM-Case 1. As another example, the control unit (1810) can also predict an optimal beam in time, as in BM-Case 2.

[0279] To this end, the control unit (1810) can input the measurement results for the actual beam into the artificial intelligence model and obtain the beam prediction results as output. For example, the beam prediction results may be the Top-K beams predicted to have the best quality and the predicted predicted measurement values. K may be configured by the base station, and although the case where K is 1 is described as an example here, the present embodiments may also be applied when K is a natural number exceeding 1.

[0280] The transmitter (1820) can transmit the predicted beam prediction result to the base station. This is to enable the terminal and the base station to perform pairing using the beam prediction result in beam management.

[0281] For example, the receiving unit (1830) may periodically or aperiodically receive a first reference signal for which resources are set in association with a first beam selected according to a beam prediction result from the base station. Since the base station knows about the first beam predicted as the optimal beam by the terminal using the beam prediction result, when performance management for the artificial intelligence model of the terminal is required, the base station may transmit a reference signal for the first beam to the terminal to perform a performance management operation for the artificial intelligence model.

[0282] The control unit (1810) can measure the actual channel quality for the received first reference signal. For example, the control unit (1810) can measure the RSRP for the first reference signal. Alternatively, the control unit (1810) can derive various channel quality measurement values, such as RSRQ and SINR.

[0283] As another example, the receiver (1830) may receive multiple reference signals, including a first reference signal, from a base station. For example, when transmitting the first reference signal predicted as the optimal beam, the base station may also transmit candidate reference signals for other candidate beams to the terminal.

[0284] The control unit (1810) can perform measurements on the received first reference signal and candidate reference signals to generate measurement results.

[0285] For example, the control unit (1810) can determine whether a preset event has occurred to manage the performance of an artificial intelligence model.

[0286] For example, the control unit (1810) may determine that an event has occurred if the measurement result of the first reference signal is less than a first threshold value. For example, the control unit (1810) receives the first reference signal associated with the optimal beam, compares the measured result with a preset first threshold value, and determines whether an event has occurred based on the comparison result. That is, if the actual measurement result for the first reference signal is lower than the first threshold value, the control unit (1810) may determine that there is a problem with the prediction of the first beam, which is the optimal beam predicted by the artificial intelligence model. Accordingly, the control unit (1810) may determine that in this case, the event condition configured in relation to artificial intelligence model management is satisfied.

[0287] As another example, the control unit (1810) may determine that an event has occurred if the measurement result of the first reference signal is less than a first threshold value, and the highest measurement result among the measurement results for reference signals excluding the first reference signal exceeds a second threshold value. For example, the control unit (1810) determines whether the actual measurement result of the first reference signal is less than a preset first threshold value. In addition, when receiving multiple reference signals including the first reference signal, the control unit (1810) also generates measurement results for the multiple received reference signals. The control unit (1810) selects one reference signal having the highest measurement result among the measurement results of reference signals excluding the first reference signal. The control unit (1810) determines whether the measurement result of the selected one reference signal exceeds the second threshold value. That is, the control unit (1810) determines that the event is satisfied when the actual measurement value of the first reference signal associated with the existing predicted optimal beam deteriorates below the first threshold value and the measurement result of another reference signal is determined to be above the second threshold value.

[0288] As another example, the control unit (1810) may determine that an event has occurred when the reliability information for the first beam is less than a third threshold value. For example, the reliability information for the first beam selected as the optimal beam according to the beam prediction result may be calculated. The reliability information may be set as the prediction accuracy, probability value, etc. for the first beam through an artificial intelligence model. For example, the artificial intelligence model applies a softmax function in the output layer of the artificial intelligence model to calculate a probability distribution for N candidate beams. Here, the candidate beam calculated with the highest probability may be selected as the optimal first beam. Therefore, when performing a prediction for the first beam, a probability value for the corresponding first beam may be calculated, and the probability value may be used as the reliability information. The control unit (1810) may also determine that the event condition is satisfied when the reliability information falls below a preset third threshold value.

[0289] Here, the first threshold value, the second threshold value, the third threshold value, etc. may be included in the measurement configuration information received from the base station and stored in the terminal.

[0290] Meanwhile, the transmitter (1820) can transmit report information to the base station when a situation occurs that satisfies the aforementioned event conditions. In other words, the satisfaction of the event conditions can act as a trigger for transmitting report information.

[0291] For example, the report information may include at least one of identification information for the first beam, measurement result information for the first reference signal, and performance monitoring result information for the artificial intelligence model. For example, the report information may include identification information for the first beam predicted by the terminal as the optimal beam and measurement result information for actual measurement performed using the first reference signal. Alternatively, the report information may include performance monitoring result information for the artificial intelligence model, which indicates whether the artificial intelligence model has deteriorated based on the result of determining whether the aforementioned event has occurred.

[0292] As another example, the report information may include identification information for the first beam, measurement result information for the first reference signal, identification information for the second beam indicating the highest measurement result, and the highest measurement result. When determining whether the above-mentioned event has occurred, the first reference signal and other additional reference signals may be used. In this case, the report information may include identification information for the first beam and an actual measurement result value for the first reference signal. In addition, the report information may further include identification information for the second beam selected as having the highest quality among the candidate reference signals and an actual measurement result value of the reference signal associated with the second beam. Through this information, the base station can determine whether a situation requires a beam change of the terminal, etc. In addition, the second beam may be identified as the target beam.

[0293] As another example, the report information may include indications indicating performance degradation of the AI ​​model. The indications may include reliability information for the first beam. Alternatively, the indications may simply indicate whether the AI ​​model is degraded.

[0294] By receiving the aforementioned report information, the base station can check the performance of the UE-side AI / ML model configured in the terminal. If the control unit (1810) determines that an event has occurred, the control unit (1810) can perform at least one of the following operations: a fallback operation for the AI ​​model, an AI model switching operation, and an AI model deactivation operation. Alternatively, the control unit (1810) can perform at least one of the following operations: a fallback operation for the AI ​​model, an AI model switching operation, and an AI model deactivation operation, according to the instruction information indicated by the base station based on the report information.

[0295] For example, the control unit (1810) may perform a fallback operation to revert the operation of performing beam management or beam prediction using an artificial intelligence model to the existing legacy beam management operation. There is no limitation on the specific fallback operation procedure. Alternatively, the control unit (1810) may perform a switching operation to change the artificial intelligence model determined to be degraded to another artificial intelligence model. Alternatively, the control unit (1810) may deactivate the artificial intelligence model determined to be degraded. This operation may be performed before the terminal transmits the aforementioned report information or after transmitting the report information. Alternatively, this operation may be performed when an instruction from the base station is received.

[0296] In addition, the receiver (1830) receives downlink control information, data, and messages from the base station through the corresponding channel.

[0297] Additionally, the control unit (1820) controls the overall operation of the user terminal (1800) according to the performance monitoring method of the AI / ML model for beam management required to perform the aforementioned disclosure.

[0298] The transmitter (1830) transmits uplink control information, data, and messages to the base station through the corresponding channel.

[0299] Figure 20 is a diagram showing a base station configuration according to one embodiment.

[0300] Referring to FIG. 20, a base station (1900) that performs a performance management operation for an artificial intelligence model of a terminal may include a transmitter (1920) that transmits to the terminal at least one reference signal including a first reference signal associated with a first beam predicted using an artificial intelligence model configured in the terminal, and a receiver (1930) that receives report information from the terminal based on a result of determining whether an event has occurred in the terminal using at least one of a measurement result of the first reference signal, reliability information for the first beam, and measurement results for reference signals excluding the first reference signal. The transmitter (1920) may transmit performance management instruction information for the artificial intelligence model based on the report information.

[0301] The receiving unit (1930) can receive information about the first beam predicted by the terminal in advance. The terminal can predict the optimal beam in space or predict the optimal beam at a future time instance in time using reference signals transmitted by the base station. For example, the beam prediction result can be a Top-K beam predicted to have the best quality and a predicted prediction measurement value. K can be configured by the base station, and although the case where K is 1 is described as an example here, the present embodiments can also be applied when K is a natural number exceeding 1.

[0302] The receiving unit (1930) can receive the predicted beam prediction result from the terminal. This is to perform pairing between the terminal and the base station using the beam prediction result in beam management.

[0303] For example, the transmitter (1920) may periodically or aperiodically transmit a first reference signal for which resources are set in association with a first beam selected according to a beam prediction result. Since the control unit (1910) knows about the first beam predicted as the optimal beam by the terminal using the beam prediction result, when performance management for the artificial intelligence model of the terminal is required, the control unit (1910) may transmit a reference signal for the first beam to the terminal to perform a performance management operation for the artificial intelligence model.

[0304] The terminal can measure the actual channel quality for the received first reference signal. For example, the terminal can measure the RSRP for the first reference signal. Alternatively, the terminal can derive various channel quality measurements, such as RSRQ and SINR.

[0305] As another example, a terminal may receive multiple reference signals, including a first reference signal, from a base station. For example, when transmitting a first reference signal predicted to be the optimal beam, the transmitter (1920) may also transmit candidate reference signals for other candidate beams to the terminal.

[0306] The terminal can perform measurements on the received first reference signal and candidate reference signals to generate measurement results.

[0307] The terminal can determine whether a preset event has occurred to manage the performance of the artificial intelligence model.

[0308] For example, the terminal may determine that an event has occurred if the measurement result of the first reference signal is less than a first threshold value. For example, the terminal receives the first reference signal associated with the optimal beam, compares the measured result with a preset first threshold value, and determines whether an event has occurred based on the comparison result. That is, if the actual measurement result for the first reference signal is lower than the first threshold value, the terminal may determine that there is a problem with the prediction of the first beam, which is the optimal beam predicted by the artificial intelligence model. Accordingly, the terminal may determine that the event condition configured in relation to artificial intelligence model management is satisfied in this case.

[0309] As another example, the terminal may determine that an event has occurred if the measurement result of the first reference signal is less than the first threshold value, and the highest measurement result among the measurement results for reference signals excluding the first reference signal exceeds the second threshold value. For example, the terminal determines whether the actual measurement result of the first reference signal is less than the preset first threshold value. In addition, when receiving multiple reference signals including the first reference signal, the terminal also generates measurement results for the multiple received reference signals. The terminal selects one reference signal having the highest measurement result among the measurement results of reference signals excluding the first reference signal. The terminal determines whether the measurement result of the selected one reference signal exceeds the second threshold value. That is, the terminal determines that the event is satisfied if the actual measurement value of the first reference signal associated with the existing predicted optimal beam deteriorates below the first threshold value, and the measurement results of other reference signals are determined to be greater than or equal to the second threshold value.

[0310] As another example, the terminal may determine that an event has occurred if the reliability information for the first beam is less than a third threshold. For example, the reliability information for the first beam selected as the optimal beam according to the beam prediction result may be calculated. The reliability information may be set as the prediction accuracy, probability value, etc. for the first beam through an artificial intelligence model. For example, the artificial intelligence model applies a softmax function in the output layer of the artificial intelligence model to calculate a probability distribution for N candidate beams. Here, the candidate beam calculated with the highest probability may be selected as the optimal first beam. Therefore, when performing a prediction for the first beam, a probability value for the first beam may be calculated, and the probability value may be used as the reliability information. The terminal may determine that the event condition is satisfied if the reliability information falls below a preset third threshold.

[0311] Here, the first threshold value, the second threshold value, the third threshold value, etc. may be included in the measurement configuration information received from the base station and stored in the terminal.

[0312] Meanwhile, the receiver (1930) can receive report information from the terminal when a situation occurs that satisfies the aforementioned event conditions. In other words, the satisfaction of the event conditions can act as a trigger for transmitting report information.

[0313] For example, the report information may include at least one of identification information for the first beam, measurement result information for the first reference signal, and performance monitoring result information for the artificial intelligence model. For example, the report information may include identification information for the first beam predicted by the terminal as the optimal beam and measurement result information for actual measurement performed using the first reference signal. Alternatively, the report information may include performance monitoring result information for the artificial intelligence model, which indicates whether the artificial intelligence model has deteriorated based on the result of determining whether the aforementioned event has occurred.

[0314] As another example, the report information may include identification information for the first beam, measurement result information for the first reference signal, identification information for the second beam indicating the highest measurement result, and the highest measurement result. When determining whether the above-mentioned event has occurred, the first reference signal and other additional reference signals may be used. In this case, the report information may include identification information for the first beam and an actual measurement result value for the first reference signal. In addition, the report information may further include identification information for the second beam selected as having the highest quality among the candidate reference signals and an actual measurement result value of the reference signal associated with the second beam. Through this information, the base station can determine whether a situation requires a beam change of the terminal, etc. In addition, the base station can also determine the suitability of the second beam as the target beam.

[0315] As another example, the report information may include indications indicating performance degradation of the AI ​​model. The indications may include reliability information for the first beam. Alternatively, the indications may simply indicate whether the AI ​​model is degraded.

[0316] By receiving the aforementioned report information, the control unit (1910) can check the performance of the UE-side AI / ML model configured in the terminal. If it is determined that the performance of the AI ​​model configured in the terminal has deteriorated based on the report information, the transmitter unit (1920) can transmit instruction information to instruct at least one of a fallback operation for the AI ​​model, an AI model switching operation, and an AI model deactivation operation.

[0317] For example, the transmitter (1920) may transmit instruction information to instruct the perform a fallback operation that returns the operation performing beam management or beam prediction using the artificial intelligence model to the existing legacy beam management operation. There is no limitation on the specific fallback operation procedure. Alternatively, the transmitter (1920) may transmit instruction information to instruct the perform a switching operation that changes the artificial intelligence model of the terminal to another artificial intelligence model. Alternatively, the transmitter (1920) may transmit deactivation instruction information to deactivate the artificial intelligence model of the terminal.

[0318] In addition, the control unit (1910) controls the overall operation of the base station (1900) according to the performance monitoring method of the AI / ML model for beam management required to perform the aforementioned disclosure.

[0319] The transmitter (1920) and receiver (1930) are used to transmit and receive signals, messages, and data necessary for performing the aforementioned embodiment to and from the terminal.

[0320] The above-described embodiments may be supported by standard documents disclosed in at least one of the wireless access systems, IEEE 802, 3GPP, and 3GPP2. That is, steps, components, and parts not described in the present embodiments to clearly illustrate the technical concepts herein may be supported by the above-described standard documents. Furthermore, all terms disclosed in this specification may be explained by the above-described standard documents.

[0321] The embodiments described above may be implemented through various means. For example, the embodiments may be implemented through hardware, firmware, software, or a combination thereof.

[0322] In the case of hardware implementation, the method according to the present embodiments may be implemented by one or more ASICs (Application Specific Integrated Circuits), DSPs (Digital Signal Processors), DSPDs (Digital Signal Processing Devices), PLDs (Programmable Logic Devices), FPGAs (Field Programmable Gate Arrays), processors, controllers, microcontrollers, or microprocessors.

[0323] When implemented using firmware or software, the methods according to the present embodiments may be implemented in the form of devices, procedures, or functions that perform the functions or operations described above. The software code may be stored in a memory unit and executed by a processor. The memory unit may be located within or outside the processor and may exchange data with the processor using various known means.

[0324] Additionally, terms such as "system," "processor," "controller," "component," "module," "interface," "model," or "unit" as described above may generally refer to a computer-related entity, such as hardware, a combination of hardware and software, software, or software in execution. For example, the aforementioned components may be, but are not limited to, a process driven by a processor, a processor, a controller, a control processor, an object, a thread of execution, a program, and / or a computer. For example, both an application running on a controller or a processor and the controller or the processor may be components. One or more components may be within a process and / or thread of execution, and the components may be located on a single device (e.g., a system, a computing device, etc.) or distributed across two or more devices.

[0325] The above description is merely an illustrative example of the technical idea of ​​the present disclosure, and those skilled in the art to which the present disclosure pertains will appreciate that various modifications and variations can be made without departing from the essential characteristics of the technical idea of ​​the present disclosure. In addition, the present embodiments are not intended to limit the technical idea of ​​the present disclosure but rather to explain it, and therefore the scope of the technical idea of ​​the present disclosure is not limited by these embodiments. The scope of protection of the present disclosure should be interpreted by the claims below, and all technical ideas within a scope equivalent thereto should be interpreted as being included within the scope of the rights of the present disclosure.

[0326]

[0327] CROSS-REFERENCE TO RELATED APPLICATION

[0328] This patent application claims priority under 35 USC §119(a) to Korean Patent Application No. 10-2024-0100887, filed in Korea on July 30, 2024, Korean Patent Application No. 10-2025-0085878, filed in Korea on June 27, 2025, and Korean Patent Application No. 10-2025-0102584, filed in Korea on July 28, 2025, the entire contents of which are incorporated herein by reference. In addition, this patent application claims priority in countries other than the United States for the same reasons, the entire contents of which are incorporated herein by reference.

Claims

1. In a method for a terminal to perform performance management operations of an artificial intelligence model, A step of generating a beam prediction result using an artificial intelligence model configured in a terminal; A step of receiving at least one reference signal including a first reference signal associated with a first beam selected according to the beam prediction result; A step of determining whether an event has occurred using at least one of the measurement result of the first reference signal, reliability information for the first beam, and measurement results for reference signals other than the first reference signal; and A method comprising a step of transmitting report information to a base station based on a determination of whether the above event has occurred.

2. In paragraph 1, The step of determining whether the above event has occurred is: A method for determining that the event has occurred when the measurement result of the first reference signal is less than the first threshold value.

3. In paragraph 2, The above report information is, A method comprising at least one of identification information for the first beam, measurement result information of the first reference signal, and performance monitoring result information of the artificial intelligence model.

4. In paragraph 1, The step of determining whether the above event has occurred is: A method for determining that the event has occurred when the measurement result of the first reference signal is less than the first threshold value and the highest measurement result among the measurement results for reference signals excluding the first reference signal exceeds the second threshold value.

5. In paragraph 4, The above report information is, A method comprising identification information for the first beam, measurement result information of the first reference signal, second beam identification information indicating the highest value measurement result, and the highest value measurement result.

6. In paragraph 1, The step of determining whether the above event has occurred is: A method for determining that the event has occurred when the reliability information for the first beam is less than a third threshold value.

7. In paragraph 6, The above report information is, A method including instruction information indicating performance degradation of the above artificial intelligence model.

8. In paragraph 1, The above terminal, A method of performing at least one of a fallback operation, an artificial intelligence model switching operation, and an artificial intelligence model deactivation operation for the artificial intelligence model when the above event is determined to have occurred.

9. In a method for a base station to perform performance management operations for an artificial intelligence model of a terminal, A step of transmitting to the terminal at least one reference signal including a first reference signal associated with a first beam predicted using an artificial intelligence model configured in the terminal; A step of receiving report information from the terminal based on a result of determining whether an event has occurred in the terminal using at least one of the measurement result of the first reference signal, reliability information for the first beam, and measurement results for reference signals excluding the first reference signal; and A method comprising a step of transmitting performance management instruction information for the artificial intelligence model based on the above reporting information.

10. In paragraph 9, The above report information is, A method for receiving information including at least one of identification information for the first beam, measurement result information for the first reference signal, and performance monitoring result information for the artificial intelligence model when the measurement result of the first reference signal is less than the first threshold value.

11. In paragraph 9, The above report information is, A method for receiving, including identification information for the first beam, measurement result information for the first reference signal, second beam identification information indicating the measurement result of the highest value, and the measurement result of the highest value, when the measurement result of the first reference signal is less than the first threshold value and the highest value measurement result of the measurement result of the reference signals excluding the first reference signal exceeds the second threshold value.

12. In paragraph 9, The above report information is, A method for receiving information including instruction information indicating performance deterioration of the artificial intelligence model when the reliability information for the first beam is below a third threshold.

13. In paragraph 9, The above performance management instructions are: A method comprising information indicating at least one of a fallback operation, an artificial intelligence model switching operation, and an artificial intelligence model deactivation operation for the artificial intelligence model.

14. In a terminal that performs performance management operations of an artificial intelligence model, A control unit that generates a beam prediction result using an artificial intelligence model configured in the terminal; and A receiving unit that receives at least one reference signal including a first reference signal associated with a first beam selected according to the beam prediction result, The above control unit, Whether an event has occurred is determined by using at least one of the measurement results of the first reference signal, reliability information for the first beam, and measurement results for reference signals other than the first reference signal. A terminal further including a transmitter that transmits report information to a base station based on whether the above event has occurred.

15. In paragraph 14, The above control unit, A terminal that determines that the event has occurred when the measurement result of the first reference signal is less than the first threshold value.

16. In paragraph 15, The above report information is, A terminal including at least one of identification information for the first beam, measurement result information of the first reference signal, and performance monitoring result information of the artificial intelligence model.

17. In paragraph 14, The above control unit, A terminal that determines that the event has occurred when the measurement result of the first reference signal is less than the first threshold value and the highest measurement result among the measurement results for reference signals excluding the first reference signal exceeds the second threshold value.

18. In paragraph 17, The above report information is, A terminal including identification information for the first beam, measurement result information of the first reference signal, second beam identification information indicating the highest value measurement result, and the highest value measurement result.

19. In paragraph 14, The above control unit, A terminal that determines that the event has occurred when the reliability information for the first beam is less than the third threshold.

20. In paragraph 14, The above control unit, A terminal that performs at least one of a fallback operation, an artificial intelligence model switching operation, and an artificial intelligence model deactivation operation for the artificial intelligence model when the above event is determined to have occurred.

Images