Method and device for collecting training data for beam management based on artificial intelligence and machine learning in wireless communication system

Abstract

The present disclosure is for signaling information related to a beam set for beam management in a wireless communication system. An operation method of a terminal may comprise: determining a first set and a second set of beams for beam management based on artificial intelligence (AI) / machine learning (ML); receiving at least one reference signal corresponding to at least one beam belonging to the second set; on the basis of the at least one reference signal, predicting information related to at least one beam included in the first set; and transmitting the information related to the at least one beam to a base station. The information related to the at least one beam included in the first set may be predicted on the basis of the result of measuring the at least one reference signal corresponding to the at least one beam included in the second set.

Description

Method and device for collecting training data for artificial intelligence and machine learning-based beam management in a wireless communication system

[0001] The present disclosure relates to AI / ML (artificial intelligence / machine learning)-based beam management in a wireless communication system, and to a method and device for collecting training data for AI / ML-based beam management.

[0002] Communication networks (e.g., 5G communication networks, 6G communication networks, etc.) are being developed to provide improved communication services compared to existing communication networks (e.g., long term evolution (LTE) and advanced LTE-A). 5G communication networks (e.g., new radio (NR) communication networks) can support frequency bands above 6 GHz as well as frequency bands below 6 GHz. That is, 5G communication networks can support FR1 bands and / or FR2 bands. 5G communication networks can support a variety of communication services and scenarios compared to LTE communication networks. For example, usage scenarios of 5G communication networks may include enhanced Mobile Broadband (eMBB), Ultra Reliable Low Latency Communication (URLLC), and massive Machine Type Communication (mMTC).

[0003] Compared to 5G, 6G communication networks can support a wider range of communication services and scenarios. 6G communication networks can meet requirements for ultra-high performance, ultra-high bandwidth, ultra-high space, ultra-high precision, ultra-intelligence, and / or ultra-reliability. 6G communication networks can support diverse and wide frequency bands and be applied to various usage scenarios (e.g., terrestrial communications, non-terrestrial communications, sidelink communications, etc.).

[0004] The use of artificial intelligence (AI) is expanding in the communications sector. AI can be utilized in a variety of areas, including network operation monitoring, predictive maintenance, network security and fraud prevention, customer service, and intelligent customer relationship management (CRM) systems.

[0005] AI can also be used to reduce the complexity of signaling between terminals and base stations or improve its accuracy. To this end, research is underway to utilize AI models to replace some or all of the signaling for channel measurement, beam management, positioning, and other tasks, or to improve performance.

[0006] Meanwhile, the technology that serves as the background for the invention is written to promote understanding of the background for the invention, and may include content that is not a prior art already known to a person with ordinary skill in the field to which the technology belongs.

[0007] The present disclosure can provide a method and device for effectively performing AI / ML (artificial intelligence / machine learning)-based beam management.

[0008] The present disclosure may provide a method and device for training a model for AI / ML-based beam management.

[0009] The present disclosure may provide a method and device for collecting training data for training a model for AI / ML-based beam management.

[0010] The present disclosure may provide a method and device for reducing overhead for transmitting training data for training a model for AI / ML-based beam management.

[0011] The technical objectives to be achieved in the present disclosure are not limited to those mentioned above, and other technical tasks not mentioned can be considered by a person having ordinary skill in the technical field to which the technical configuration of the present disclosure is applied from the embodiments of the present disclosure described below.

[0012] According to one embodiment of the present disclosure, a method of operating a terminal in a wireless communication system includes receiving reference signals for generating beam information for model training, transmitting a data report including at least a portion of measurement data generated based on measurements of the terminal for the reference signals, and receiving a beam indication generated by inference using a model trained based on the data report, wherein the data report may include the remainder of beam reports for beams measured within a corresponding measurement interval, excluding at least one other beam report included in at least one other data report for inference transmitted within the measurement interval.

[0013] According to one embodiment of the present disclosure, a method of operating a base station in a wireless communication system includes transmitting reference signals for generating beam information for model training, receiving a data report including at least a portion of measurement data generated based on measurement of a terminal for the reference signals, performing training on a model based on the data report, and transmitting a beam indication generated by inference using the model, wherein the data report may include the remainder of beam reports for beams measured within a corresponding measurement interval, excluding at least one other beam report included in at least one other data report for inference transmitted within the measurement interval.

[0014] According to one embodiment of the present disclosure, in a wireless communication system, a terminal includes at least one transceiver, at least one processor, and at least one memory operably connected to the at least one processor and storing instructions that, when executed by the processor, control the terminal to perform operations, the operations including receiving reference signals for generating beam information for model training, transmitting a data report including at least a portion of measurement data generated based on measurements of the terminal for the reference signals, and receiving a beam indication generated by inference using a model trained based on the data report, wherein the data report may include beam reports for beams measured within a corresponding measurement interval, except for at least one other beam report included in at least one other data report for inference transmitted within the measurement interval.

[0015] According to one embodiment of the present disclosure, in a wireless communication system, a base station includes at least one transceiver, at least one processor, and at least one memory operably connected to the at least one processor and storing instructions that, when executed by the processor, control the terminal to perform operations, the operations including transmitting reference signals for generating beam information for model training, receiving a data report including at least a portion of measurement data generated based on measurements of the terminal for the reference signals, performing training on the model based on the data report, and transmitting a beam indication generated by inference using the model, wherein the data report may include beam reports for beams measured within a corresponding measurement interval, except for at least one other beam report included in at least one other data report for inference transmitted within the measurement interval.

[0016] According to the present disclosure, it is possible to perform beam management effectively.

[0017] The effects that can be obtained from the embodiments of the present disclosure are not limited to the effects mentioned above, and other effects not mentioned can be clearly derived and understood by those skilled in the art to which the technical configuration of the present disclosure is applied, from the description of the embodiments of the present disclosure below. In other words, unintended effects resulting from implementing the configuration described in the present disclosure can also be derived by those skilled in the art from the embodiments of the present disclosure.

[0018] Figure 1 illustrates a wireless communication system according to an embodiment of the present disclosure.

[0019] FIG. 2 illustrates a block diagram of a communication node according to an embodiment of the present disclosure.

[0020] FIG. 3 illustrates a device for inferring or learning an AI / ML (artificial intelligence / machine learning) model according to an embodiment of the present disclosure.

[0021] FIGS. 4A and 4B illustrate block diagrams of a transmission path and a reception path of a communication node according to an embodiment of the present disclosure.

[0022] Figure 5 illustrates the structure of a neural network according to an embodiment of the present disclosure.

[0023] FIG. 6 illustrates an AI / ML framework according to an embodiment of the present disclosure.

[0024] FIG. 7 illustrates the life cycle management (LCM) of the NW (network)-side model according to an embodiment of the present disclosure.

[0025] FIG. 8 illustrates a procedure for AI / ML-based beam management in a wireless communication system according to an embodiment of the present disclosure.

[0026] FIG. 9 illustrates a procedure for transmitting training data without duplication in a wireless communication system according to an embodiment of the present disclosure.

[0027] FIG. 10 illustrates an example of non-duplicated training data transmission in a wireless communication system according to an embodiment of the present disclosure.

[0028] FIG. 11 illustrates a procedure for transmitting training data based on priority in a wireless communication system according to an embodiment of the present disclosure.

[0029] FIG. 12 illustrates an example of priority-based training data transmission in a wireless communication system according to an embodiment of the present disclosure.

[0030] FIG. 13 illustrates a procedure for performing training on a model using training data in a wireless communication system according to an embodiment of the present disclosure.

[0031] This disclosure may be subject to various modifications and various embodiments. Specific embodiments are illustrated and described in detail in the drawings. However, this is not intended to limit the disclosure to specific embodiments, but rather to encompass all modifications, equivalents, and alternatives falling within the spirit and technical scope of the disclosure.

[0032] While terms such as "first" and "second" may be used to describe various components, these components should not be limited by these terms. These terms are used solely to distinguish one component from another. For example, without departing from the scope of the present disclosure, a first component could be referred to as a "second component," and similarly, a second component could also be referred to as a "first component." The term "and / or" may refer to a combination of multiple related items described herein or to any of multiple related items described herein.

[0033] In the present disclosure, “at least one of A and B” may mean “at least one of A or B” or “at least one of combinations of one or more of A and B.” Additionally, in the present disclosure, “at least one of A and B” may mean “at least one of A or B” or “at least one of combinations of one or more of A and B.”

[0034] In the present disclosure, (re)transmission may mean “transmission,” “retransmission,” or “transmission and retransmission,” (re)setting may mean “setting,” “resetting,” or “setting and resetting,” (re)connection may mean “connection,” “reconnection,” or “connection and reconnection,” and (re)connection may mean “connection,” “reconnection,” or “connection and reconnection.”

[0035] When a component is referred to as being "connected" or "connected" to another component, it should be understood that it may be directly connected or connected to that other component, but that there may be other components intervening. Conversely, when a component is referred to as being "directly connected" or "connected" to another component, it should be understood that there are no other components intervening.

[0036] The terminology used in this disclosure is only used to describe specific embodiments and is not intended to limit the present disclosure. The singular expression includes the plural expression unless the context clearly indicates otherwise. In this disclosure, it should be understood that the terms "comprises" or "has" indicate the presence of a feature, number, step, operation, component, part, or combination thereof described in the specification, but do not preclude the presence or addition of one or more other features, numbers, steps, operations, components, parts, or combinations thereof.

[0037] Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by a person of ordinary skill in the art to which this disclosure pertains. Terms defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant technology, and shall not be interpreted in an idealized or overly formal sense unless explicitly defined herein.

[0038] Hereinafter, preferred embodiments of the present disclosure will be described in more detail with reference to the attached drawings. In order to facilitate an overall understanding in describing the present disclosure, the same reference numerals will be used for identical components in the drawings, and redundant descriptions of identical components will be omitted. In addition to the embodiments explicitly described in the present disclosure, operations may be performed according to combinations of embodiments, extensions of embodiments, and / or modifications of embodiments. The performance of some operations may be omitted, and the order of operation may be changed.

[0039] In an embodiment, even if a method (e.g., transmitting or receiving a signal) performed by a first communication node among communication nodes is described, a corresponding second communication node can perform a method (e.g., receiving or transmitting a signal) corresponding to the method performed by the first communication node. That is, if an operation of a UE (user equipment) is described, a corresponding base station can perform an operation corresponding to the operation of the UE. Conversely, if an operation of a base station is described, a corresponding UE can perform an operation corresponding to the operation of the base station.

[0040] A base station may be referred to as a NodeB, an evolved NodeB, a gNodeB (next generation node B), a gNB, a device, an apparatus, a node, a communication node, a BTS (base transceiver station), a RRH (radio remote head), a TRP (transmission reception point), a RU (radio unit), an RSU (road side unit), a radio transceiver, an access point, an access node, etc. A UE may be referred to as a terminal, a device, an apparatus, a node, a communication node, an end node, an access terminal, a mobile terminal, a station, a subscriber station, a mobile station, a portable subscriber station, an OBU (on-broad unit), etc.

[0041] In the present disclosure, signaling may be at least one of upper layer signaling, MAC signaling, or PHY (physical) signaling. A message used for upper layer signaling may be referred to as an "upper layer message" or an "upper layer signaling message." A message used for MAC signaling may be referred to as a "MAC message" or a "MAC signaling message." A message used for PHY signaling may be referred to as a "PHY message" or a "PHY signaling message." Upper layer signaling may refer to a transmission and reception operation of system information (e.g., a master information block (MIB), a system information block (SIB)) and / or an RRC message. MAC signaling may refer to a transmission and reception operation of a MAC control element (CE). PHY signaling may refer to a transmission and reception operation of control information (e.g., downlink control information (DCI), uplink control information (UCI), sidelink control information (SCI)).

[0042] In the present disclosure, “an operation (e.g., a transmission operation) is set” may mean that “setting information for the operation (e.g., an information element, a parameter)” and / or “information instructing the performance of the operation” is signaled. “An information element (e.g., a parameter) is set” may mean that the information element is signaled. In the present disclosure, “a signal and / or a channel” may mean a signal, a channel, or “a signal and a channel,” and a signal may be used to mean “a signal and / or a channel.”

[0043] The communication networks to which the embodiments are applied are not limited to those described below, and the embodiments may be applied to various communication networks (e.g., 4G communication networks, 5G communication networks, and / or 6G communication networks). Here, the term "communication network" may be used interchangeably with the term "communication system."

[0044] Figure 1 illustrates a wireless communication system according to an embodiment of the present disclosure.

[0045] Referring to FIG. 1, the communication system (100) may include a plurality of communication nodes (110-1, 110-2, 110-3, 120-1, 120-2, 130-1, 130-2, 130-3, 130-4, 130-5, 130-6). In addition, the communication system (100) may further include a core network (e.g., S-GW (serving-gateway), P-GW (PDN (packet data network)-gateway), MME (mobility management entity)). If the communication system (100) is a 5G communication system (e.g., NR (new radio) system), the core network may include an AMF (access and mobility management function), UPF (user plane function), SMF (session management function), etc.

[0046] A plurality of communication nodes (110 to 130) can support a communication protocol (e.g., LTE communication protocol, LTE-A communication protocol, NR communication protocol, etc.) specified in the 3GPP (3rd generation partnership project) standard. The plurality of communication nodes (110 to 130) may support CDMA (code division multiple access) technology, WCDMA (wideband CDMA) technology, TDMA (time division multiple access) technology, FDMA (frequency division multiple access) technology, OFDM (orthogonal frequency division multiplexing) technology, Filtered OFDM technology, CP (cyclic prefix)-OFDM technology, DFT-s-OFDM (discrete Fourier transform-spread-OFDM) technology, OFDMA (orthogonal frequency division multiple access) technology, SC (single carrier)-FDMA technology, NOMA (non-orthogonal multiple access) technology, GFDM (generalized frequency division multiplexing) technology, FBMC (filter bank multi-carrier) technology, UFMC (universal filtered multi-carrier) technology, SDMA (space division multiple access) technology, etc. Each of the plurality of communication nodes may have the following structure.

[0047] FIG. 2 illustrates a block diagram of a communication node according to an embodiment of the present disclosure. FIG. 2 illustrates an example of a wireless device (200) in a wireless communication system according to an embodiment of the present disclosure. The wireless device (200) according to an embodiment of the present disclosure may be a mobile terminal such as a smartphone, tablet PC, or wearable device, but is not limited thereto.

[0048] Referring to FIG. 2, the wireless device (200) may include at least one control unit (210), at least one memory (220), at least one power unit (230), at least one transceiver unit (240), at least one input unit (250), at least one output unit (260), and / or at least one antenna (270).

[0049] The control unit (210) can control the memory (220) and / or the transceiver unit (240), and can be configured to implement the descriptions, functions, procedures, proposals, methods, and / or operation flowcharts disclosed in the present disclosure. The memory (220) can be connected to the control unit (210) and can store various information related to the operation of the control unit (210). For example, the memory (220) can perform some or all of the controls controlled by the control unit (210), or store software code including commands for performing the descriptions, functions, procedures, proposals, methods, and / or operation flowcharts disclosed in the present disclosure. The configuration of the memory is not limited in a specific manner. For example, it can be configured as at least one of a read-only memory (ROM) and a random access memory (RAM).

[0050] At least one control unit (210) may be referred to as a controller, a microcontroller, a microprocessor, or a microcomputer. The descriptions, functions, procedures, proposals, methods, and / or operational flowcharts disclosed in the present disclosure may be implemented using firmware or software in the form of codes, instructions, and / or a set of instructions. Here, the firmware or software may execute another program stored in a memory (220), such as an OS. The control unit (210) may be implemented to support beamforming or directional routing operations in which signals from at least one antenna (270) are differently weighted to effectively steer signals outgoing in a desired direction.

[0051] Additionally, at least one control unit (210) may be coupled to a backhaul or network interface. The wireless device (200) may communicate with other wireless devices through the backhaul or network interface. The control unit (210) may include at least one processor. The processor may refer to a central processing unit (CPU), a graphics processing unit (GPU), or a dedicated processor on which methods according to embodiments of the present disclosure are performed.

[0052] At least one transceiver (240) may be connected to the control unit (210) and may transmit and / or receive a wireless signal via at least one antenna (270). The transceiver (240) may include a transmitter and / or a receiver. The at least one transceiver (240) may transmit user data, control information, wireless signals / channels, etc. mentioned in the methods and / or operation flowcharts of the present disclosure to at least one other device. For example, the at least one transceiver (240) may be connected to at least one control unit (210) and may transmit and receive wireless signals. In addition, the at least one control unit (210) may control the at least one transceiver (240) to transmit user data, control information, or a wireless signal to at least one other device. The at least one transmitter (240) may receive a signal transmitted by another wireless device from at least one antenna (270). Additionally, at least one transceiver (240) may downconvert or upconvert the received signal to generate a baseband signal. At least one antenna (270) may be a plurality of physical antennas or a plurality of logical antennas (e.g., antenna ports).

[0053] The input unit (250) can obtain information such as user input, video, and audio, and can include various input means such as various mechanical / electronic input means, cameras, and microphones. The output unit (260) is for providing information to users by generating output related to sight, hearing, or touch, and can include a display, a speaker, a vibration module, and the like. The wireless device (200) supplies power through the power supply unit (230), and the power supply unit (230) can include a wired / wireless charging circuit, a battery, and the like.

[0054] Referring again to FIG. 1, the communication system (100) may include a plurality of base stations (110-1, 110-2, 110-3, 120-1, 120-2) and a plurality of terminals (130-1, 130-2, 130-3, 130-4, 130-5, 130-6). Each of the first base station (110-1), the second base station (110-2), and the third base station (110-3) may form a macro cell. Each of the fourth base station (120-1) and the fifth base station (120-2) may form a small cell. The fourth base station (120-1), the third terminal (130-3), and the fourth terminal (130-4) may be within the cell coverage of the first base station (110-1). The second terminal (130-2), the fourth terminal (130-4), and the fifth terminal (130-5) may be within the cell coverage of the second base station (110-2). The fifth base station (120-2), the fourth terminal (130-4), the fifth terminal (130-5), and the sixth terminal (130-6) may be within the cell coverage of the third base station (110-3). The first terminal (130-1) may be within the cell coverage of the fourth base station (120-1). The sixth terminal (130-6) may be within the cell coverage of the fifth base station (120-2).

[0055] Here, each of the plurality of base stations (110-1, 110-2, 110-3, 120-1, 120-2) may be referred to as a NodeB (NB), an evolved NodeB (eNB), a gNB, an advanced base station (ABS), a high reliability-base station (HR-BS), a base transceiver station (BTS), a radio base station, a radio transceiver, an access point, an access node, a radio access station (RAS), a mobile multihop relay-base station (MMR-BS), a relay station (RS), an advanced relay station (ARS), a high reliability-relay station (HR-RS), a home NodeB (HNB), a home eNodeB (HeNB), a road side unit (RSU), a radio remote head (RRH), a transmission point (TP), a transmission and reception point (TRP), etc.

[0056] Each of the plurality of terminals (130-1, 130-2, 130-3, 130-4, 130-5, 130-6) may be referred to as a user equipment (UE), terminal equipment (TE), advanced mobile station (AMS), high reliability-mobile station (HR-MS), terminal, access terminal, mobile terminal, station, subscriber station, mobile station, portable subscriber station, node, device, on board unit (OBU), etc.

[0057] Meanwhile, each of the plurality of base stations (110-1, 110-2, 110-3, 120-1, 120-2) may operate in a different frequency band or may operate in the same frequency band. Each of the plurality of base stations (110-1, 110-2, 110-3, 120-1, 120-2) may be connected to each other via an ideal backhaul link or a non-ideal backhaul link, and may exchange information with each other via the ideal backhaul link or the non-ideal backhaul link. Each of the plurality of base stations (110-1, 110-2, 110-3, 120-1, 120-2) may be connected to the core network via the ideal backhaul link or the non-ideal backhaul link. Each of the plurality of base stations (110-1, 110-2, 110-3, 120-1, 120-2) can transmit a signal received from the core network to the corresponding terminal (130-1, 130-2, 130-3, 130-4, 130-5, 130-6), and can transmit a signal received from the corresponding terminal (130-1, 130-2, 130-3, 130-4, 130-5, 130-6) to the core network.

[0058] Additionally, each of the multiple base stations (110-1, 110-2, 110-3, 120-1, 120-2) can support MIMO transmission (e.g., single user (SU)-MIMO, multi user (MU)-MIMO, massive MIMO, etc.), coordinated multipoint (CoMP) transmission, carrier aggregation (CA) transmission, transmission in an unlicensed band, sidelink communication (e.g., device to device communication (D2D), proximity services (ProSe)), Internet of Things (IoT) communication, dual connectivity (DC), etc. Here, each of the plurality of terminals (130-1, 130-2, 130-3, 130-4, 130-5, 130-6) can perform an operation corresponding to the base station (110-1, 110-2, 110-3, 120-1, 120-2) and an operation supported by the base station (110-1, 110-2, 110-3, 120-1, 120-2). For example, the second base station (110-2) can transmit a signal to the fourth terminal (130-4) based on the SU-MIMO scheme, and the fourth terminal (130-4) can receive a signal from the second base station (110-2) by the SU-MIMO scheme. Alternatively, the second base station (110-2) can transmit signals to the fourth terminal (130-4) and the fifth terminal (130-5) based on the MU-MIMO method, and each of the fourth terminal (130-4) and the fifth terminal (130-5) can receive signals from the second base station (110-2) based on the MU-MIMO method.

[0059] Each of the first base station (110-1), the second base station (110-2), and the third base station (110-3) can transmit a signal to the fourth terminal (130-4) based on the CoMP scheme, and the fourth terminal (130-4) can receive a signal from the first base station (110-1), the second base station (110-2), and the third base station (110-3) based on the CoMP scheme. Each of the plurality of base stations (110-1, 110-2, 110-3, 120-1, 120-2) can transmit and receive a signal with terminals (130-1, 130-2, 130-3, 130-4, 130-5, 130-6) within its cell coverage based on the CA scheme. Each of the first base station (110-1), the second base station (110-2), and the third base station (110-3) can control sidelink communication between the fourth terminal (130-4) and the fifth terminal (130-5), and each of the fourth terminal (130-4) and the fifth terminal (130-5) can perform sidelink communication under the control of the second base station (110-2) and the third base station (110-3), respectively.

[0060]

[0061] FIG. 3 illustrates a device for inferring or learning an AI / ML (artificial intelligence / machine learning) model according to one embodiment of the present disclosure.

[0062] The device for inferring or learning the AI / ML model of FIG. 3 may be an embodiment of the wireless device (200) of the device of FIG. 2. Accordingly, the communication unit, the control unit, and the storage unit of FIG. 3 may each correspond to the transceiver unit (240), the control unit (210), and the storage unit (220) of FIG. 2. For convenience of explanation, the device for inferring or learning the AI / ML model will be described below assuming that it is a server. In addition, when the server is connected to the outside world via wireless communication, the communication unit may include the antenna (270) of FIG. 2. The control unit may perform AL / ML data learning or inference. The AI ​​model may be implemented in various forms. For example, the AI ​​model may be a model based on a neural network. The neural network model may include a deep neural network (DNN), a recurrent neural network (RNN), a bidirectional recurrent deep neural network (BRDNN), and the like.

[0063] The communication unit can communicate with external devices, and the server can receive various information through the communication unit. Therefore, the communication unit can receive monitoring information, user input information, etc. for training AI / ML models, and transmit them to the control unit. To this end, the communication unit can communicate with external devices via wired or wireless communication. The server can transmit and receive signals to external devices such as artificial satellites, mobile devices, and autonomous vehicles through the communication unit. Wireless communication may include cellular communication, short-range wireless communication, or global navigation satellite system (GNSS) communication.

[0064] The storage unit can store learning models, input data, output data, etc. Therefore, the storage unit can store AI models learned by the control unit or updated AI, and can provide data when necessary according to the control unit's commands.

[0065] The control unit may include an AL / ML management unit, a model learning unit, and a model inference unit. The AL / ML management unit can manage the communication unit, the model inference unit, and the model learning unit so that the inference or learning task can be performed efficiently. For example, the management unit can receive monitoring information through the communication unit, receive output from the model inference unit, and then check the performance of the inference. The management unit can transmit the checked performance to the model learning unit and instruct the model learning unit to learn the AI / ML model based on the performance. In other words, the management unit can transmit performance feedback information to the model inference unit, and the model inference unit can use the performance feedback information to instruct the learning goal or as a reward for reinforcement learning. In addition, the control unit can instruct the AI / ML model to be used by the server or an external device, or instruct the activation / deactivation of the use of the AI / ML model. For convenience of explanation, Fig. 3 illustrates the server as including both the model learning unit and the model inference unit. However, depending on the purpose of the server, either the model learning unit or the model inference unit may be included. In addition, the structure of the aforementioned FIG. 3 can be applied not only to servers but also to other devices such as terminals, wireless devices, autonomous vehicles, and mobile devices where inference or learning occurs.

[0066]

[0067] FIGS. 4A and 4B illustrate block diagrams of a transmission path and a reception path of a communication node according to an embodiment of the present disclosure. The transmission path (410) illustrated in FIG. 4A may be implemented in a communication node transmitting a signal, and the reception path (420) illustrated in FIG. 4B may be implemented in a communication node receiving a signal.

[0068] Referring to FIGS. 4A and 4B, the transmission path (410) may include a channel coding and modulation block (411), a serial-to-parallel (S-to-P) block (512), an Inverse Fast Fourier Transform (N IFFT) block (413), a parallel-to-serial (P-to-S) block (414), a cyclic prefix (CP) addition block (415), and an up-converter (UC) (416). The reception path (420) may include a down-converter (DC) (421), a CP removal block (422), an S-to-P block (423), an N FFT block (424), a P-to-S block (425), and a channel decoding and demodulation block (426). Here, N may be a natural number.

[0069] In the transmission path (410), information bits may be input to a channel coding and modulation block (411). The channel coding and modulation block (411) may perform a coding operation (e.g., low-density parity check (LDPC) coding operation, polar coding operation, etc.) and a modulation operation (e.g., quadrature phase shift keying (QPSK), quadrature amplitude modulation (QAM), etc.) on the information bits. The output of the channel coding and modulation block (411) may be a sequence of modulation symbols.

[0070] The S-to-P block (412) can convert modulation symbols in the frequency domain into parallel symbol streams to generate N parallel symbol streams. N can be an IFFT size or an FFT size. The N IFFT block (413) can perform an IFFT operation on the N parallel symbol streams to generate signals in the time domain. The P-to-S block (414) can convert the output (e.g., parallel signals) of the N IFFT block (413) into a serial signal to generate a serial signal.

[0071] The CP addition block (415) can insert a CP into a signal. The UC (416) can up-convert the frequency of the output of the CP addition block (415) to an RF (radio frequency) frequency. Additionally, the output of the CP addition block (415) can be filtered at the baseband before up-conversion.

[0072] A signal transmitted from a transmission path (410) may be input to a reception path (420). An operation in the reception path (420) may be the reverse operation of the operation in the transmission path (410). A DC (421) may down-convert the frequency of the received signal to a baseband frequency. A CP removal block (422) may remove a CP from a signal. The output of the CP removal block (422) may be a serial signal. An S-to-P block (423) may convert the serial signal into parallel signals. An N FFT block (424) may perform an FFT algorithm to generate N parallel signals. A P-to-S block (425) may convert the parallel signals into a sequence of modulation symbols. A channel decoding and demodulation block (426) may perform a demodulation operation on the modulation symbols and perform a decoding operation on the result of the demodulation operation to restore data.

[0073] In FIGS. 4A and 4B , Discrete Fourier Transform (DFT) and Inverse DFT (IDFT) may be used instead of FFT and IFFT. Each of the blocks (e.g., components) in FIGS. 4A and 4B may be implemented by at least one of hardware, software, or firmware. For example, in FIGS. 4A and 4B , some blocks may be implemented by software, and the remaining blocks may be implemented by hardware or a “combination of hardware and software.” In FIGS. 4A and 4B , a single block may be subdivided into multiple blocks, multiple blocks may be integrated into a single block, some blocks may be omitted, and blocks supporting other functions may be added.

[0074]

[0075] Figure 5 illustrates the structure of a neural network according to an embodiment of the present disclosure. In a neural network, the input layer is the first layer that receives external data. The input layer can receive raw data in various forms and pass it on after feature processing. Feature processing here refers to extracting characteristic parts of the raw data. The number of neurons in the input layer can be determined based on the characteristics and requirements of the given data.

[0076] In a neural network, hidden layers can consist of one or more layers located between the input and output layers. Therefore, unlike Figure 5, hidden layers can consist of two or more layers. Neurons in the hidden layers can represent transformed forms of input data, and the depth and width of the hidden layers can determine the complexity of the AI / ML model. Hidden layers can be used to implement nonlinear relationships between the input and output layers.

[0077] The output layer is the final layer of the neural network and provides the final output value. Based on the final output value of the output layer, the error between the neural network's prediction and the actual target value can be calculated.

[0078] Therefore, data input through the input layer can pass through one or more hidden layers before being passed to the output layer. During this process, features of the input data can be extracted. Neural networks can be trained using algorithms such as backpropagation and gradient descent. Learning can be fed back by rewarding the errors calculated in the output layer. Based on this feedback, the weights of each layer can be updated. This training process can be repeated to optimize the inference of AI / ML models.

[0079]

[0080] Terms related to AI to be used in this disclosure may be defined as follows.

[0081] AI / ML-enabled features refer to specific functions that utilize AI / ML. Examples of AI / ML-enabled features include CSI measurements, beam management, and positioning.

[0082] An AI / ML model (AI / ML Model) refers to an algorithm that applies AI / ML technology to generate a set of outputs based on a set of inputs. In this disclosure, the AI / ML model may not refer to the algorithm itself, but rather to a set of parameter values ​​that specify the AI / ML model. For example, an AI / ML model may be expressed as a set of weights in a neural network.

[0083] AI / ML model delivery refers to the transfer of an AI / ML model from one entity to another. The entities can include network nodes / functions (e.g., gNBs, LMFs, etc.), UEs, standalone servers, etc. Therefore, an AI / ML model trained by a first entity can be transferred to a second entity.

[0084] AI / ML model inference refers to the process of generating a set of outputs based on a set of inputs using a trained AI / ML model within a single entity.

[0085] AI / ML model testing is the process of evaluating the performance of the final AI / ML model using a dataset different from the one used for training and validation. It can be included as a sub-step of training. Unlike AI / ML model validation, testing involves further tuning the model.

[0086] AI / ML model training refers to the process of training an AI / ML model in a data-driven manner to learn input / output relationships and obtain a trained AI / ML model that can be used for inference.

[0087] AI / ML model transfer refers to the transfer of an AI / ML model from a sender to a receiver. Transfer can be wireless or wired, and can include parameters of a known model structure or parameters related to a new model. The transferred parameters can include parameters for the entire model or a portion of the model.

[0088] AI / ML model validation is a sub-process of training that evaluates the quality of an AI / ML model using a dataset different from the one used for training. It can be used to determine model parameters for not only the training dataset but also generalized datasets.

[0089] Data collection refers to the collection of data by network nodes, management entities, or UEs for AI / ML model training, data analysis, and inference.

[0090] Federated learning (or federated training) refers to a machine learning technique that trains AI / ML models by performing local model training on multiple distributed edge nodes (e.g., UEs, gNBs) using local data. Federated learning may require multiple interactions between nodes involved in edge AI / ML models. Local data exchange is not essential.

[0091] Functionality identification refers to the process of identifying AI / ML functionality so that it can be shared between the network and the UE. Information about AI / ML functionality can be shared during the functionality identification process.

[0092] Management instructions refer to the information necessary to ensure proper inference operations. Management instructions may include selecting / deactivating / switching AI / ML models or AI / ML functions, performing AI / ML tasks, and performing other alternative operations.

[0093] Model activation means activating an AI / ML model for a specific AI / ML supported function, and model deactivation means deactivating an AI / ML model for a specific AI / ML supported function.

[0094] Model download refers to the transmission of an AI / ML model from the network to the UE, and the UE receives the AI / ML model. Model identification refers to the identification of an AI / ML model based on a shared functional identifier between the network and the UE. Information about the AI / ML model may be shared during model identification. Model upload refers to the transmission of a model from the UE to the network.

[0095] Model monitoring refers to the process of monitoring the inference performance of an AI / ML model, and model parameter update refers to the process of updating the model's parameters.

[0096] Model selection refers to selecting one of several models to activate for a single AI / ML-enabled function. Model selection can occur simultaneously with model activation. Model switching refers to deactivating the currently activated AI / ML model for a specific AI / ML-enabled function and activating a different AI / ML model. Model updating refers to the process of updating model parameters and / or model structure.

[0097] AI / ML models can be classified based on the location of inference. A network-side AI / ML model refers to an AI / ML model in which inference is performed entirely within the network. A two-sided AI / ML model refers to a pair of AI / ML models in which inference is performed jointly by the UE and the network. Joint inference means that some inference is performed first in the UE, and the remaining portion is performed in the gNB. A UE-side AI / ML model refers to an AI / ML model in which inference is performed entirely within the UE.

[0098] The following learning methods can be used in AI / ML learning. Reinforcement learning (RL) refers to the process of training an AI / ML model based on inputs (i.e., states) and feedback signals (e.g., rewards) generated by the model's output (e.g., actions) in an environment in which the model interacts. Semi-supervised learning refers to the process of training a model using a mixture of labeled and unlabeled data. Supervised learning refers to the process of training a model based on inputs and their corresponding labels. Unsupervised learning refers to the process of training a model using unlabeled data.

[0099]

[0100] Hereinafter, the initial connection procedure between a terminal and a base station will be described. If the initial connection procedure is performed with the base station due to reasons such as the terminal's power on / off operation or loss of coverage, an identification procedure between the base station and the terminal may be required. First, the terminal may perform an initial cell search operation with the base station. The terminal may perform monitoring to receive a synchronization signal. The synchronization signal may be at least one of a primary synchronization signal (PSS) and a secondary synchronization signal (SSS). The terminal may receive a physical broadcast channel (PBCH) signal from the base station to obtain broadcast information within the cell. Based on the physical broadcast channel, the terminal may obtain information about the cell using at least one of the MIB or SIB. A block including all of the PSS, SSS, and PBCH may be referred to as a synchronization signal block (SSB).

[0101] A terminal can perform a random access procedure. The terminal can transmit a preamble to the base station and receive a random access response (RAR) from the base station. The RAR message can include a temporary identifier. The terminal can transmit MSG3 (or an RRC connection request message) using the scheduling information in the RAR, and the base station can perform a contention resolution procedure by transmitting MSG4 (or a contention resolution message) to the terminal in response to MSG3.

[0102] The base station can perform beam management based on the RACH occasion used for transmitting the preamble used in the random access procedure. For example, the base station can determine the beam on which the terminal received the synchronization signal based on the RACH occasion in which the preamble was transmitted. A synchronization signal can also be included as a reference signal for indicating the QCL relationship, and the QCL relationship can be established based on the SSB received during the initial access procedure.

[0103] Additionally, a channel measurement procedure may be performed for beam management. The terminal may receive a reference signal from the base station. Based on this, the terminal may report channel state information (CSI) to the base station. The channel state information may include at least one of reference signal received power (RSRP), reference signal received quality (RSRQ), and signal-to-noise ratio (SNR). The base station may use the received channel state information to adjust beamforming for the terminal or optimize radio resource allocation. For channel measurement, the base station may transmit configuration information for channel measurement to the terminal. The configuration information for channel measurement may include information related to a measurement target, a measurement cycle, and the like.

[0104]

[0105] Enhancing CSI Feedback Using AI

[0106] Terminals and base stations can use AI to perform channel state estimation and channel state reporting procedures. Using AI can enhance the CSI feedback process. For example, an AI model can be used for spatial-frequency domain CSI compression. A two-sided model can be used for spatial-frequency domain CSI compression. Preprocessing, postprocessing, quantization, and dequantization can be included in the CSI compression process. AI / ML-based CSI compression can be performed based on existing frameworks, such as the aforementioned CSI feedback procedure. Additionally, an AI model can be used for time-domain CSI prediction. Since CSI measurements are performed at the terminal, a UE-side model can be used.

[0107] For CSI compression using a two-sided model, three types of AI / ML model training collaboration can be considered. The first type of training collaboration refers to joint training of two-sided models on a single side or entity, and can be performed on either the UE side or the network side. The decision on which side to train can be based on at least one of the following factors: whether the model can be maintained exclusively, privacy protection, whether specific models are supported, whether device-specific optimization is required, flexibility in model updates, the possibility of developing / updating each model, whether to build a unified CSI reconfiguration model for different UEs, whether to build a unified CSI generation model for different networks, scalability, and model performance.

[0108] The second type of collaborative training involves jointly training two-sided models on both the network and UE sides. The third type of collaborative training involves separate training on the network and UE sides, with the UE training the CSI generation portion and the network training the CSI reconfiguration portion. Joint training requires that the generation and reconstruction models be trained in the same loop through forward and backpropagation, and can be performed on at least one node. Separate training refers to sequential training initiated on either the UE or network side.

[0109] In the second training collaboration type, joint training can include both simultaneous and sequential training. Sequential training in the second training collaboration type can begin with network-side training.

[0110] In the second training collaboration type, whether to choose concurrent or sequential training can be determined based on at least one of the following: whether the model can be maintained exclusively, whether the model can be maintained exclusively, privacy protection, flexibility in whether to support a specific model, whether to optimize for each device, flexibility in model update, possibility of development / update of each model, whether to build a unified CSI reset model for different UEs, whether to build a unified CSI generation model for different networks, scalability, whether the training data distribution can match the inference device, software / hardware compatibility, and model performance. In the third training collaboration type, whether the training is performed first on the network or on the UE can be determined based on the same conditions as those considered in the second training collaboration type, such as whether to maintain exclusivity and whether to protect privacy.

[0111] Data generation for AI models can be implemented in various ways. For example, in the CSI compression use case, training data for model training can be generated by the UE / gNB. For the network portion of two-sided model inference, input data can be generated by the UE and terminated at the base station. For the UE portion of two-sided model inference, input data can be used internally within the UE.

[0112] Additionally, for network-side performance monitoring, calculated performance metrics or data for performance metric calculations, if necessary, can be generated by the UE and terminated at the base station. In CSI compression use cases using a double-sided model, pairing information based on model identification can be established to select a CSI generation model compatible with the CSI reconstruction model used at the base station.

[0113] The following can be applied to CSI prediction using AI. Training data for model training can be generated by the UE. For UE-side model inference, input data can be used internally within the UE. Data for performance metrics or performance metric calculations required for network-side performance monitoring can be generated by the UE and terminated at the gNB.

[0114]

[0115] Beam management using AI

[0116] The beams of base stations and terminals can manage beam-related settings through AI. For convenience, beam set B refers to the beam set where measurements are performed as input to the AI / ML model, and set A refers to the beam set determined based on the inference of the AI / ML model. Beam sets A and B may contain beam information for the same frequency range.

[0117] AI can be used to infer spatial domain downlink beams for beam set A based on measurements of beam set B. As another example, AI can be used to infer temporal downlink beams for beam set A based on past measurements of beam set B, where beam set A and beam set B may be different sets or beam set B may be a subset of beam set A.

[0118] Additionally, the input of the AI ​​model can be formed from various combinations. For example, the input of the AI ​​can include at least one of an L1-RSRP measurement based on beam set B, other auxiliary information, a channel impulse response (CIR) based on beam set B, and a downlink Tx / Rx beam identifier (ID) associated with the L1-RSRP measurement of beam set B.

[0119] The aforementioned AI model can be designed to infer a beam including at least one of a downlink reception beam and a downlink transmission beam. In addition, the output of the AI ​​model can include at least one of a transmission beam, a reception beam, an L1-RSRP of the transmission beam, an L1-RSRP of the reception beam, an angle of the transmission beam, an angle of the reception beam, and other information.

[0120] Beam management methods using AI models are not limited to the aforementioned methods. AI models can be configured in various ways by configuring inputs and outputs with various combinations of settings for beam sets A and B, performance monitoring, data collection, and auxiliary information.

[0121] Learning and inference methods can also be implemented in various ways. For example, AI can be learned or trained using AI / ML models. Learning and training can be performed by the network or the UE. Furthermore, learning and inference can be performed on different devices. For example, learning can be performed on the network and inference on the terminal. Split learning can be performed in such a way that some of the learning is performed on a first device and some on a second device. Similarly to learning, split inference can be performed using multiple devices. Input data for inference can also be generated in various ways. For example, input data can be generated on the UE, and inference can be performed using the input data on the network. Furthermore, input data generated on the UE can be used for inference within the UE.

[0122]

[0123] Enhanced positioning method using AI

[0124] Terminals and base stations can perform positioning procedures that use AI to determine the terminal's location. To improve positioning accuracy, methods that directly determine location through AI / ML models and methods that determine auxiliary location can be considered.

[0125] The method of directly determining location through an AI / ML model outputs the UE's location, and channel fingerprinting based on channel observations can be used as input for the AI / ML model. Fingerprinting is a positioning technique based on probabilistic modeling that utilizes noise and surrounding environmental information for location tracking. Therefore, fingerprinting utilizes existing devices, such as wireless access points (APs), to construct a fingerprint map based on signal strength values, and the terminal's location can be determined based on the channel fingerprint generated by the terminal's channel observation.

[0126] The way AI / ML models determine auxiliary positions can be output as new and / or enhanced values ​​of existing measurements, and inputs can include LOS / NLOS identification, timing and / or angle of measurements, and probability of measurements.

[0127] Specifically, the following methods may be considered:

[0128] - UE-based positioning, direct AI / ML or AI / ML-assisted positioning using UE-side models

[0129] - UE-assisted / LMF-based positioning, AI / ML-assisted positioning using UE-side models

[0130] - UE-assisted / LMF-based positioning, direct AI / ML positioning using LMF-side models

[0131] - NG-RAN node auxiliary positioning, AI / ML auxiliary positioning using gNB-side model

[0132] - NG-RAN node auxiliary positioning, direct AI / ML positioning using LMF side model

[0133] Data generation for AI models can be implemented in various ways. For example, for training AI / ML models for positioning, training data can be generated by the UE / PRU / gNB / LMF. For LMF-side model inference, input data can be generated by the UE / gNB and terminated in the LMF. For gNB-side model inference, input data can be used internally within the gNB. For UE-side model inference, input data can be used internally within the UE. For LMF-side performance monitoring, if necessary, calculated performance metrics or data for performance metric calculation can be generated by the UE / gNB and terminated in the LMF. For gNB-side performance monitoring, if necessary, calculated performance metrics or data for performance metric calculation can be generated at least by the gNB.

[0134]

[0135]

[0136] Among the agenda items for discussion on AI / ML for NR air interface to be discussed in Rel-19 RAN (radio access network) WG1 (working group 1), the contents of the WID (work item description) document on AI / ML based beam management are as shown in [Table 1] below.

[0137] Provide specification support for the following aspects:....- Beam management - DL Tx beam prediction for both UE-sided model and NW-sided model, encompassing [RAN1 / RAN2]:o Spatial-domain DL Tx beam prediction for Set A of beams based on measurement results of Set B of beams ("BM-Case1")o Temporal DL Tx beam prediction for Set A of beams based on the historic measurement results of Set B of beams ("BM-Case2")o Specify necessary signalling / mechanism(s) to facilitate LCM operations specific to the Beam Management use cases, if anyo Enabling method(s) to ensure consistency between training and inference regarding NW-side additional conditions (if identified) for inference at UENOTE: Strive for common framework design to support both BM-Case1 and BM-Case2

[0138] In 3GPP Rel-18 RAN WG1, discussions were held at the SI (study item) stage to develop standards utilizing AI / ML. Specifically, studies were conducted on three major technology categories: CSI feedback, beam management, and positioning accuracy enhancement. Among these, beam management and positioning accuracy enhancement will be discussed at the WI (working item) stage in Rel-19. [Table 1] shows the development scope of AI / ML-based beam management to be discussed in Rel-19. As presented above, Rel-19 considers DL transmit (Tx) beam prediction, and the AI / ML models used here are considered as UE-sided models and NW-sided models. Here, the UE-side model and the NW-side model may have a structure in which the UE and NW each perform at least one operation for an AI / ML procedure, rather than utilizing an AI / ML model coupled between the UE and NW. The detailed development scope includes the following four topics.

[0139] - Spatial domain beam prediction for set A beam based on set B beam measurement results: BM-Case1

[0140] - Time domain beam prediction for set A beam based on set B beam measurement results: BM-Case2

[0141] - Signaling and mechanisms required for beam management-related LCM (life cycle management) operations.

[0142] - A method to ensure continuity between training and inference related to additional conditions of NW for inference at the UE level.

[0143] Additionally, standardization aims to develop a common framework that can support both BM-Case1 and BM-Case2, if possible.

[0144] Research on AI / ML in 3GPP Rel-18 led to the TR 38.843 document, which covers the NR air interface based on AI / ML. Specifically, the standard defines an AI / ML framework that will be commonly used in AI / ML for the NR air interface, and includes descriptions of three representative use cases (e.g., CSI feedback, positioning accuracy enhancement, beam management, etc.), performance evaluation results, and expected specification changes. Subsequently, Rel-19 RAN WG1 will develop standards for positioning accuracy enhancement and beam management, and conduct further research on CSI-RS feedback. The aforementioned AI / ML framework is as shown in Fig. 6.

[0145] FIG. 6 illustrates an AI / ML framework according to an embodiment of the present disclosure. Depending on the physical entity where operations for the AI / ML framework as illustrated in FIG. 6 are performed, AI / ML models are classified into NW-side AI / ML models, UE-side AI / ML models, and two-side AI / ML models. Among these, the NW-side AI / ML model and the UE-side AI / ML model process all AI / ML operations at either the base station or the terminal, while the two-side AI / ML model jointly performs AI / ML operations at the base station and the terminal. Therefore, the two-side AI / ML model requires significantly more data transmission for AI / ML than the NW-side AI / ML model and the UE-side AI / ML model, and may have significantly greater complexity. Therefore, standardization of the NW-side AI / ML model and the UE-side AI / ML model is expected in Rel-19.

[0146] In AI / ML-based beam management, beams are categorized into multiple sets. Set A of beams, Set B of beams, and Set C of beams can be defined, and the technical meaning of each can be defined as shown in [Table 2] below.

[0147] Clause 5.2.1 in TR 38.843… The following are selected as representative sub-use cases:- BM-Case1: Spatial-domain Downlink beam prediction for Set A of beams based on measurement results of Set B of beams...- BM-Case2: Temporal Downlink beam prediction for Set A of beams based on the historic measurement results of Set B of beams...Set B is a set of beams whose measurements are taken as inputs of the AI / ML model.… Clause 6.3.2.3 in TR38.843… - (Opt 2D) For the case that Set B of beams (pairs) is a subset of measured beams (pairs) Set C (where Set C is fixed across training and inference), compared to the case with all measurements of measured beam Set C as AI inputs- with Top K=1 / 2 of the measurements of Set C,…

[0148] To summarize, a set A of beams (hereinafter referred to as 'beam set A') includes a set of beams to be predicted via AI / ML, a set B of beams (hereinafter referred to as 'beam set B') includes a set of beams to be used as input data to AI / ML to predict beams within beam set A, and a set C of beams (hereinafter referred to as 'beam set C') includes a set of beams on which beam measurement is performed. Here, the beams included in beam set C may or may not be used as input data for AI / ML.

[0149] For the aforementioned beam sets A, B, and C, beam set B may be a subset of beam set C. The specification does not constrain the inclusion relationship between beam sets A, B, and C, but a case may be considered where beam set A includes beam set B and / or C.

[0150] At the 3GPP RAN1 #116 meeting, the approvals for the NW-side model were decided as shown in [Table 3] below.

[0151] AgreementFor NW-sided model, for inference, in a beam report initiated by network, based on one measurement resource set, support the report of more than 4 beam related information in L1 signaling· Note: Purpose, such as above "For NW-sided model, for inference", will not be specified in RAN 1 specifications· FFS on the report content for beam related information· FFS on max number of reported beam related information in one reportAgreementFor NW-sided model and for UE-sided model, beam indication is based on unified TCI state framework· FFS on whether / how potential enhancement is needed

[0152] Among the approvals shown in [Table 3], the first approval relates to data collection for inference of the NW-side model. When a UE transmits a beam report under NW control, the UE reports information on at least four beams in a single report, which is based on L1 signaling. Furthermore, as FFS (for further study), the content included in the report and the maximum number of beams included in a single report were mentioned. Regarding the second approval, for both the NW-side model and the UE-side model, beam indication is performed based on the existing unified TCI state framework. Furthermore, further improvements to FFS will be discussed.

[0153] This disclosure addresses data collection techniques in beam management based on a network-side model. In AI / ML-based beam management using a network-side model, information is transferred between the network and the UE, as illustrated in FIG. 7 . FIG. 7 illustrates life cycle management (LCM) of the network-side model according to an embodiment of the present disclosure. In FIG. 7 , the network can be understood to include a base station, gNB, etc.

[0154] Referring to FIG. 7, the NW transmits a reference signal (RS) to the UE, the UE transmits a report of collected data to the NW, and the NW transmits a beam indication. At this time, the NW can perform model training, performance monitoring, and model inference using the collected data. That is, since it is a NW-side model, there is no AI / ML-related operation in the UE, and the UE is responsible for receiving the reference signal transmitted by the NW and performing the report. On the other hand, the NW performs model training, performance monitoring, and model inference by distributing the data reported from the terminal, and transmits a beam indication to the terminal based on the beam prediction, which is the result of the model inference.

[0155] For data collection, the NW can configure reference signal resources (e.g., SSB resources or CSI-RS resources) for L1-RSRP measurements. For example, the NW can configure reference signal resources for beam set A and / or beam set B. The UE performs measurements based on the configured reference signal resources and reports the collected data under the control of the NW. From the NW's perspective, the collected data can be used for model training, model inference, or performance monitoring. For model training, the collected data can be forwarded or processed for offline learning or fine tuning. For model inference, the NW can use the collected data as model input and generate output, and the output (e.g., beam prediction result) can be further used for beam direction. The collected data can be used to monitor the performance of the model, and the NW can decide whether to enable / disable / fallback AI / ML operations based on the performance monitoring.

[0156] AI / ML performed on the NW operates based on reports from the terminal regarding beam measurements. The NW can classify the information reported from the terminal into data for model training, data for model inference, and data for performance monitoring. Model training data is data for training AI / ML models and is characterized by a relatively large amount of data for reporting beam measurement results and a long latency requirement. On the other hand, model inference data requires a relatively short latency requirement to generate beam prediction results through AI / ML in near real-time. The amount of data for reporting beam measurement results based on model inference can be less than or equal to the amount of data for the model training data. Finally, performance monitoring data can use as little as model inference data or as much as AI / ML model training data or more.

[0157]

[0158] FIG. 8 illustrates a procedure for AI / ML-based beam management in a wireless communication system according to an embodiment of the present disclosure. FIG. 8 illustrates signal exchange between a base station and a terminal for beam management.

[0159] Referring to FIG. 8, in step S801, the base station transmits reference signals to the terminal. Prior to this, the base station may transmit configuration information or control information including the configuration of the reference signals to the terminal. Furthermore, the configuration information may further include configuration for measurement and reporting. In the example of FIG. 8, the transmission of the reference signals is expressed as being performed once, but the base station may transmit the reference signals on multiple occasions (e.g., after transmitting data reports, after model training, etc.).

[0160] In step 803, the terminal transmits a data report to the base station. The data report is a message containing data for model training. The data report includes the results of measurements for reference signals. For example, the data report may include beam reports for multiple beams (e.g., at least one beam belonging to beam set A or beam set C). Each beam report includes identification information and measurement values ​​(e.g., RSRP) for each beam.

[0161] In step S805, the base station performs model training. That is, the base station can perform training based on data reports received from the terminal. Through this, the base station acquires a trained model for beam management. However, in other embodiments, the model training may be performed by a device other than the base station. In this case, the base station may forward the beam reports included in the received data reports to the other device.

[0162] In step S807, the terminal transmits a data report to the base station. Here, the data report is a message containing information used as input data, i.e., data for model-based inference. For example, the data report may include beam reports for multiple beams (e.g., at least one beam belonging to beam set B). For example, the data report may include at least one measurement value (e.g., RSRP) for each of the multiple beams.

[0163] In step S809, the base station performs inference. That is, the base station can use the trained model to predict measurement values ​​for other beams from at least one beam report included in the data report. Accordingly, the base station can determine the optimal beam or preferred beam to use for the terminal.

[0164] In step S811, the base station transmits a beam instruction. The beam instruction may indicate the optimal beam or preferred beam to be used for the terminal. For example, the beam instruction may indicate a TCI state associated with at least one beam. Accordingly, the terminal can perform communication using the beam indicated by the beam instruction.

[0165] In the embodiment described with reference to FIG. 8, one model training operation and one inference operation are exemplified. However, the inference operation and the model training operation may be performed repeatedly. In this case, the model training operation is performed based on data collected during a given interval, and the inference operation may be performed at least once during the interval. Accordingly, at least one data report for training and at least one data report for inference may be transmitted and received during the interval.

[0166]

[0167] In the present disclosure below, beam set A, beam set B, and beam set C may be understood as follows.

[0168] - Beam Set A: The set of all beams to be analyzed and / or predicted via model inference in NW-side AI / ML.

[0169] - Beam set B: A set of at least one beam corresponding to information to be used in the NW, which is information (e.g., measurement results) about at least some of the beams measured by the UE. This set may include information about different elements (e.g., beams) depending on the time point. For example, {beam1, beam2, beam3, beam4} may be included in beam set B at a certain time point, and {beam1, beam3, beam5, beam7} may be included in beam set B at another time point.

[0170] Beam Set C: A set of beams measured by the UE or a set of beams used by the NW to transmit reference signals for beam measurement to the UE. Similar to Beam Set B, the beams included in Beam Set C may vary depending on the time point.

[0171] Beam set A, beam set B, and beam set C may or may not have the inclusion relationship of beam set B ⊆ beam set C ⊆ beam set A.

[0172]

[0173] In addition, the present disclosure defines the following MTI (model training information) as information elements included in data reporting for training. As information to be included in the aforementioned model training data (hereinafter referred to as "model training data information"), at least one of the following MTI1 to MTI5 may be included.

[0174] - MTI1: L1-RSRP values ​​for all or part of beam set B.

[0175] - MTI2: L1-RSRP values ​​for all or part of the beams for model training (e.g. beam set C or beam set A).

[0176] - MTI3: Identification information of the top K beam(s) for measurement values ​​among beams for model training (e.g. beam set C or beam set A)

[0177] - MTI4: Point-in-time information for MTI1 / MTI2 / MTI3 (e.g. timestamp)

[0178] - MTI5: Assistance information for terminals that helps train NW-side AI / ML models.

[0179] In one embodiment, MTI1 and MTI2 include L1-RSRP. In another embodiment, MTI1 and MTI2 may include other measurement values ​​instead of L1-RSRP. For example, a measurement value such as SINR may be used instead of L1-RSRP. In another example, L1-RSRP may be included in the general case, and SINR may be included in the case of high interference. The types of the aforementioned measurement values ​​may be configured through higher layer signaling by the NW (e.g., MAC CE, RRC reconfiguration).

[0180] According to one embodiment, for reporting MTI1 to MTI5, the UE may buffer values ​​measured over multiple time instances and report the data either all at once or in a distributed manner at a specific time point. The reporting time point may be configured by the NW and may be, for example, periodic / semi-periodic / aperiodic. The reporting time point may be configured by the NW via upper layer signaling or L1 signaling. When configuring periodic beam reporting (e.g., reporting at least one of MTI1 to MTI5), the NW may also indicate the beam reporting periodicity to the UE. When configuring semi-periodic beam reporting, the NW can indicate beam reporting periodicity information to the UE through higher layer signaling (e.g., MAC CE or RRC reconfiguration) and enable / disable beam reporting through MAC CE or DCI. When configuring aperiodic beam reporting, the NW can indicate beam reporting through MAC CE or DCI.

[0181] In one embodiment, if the measurement time for MTI1, MTI2, and / or MTI3 is self-evident, i.e., if the measurement time for MTI1, MTI2, and / or MTI3 can be determined without explicit signaling, MTI4 may be omitted. For example, there may be a case where the only time a beam measurement was performed is after the most recent beam report. As another example, when periodic beam reporting is performed, the time at which the reported beam information was measured can be determined without a separate timestamp.

[0182] In one embodiment, MTI5 represents information that can inform the terminal's instantaneous status. For example, the terminal may include its current movement speed information in MTI5. This movement speed information may be used by the NW to determine which data to utilize as input for AI / ML. As another example, the terminal's location information may be included in MTI5. This location information may be used to determine whether the terminal's proximity to the base station leads to more frequent changes in the optimal beam.

[0183] In one embodiment, all of MTI1 to MTI5 may be reported, or some of them may be selectively reported. Full reporting or selective reporting of some of them may be selected depending on the type of AI model. For example, for a classification-based model, the training data content may include {MTI1, MTI3, MTI4}. As another example, the training data content for a regression-based model may include {MTI1, MTI2, MTI4}. As in the examples described above, various training data content sets that utilize all or some of the MTIs may be defined depending on the model. In this case, which MTI(s) to include as training data content may be indicated to the terminal through the RRC configuration / reconfiguration of the base station.

[0184] In one embodiment, when MTI1 and MTI2 are used simultaneously as training data for a model, all or part of MTI1 may overlap with part of MTI2. In this case, by omitting the duplicated data in MTI1 or MTI2, the overhead of beam reporting can be reduced.

[0185] In one embodiment, to reduce the overhead of reporting MTI1 and MTI2, the terminal may not report measurement values ​​for all beams, but may report measurement results for only a subset of beams that are meaningful from an AI / ML perspective to the NW via L1-RSRP. The subset of beams that are meaningful may be determined based on various criteria, some examples of which are as follows:

[0186] - Method 1: NW sets a parameter (e.g. nrBeamReport) for the number of beams to be reported in advance, and the terminal can perform reporting on the upper nrBeamReport number of beams in descending order of L1_RSRP.

[0187] - Method 2: NW can set a parameter for the offset of channel quality in advance (e.g. L1_RSRP_Offset_forBeamReportModelTraining or RSRP_offset), and the terminal can perform reporting on beam(s) that have a measurement value greater than or equal to L1_RSRP_max - RSRP_offset based on the maximum L1-RSRP (e.g. L1_RSRP_max) among the beams to be reported.

[0188] - Scheme 3: In addition to Scheme 2, the NW can configure a parameter for the minimum number (e.g., nrBeamReport_min) and / or a parameter for the maximum number (e.g., nrBeamReport_max) for beam reporting via upper layer signaling. For example, if the number of beams having a measurement value greater than or equal to L1_RSRP_max - RSRP_offset is less than nrBeamReport_min, the terminal can further report information about at least one beam having a higher L1-RSRP among the beam(s) having a measurement value less than or equal to L1_RSRP_max - RSRP_offset or smaller. As another example, if the number of beams having a measurement value greater than or equal to L1_RSRP_max - RSRP_offset is greater than nrBeamReport_max, the terminal can perform reporting for nrBeamReport_max beams having a higher L1-RSRP.

[0189] - Option 4: L1_RSRP_max and L1_RSRP_min can be set identically to Option 2. In this case, the terminal reports on the number of beams set according to Option 1. In this case, the RSRP_offset value becomes meaningless, so the setting can be omitted.

[0190] - Option 5: Reporting can be performed on beams measured at multiple time instances. In this case, Options 1 and 2 can each be applied independently for each time instance. For example, if nrBeamReport=8 in Option 1, beam reporting can be performed for 8 beams for a specific time instance and 8 beams for another time instance.

[0191] In the above-described methods, a time instance can be defined in various ways. A time instance can be defined as any predefined or configured time interval, for example, as a slot, multiple slots, a subframe, or multiple subframes.

[0192]

[0193] According to one embodiment of the present disclosure, a method for eliminating duplication of beam reports may be applied to reduce overhead. An example of a procedure for transmitting training data without duplication of beam reports is shown in FIG. 9.

[0194] Figure 9 illustrates a procedure for transmitting training data without duplication in a wireless communication system according to an embodiment of the present disclosure. Figure 9 also illustrates an operating method of a terminal.

[0195] Referring to FIG. 9, in step S901, the terminal transmits at least one data report for inference. That is, the terminal transmits at least one data report including measurement values ​​for at least one beam for input data for inference of the base station. In other words, the terminal generates at least one beam report based on at least one reference signal transmitted through at least one beam for inference, and transmits at least one data report including at least one beam report. At this time, the terminal may transmit one data report in one opportunity or transmit multiple data reports in multiple opportunities.

[0196] In step S903, the terminal generates measurement data for training. To this end, the terminal may receive multiple reference signals and perform measurements. At this time, measurement values ​​for beams measured during a given measurement interval may be included in the measurement data. In other words, the terminal may generate measurement data that includes measurement values ​​for beams received during a corresponding measurement interval.

[0197] In step S905, the terminal omits duplicated data. Here, duplication may occur between beams included in at least one data report for inference and beams included in a data report for training. That is, reference signals for generating data for inference may be transmitted during a measurement interval, and thus at least one data report for inference and a data report for training may be generated during a single measurement interval. In this case, since beam duplication may occur between at least one data report for inference and a data report for training, the terminal may exclude measurement values ​​for the duplicated beams, i.e., beams previously reported for inference.

[0198] In step S907, the terminal transmits a data report for training. The terminal may omit redundant data and transmit a data report that includes at least one remaining beam report. That is, the terminal transmits a data report that includes measurement values ​​for the remaining beams, excluding the beams included in the data report for inference during the corresponding measurement interval. Through this, the base station can perform model training.

[0199]

[0200] A specific example of transmitting data reports for training according to the embodiment described with reference to FIG. 9 is as shown in FIG. 10. FIG. 10 illustrates an example of non-duplicated training data transmission in a wireless communication system according to an embodiment of the present disclosure. In FIG. 10, Bx denotes a beam report for a beam having an index x.

[0201] Referring to Fig. 10, to reduce the overhead of beam reporting for training, beam reporting may be omitted for at least one beam that has been reported as data for other AI / ML operations (e.g., inference, performance monitoring). In the example of Fig. 10, for data collection for inference, beam reporting is performed with a small delay through L1 signaling. Therefore, data reporting for inference may be transmitted earlier than data reporting for data collection for model training. In this case, beams already reported for data collection for inference may overlap with beams included in data for model training.

[0202] Specifically, based on the measurement at time instance 0, a data report for inference including {B0, B1, B2, B3} is transmitted. Then, based on the measurement at time instance 1, a data report for inference including {B2, B3, B4, B6} is transmitted. Furthermore, based on the measurement at time instance 2, a data report for inference including {B1, B2, B3, B5} is transmitted. Then, at time instance 3, a data report for model training including beam reports for beams measured at time instances 0 to 2 is transmitted.

[0203] Here, in the data report for model training, it may be necessary to label which point in time each beam report is based on. That is, the data report for model training may include a set of beam reports labeled as time instance 0, a set of beam reports labeled as time instance 1, and a set of beam reports labeled as time instance 2. In the case of Fig. 10, {B0, B1, B2, B3, B4, B5, B6, …} are generated for each time instance. However, beam reports transmitted for inference for the same time instance are excluded from the report data, and {B4, B5, B6, …} for time instance 0, {B0, B1, B5, …} for time instance 1, and {B0, B4, B6, …} for time instance 2 are included in the data report.

[0204] Accordingly, the base station can receive data reports for model training and restore a complete training data report by adding beam reports included in the previously received data reports for inference. Therefore, the base station can train the model using data collected for inference and data for model training (e.g., data with redundant data omitted from the data for inference) without compromising AI / ML performance.

[0205]

[0206] According to one embodiment of the present disclosure, a method of applying different priorities for each time instance can be applied to reduce overhead. In other words, in the data collection for the aforementioned periodic / semi-periodic / aperiodic model training, when collecting buffered beam measurement information, if there is a large amount of buffered data, it is possible to assign priorities based on beam measurement time and reduce the amount of information used for beam reporting. An example of a procedure for transmitting training data based on priorities for each time instance is shown in FIG. 11.

[0207] FIG. 11 illustrates a procedure for transmitting training data based on priority in a wireless communication system according to an embodiment of the present disclosure. FIG. 11 also illustrates an operating method of a terminal.

[0208] Referring to Figure 11, in step S1101, the terminal collects measurement data for training. Specifically, the terminal performs measurements on reference signals received from the base station and obtains measurement values ​​for each reference signal, i.e., each beam. At this time, the collection of measurement data can be performed during a configured measurement interval.

[0209] In step S1103, the terminal determines the priority for each time instance within the measurement interval. The measurement interval includes multiple time instances, and each time instance is assigned a priority. Information regarding the priority can be received from the base station.

[0210] In step S1105, the terminal generates a set of beam reports for each time instance according to priority. Depending on the priority of the time instance, some of the measurement data acquired in each time instance may be omitted or expressed differently. For example, a time instance with a higher priority may include a larger number of beam reports compared to a time instance with a lower priority. In another example, a time instance with a higher priority may include beam reports expressed at a higher quantization level compared to a time instance with a lower priority.

[0211] In step S1107, the terminal transmits a data report. The terminal may transmit a data report including beam report sets generated according to priority. Each beam report set may be transmitted with information about the time instance. Furthermore, each beam report set may be transmitted as generated in step S1105 or after undergoing additional processing. For example, the additional processing may include selecting some of the beam report sets.

[0212]

[0213] FIG. 12 illustrates an example of priority-based training data transmission in a wireless communication system according to an embodiment of the present disclosure. Referring to FIG. 12 , transmission of data reports for training based on measurements at time instances 10 to 17 is illustrated. Referring to FIG. 12 , when reporting data for model training corresponding to time instances 10 to 17, two or more priority regions can be set and managed. In the case of FIG. 12 , three priority regions are set. The size of each priority region (e.g., the number of time instances) can be pre-configured via higher-layer signaling of the base station. Additionally, instructions are required for each priority region on how to process and transmit data for model training. In the example of FIG. 12 , data collection is performed for 16 beams and 8 beams per time instance in the first priority region and the second priority region, respectively, and no data collection is performed in the third priority region. Here, the instruction to not collect data can be performed by instructing data collection for 0 beams. In this way, by setting the number of beams subject to data collection for each priority area, differential operation is possible for each time instance.

[0214] As another embodiment related to priorities, a method of varying the quantization level for measurement values ​​for beams (e.g., L1-RSRP) instead of the number of beams for each priority region may be applied. For example, the terminal may apply a quantization level of 2 dBm intervals to beam reports reported in the first priority region, and a quantization level of 4 dBm or 6 dBm to beam reports reported in the second priority region. This may reduce the amount of information in beam reports. In this case, by setting the quantization level to a specific value (e.g., setting the measurement value to be expressed with 0 bits), it is possible to omit data collection in regions with lower priorities.

[0215]

[0216]

[0217] Figure 13 illustrates a procedure for training a model using training data in a wireless communication system according to an embodiment of the present disclosure. Figure 13 also illustrates an operating method of a base station. Here, the base station may be understood as a mobile network (MW), gNB, or the like.

[0218] Referring to FIG. 13, in step S1301, the base station determines a configuration for data reporting for training. For example, the base station may determine a configuration for a measurement interval and the content of the report. For example, the measurement interval may indicate at least one time instance. For example, the content of the report may include at least one of a method for selecting some of the measured beams (e.g., a measurement value threshold, a number of beams, etc.) or a method for expressing the measurement value (e.g., a quantization level, a differential reporting method, etc.). In addition, the configuration for the data report may also include a configuration of reference signals for measurement.

[0219] In step S1303, the base station transmits configuration information for data reporting. The base station signals the configuration determined in step S1301. For example, the base station may transmit the configuration information via RRC signaling or MAC CE.

[0220] In step S1305, the base station receives a data report for training. For example, the data report may include a beam report for at least one beam measured by the terminal that has a measurement value greater than a threshold. As another example, the data report may include beam reports for a specified number of beams measured by the terminal that have measurement values ​​higher than a threshold.

[0221] In step S1307, the base station performs training on the model. The base station may perform the training using the received data reports. At this time, the training may be performed using the beam reports included in the data reports received in step S1305 together with at least one other beam report included in at least one other data report for inference received within the corresponding measurement interval. The base station may perform the training by using some beam reports as labels and the remaining beam reports as input data to perform inference and backpropagation.

[0222]

[0223] According to the various embodiments described above, data including beam reports for model training can be provided to the NW. Various methods for reducing overhead in reporting measurement values ​​through data reporting can be broadly categorized into partial reporting and expression control, and further categorized into sub-methods.

[0224] The first method involves reporting some of the measured values. Specifically, there are several possible methods, as follows:

[0225] [Method 1-1] Threshold-based selective beam reporting: The terminal reports at least one beam having a measurement value exceeding a certain threshold based on measurement values ​​for the beams (e.g., RSRP or similar measurement values). For example, the threshold is configured by the NW and can be configured via RRC signaling or MAC CE. That is, the NW transmits configuration information for beam reporting to the terminal, and the configuration information can include information about the threshold of the measurement value used to determine whether to report each beam.

[0226] [Method 1-2] Beam reporting for the top M beams based on the size of the measurement value (e.g., RSRP): The terminal sorts the beams in order of the measurement value (e.g., RSRP) for the beams, i.e., in descending order of the measurement value, and reports the top M beams. Here, the M value is configured by the NW and can be configured via RRC signaling, MAC CE, or DCI. That is, the NW transmits configuration information for the beam report to the terminal, and the configuration information can include information on the number of beams reported via one data report. In this case, if the M value is not configured by the NW, it can be operated with a predefined specific value (e.g., 4).

[0227] [Method 1-3] Channel status-based beam reporting: The terminal reports at least one beam with a good channel based on the channel status and statistics for the beams. Here, the reported information is a measurement value of at least one beam, which may include RSRP or another value that can replace it. The channel status and statistics may be the magnitude of the RSRP, or the stability of the channel in the time and / or frequency domain (e.g., deviation, etc.). In this method, a kind of threshold may be used. For example, for a first beam, if N recently measured RSRP values ​​are consistently greater than the threshold, the terminal reports the first beam. On the other hand, for another beam, a second beam, if only L RSRP values ​​among N recently measured RSRP values ​​are greater than the threshold (L≤N), the terminal may not report the second beam.

[0228] [Method 1-4] Variance-based Beam Reporting: For each beam, if the currently measured RSRP is significantly different from the most recently reported past measurement value (e.g., RSRP or similar measurement value), the terminal reports the measurement value for that beam. For example, if the last reported past RSRP of the first beam is -50 [dBm] and the currently measured RSRP is -45 [dBm], and the last reported past RSRP of the second beam is -40 [dBm] and the currently measured RSRP is -39 [dBm], it may be more efficient to report the first beam first because the difference between the RSRP of the previously reported beams is not significant even though the second beam has a larger RSRP. Since this method basically assumes that the past RSRP value for each beam has been reported to the NW, it can be mixed with at least one of the other methods described above to report the initial RSRP for each beam.

[0229] The above-described [Method 1-1], [Method 1-2], [Method 1-3], and [Method 1-4] methods can be applied in combination with each other with some modifications. For example, [Method 1-1] can be applied to the first few reports, and [Method 1-4] can be applied to subsequent reports. As another example, by applying [Method 1-1] and [Method 1-2] simultaneously, the top M beams can be reported, and in addition, at least one beam among the remaining beams, that is, the beams other than the top M beams, having a measurement value greater than a threshold can be reported.

[0230] The second method involves controlling the representation of measurement values. Specifically, there are several possible methods, as follows:

[0231] [Method 2-1] Beam reporting based on change in measurement value (e.g., RSRP) over time: For each beam, if at least one report has been made, the terminal reports the difference value based on the previously reported measurement value. For example, if an RSRP of -40 [dBm] is obtained for the first beam, and the previously reported RSRP for the first beam was -43 [dBm], the difference, +3, is reported. This can be applied independently to each beam.

[0232] [Method 2-2] Beam reporting based on difference value compared to maximum measurement value: When performing reporting for multiple beams at any reporting point, the terminal reports the measurement value of the beam with the maximum measurement value, and reports the difference value compared to the maximum measurement value for the other beams. That is, a differential report method can be used. For example, when reporting -50[dBm], -45[dBm], -48[dBm], and -52[dBm] for the first to fourth beams, respectively, the terminal can report the second RSRP with the maximum RSRP, and based on this, report the difference values ​​of the RSRP of the remaining beams of -5, -3, and -7.

[0233] [Method 2-3] Beam reporting based on quantization level adjustment: Instead of a fixed quantization level, a controllable quantization level can be applied based on the importance of the beam report. Here, the importance of the beam report can be determined based on how long ago the measurement value was acquired from the reporting time. That is, the importance of the beam report can be determined based on the time interval between the transmission time of the beam report and the measurement time of the information reported through the beam report. Alternatively, the importance can be determined based on the purpose of the data report to be transmitted (e.g., inference, model training). Depending on the importance, the quantization level can be adjusted, thereby changing the number of bits representing the measurement value.

[0234] The aforementioned [Method 2-1], [Method 2-2], and [Method 2-3] methods can be applied differently or in combination with slight modifications. That is, it is possible to modify the methods described above to reduce overhead, or to apply two or more methods simultaneously.

[0235]

[0236] The operations of the method according to the present disclosure can be implemented as a computer-readable program or code on a computer-readable recording medium. A computer-readable recording medium includes any type of recording device that stores information readable by a computer system. Furthermore, a computer-readable recording medium can be distributed across network-connected computer systems, allowing the computer-readable program or code to be stored and executed in a distributed manner.

[0237] Additionally, the computer-readable recording medium may include hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, flash memory, etc. The program instructions may include not only machine language codes produced by a compiler, but also high-level language codes that can be executed by a computer using an interpreter, etc.

[0238] While some aspects of the present disclosure have been described in the context of a device, they may also represent a description of a corresponding method, wherein a block or device corresponds to a method step or a feature of a method step. Similarly, aspects described in the context of a method may also be described as a corresponding block or item or a feature of a corresponding device. Some or all of the method steps may be performed by (or using) a hardware device, such as, for example, a microprocessor, a programmable computer, or an electronic circuit. In some embodiments, at least one or more of the most significant method steps may be performed by such a device.

[0239] A programmable logic device (e.g., a field-programmable gate array) may be used to perform some or all of the functions of the methods described in this disclosure. The field-programmable gate array may operate in conjunction with a microprocessor to perform one of the methods described in this disclosure. In general, the methods are preferably performed by some hardware device.

[0240] Although the present disclosure has been described with reference to preferred embodiments thereof, it will be understood by those skilled in the art that various modifications and changes may be made to the present disclosure without departing from the spirit and scope of the present disclosure as set forth in the claims below.

Claims

1. In a method of operating a terminal in a wireless communication system, Receiving reference signals for generating beam information for model training; Transmitting a data report including at least a portion of measurement data generated based on the terminal's measurement of the above reference signals; and Receiving a beam instruction generated by inference using a trained model based on the above data report, A method wherein the data report includes beam reports for beams measured within a corresponding measurement interval, excluding at least one other beam report included in at least one other data report for inference transmitted within the measurement interval.

2. In claim 1, The data report includes a first set of beam reports for at least one beam measured at a first time instance within the measurement interval and a second set of beam reports for at least one beam measured at a second time instance, A method wherein the first beam report set includes beam reports for a greater number of beams than the second beam report set.

3. In claim 2, A method wherein the first time instance comprises one slot, a plurality of slots, one subframe, or a plurality of subframes.

4. In claim 1, The data report includes a first set of beam reports for at least one beam measured at a first time instance within the measurement interval and a second set of beam reports for at least one beam measured at a second time instance, A method wherein the first beam report set includes at least one beam report generated at a greater quantization level than the second beam report set.

5. In claim 1, The method of claim 1, wherein the data report comprises beam reports for a predefined or network-configured number of beams that fall above the measured values ​​among the measured beams.

6. In claim 1, The above data report is, First model training information including measurement values ​​for all or part of the first beam set used as input data of the above model, Second model training information including measurement values ​​for all or part of a second beam set that is the target of beam measurement or a third beam set including beams predicted through the model; Third model training information including identification information of a specific number of beams belonging to the upper portion of the measured values ​​among the beams included in the second beam set or the third beam set; Fourth model training information including a time stamp indicating the time at which the first model training information, the second model training information, or the third model training information was generated, or A method comprising at least one of the fifth model training information including assistance information for the terminal that is helpful in training the model.

7. In claim 6, A method wherein the fifth model training information includes at least one of a moving speed of the terminal, a location of the terminal, or a distance between the terminal and the base station.

8. In claim 6, Further comprising transmitting configuration information for transmission of the above data report to the terminal, A method in which the configuration information indicates at least one item included in the data report among the first model training information, the second model training information, the third model training information, the fourth model training information, or the fifth model training information.

9. In claim 6, A method wherein, when the target of the model training is a classification-based model, the data report includes the first model training information, the third model training information, and the fourth model training information.

10. In claim 6, A method wherein, when the target of the model training is a regression-based model, the data report includes the first model training information, the second model training information, and the fourth model training information.

11. In claim 1, Further comprising receiving configuration information for transmission of the above data report, The above configuration information is a method for configuring a periodic report, a semi-periodic report, or an aperiodic report.

12. In claim 1, Further comprising receiving configuration information for transmission of the above data report, A method wherein the above configuration information includes at least one of information about the number of beams being reported or information about a threshold of a measurement value for determining whether to report each beam.

13. In claim 12, The above threshold value includes a threshold value of a measurement value or a threshold value of a change amount.

14. In a method of operating a base station in a wireless communication system, Transmitting reference signals to generate beam information for model training; Receiving a data report comprising at least a portion of measurement data generated based on measurements of the terminal for the above reference signals; Performing training for the model based on the above data report; and Including transmitting a beam instruction generated by inference using the above model, A method wherein the data report includes beam reports for beams measured within a corresponding measurement interval, excluding at least one other beam report included in at least one other data report for inference transmitted within the measurement interval.

15. In claim 14, A method wherein the training is performed using beam reports included in the data report and at least one other beam report included in at least one other data report for inference received within the measurement interval.

16. In claim 14, A step of transmitting configuration information for the above data report is included, A method wherein the above configuration information includes at least one of a method for selecting some of the beams measured by the terminal or a method for expressing the measurement value.

17. In claim 14, A method wherein the above data report includes a beam report for at least one beam having a measurement value greater than a threshold among beams measured by the terminal.

18. In claim 14, The above data report includes beam reports for a specified number of beams having measurement values ​​higher than a threshold among beams measured by the terminal, The above specified number is configured or predefined by the base station.

19. In a wireless communication system, at a terminal, At least one transmitter / receiver; at least one processor; and At least one memory operably connected to said at least one processor and storing instructions that, when executed by said processor, control said terminal to perform operations; The above actions are, Receiving reference signals for generating beam information for model training; Transmitting a data report including at least a portion of measurement data generated based on the terminal's measurement of the above reference signals; and Receiving a beam instruction generated by inference using a trained model based on the above data report, The above data report is a terminal that includes beam reports for beams measured within a corresponding measurement interval, excluding at least one other beam report included in at least one other data report for inference transmitted within the measurement interval.

20. In a base station in a wireless communication system, At least one transmitter / receiver; at least one processor; and At least one memory operably connected to said at least one processor and storing instructions that, when executed by said processor, control said terminal to perform operations; The above actions are, Transmitting reference signals to generate beam information for model training; Receiving a data report comprising at least a portion of measurement data generated based on measurements of the terminal for the above reference signals; Performing training for the model based on the above data report; and Including transmitting a beam instruction generated by inference using the above model, A base station wherein the above data report includes beam reports for beams measured within a corresponding measurement interval, excluding at least one other beam report included in at least one other data report for inference transmitted within the measurement interval.

Images