Method by which device performs communication in wireless communication system and device therefor

Abstract

According to various embodiments, disclosed are a method for performing communication in a wireless communication system and a device therefor. The device may: receive CSI reporting configuration information; measure CSI on the basis of the CSI reporting configuration information; and transmit output information about an AI model having the CSI as an input.

Description

A method for a device to perform communication in a wireless communication system and a device for the same

[0001] The present invention relates to a method for a terminal to transmit feedback information related to channel measurement information in a wireless communication system and an apparatus for the same.

[0002] A wireless communication system is a multiple access system that supports communication with multiple users by sharing available system resources (e.g., bandwidth, transmission power, etc.). Examples of multiple access systems include CDMA (code division multiple access), FDMA (frequency division multiple access), TDMA (time division multiple access), OFDMA (orthogonal frequency division multiple access), SC-FDMA (single carrier frequency division multiple access), and MC-FDMA (multi carrier frequency division multiple access) systems.

[0003] Sidelink (SL) refers to a communication method in which User Equipment (UE) establishes a direct link to directly exchange voice or data between terminals without passing through a Base Station (BS). SL is being considered as a solution to address the burden on base stations caused by rapidly increasing data traffic.

[0004] V2X (vehicle-to-everything) refers to a communication technology that exchanges information with other vehicles, pedestrians, and infrastructure-equipped objects through wired or wireless communication. V2X can be classified into four types: V2V (vehicle-to-vehicle), V2I (vehicle-to-infrastructure), V2N (vehicle-to-network), and V2P (vehicle-to-pedestrian). V2X communication can be provided through PC5 interfaces and / or Uu interfaces.

[0005] Meanwhile, as more communication devices require larger communication capacities, the need for improved mobile broadband communication compared to existing Radio Access Technology (RAT) is emerging. Accordingly, communication systems considering services or terminals sensitive to reliability and latency are being discussed; next-generation radio access technology that incorporates improved mobile broadband communication, Massive Machine Type Communication (MTC), and Ultra-Reliable and Low Latency Communication (URLC) can be referred to as new radio access technology (new RAT) or new radio (NR). Vehicle-to-everything (V2X) communication can also be supported in NR.

[0006] Figure 1 is a diagram illustrating a comparison between V2X communication based on RAT prior to NR and V2X communication based on NR.

[0007] Regarding V2X communication, prior to NR, RATs mainly discussed methods for providing safety services based on V2X messages such as BSM (Basic Safety Message), CAM (Cooperative Awareness Message), and DENM (Decentralized Environmental Notification Message). V2X messages can include location information, dynamic information, attribute information, etc. For example, a terminal can transmit a CAM of the periodic message type and / or a DENM of the event-triggered message type to another terminal.

[0008] For example, the CAM may include basic vehicle information such as dynamic state information of the vehicle, such as direction and speed, static data of the vehicle, such as dimensions, external lighting conditions, and route history. For example, a terminal may broadcast the CAM, and the latency of the CAM may be less than 100ms. For example, in the event of an unexpected situation such as a vehicle breakdown or accident, the terminal may generate a DENM and transmit it to other terminals. For example, all vehicles within the transmission range of the terminal may receive the CAM and / or DENM. In this case, the DENM may have a higher priority than the CAM.

[0009] Since then, various V2X scenarios regarding V2X communication have been presented in NR. For example, various V2X scenarios may include vehicle platooning, advanced driving, extended sensors, remote driving, etc.

[0010] For example, based on vehicle platooning, vehicles can dynamically form groups and move together. For example, to perform platoon operations based on vehicle platooning, vehicles belonging to said group can receive periodic data from the lead vehicle. For example, vehicles belonging to said group can use said periodic data to reduce or increase the distance between vehicles.

[0011] For example, based on enhanced driving, vehicles can be semi-automated or fully automated. For example, each vehicle can adjust trajectories or maneuvers based on data acquired from local sensors of nearby vehicles and / or nearby logical entities. Additionally, for example, each vehicle can mutually share driving intentions with nearby vehicles.

[0012] For example, based on extended sensors, raw data or processed data or live video data acquired through local sensors can be exchanged between vehicles, logical entities, pedestrian terminals and / or V2X application servers. Thus, for example, a vehicle can perceive an environment that is enhanced compared to the environment it can detect using its own sensors.

[0013] For example, based on remote driving, a remote driver or V2X application can operate or control a remote vehicle for a person unable to drive or for a remote vehicle located in a dangerous environment. For example, in cases where the route is predictable, such as in public transportation, cloud computing-based driving can be used for the operation or control of the remote vehicle. Additionally, access to a cloud-based back-end service platform, for example, can be considered for remote driving.

[0014] Meanwhile, methods to specify service requirements for various V2X scenarios, such as vehicle platooning, enhanced driving, extended sensors, and remote driving, are being discussed in NR-based V2X communication.

[0015] The technical problem that the present invention aims to solve is to provide a method for transmitting feedback information and monitoring information related to CSI more accurately and efficiently.

[0016] The technical problems are not limited to those mentioned above, and other technical problems not mentioned will be clearly understood by those skilled in the art to which the present invention belongs from the description below.

[0017] A method by a UE (User Equipment) according to one aspect comprises the steps of: receiving CSI (channel state information) reporting setting information; measuring CSI based on the CSI reporting setting information; and transmitting output information of an AI (Artificial Intelligence) model having the CSI as input, wherein the payload size of the output information of the AI ​​model may be determined based on whether the output information of the AI ​​model is related to performance monitoring of the AI ​​model.

[0018] Alternatively, based on the fact that the output information of the AI ​​model is output information for performance monitoring of the AI ​​model, the payload size of the output information of the AI ​​model is determined to be a first payload size, and based on the fact that the output information of the AI ​​model is output information for CSI inference based on the CSI, the payload size of the output information of the AI ​​model may be determined to be a second payload size different from the first payload size.

[0019] Alternatively, the first payload size may be larger than the second payload size.

[0020] Alternatively, the output information of the AI ​​model having the first payload size may have a finer quantization grain than the output information of the AI ​​model having the second payload size.

[0021] Alternatively, the UE transmits output information of the AI ​​model based on the CSI report setting information N times, and one of the N transmissions may be a transmission of output information for performance monitoring of the AI ​​model.

[0022] Alternatively, the above CSI report setting information may include information for setting at least one of a plurality of reporting opportunities as a reporting opportunity related to performance monitoring of the AI ​​model.

[0023] Alternatively, the method may further include the step of receiving a control signal from a base station to trigger or activate at least one CSI report setting among a plurality of CSI report settings included in the CSI report setting information for performance monitoring of the AI ​​model.

[0024] Alternatively, output information for performance monitoring of the AI ​​model and output information for CSI inference may be latent information.

[0025] Alternatively, based on the fact that the output information of the AI ​​model is output information for monitoring the performance of the AI ​​model, the output information of the AI ​​model may be transmitted together with additional information indicating the difference in data distribution between the training CSI included as training data of the AI ​​model and the CSI.

[0026] According to another aspect, at least one non-transient computer-readable recording medium comprises instructions for performing operations when executed by at least one processor, said operations include receiving CSI (channel state information) reporting setting information; measuring CSI based on said CSI reporting setting information; and transmitting output information of an AI (Artificial Intelligence) model having said CSI as input, and the payload size of said AI model output information may be determined based on whether said AI model output information is related to performance monitoring of said AI model.

[0027] According to another aspect, a UE (User Equipment) includes an RF (Radio Frequency) transceiver; and a processor connected to the RF transceiver, wherein the processor controls the RF transceiver to receive CSI (channel state information) reporting setting information, measures the CSI based on the CSI reporting setting information, transmits output information of an AI (Artificial Intelligence) model with the CSI as input, and the payload size of the output information of the AI ​​model may be determined based on whether the output information of the AI ​​model is related to performance monitoring of the AI ​​model.

[0028] According to another aspect, a processing device controlling a UE (User Equipment) comprises at least one processor; and at least one memory connected to the at least one processor and storing instructions that perform operations when executed by the at least one processor, wherein the operations include causing the UE to: receive CSI (channel state information) reporting setting information; measure CSI based on the CSI reporting setting information; and transmit output information of an AI (Artificial Intelligence) model having the CSI as input, and the payload size of the output information of the AI ​​model may be determined based on whether the output information of the AI ​​model is related to performance monitoring of the AI ​​model.

[0029] A method by a base station according to another aspect comprises the steps of: transmitting CSI (channel state information) reporting setting information; and receiving output information of an AI (Artificial Intelligence) model from a UE (User Equipment) that takes a measured CSI based on the CSI reporting setting information as input, wherein the CSI reporting setting information may include information for setting the payload size of the output information of the AI ​​model differently based on whether the output information of the AI ​​model is related to performance monitoring of the AI ​​model.

[0030] A base station according to another aspect includes an RF (Radio Frequency) transceiver; and a processor connected to the RF transceiver, wherein the processor controls the RF transceiver to transmit CSI (channel state information) reporting setting information and receives output information of an AI (Artificial Intelligence) model from a UE (User Equipment) that takes a measured CSI based on the CSI reporting setting information as input, and the CSI reporting setting information may include information for setting the payload size of the output information of the AI ​​model differently based on whether the output information of the AI ​​model is related to performance monitoring of the AI ​​model.

[0031] According to one embodiment of the present invention, the transmission of feedback information and monitoring information related to CSI in a wireless communication system can be performed more accurately and efficiently. For example, even without separately reporting raw target CSI or ground-truth CSI, a network-side estimator can calculate intermediate KPIs or performance monitoring outputs with significant accuracy based on latent information of the CSI compression model reported or transmitted by the UE, particularly latent information expressed with a higher or finer granularity.

[0032] The effects obtainable from various embodiments are not limited to those mentioned above, and other unmentioned effects will be clearly understood by those skilled in the art from the description below.

[0033] The drawings attached to this specification are intended to provide an understanding of the present invention, to illustrate various embodiments of the invention, and to explain the principles of the invention together with the description in the specification.

[0034] Figure 1 is a diagram illustrating a comparison between V2X communication based on RAT prior to NR and V2X communication based on NR.

[0035] Figure 2 shows the structure of an LTE system.

[0036] Figure 3 shows the structure of the NR system.

[0037] Figure 4 shows the structure of a wireless frame of NR.

[0038] Figure 5 shows the slot structure of an NR frame.

[0039] FIG. 6 shows a communication structure that can be provided in a 6G system according to one embodiment of the present disclosure.

[0040] FIG. 7 shows an electromagnetic spectrum according to one embodiment of the present disclosure.

[0041] Figure 8 shows the radio protocol architecture for SL communication.

[0042] Figure 9 shows a terminal performing V2X or SL communication.

[0043] Figure 10 shows a resource unit for V2X or SL communication.

[0044] FIG. 11 shows an example of a BWP according to one embodiment of the present disclosure.

[0045] FIG. 12 illustrates a procedure in which a terminal performs V2X or SL communication according to a resource allocation mode, according to one embodiment of the present disclosure.

[0046] FIGS. 13 and FIGS. 14 illustrate an example of a sensing operation according to an embodiment of the present disclosure.

[0047] FIG. 15 illustrates a time / frequency resource for a sensing operation according to one embodiment of the present specification.

[0048] Figure 16 illustrates a general functional architecture for an AI / ML model.

[0049] Figure 17 is a diagram illustrating an inference method during AI / ML-based CSI compression.

[0050] Figure 18 is a diagram illustrating a method for performing AI / ML-based CSI prediction.

[0051] Figure 19 is a diagram illustrating a method for performing performance monitoring in CSI compression.

[0052] FIG. 20 is a diagram illustrating a method for a UE to transmit output information for monitoring an AI model to a base station.

[0053] FIG. 21 is a diagram illustrating how a base station receives output information of an AI model from a UE.

[0054] FIG. 22 illustrates a communication system to which the present invention is applied.

[0055] FIG. 23 illustrates a wireless device that can be applied to the present invention.

[0056] FIG. 24 illustrates another example of a wireless device to which the present invention applies. The wireless device may be implemented in various forms depending on the use-example / service.

[0057] FIG. 25 illustrates a vehicle or autonomous vehicle to which the present invention is applied.

[0058] A wireless communication system is a multiple access system that supports communication with multiple users by sharing available system resources (e.g., bandwidth, transmission power, etc.). Examples of multiple access systems include CDMA (code division multiple access), FDMA (frequency division multiple access), TDMA (time division multiple access), OFDMA (orthogonal frequency division multiple access), SC-FDMA (single carrier frequency division multiple access), and MC-FDMA (multi carrier frequency division multiple access) systems.

[0059] Sidelink refers to a communication method in which User Equipment (UE) establishes a direct link to directly exchange voice or data between terminals without passing through a Base Station (BS). Sidelink is being considered as a solution to address the burden on base stations caused by rapidly increasing data traffic.

[0060] V2X (vehicle-to-everything) refers to a communication technology that exchanges information with other vehicles, pedestrians, and infrastructure-equipped objects through wired or wireless communication. V2X can be classified into four types: V2V (vehicle-to-vehicle), V2I (vehicle-to-infrastructure), V2N (vehicle-to-network), and V2P (vehicle-to-pedestrian). V2X communication can be provided through PC5 interfaces and / or Uu interfaces.

[0061] Meanwhile, as more communication devices require larger communication capacities, the need for improved mobile broadband communication compared to existing Radio Access Technology (RAT) is emerging. Accordingly, communication systems considering services or terminals sensitive to reliability and latency are being discussed; next-generation radio access technology that incorporates improved mobile broadband communication, Massive MTC, and URLLC (Ultra-Reliable and Low Latency Communication) can be referred to as new radio access technology (new RAT) or new radio (NR). Vehicle-to-everything (V2X) communication can also be supported in NR.

[0062] The following technologies can be used in various wireless communication systems such as CDMA (code division multiple access), FDMA (frequency division multiple access), TDMA (time division multiple access), OFDMA (orthogonal frequency division multiple access), and SC-FDMA (single carrier frequency division multiple access). CDMA can be implemented using wireless technologies such as UTRA (universal terrestrial radio access) or CDMA2000. TDMA can be implemented using wireless technologies such as GSM (global system for mobile communications), GPRS (general packet radio service), and EDGE (enhanced data rates for GSM evolution). OFDMA can be implemented using wireless technologies such as IEEE (institute of electrical and electronics engineers) 802.11 (Wi-Fi), IEEE 802.16 (WiMAX), IEEE 802-20, and E-UTRA (evolved UTRA). IEEE 802.16m is an evolution of IEEE 802.16e and provides backward compatibility with systems based on IEEE 802.16e. UTRA is part of UMTS (universal mobile telecommunications system). 3GPP (3rd generation partnership project) LTE (long term evolution) is part of E-UMTS (evolved UMTS) which uses E-UTRA (evolved-UMTS terrestrial radio access), employing OFDMA in the downlink and SC-FDMA in the uplink.LTE-A (advanced) is an evolution of 3GPP LTE.

[0063] 5G NR is a successor technology to LTE-A and is a new clean-slate type mobile communication system with characteristics such as high performance, low latency, and high availability. 5G NR can utilize all available spectrum resources, ranging from low frequency bands below 1 GHz to mid-frequency bands from 1 GHz to 10 GHz, and high frequency (millimeter wave) bands above 24 GHz.

[0064] For clarity of explanation, the description focuses on LTE-A or 5G NR, but the technical concept of the embodiment(s) is not limited thereto.

[0065] Figure 2 shows the structure of an applicable LTE system. This can be called an E-UTRAN (Evolved-UMTS Terrestrial Radio Access Network), or an LTE (Long Term Evolution) / LTE-A system.

[0066] Referring to FIG. 2, the E-UTRAN includes a base station (20; Base Station, BS) that provides a control plane and a user plane to a terminal (10). The terminal (10) may be fixed or mobile and may be referred to by other terms such as MS (Mobile Station), UT (User Terminal), SS (Subscriber Station), MT (Mobile Terminal), or Wireless Device. The base station (20) refers to a fixed station that communicates with the terminal (10) and may be referred to by other terms such as eNB (evolved-NodeB), BTS (Base Transceiver System), or Access Point.

[0067] Base stations (20) can be connected to each other through an X2 interface. The base station (20) is connected to the EPC (Evolved Packet Core, 30) through the S1 interface, more specifically to the MME (Mobility Management Entity) through the S1-MME and to the S-GW (Serving Gateway) through the S1-U.

[0068] The EPC (30) consists of an MME, an S-GW, and a P-GW (Packet Data Network-Gateway). The MME holds information regarding the terminal's connection information or capabilities, and this information is primarily used for managing the terminal's mobility. The S-GW is a gateway with an E-UTRAN as its endpoint, and the P-GW is a gateway with a PDN as its endpoint.

[0069] The layers of the Radio Interface Protocol between a terminal and a network can be classified into L1 (Layer 1), L2 (Layer 2), and L3 (Layer 3) based on the lower three layers of the Open System Interconnection (OSI) model, which is widely known in communication systems. Among these, the Physical Layer, belonging to Layer 1, provides Information Transfer Services using a physical channel, while the Radio Resource Control (RRC) layer, located at Layer 3, performs the role of controlling radio resources between the terminal and the network. To this end, the RRC layer exchanges RRC messages between the terminal and the base station.

[0070] Figure 3 shows the structure of the NR system.

[0071] Referring to FIG. 3, the NG-RAN may include gNBs and / or eNBs that provide user plane and control plane protocol termination to terminals. FIG. 7 illustrates a case where only gNBs are included. The gNBs and eNBs are connected to each other via Xn interfaces. The gNBs and eNBs are connected to the 5G Core Network (5GC) via NG interfaces. More specifically, they are connected to the access and mobility management function (AMF) via NG-C interfaces and to the user plane function (UPF) via NG-U interfaces.

[0072] Figure 4 shows the structure of a wireless frame of NR.

[0073] Referring to FIG. 4, radio frames can be used for uplink and downlink transmission in NR. The radio frame has a length of 10 ms and can be defined as two 5 ms half-frames (HF). A half-frame may contain five 1 ms subframes (SF). A subframe may be divided into one or more slots, and the number of slots within a subframe may be determined by the subcarrier spacing (SCS). Each slot may contain 12 or 14 OFDM(A) symbols according to the cyclic prefix (CP).

[0074] When normal CP is used, each slot may contain 14 symbols. When extended CP is used, each slot may contain 12 symbols. Here, the symbols may include OFDM symbols (or CP-OFDM symbols) and SC-FDMA (Single Carrier - FDMA) symbols (or DFT-s-OFDM (Discrete Fourier Transform-spread-OFDM) symbols).

[0075] Table 1 below shows the number of symbols per slot ((N) according to the SCS setting (u) when normal CP is used. slot symb ), number of slots per frame((N frame,u slot ) and the number of slots per subframe((N subframe,u slot ) exemplifies.

[0076] SCS (15*2 u )N slot symb N frame,u slot N subframe,u slot 15KHz (u=0)1410130KHz (u=1)1420260KHz (u=2)14404120KHz (u=3)14808240KHz (u=4)1416016

[0077] Table 2 shows the number of symbols per slot, the number of slots per frame, and the number of slots per subframe according to the SCS when an extended CP is used.

[0078] SCS (15*2 u )N slot symb N frame,u slot N subframe,u slot 60KHz (u=2)12404

[0079] In an NR system, the OFDM(A) numerology (e.g., SCS, CP length, etc.) can be configured differently among multiple cells that are merged into a single terminal. Accordingly, the (absolute time) interval of a time resource (e.g., subframe, slot, or TTI) (collectively referred to as TU (Time Unit) for convenience) composed of the same number of symbols can be configured differently among the merged cells.

[0080] In NR, multiple numerologies or SCSs may be supported to support various 5G services. For example, if the SCS is 15 kHz, a wide area in traditional cellular bands may be supported, and if the SCS is 30 kHz / 60 kHz, dense-urban, lower latency, and wider carrier bandwidth may be supported. If the SCS is 60 kHz or higher, a bandwidth greater than 24.25 GHz may be supported to overcome phase noise.

[0081] The NR frequency band can be defined by two types of frequency ranges. The two types of frequency ranges may be FR1 and FR2. The numerical values ​​of the frequency ranges may change, for example, as shown in Table 3 below. Among the frequency ranges used in an NR system, FR1 may mean "sub 6GHz range" and FR2 may mean "above 6GHz range" and may be referred to as millimeter wave (mmW).

[0082] Frequency Range designationCorresponding frequency rangeSubcarrier Spacing (SCS)FR1450MHz - 6000MHz15, 30, 60kHzFR224250MHz - 52600MHz60, 120, 240kHz

[0083] As described above, the numerical value of the frequency range of the NR system may change. For example, FR1 may include a band of 410 MHz to 7125 MHz as shown in Table 4 below. That is, FR1 may include a frequency band of 6 GHz (or 5850, 5900, 5925 MHz, etc.) or higher. For example, the frequency band of 6 GHz (or 5850, 5900, 5925 MHz, etc.) or higher included within FR1 may include an unlicensed band. The unlicensed band may be used for various purposes, for example, for communication for vehicles (e.g., autonomous driving).

[0084] Frequency Range designationCorresponding frequency rangeSubcarrier Spacing (SCS)FR1410MHz - 7125MHz15, 30, 60kHzFR224250MHz - 52600MHz60, 120, 240kHz

[0085] Figure 5 shows the slot structure of an NR frame.

[0086] Referring to FIG. 5, a slot contains multiple symbols in the time domain. For example, in the case of a normal CP, one slot may contain 14 symbols, but in the case of an extended CP, one slot may contain 12 symbols. Alternatively, in the case of a normal CP, one slot may contain 7 symbols, but in the case of an extended CP, one slot may contain 6 symbols.

[0087] A carrier includes multiple subcarriers in the frequency domain. A Resource Block (RB) can be defined as multiple (e.g., 12) consecutive subcarriers in the frequency domain. A Bandwidth Part (BWP) can be defined as multiple consecutive (P)RBs ((Physical) Resource Blocks) in the frequency domain and can correspond to a single numerology (e.g., SCS, CP length, etc.). A carrier can include up to N (e.g., 5) BWPs. Data communication can be performed through the active BWPs. Each element can be referred to as a Resource Element (RE) in a resource grid and can be mapped to a single complex symbol.

[0088] Meanwhile, a wireless interface between terminals or a wireless interface between a terminal and a network may be composed of L1, L2, and L3 layers. In various embodiments of the present disclosure, L1 layer may refer to the physical layer. Additionally, for example, L2 layer may refer to at least one of the MAC layer, RLC layer, PDCP layer, and SDAP layer. Additionally, for example, L3 layer may refer to the RRC layer.

[0089] FIG. 6 illustrates a communication structure that can be provided in a 6G system according to one embodiment of the present disclosure. The embodiment of FIG. 6 can be combined with various embodiments of the present disclosure.

[0090] New network characteristics in 6G may be as follows.

[0091] - Satellite Integrated Network

[0092] - Connected Intelligence: Unlike previous generations of wireless communication systems, 6G is innovative and will update wireless evolution from "connected things" to "connected intelligence." AI can be applied at each stage of the communication process (or at each step of the signal processing described below).

[0093] - Seamless integration of wireless information and energy transfer

[0094] - Ubiquitous Super 3D Connectivity: Connectivity to the network and core network functions of drones and very low Earth orbit satellites will create Super 3D connectivity in 6G ubiquitous.

[0095] Some general requirements regarding the new network characteristics of 6G mentioned above may be as follows.

[0096] - Small cell networks

[0097] - Ultra-dense heterogeneous network

[0098] - High-capacity backhaul

[0099] - Radar technology integrated with mobile technology: High-precision localization (or location-based services) through communication is one of the functions of 6G wireless communication systems. Therefore, radar systems will be integrated with 6G networks.

[0100] - Softwarization and virtualization

[0101] The core implementation technologies of the 6G system are described below.

[0102] - Artificial Intelligence: Introducing AI into communications can streamline and enhance real-time data transmission. AI can determine how complex target tasks are performed using numerous analyses. In other words, AI can increase efficiency and reduce processing latency. Time-consuming tasks such as handover, network selection, and resource scheduling can be performed instantly using AI. AI can also play a significant role in M2M, machine-to-human, and human-to-machine communication. Furthermore, AI can enable rapid communication in Brain-Computer Interfaces (BCI). AI-based communication systems can be supported by metamaterials, intelligent structures, intelligent networks, intelligent devices, intelligent cognitive radios, self-sustaining wireless networks, and machine learning.

[0103] - THz Communication: Data transmission rates can be increased by expanding bandwidth. This can be achieved by using sub-THz communication with wide bandwidth and applying advanced large-scale MIMO technology. THz waves, also known as sub-millimeter radiation, generally refer to a frequency band between 0.1 THz and 10 THz with corresponding wavelengths ranging from 0.03 mm to 3 mm. The 100 GHz-300 GHz band range (Sub-THz band) is considered the primary portion of the THz band for cellular communication. Adding the Sub-THz band to the mmWave band increases 6G cellular communication capacity. Among the defined THz bands, the 300 GHz-3 THz band is located in the far-infrared (IR) frequency band. Although the 300 GHz-3 THz band is part of the optical band, it lies at the boundary of the optical band and immediately following the RF band. Therefore, this 300 GHz-3 THz band exhibits similarities to RF.

[0104] FIG. 7 illustrates an electromagnetic spectrum according to one embodiment of the present disclosure. The embodiment of FIG. 7 may be combined with various embodiments of the present disclosure. Key characteristics of THz communication include (i) a widely available bandwidth to support very high data transmission rates, and (ii) high path loss occurring at high frequencies (highly directional antennas are indispensable). The narrow beam width generated by highly directional antennas reduces interference. The small wavelength of THz signals allows a much larger number of antenna elements to be integrated into devices and BSs operating in this band. This enables the use of advanced adaptive array techniques that can overcome range limitations.

[0105] - Large-scale MIMO technology

[0106] - Hologram beamforming (HBF)

[0107] - Optical wireless technology

[0108] - Free Space Optical Transmission Backhaul Network (FSO backhaul network)

[0109] - Quantum communication

[0110] - Cell-free communication

[0111] - Integration of wireless information and power transmission

[0112] - Integration of wireless communication and sensing

[0113] - Integrated access and backhaul network

[0114] - Big data analysis

[0115] - Reconfigurable intelligent metasurface

[0116] - Metaverse

[0117] - blockchain

[0118] - Unmanned Aerial Vehicle (UAV): UAVs or drones will be a critical element in 6G wireless communication. In most cases, high-speed data wireless connectivity can be provided using UAV technology. Base station (BS) entities can be installed on UAVs to provide cellular connectivity. UAVs can possess specific features not found in fixed BS infrastructure, such as easy deployment, robust line-of-sight links, and controlled degrees of freedom for mobility. During emergencies, such as natural disasters, the deployment of ground communication infrastructure is not economically feasible, and sometimes services cannot be provided in volatile environments. UAVs can easily handle these situations. UAVs will become a new paradigm in the field of wireless communication. This technology facilitates the three fundamental requirements of wireless networks: eMBB, URLLC, and mMTC. UAVs can also support various purposes, such as enhancing network connectivity, fire detection, disaster emergency services, security and surveillance, pollution monitoring, parking monitoring, and accident monitoring. Therefore, UAV technology is recognized as one of the most critical technologies for 6G communication.

[0119] - Autonomous Driving (Self-Driving): V2X (Vehicle to Everything), a core element in building autonomous driving infrastructure, refers to technologies that enable vehicles to communicate and share with various elements on the road for autonomous driving, such as wireless communication between vehicles (Vehicle to Vehicle, V2V) and between vehicles and infrastructure (Vehicle to Infrastructure, V2I). Fast transmission speeds and low-latency technologies are essential to maximize autonomous driving performance and ensure high safety. Furthermore, future autonomous driving may go beyond simply delivering warning or guidance messages to the driver to actively intervene in vehicle operation and directly control the vehicle in dangerous situations. Since the amount of information to be transmitted and received may become massive for this purpose, it is expected that 6G will be able to maximize autonomous driving through faster transmission speeds and lower latency compared to 5G.

[0120] FIG. 8 illustrates a radio protocol architecture for SL communication. Specifically, FIG. 8 (a) shows the user plane protocol stack of NR, and FIG. 8 (b) shows the control plane protocol stack of NR.

[0121] The Sidelink Synchronization Signal (SLSS) and synchronization information are described below.

[0122] SLSS is an SL-specific sequence that may include PSSS (Primary Sidelink Synchronization Signal) and SSSS (Secondary Sidelink Synchronization Signal). The PSSS may be referred to as S-PSS (Sidelink Primary Synchronization Signal), and the SSSS may be referred to as S-SSS (Sidelink Secondary Synchronization Signal). For example, length-127 M-sequences may be used for S-PSS, and length-127 Gold sequences may be used for S-SSS. For example, a terminal may use S-PSS to detect a primary signal and obtain synchronization. For example, a terminal may use S-PSS and S-SSS to obtain detailed synchronization and detect a synchronization signal ID.

[0123] PSBCH (Physical Sidelink Broadcast Channel) may be a (broadcast) channel through which basic (system) information that a terminal must know first is transmitted before transmitting or receiving SL signals. For example, the basic information may include information related to SLSS, Duplex Mode (DM), TDD UL / DL (Time Division Duplex Uplink / Downlink) configuration, information related to resource pools, types of applications related to SLSS, subframe offsets, broadcast information, etc. For example, to evaluate PSBCH performance, in NR V2X, the payload size of PSBCH may be 56 bits, including a 24-bit CRC.

[0124] S-PSS, S-SSS, and PSBCH may be included in a block format that supports periodic transmission (e.g., SL SS (Synchronization Signal) / PSBCH block, hereinafter S-SSB (Sidelink-Synchronization Signal Block)). The S-SSB may have the same numerology (i.e., SCS and CP lengths) as the PSCCH (Physical Sidelink Control Channel) / PSSCH (Physical Sidelink Shared Channel) within the carrier, and the transmission bandwidth may be within a (pre-)set SL BWP (Sidelink BWP). For example, the bandwidth of the S-SSB may be 11 RB (Resource Block). For example, the PSBCH may span 11 RB. Additionally, the frequency position of the S-SSB may be (pre-)set. Therefore, the terminal does not need to perform hypothesis detection at the frequency to discover the S-SSB in the carrier.

[0125] Meanwhile, in an NR SL system, multiple numerologies having different SCS and / or CP lengths may be supported. In this case, as the SCS increases, the length of the time resource for the transmitting terminal to transmit S-SSBs may decrease. Consequently, the coverage of S-SSBs may decrease. Therefore, to ensure S-SSB coverage, the transmitting terminal may transmit one or more S-SSBs to the receiving terminal within a single S-SSB transmission cycle according to the SCS. For example, the number of S-SSBs transmitted by the transmitting terminal to the receiving terminal within a single S-SSB transmission cycle may be pre-configured or configured for the transmitting terminal. For example, the S-SSB transmission cycle may be 160ms. For example, an S-SSB transmission cycle of 160ms may be supported for all SCSs.

[0126] For example, if the SCS is 15 kHz at FR1, the transmitting terminal may transmit one or two S-SSBs to the receiving terminal within one S-SSB transmission cycle. For example, if the SCS is 30 kHz at FR1, the transmitting terminal may transmit one or two S-SSBs to the receiving terminal within one S-SSB transmission cycle. For example, if the SCS is 60 kHz at FR1, the transmitting terminal may transmit one, two, or four S-SSBs to the receiving terminal within one S-SSB transmission cycle.

[0127] For example, if the SCS is 60 kHz at FR2, the transmitting terminal can transmit 1, 2, 4, 8, 16, or 32 S-SSBs to the receiving terminal within one S-SSB transmission cycle. For example, if the SCS is 120 kHz at FR2, the transmitting terminal can transmit 1, 2, 4, 8, 16, 32, or 64 S-SSBs to the receiving terminal within one S-SSB transmission cycle.

[0128] Meanwhile, when the SCS is 60 kHz, two types of CP may be supported. Additionally, depending on the CP type, the structure of the S-SSB transmitted by the transmitting terminal to the receiving terminal may differ. For example, the CP type may be Normal CP (NCP) or Extended CP (ECP). Specifically, for example, if the CP type is NCP, the number of symbols mapping PSBCH within the S-SSB transmitted by the transmitting terminal may be 9 or 8. On the other hand, for example, if the CP type is ECP, the number of symbols mapping PSBCH within the S-SSB transmitted by the transmitting terminal may be 7 or 6. For example, PSBCH may be mapped to the first symbol within the S-SSB transmitted by the transmitting terminal. For example, the receiving terminal receiving the S-SSB may perform Automatic Gain Control (AGC) operation during the first symbol interval of the S-SSB.

[0129] Figure 9 shows a terminal performing V2X or SL communication.

[0130] Referring to FIG. 9, in V2X or SL communication, the term terminal may primarily refer to a user's terminal. However, if network equipment such as a base station transmits and receives signals according to the communication method between terminals, the base station may also be considered a type of terminal. For example, terminal 1 may be a first device (100), and terminal 2 may be a second device (200).

[0131] For example, terminal 1 can select a resource unit corresponding to a specific resource within a resource pool, which represents a set of resources. Then, terminal 1 can transmit an SL signal using the said resource unit. For example, terminal 2, which is a receiving terminal, can be configured with a resource pool in which terminal 1 can transmit a signal, and can detect terminal 1's signal within said resource pool.

[0132] Here, if terminal 1 is within the connection range of the base station, the base station may inform terminal 1 of the resource pool. On the other hand, if terminal 1 is outside the connection range of the base station, another terminal may inform terminal 1 of the resource pool, or terminal 1 may use a pre-configured resource pool.

[0133] Generally, a resource pool can be composed of multiple resource units, and each terminal can select one or more resource units to use for its SL signal transmission.

[0134] Figure 10 shows a resource unit for V2X or SL communication.

[0135] Referring to FIG. 10, the total frequency resources of the resource pool can be divided into NF units, and the total time resources of the resource pool can be divided into NT units. Thus, a total of NF * NT resource units can be defined within the resource pool. FIG. 10 illustrates an example where the resource pool is repeated in a period of NT subframes.

[0136] As shown in FIG. 10, a single resource unit (e.g., Unit #0) may appear repeatedly over time. Alternatively, to obtain diversity effects in the time or frequency dimension, the index of the physical resource unit to which a single logical resource unit is mapped may change in a predetermined pattern over time. In this structure of resource units, a resource pool may refer to a set of resource units that a terminal intending to transmit an SL signal can use for transmission.

[0137] Resource pools can be subdivided into several types. For example, depending on the content of the SL signals transmitted from each resource pool, resource pools can be classified as follows.

[0138] (1) A Scheduling Assignment (SA) may be a signal containing information such as the location of the resource used by the transmitting terminal for transmission of the SL data channel, the Modulation and Coding Scheme (MCS) or Multiple Input Multiple Output (MIMO) transmission method required for demodulation of the data channel, and Timing Advance (TA). The SA may also be multiplexed and transmitted together with the SL data on the same resource unit, in which case the SA resource pool may refer to a resource pool in which the SA is multiplexed and transmitted together with the SL data. The SA may also be called the SL control channel.

[0139] (2) A Physical Sidelink Shared Channel (PSSCH) may be a resource pool used by a transmitting terminal to transmit user data. If SA is multiplexed and transmitted along with SL data on the same resource unit, only the form of the SL data channel excluding SA information can be transmitted from the resource pool for the SL data channel. In other words, REs (Resource Elements) that were used to transmit SA information on individual resource units within the SA resource pool can still be used to transmit SL data in the resource pool of the SL data channel. For example, the transmitting terminal can transmit by mapping the PSSCH to a succession of PRBs.

[0140] (3) The discovery channel may be a resource pool for a transmitting terminal to transmit information such as its ID. Through this, the transmitting terminal can enable adjacent terminals to discover it.

[0141] Even if the content of the SL signal described above is the same, different resource pools may be used depending on the transmission and reception attributes of the SL signal. For example, even if the same SL data channel or discovery message is used, it may be divided into different resource pools depending on the method of determining the transmission timing of the SL signal (e.g., whether it is transmitted at the time of reception of the synchronization reference signal or whether it is transmitted by applying a certain timing advance at the time of reception), the method of resource allocation (e.g., whether the base station assigns the transmission resource of an individual signal to the individual transmission terminal or whether the individual transmission terminal selects the individual signal transmission resource itself from within the resource pool), the signal format (e.g., the number of symbols occupied by each SL signal in one subframe, or the number of subframes used for the transmission of one SL signal), the signal strength from the base station, the transmission power strength of the SL terminal, etc.

[0142] FIG. 11 illustrates an example of a BWP according to an embodiment of the present disclosure. The embodiment of FIG. 11 may be combined with various embodiments of the present disclosure. In the embodiment of FIG. 11, it is assumed that there are three BWPs.

[0143] Referring to FIG. 11, the common resource block (CRB) may be a numbered carrier resource block extending from one end of the carrier band to the other. And, the PRB may be a numbered resource block within each BWP. Point A may indicate a common reference point for the resource block grid.

[0144] A BWP can be configured by point A, an offset from point A (NstartBWP), and a bandwidth (NsizeBWP). For example, point A may be an external reference point of the PRB of a carrier where the subcarrier 0 of all numerologies (e.g., all numerologies supported by the network on that carrier) is aligned. For example, the offset may be the PRB interval between the lowest subcarrier in a given numerology and point A. For example, the bandwidth may be the number of PRBs in a given numerology.

[0145] SLSS (Sidelink Synchronization Signal) is a sidelink-specific sequence and may include PSSS (Primary Sidelink Synchronization Signal) and SSSS (Secondary Sidelink Synchronization Signal). The PSSS may be referred to as S-PSS (Sidelink Primary Synchronization Signal), and the SSSS may be referred to as S-SSS (Sidelink Secondary Synchronization Signal). For example, length-127 M-sequences may be used for S-PSS, and length-127 Gold sequences may be used for S-SSS. For example, a terminal may use S-PSS to detect the initial signal and obtain synchronization. For example, a terminal may use S-PSS and S-SSS to obtain detailed synchronization and detect the synchronization signal ID.

[0146] The PSBCH (Physical Sidelink Broadcast Channel) may be a (broadcast) channel through which basic (system) information that the terminal must know first is transmitted before transmitting or receiving SL signals. For example, the basic information may include information related to SLSS, Duplex Mode (DM), TDD UL / DL (Time Division Duplex Uplink / Downlink) configuration, information related to resource pools, types of applications related to SLSS, subframe offsets, broadcast information, etc. For example, to evaluate PSBCH performance, in NR V2X, the payload size of the PSBCH may be 56 bits, including a 24-bit CRC (Cyclic Redundancy Check).

[0147] S-PSS, S-SSS, and PSBCH may be included in a block format that supports periodic transmission (e.g., SL SS (Synchronization Signal) / PSBCH block, hereinafter S-SSB (Sidelink-Synchronization Signal Block)). The S-SSB may have the same numerology (i.e., SCS and CP lengths) as the PSCCH (Physical Sidelink Control Channel) / PSSCH (Physical Sidelink Shared Channel) within the carrier, and the transmission bandwidth may be within a (pre-)set SL BWP (Sidelink BWP). For example, the bandwidth of the S-SSB may be 11 RB (Resource Block). For example, the PSBCH may span 11 RB. Additionally, the frequency position of the S-SSB may be (pre-)set. Therefore, the terminal does not need to perform hypothesis detection at the frequency to discover the S-SSB in the carrier.

[0148] FIG. 12 illustrates a procedure in which a terminal performs V2X or SL communication according to a resource allocation mode, according to one embodiment of the present disclosure. The embodiment of FIG. 12 may be combined with various embodiments of the present disclosure.

[0149] Referring to FIG. 12(a), in resource allocation mode 1, the base station may schedule SL resources to be used by the terminal for SL transmission. For example, in step S1200, the base station may transmit information related to SL resources and / or information related to UL resources to the first terminal. For example, the UL resources may include PUCCH resources and / or PUSCH resources. For example, the UL resources may be resources for reporting SL HARQ feedback to the base station.

[0150] For example, the first terminal may receive information related to a dynamic grant (DG) resource and / or information related to a configured grant (CG) resource from the base station. For example, the CG resource may include a CG type 1 resource or a CG type 2 resource. In this specification, the DG resource may be a resource that the base station sets / assigns to the first terminal via downlink control information (DCI). In this specification, the CG resource may be a (periodic) resource that the base station sets / assigns to the first terminal via DCI and / or RRC messages. For example, in the case of a CG type 1 resource, the base station may transmit an RRC message containing information related to the CG resource to the first terminal. For example, in the case of a CG type 2 resource, the base station may transmit an RRC message containing information related to the CG resource to the first terminal, and the base station may transmit DCI related to the activation or release of the CG resource to the first terminal.

[0151] In step S1210, the first terminal may transmit a PSCCH (e.g., Sidelink Control Information or 1st-stage SCI) to the second terminal based on the resource scheduling. In step S1220, the first terminal may transmit a PSSCH (e.g., 2nd-stage SCI, MAC PDU, data, etc.) associated with the PSCCH to the second terminal. In step S1230, the first terminal may receive a PSFCH associated with the PSCCH / PSSCH from the second terminal. For example, HARQ feedback information (e.g., NACK information or ACK information) may be received from the second terminal via the PSFCH. In step S1240, the first terminal may transmit / report the HARQ feedback information to the base station via a PUCCH or PUSCH. For example, the HARQ feedback information reported to the base station may be information generated by the first terminal based on HARQ feedback information received from the second terminal. For example, the HARQ feedback information reported to the base station may be information generated by the first terminal based on a pre-set rule. For example, the DCI may be a DCI for scheduling SL.

[0152] Referring to FIG. 12(b), in resource allocation mode 2, the terminal can determine an SL transmission resource within an SL resource set by the base station / network or a preset SL resource. For example, the set SL resource or the preset SL resource may be a resource pool. For example, the terminal may autonomously select or schedule a resource for SL transmission. For example, the terminal may perform SL communication by selecting a resource itself within the set resource pool. For example, the terminal may select a resource itself within a selection window by performing a sensing and resource (re)selection procedure. For example, the sensing may be performed on a subchannel basis. For example, in step S1210, the first terminal, having selected a resource itself within the resource pool, may use the resource to transmit PSCCH (e.g., SCI (Sidelink Control Information) or 1st-stage SCI) to the second terminal. In step S1220, the first terminal can transmit PSSCH (e.g., 2nd-stage SCI, MAC PDU, data, etc.) associated with the PSCCH to the second terminal. In step S1230, the first terminal can receive PSFCH associated with the PSCCH / PSSCH from the second terminal.

[0153] Referring to FIG. 12 (a) or (b), for example, the first terminal may transmit an SCI to the second terminal over the PSCCH. Or, for example, the first terminal may transmit two consecutive SCIs (e.g., 2-stage SCIs) to the second terminal over the PSCCH and / or PSSCH. In this case, the second terminal may decode the two consecutive SCIs (e.g., 2-stage SCIs) to receive the PSSCH from the first terminal. In this specification, an SCI transmitted over the PSCCH may be referred to as the 1st SCI, the 1st SCI, the 1st-stage SCI, or the 1st-stage SCI format, and an SCI transmitted over the PSSCH may be referred to as the 2nd SCI, the 2nd SCI, the 2nd-stage SCI, or the 2nd-stage SCI format.

[0154] Referring to FIG. 12 (a) or (b), in step S1230, the first terminal can receive PSFCH. For example, the first terminal and the second terminal can determine a PSFCH resource, and the second terminal can use the PSFCH resource to transmit HARQ feedback to the first terminal.

[0155] Referring to FIG. 12(a), in step S1240, the first terminal can transmit SL HARQ feedback to the base station via PUCCH and / or PUSCH.

[0156] CSI Measurement / Reporting

[0157] In NR (New Radio) systems, CSI-RS (channel state information-reference signal) is used for time and / or frequency tracking, CSI computation, L1 (layer 1)-RSRP (reference signal received power) computation, and mobility. Here, CSI computation is related to CSI acquisition, and L1-RSRP computation is related to beam management (BM). CSI (channel state information) is a general term for information that can indicate the quality of the radio channel (also called a link) formed between a terminal and an antenna port.

[0158] A base station can transmit CSI-RS to a terminal to determine the characteristics of the downlink channel, and can receive feedback from the terminal regarding channel measurement results based on CSI-RS.

[0159] CSI-RS may be configured for one or more terminals. Different CSI-RS configurations may be provided for each terminal, or the same CSI-RS configuration may be provided for multiple terminals. CSI-RS may support up to 32 antenna ports. CSI-RS corresponding to N (N is 1 or more) antenna ports may be mapped to N RE positions within a time-frequency unit corresponding to one slot and one RB. If N is 2 or more, N-port CSI-RS may be multiplexed in CDM, FDM, and / or TDM modes. A CDM group may include two antenna ports (CDM2) distinguished by code resources on the same two adjacent subcarriers, four antenna ports (CDM4) distinguished by code resources on the same two adjacent subcarriers and the same two adjacent OFDM slots, or eight antenna ports (CDM8) distinguished by code resources on the same two adjacent subcarriers and the same four adjacent OFDM slots. If multiple CDM groups exist, the CDM groups may not be mapped to adjacent subcarriers and / or adjacent OFDM symbols. CSI-RS antenna ports can be indexed in the order of CDM groups, frequency domain, and time domain. CSI-RS can be mapped to REs other than the REs to which CORESET, DMRS, and SSB are mapped.

[0160] In the frequency domain, CSI-RS can be configured for the entire bandwidth, a portion of the bandwidth (BWP), or a portion of the bandwidth. CSI-RS may be transmitted at each RB within the configured bandwidth (i.e., density=1), or at every second RB (e.g., the even or odd RB) (i.e., density=1 / 2). When CSI-RS is used as a Tracking Reference Signal (TRS), a single-port CSI-RS may be mapped onto three subcarriers in each resource block (i.e., density=3).

[0161] One or more CSI-RS resource sets may be configured for the terminal in the time domain. Each CSI-RS resource set may include one or more CSI-RS settings.

[0162] Each CSI-RS resource set can be configured to be periodic, semipersistent, or non-periodic. For periodic CSI-RS resource sets, the period can be configured to a number of slots ranging from 4 to 640. Additionally, a starting offset value for periodic CSI-RS resource sets can be configured. For semipersistent CSI-RS resource sets, an offset and period for CSI-RS resource set candidates can be configured. Here, actual CSI-RS transmission can be activated / deactivated based on a MAC Control Element (CE). When a CSI-RS resource set is activated, CSI-RS transmission may be performed according to the configured offset and period until it is deactivated. When a CSI-RS resource set is deactivated, CSI-RS transmission may not be performed until it is explicitly reactivated. For non-periodic CSI-RS transmission, information regarding each CSI-RS resource set may be explicitly provided by the DCI.

[0163] A CSI-IM resource may be configured for Interference Measurement (IM) of a terminal. A CSI-IM resource may contain four REs within one slot and one resource block. The four REs may correspond to two consecutive OFDM symbols and two consecutive subcarriers, or to one OFDM symbol and four consecutive subcarriers. In the frequency domain, the location of the CSI-IM REs may be determined by the CSI-IM configuration. In the time domain, the CSI-IM resource set may be configured periodic, semi-persistent, or non-periodically, similar to the CSI resource set. Generally, transmission may not be performed in the corresponding cell but may be performed in a neighboring cell. As such, the CSI-IM resource may be configured as Zero Power (ZP)-CSI-RS for the terminal.

[0164] ZP-CSI-RS can be configured to be distinct from Non-Zero Power (NZP)-CSI-RS. When a PDSCH is scheduled on a resource containing a CSI-RS RE, the first terminal may assume that rate matching considering the CSI-RS RE is applied to the PDSCH, so that the PDSCH is not mapped to the CSI-RS RE. Here, the CSI-RS may be configured for the first terminal or for the second terminal. In this case, the CSI-RS for the first terminal may be configured as NZP-CSI-RS for the first terminal, and an NZP-CSI-RS resource set may be configured for the first terminal. Meanwhile, the CSI-RS for the second terminal may be configured as ZP-CSI-RS for the first terminal, and a ZP-CSI-RS resource set may be configured for the first terminal. The NZP-CSI-RS resource set can be used for the CSI report configuration of the terminal. The NZP-CSI-RS resource set may also be associated with a CSI-RS or an SSB. Additionally, multiple periodic NZP-CSI-RS resource sets can be configured as TRS resource sets.

[0165] CSI-related operations can be summarized as follows.

[0166] To perform one of the uses of CSI-RS, a terminal (e.g., user equipment, UE) receives configuration information related to CSI from a base station (e.g., general Node B, gNB) via radio resource control (RRC) signaling. The configuration information related to CSI may include at least one of information related to CSI-IM (interference management) resources, information related to CSI measurement configuration, information related to CSI resource configuration, information related to CSI-RS resources, or information related to CSI report configuration.

[0167] - CSI-IM resource-related information may include CSI-IM resource information, CSI-IM resource set information, etc. A CSI-IM resource set is identified by a CSI-IM resource set ID (identifier), and one resource set includes at least one CSI-IM resource. Each CSI-IM resource is identified by a CSI-IM resource ID.

[0168] - Information related to CSI resource configuration can be expressed as CSI-ResourceConfig IE. Information related to CSI resource configuration defines a group that includes at least one of an NZP (non-zero power) CSI-RS resource set, a CSI-IM resource set, or a CSI-SSB resource set. That is, the information related to CSI resource configuration includes a CSI-RS resource set list, and the CSI-RS resource set list may include at least one of an NZP CSI-RS resource set list, a CSI-IM resource set list, or a CSI-SSB resource set list. A CSI-RS resource set is identified by a CSI-RS resource set ID, and one resource set includes at least one CSI-RS resource. Each CSI-RS resource is identified by a CSI-RS resource ID. Through the NZP CSI-RS resource set IE, parameters indicating the use of CSI-RS (e.g., a BM-related 'repetition' parameter, a tracking-related 'trs-Info' parameter) can be set for each NZP CSI-RS resource set. Also, the repetition parameter corresponding to the higher layer parameter corresponds to 'CSI-RS-ResourceRep' of the L1 parameter.

[0169] - Information related to CSI report configuration includes a reportConfigType parameter representing time domain behavior and a reportQuantity parameter representing the CSI-related quantity to be reported. The time domain behavior may be periodic, aperiodic, or semi-persistent. Information related to CSI report configuration may be expressed as CSI-ReportConfig IE.

[0170] The terminal measures the CSI based on configuration information related to the above CSI. The terminal's CSI measurement may include a process of receiving CSI-RS on a CSI-RS RE specified based on the higher layer parameter CSI-RS-ResourceMapping, and computing the CSI through the received CSI-RS. By CSI-RS-ResourceMapping, the RE (resource element) mapping of the CSI-RS resource in the time and frequency domains is set for the terminal. In CSI-RS-ResourceMapping, density (D) represents the density of the CSI-RS resource measured in the RE / port / PRB (physical resource block), and nrofPorts represents the number of antenna ports.

[0171] The terminal reports the measured CSI to the base station. If the quantity of CSI-ReportConfig is set to 'none (or No report)', the terminal may omit the CSI report. However, even if the quantity is set to 'none (or No report)', the terminal may still report to the base station. The case where the quantity is set to 'none' is when the aperiodic TRS is triggered or when repetition is enabled. Here, the terminal may omit the report only when repetition is set to 'ON'.

[0172] As time domain behaviors for CSI measurement and reporting, aperiodic / semi-persistent / periodic CM (channel measurement) and IM (interference measurement) are supported. A 4-port NZP CSI-RS RE pattern is used for CSI-IM configuration.

[0173] NR's CSI-IM-based Interference Measurement Resource (IMR) has a design similar to LTE's CSI-IM and is configured independently of ZP CSI-RS resources for PDSCH rate matching. In addition, in the NZP CSI-RS-based IMR, each port emulates an interference layer with (desired channel and) precoded NZP CSI-RS. This is for intra-cell interference measurement in the multi-user case and primarily targets MU interference.

[0174] The base station transmits precoded NZP CSI-RS to the terminal on each port of the configured NZP CSI-RS-based IMR.

[0175] The terminal assumes a channel / interference layer for each port in the resource set and measures interference.

[0176] For a channel, if there is no PMI and RI feedback, multiple resources are set, and the base station or network indicates a subset of NZP CSI-RS resources for channel / interference measurement via DCI.

[0177] Each CSI resource setting 'CSI-ResourceConfig' contains a configuration for S≥1 CSI resource sets (given by the higher layer parameter csi-RS-ResourceSetList). The CSI resource setting corresponds to the CSI-RS-resourcesetlist, where S represents the number of configured CSI-RS resource sets. Here, the list of S≥1 CSI resource sets includes either or both of the NZP CSI-RS resource set(s) and the SS / PBCH block (SSB) set(s) used for L1-RSRP computation, or includes CSI-IM resource set(s).

[0178] Each CSI resource setting is located in a DL BWP (bandwidth part) identified by a higher layer parameter BWP-id. Additionally, all CSI resource settings linked to a CSI reporting setting have the same DL BWP.

[0179] Within the CSI resource setting included in CSI-ResourceConfig IE, the time domain behavior of the CSI-RS resource is dictated by the higher layer parameter resourceType and can be set to aperiodic, periodic, or semi-persistent. For periodic and semi-persistent CSI resource settings, the number of configured CSI-RS resource sets (S) is limited to '1'. For periodic and semi-persistent CSI resource settings, the configured periodicity and slot offset are given from the numerology of the associated DL BWP, as given by the BWP-id.

[0180] When a UE is configured with multiple CSI-ResourceConfigs containing the same NZP CSI-RS resource ID, the same time domain behavior is configured for the multiple CSI-ResourceConfigs.

[0181] When a UE is configured with multiple CSI-ResourceConfigs containing the same CSI-IM resource ID, the same time domain behavior is configured for the multiple CSI-ResourceConfigs.

[0182] Next, one or more CSI resource settings for channel measurement (CM) and interference measurement (IM) are established through higher layer signaling.

[0183] - CSI-IM resource for interference measurement.

[0184] - NZP CSI-RS resources for interference measurement.

[0185] - NZP CSI-RS resources for channel measurement.

[0186] That is, the CMR (channel measurement resource) may be an NZP CSI-RS for CSI acquisition, and the IMR (Interference measurement resource) may be an NZP CSI-RS for CSI-IM and IM.

[0187] Here, CSI-IM (or ZP CSI-RS for IM) is primarily used for inter-cell interference measurements.

[0188] Also, the NZP CSI-RS for IM is mainly used for intra-cell interference measurement from multi-users.

[0189] A UE can assume that the CSI-RS resource(s) for channel measurement set for one CSI reporting and the CSI-IM / NZP CSI-RS resource(s) for interference measurement (when NZP CSI-RS resource(s) are used for interference measurement) have a QCL relationship with respect to 'QCL-TypeD' on a resource-by-resource basis.

[0190] As examined, resource setting can refer to a resource set list.

[0191] For aperiodic CSI, each trigger state set using the higher layer parameter CSI-AperiodicTriggerState is associated with one or more CSI-ReportConfigs, and each CSI-ReportConfig is linked to a periodic, semi-persistent, or aperiodic resource setting.

[0192] One reporting setting can be linked to up to three resource settings.

[0193] - When a resource setting is set, that resource setting (given by the higher layer parameter resourcesForChannelMeasurement) is for channel measurement for L1-RSRP computation or channel and interference measurement for L1-SINR computation.

[0194] - When two resource settings are set, the first resource setting (given by the higher layer parameter resourcesForChannelMeasurement) is for channel measurement, and the second resource setting (given by csi-IM-ResourcesForInterference or nzp-CSI-RS-ResourcesForInterference) is for interference measurement performed on CSI-IM or NZP CSI-RS.

[0195] - When three resource settings are set, the first resource setting (given by resourcesForChannelMeasurement) is for channel measurement, the second resource setting (given by csi-IM-ResourcesForInterference) is for CSI-IM based interference measurement, and the third resource setting (given by nzp-CSI-RS-ResourcesForInterference) is for NZP CSI-RS based interference measurement.

[0196] For semi-persistent or periodic CSI, each CSI-ReportConfig is linked to periodic or semi-persistent resource setting(s).

[0197] - When a resource setting (given by resourcesForChannelMeasurement) is configured, said resource setting is for channel measurement for L1-RSRP computation or channel and interference measurement for L1-SINR computation.

[0198] - When two resource settings are configured, the first resource setting (given by resourcesForChannelMeasurement) is for channel measurement, and the second resource setting (given by higher layer parameter csi-IM-ResourcesForInterference or nzp-CSI-RS-ResourcesForInterference) is used for interference measurement performed on CSI-IM or NZP CSI-RS.

[0199] When interference measurements are performed on CSI-IM, each CSI-RS resource for channel measurement is associated with a CSI-IM resource by resource in the order of CSI-RS resources and CSI-IM resources within the corresponding resource set. The number of CSI-RS resources for channel measurement is equal to the number of CSI-IM resources.

[0200] And, when interference measurement is performed in NZP CSI-RS, the UE does not expect to be set to one or more NZP CSI-RS resources in the associated resource set within the resource setting for channel measurement.

[0201] A terminal with the Higher layer parameter nzp-CSI-RS-ResourcesForInterference configured does not expect more than 18 NZP CSI-RS ports to be configured within a single NZP CSI-RS resource set.

[0202] For CSI measurement(s) other than L1-SINR, the terminal assumes the following:

[0203] - Each NZP CSI-RS port configured for interference measurement corresponds to the interference transport layer.

[0204] - All interference transmission layers of the NZP CSI-RS port for interference measurement consider the associated EPRE (energy per resource element) ratio.

[0205] - Other interference signals on the RE(s) of the NZP CSI-RS resource for channel measurement, the NZP CSI-RS resource for interference measurement, or the CSI-IM resource for interference measurement.

[0206] The terminal can perform measurements of channel characteristics based on CSI-RS and feed back a CSI report to the base station as a result. To this end, a CSI report configuration may be provided to the terminal. Each CSI report configuration may include settings for feedback type, measurement resource, report type, etc.

[0207] Feedback types may include Channel Quality Indicator (CQI), Precoding Matrix Indicator (PMI), CSI-RS Resource Indicator (CRI), SSBRI (SSB Resource block Indicator), Layer Indicator (LI), Rank Indicator (RI), Layer 1-Reference Signal Received Strength (RSRP), etc.

[0208] The measurement resource may include settings for downlink signals and / or downlink resources for which the terminal will perform measurements to determine feedback information. The measurement resource may be set as a set of ZP and / or NZP CSI-RS resources associated with a CSI reporting setting. The NZP CSI-RS resource set may include a CSI-RS set or an SSB set. For example, L1-RSRP may be measured against a CSI-RS set or against an SSB set.

[0209] The report type may include settings for the timing at which the terminal performs the report and the uplink channel, etc. The reporting timing may be set to periodic, semi-persistent, or non-periodic. Periodic CSI reports may be transmitted over the PUCCH. Semi-persistent CSI reports may be transmitted over the PUCCH or PUSCH based on MAC CEs indicating activation / deactivation. Non-periodic CSI reports may be indicated by DCI signaling. For example, the CSI request field of an uplink grant may indicate one of various report trigger sizes. Non-periodic CSI reports may be transmitted over the PUSCH.

[0210] CSI can be defined as two types. Type 1 CSI may be associated with cases where a single user is scheduled, and Type 2 CSI may be associated with cases where multiple users are scheduled simultaneously on the same resource. Type 1 CSI may include single-panel CSI and multi-panel CSI, each of which may correspond to a different codebook. The precoding matrix in the codebook may be specified by a combination of w1 and w2. Long-term and wideband characteristics may correspond to w1, and short-term and subband characteristics may correspond to w2. While Type 1 CSI reports one precoding matrix selected by the terminal, Type 2 CSI may report information on the magnitude and phase of up to four beams.

[0211] For CSI reporting, the time and frequency resources available to the UE are controlled by the base station.

[0212] Channel state information (CSI) may include at least one of a channel quality indicator (CQI), a precoding matrix indicator (PMI), a CSI-RS resource indicator (CRI), an SS / PBCH block resource indicator (SSBRI), a layer indicator (LI), a rank indicator (RI), L1-RSRP, and / or L-SINR.

[0213] For CQI, PMI, CRI, SSBRI, LI, RI, and L1-RSRP, the terminal is configured by a higher layer with N≥1 CSI-ReportConfig reporting settings, M≥1 CSI-ResourceConfig resource settings, and a list of one or two trigger states (provided by CSI-AperiodicTriggerStateList and CSI-SemiPersistentOnPUSCH-TriggerStateList). In the CSI-AperiodicTriggerStateList, each trigger state includes an associated list of CSI-ReportConfigs indicating resource set IDs for the channel and optionally interference. In the CSI-SemiPersistentOnPUSCH-TriggerStateList, each trigger state includes one associated CSI-ReportConfig.

[0214] In addition, the time domain behavior of CSI reporting supports periodic, semi-persistent, and aperiodic.

[0215] i) Periodic CSI reporting is performed on short PUCCH and long PUCCH. The periodicity and slot offset of Periodic CSI reporting can be set to RRC, and refer to CSI-ReportConfig IE.

[0216] ii) SP (semi-periodic) CSI reporting is performed on short PUCCH, long PUCCH, or PUSCH.

[0217] In the case of SP CSI on Short / long PUCCH, periodicity and slot offset are set to RRC, and CSI reporting is activated / deactivated with separate MAC CE / DCI.

[0218] In the case of SP CSI on PUSCH, the periodicity of SP CSI reporting is set to RRC, but the slot offset is not set to RRC, and SP CSI reporting is activated / deactivated by DCI(format 0_1). For SP CSI reporting on PUSCH, a separate RNTI (SP-CSI C-RNTI) is used.

[0219] The initial CSI reporting timing follows the PUSCH time domain allocation value specified by DCI, and subsequent CSI reporting timing follows the period set by RRC.

[0220] DCI format 0_1 ​​includes a CSI request field and can activate / deactivate a specific configured SP-CSI trigger state. SP CSI reporting has the same or similar activation / deactivation mechanism as the data transmission mechanism on SPS PUSCH.

[0221] iii) aperiodic CSI reporting is performed on PUSCH and triggered by DCI. In this case, information related to the trigger of aperiodic CSI reporting can be transmitted / instructed / set via MAC-CE.

[0222] In the case of an AP CSI with AP CSI-RS, the AP CSI-RS timing is set by the RRC, and the timing for AP CSI reporting is dynamically controlled by the DCI.

[0223] NR does not apply the method of splitting CSIs across multiple reporting instances (e.g., transmitting in the order of RI, WB PMI / CQI, SB PMI / CQI) that was applied to PUCCH-based CSI reporting in LTE. Instead, NR restricts the setting of specific CSI reports in short / long PUCCHs, and CSI omission rules are defined. Regarding AP CSI reporting timing, the PUSCH symbol / slot location is dynamically determined by the DCI, and candidate slot offsets are set by the RRC. For CSI reporting, the slot offset (Y) is set per reporting setting. For UL-SCH, the slot offset K2 is set separately.

[0224] Two CSI latency classes (low latency class and high latency class) are defined in terms of CSI computation complexity. Low-latency CSI refers to WB CSIs that include up to 4-port Type-I codebooks or up to 4-port non-PMI feedback CSIs. High-latency CSI refers to any CSI other than low-latency CSIs. For a normal terminal, (Z, Z') is defined in the unit of OFDM symbols. Here, Z represents the minimum CSI processing time from receiving an Aperiodic CSI-triggering DCI to performing a CSI report. Additionally, Z' represents the minimum CSI processing time from receiving a CSI-RS for channel / interference to performing a CSI report.

[0225] Additionally, the terminal reports the number of CSIs that can be calculated simultaneously.

[0226] The activation / deactivation of a semi-persistent CSI-RS / CSI-IM resource set is directed by the network via a specific MAC CE. The configured semi-persistent CSI-RS / CSI-IM resource set is initially deactivated at the time of configuration and after handover. The MAC entity receiving the corresponding MAC CE instructs the lower layer (e.g., PHY) with the information related to that MAC CE.

[0227] Aperiodic CSI Trigger State subselection is directed from the network via a specific MAC CE, and an Aperiodic CSI Trigger State can be selected from the configured AP CSI trigger states of the serving cell. The MAC entity receiving the corresponding MAC CE instructs the lower layer (e.g., PHY) with information related to that MAC CE.

[0228] Integrated Sensing and Communication (ISAC)

[0229] Wireless sensing is a technology that utilizes radio frequencies to determine the instantaneous linear velocity, angle, and distance (range) of an object, thereby obtaining information about the characteristics of the environment and / or objects within that environment. Since radio frequency sensing capabilities do not require connecting to objects via devices within a network, they can provide services for determining object locations without the need for devices. The ability to obtain range, velocity, and angle information from radio frequency signals can provide a wide range of new functions, such as various object detection, object recognition (e.g., vehicles, humans, animals, UAVs), and high-precision localization, tracking, and activity recognition. Wireless sensing services can provide information to various industries (e.g., unmanned aerial vehicles, smart homes, V2X, factories, railways, public safety, etc.) that enable applications such as intruder detection, assisted vehicle steering and navigation, trajectory tracking, collision avoidance, traffic management, and health and traffic management. In some cases, wireless sensing may utilize non-3GPP type sensors (e.g., radar, cameras) to further support 3GPP-based sensing. For example, the operation of a wireless sensing service, that is, the sensing operation, may depend on the transmission, reflection, and scattering processing of wireless sensing signals. Therefore, wireless sensing can provide an opportunity to enhance existing communication systems from communication networks into wireless communication and sensing networks.

[0230] FIGS. 13 and FIGS. 14 illustrate examples of sensing operations according to an embodiment of the present disclosure. Specifically, FIG. 13 illustrates an example of sensing using a sensing receiver and a sensing transmitter located at the same position (e.g., monostatic sensing), and FIG. 14 illustrates an example of sensing using a separated sensing receiver and a sensing transmitter (e.g., bistatic sensing).

[0231] For example, in a wireless communication system based on a 6G network of the present specification, referring to FIG. 13, the sensing transmitter and the sensing receiver may be configured to be included in a single base station (i.e., the same base station) or a single terminal (i.e., the same terminal). Alternatively, referring to FIG. 14, the sensing transmitter and the sensing receiver may be configured to be included in different base stations, in different terminals, or in a terminal and a base station, respectively.

[0232] In this regard, based on whether the sensing transmitter and the sensing receiver are each included in a base station or a terminal, the following six types of sensing modes can be defined.

[0233] - Mode 1: A mode in which the sensing transmitter and sensing receiver are included in a single base station (e.g., base station-based sensing mode in monostatic mode)

[0234] - Second mode: A mode in which the sensing transmitter is included in the first base station and the sensing receiver is included in a second base station different from the first base station (e.g., base station-based sensing mode in bistatic mode)

[0235] - 3rd Mode: A mode in which the sensing transmitter is included in the base station and the sensing receiver is included in the terminal (e.g., base station-terminal sensing mode)

[0236] - 4th Mode: A mode in which the sensing transmitter is included in the terminal and the sensing receiver is included in the base station (e.g., terminal-base station sensing mode)

[0237] - 5th Mode: A mode in which the sensing transmitter and the sensing receiver are contained in a single terminal (e.g., terminal-based sensing mode in monostatic mode)

[0238] - 6th mode: A mode in which the sensing transmitter is included in the first terminal and the sensing receiver is included in a second terminal different from the first terminal (e.g., terminal-based sensing mode in bistatic mode)

[0239] In a wireless communication system based on a 6G network of the present specification, one or more of the six types of sensing modes described above may be utilized independently or in combination.

[0240] In relation to the sensing operation in FIGS. 13 and 14, a sensing transmitter may transmit a sensing signal for sensing one or more objects (and / or the environment surrounding the objects). For example, the sensing signal may correspond to a radio (frequency) signal defined to be transmittable by a base station / terminal in a wireless communication system based on a 6G network of the present specification. A sensing receiver may receive a signal that is scattered / reflected by one or more objects (and / or the environment surrounding the objects) from the sensing signal transmitted from the sensing transmitter. In the sensing receiver, sensing data may be derived from the scattered / reflected signal, and sensing results may be generated / obtained through processing of the sensing data. Here, the sensing result may include characteristic information (e.g., location, distance, speed, angle, etc.) about one or more objects (and / or the environment surrounding the objects). The sensing result thus generated / acquired may be utilized for wireless sensing services (e.g., detection, tracking, etc. of objects and / or environments) provided by a wireless communication system based on a 6G network of the present specification, or may be provided / disclosed to a trusted third party.

[0241] Additionally, the sensing operation in FIGS. 13 and 14 is described using a wireless communication system based on a 6G network as a representative example, but it can be extended and applied to cases where terminals / base stations / signals based on previous generations (e.g., 4G, 5G, etc.) networks are utilized.

[0242] Additionally, with respect to the wireless sensing described in this specification, in a wireless communication system based on a 6G network of this specification, time / frequency resources for sensing operations and time / frequency resources for general communication (e.g., UL / DL / sidelink-based communication, etc.) may be scheduled / configured separately.

[0243] FIG. 15 illustrates a time / frequency resource for a sensing operation according to one embodiment of the present specification. The embodiment of FIG. 15 may be combined with various embodiments of the present disclosure.

[0244] Referring to FIG. 15, the time / frequency resources (hereinafter, sensing resources) for the aforementioned sensing operation (e.g., sensing operation based on FIG. 13 and FIG. 14) can be set / assigned separately from the time / frequency resources (hereinafter, communication resources) for general communication.

[0245] For example, as illustrated in FIG. 15, sensing resources may be configured / assigned in units of symbols in the time domain and / or in units of resource blocks in the frequency domain. Resources other than those configured / assigned to the sensing resources may be utilized as resources for general communication. That is, sensing resources and communication resources may be configured / assigned based on time-division multiplexing (TDM) and / or frequency-division multiplexing (FDM) methods in terms of base station / terminal operation. Additionally or alternatively, unlike those illustrated in FIG. 13 and FIG. 14, sensing resources may also be configured / assigned based on other units in the time domain (e.g., slot, frame, absolute time (ms, us), etc.) and / or other units in the frequency domain (e.g., subcarrier, carrier, absolute frequency (MHz, GHz), etc.).

[0246] Additionally or alternatively, in relation to the setup / allocation / scheduling of resources for general communication described herein, it may be necessary to consider the relationship between said resources and the aforementioned sensing resources. For example, when setting / allocating resources for general communication according to the embodiments of the present disclosure, said resources may be set / allocated to rate-match or puncturing resource areas corresponding to the sensing resources. For example, when scheduling resources for general communication according to the embodiments of the present disclosure, said resources may be scheduled so as not to overlap with resource areas corresponding to the sensing resources. If resources for general communication and resource areas corresponding to the sensing resources are set / allocated / scheduled to overlap according to the embodiments of the present disclosure, either one or both operations may be dropped, skipped, or postponed based on priority, predefined rules, etc. That is, in the embodiments of this specification, resources related to general communication (e.g., resources for signals / channels related to UL / DL / Sidelink-based data / control, etc.) may be configured / assigned / scheduled so as not to overlap with the aforementioned sensing resources.

[0247] Additionally, various channel modeling methods may be applied in relation to the wireless sensing described herein. Channel modeling related to sensing may mean constructing a path for transmitting and receiving sensing signals and / or scattered / reflected signals by considering the object to be sensed and / or the environment to which the object belongs. Since channel modeling may be related to the performance / requirements of sensing in a wireless communication system, it may be an important matter for verifying the validity of the sensing function.

[0248] Channels related to sensing can be classified into channels between an object (e.g., target of interest) and a sensing transmitter / receiver, and channels between the environment to which the object belongs and a sensing transmitter / receiver. In this regard, channel modeling related to sensing can be classified based on the sensing mode (e.g., the six types of modes mentioned above), whether it is an object or an environment, and / or sensing scenarios. For example, channel modeling for a target in a base station / terminal-based monostatic sensing mode, channel modeling for a target in a base station / terminal-based bistatic sensing mode, channel modeling for an environment in a base station / terminal-based monostatic sensing mode, and channel modeling for an environment in a base station / terminal-based bistatic sensing mode can be optimized and configured differently. For example, when various sensing scenarios are classified, they can be divided into channel modeling for detection, location, and tracking scenarios, channel modeling for motion recognition, and channel modeling for imaging / environment reconstruction scenarios. Additionally, channel modeling related to sensing may be based on statistical channel modeling techniques and / or deterministic channel modeling techniques. For example, modeling for sensing in a wireless communication system based on a 6G network of this specification may be based on stochastic geometry channel modeling techniques and / or hybrid with ray tracing channel modeling techniques. Here, the stochastic geometry channel model may be based on various statistical characteristics of the channel state. Furthermore, the hybrid channel model may be based on both ray tracing techniques and stochastic techniques.In the case of a hybrid approach, channels for objects requiring high accuracy and consistency (e.g., targets of interest) can be modeled using ray tracing techniques, while channels for the environment can be modeled using probabilistic techniques.

[0249] artificial intelligence

[0250] The introduction of AI into communications can streamline and enhance real-time data transmission. AI can determine how complex target tasks are performed using numerous analyses. In other words, AI can increase efficiency and reduce processing latency. Time-consuming tasks such as handover, network selection, and resource scheduling can be performed instantly using AI. AI can also play a significant role in M2M, machine-to-human, and human-to-machine communication. Furthermore, AI can enable rapid communication in Brain-Computer Interfaces (BCI). AI-based communication systems can be supported by metamaterials, intelligent structures, intelligent networks, intelligent devices, intelligent cognitive radios, self-sustaining wireless networks, and machine learning.

[0251] The following describes a functional framework for AI / ML operations.

[0252] Below, to provide a more specific explanation of AI (or AI / ML), terms may be defined as follows.

[0253] - Data collection: Data collected from network nodes, management entities, or terminals, serving as a basis for AI model training, data analysis, and inference.

[0254] - AI Model: A data-driven algorithm that applies AI technology to generate a set of outputs containing predictive information and / or decision parameters based on a set of inputs.

[0255] - AI / ML Training: An online or offline process of training an AI model by learning features and patterns that best represent data and acquire an AI / ML model trained for inference.

[0256] - AI / ML Inference: A process of making predictions or deriving decisions based on collected data and AI models using trained AI models.

[0257] Life Cycle Management (LCM) procedures for AI / ML models (i.e., model training, model deployment, model inference, model monitoring, model updating, etc.) can be classified into functionality-based LCM and model-based LCM. In functionality-based LCM, AI / ML models may not be identifiable within the network, and the network can direct the activation, deactivation, fallback, or switching of AI / ML functionality. In model-ID (identifier)-based LCM, AI / ML models can be identified within the network, and the network or terminal can activate, deactivate, select, or switch AI / ML models via the model ID.

[0258] Figure 16 illustrates a general functional architecture for an AI / ML model.

[0259] In particular, FIG. 16 illustrates a general functional architecture related to both Functionality-based LCM and Model-based LCM. Some functions or some data / information / command flows (i.e., arrows) illustrated in FIG. 16 may be omitted.

[0260] Referring to FIG. 16, a general functional framework may be configured to include a data collection function (10), a model training function (20), a management function (30), an inference function (40), and a model storage function (50).

[0261] The Data Collection function (10) is a function that provides input data to the Model Training function (20), Management function (30), and Inference function (40). The Data Collection function (10) performs data preparation based on raw data and can provide input data processed through data preparation. Examples of raw data may include received data / measurement data from terminals or other network entities, inference / output of AI / ML models, etc. The Data Collection function (10) may be performed by a single entity (e.g., terminal, network node, etc.) but may also be performed by multiple entities.

[0262] Here, training data (11) refers to data required as input for the AI / ML model training function (20). monitoring data (12) refers to data required as input for the management (30) of the AI / ML model or AI / ML function. inference data (13) refers to data required as input for the AI / ML inference function (30).

[0263] The Model Training function (20) is a function that performs AI / ML model training, validation, and testing, which can generate model performance metrics that can be used as part of the AI / ML model testing procedure. If necessary, the Model Training function (20) can perform data preparation (e.g., data pre-processing and cleaning, forming and transformation) based on the Training Data (11) delivered from the Data Collection function (10).

[0264] Trained / Updated Model (21): If there is a Model Storage function (50), it is used to transfer trained, validated, and tested AI / ML models to the Model Storage function (50) or to transfer updated versions of the models to the Model Storage function (50).

[0265] The Management function (30) is a function that supervises the operation of an AI / ML model or an AI / ML function. Additionally, the Management function (30) may make decisions to ensure appropriate inference operations based on data received from the Data Collection function (10) (i.e., Monitoring Data (12)) and / or data received from the Inference function (40) (i.e., Inference Output (41)).

[0266] Management Instruction (32) is information required as input to manage the Inference function (40). The relevant information may include the selection / (de)activation / switching of an AI / ML model or an AI / ML-based function, and may also include a fallback to a non-AI / ML operation (i.e., not relying on the inference process).

[0267] A Model Transfer / Delivery Request (33) can be used to request model(s) from Model Storage (50).

[0268] Performance Feedback / Retraining Request (31) refers to information required as input to Model Training function (20) (e.g., for the purpose of retraining or updating the model).

[0269] The inference function (40) is a function that provides output from the process of applying an AI / ML model or AI / ML function using data (i.e., inference data (13)) provided by the data collection (10) as input. Data preparation (e.g., data preprocessing and cleaning, formatting and transformation) may also be performed based on the inference data (13) delivered by the data collection (10). If necessary, the inference function (40) may also perform data preparation (e.g., data pre-processing and cleaning, forming and transformation) based on the inference data (13) provided by the data collection function (10).

[0270] Inference Output (41) is data used in the Management function (30) to monitor the performance of an AI / ML model or AI / ML function. Inference Output (41) may include the inference output of an AI / ML model generated by the Inference function (30), and the details of the inference output may vary depending on the use case.

[0271] The Model Storage function (50) is a function that stores a trained / updated model that can be used to perform the Inference function (40). The Model Storage function (50) exemplified in FIG. 15 can be used as a reference point (if any) applicable to protocol termination, model transmission / delivery, and related processes. Additionally, the Model Storage function (50) is an example and is not intended to restrict the storage location of the actual AI / ML model and may be omitted.

[0272] Model Transfer / Delivery (51) is used to transfer an AI / ML model to an inference function.

[0273] Cooperation levels can be defined as follows based on the capability of AI / ML functions among multiple nodes, and variations resulting from the combination of multiple levels or the separation of any one level are also possible.

[0274] Cat 0a) No collaboration framework: AI / ML algorithms are based on pure implementation and do not require changes to the wireless interface.

[0275] Cat 0b) This level corresponds to a framework that involves a wireless interface modified to fit efficient implementation-based AI / ML algorithms but without cooperation.

[0276] Cat 1) Inter-node support is involved to improve the AI / ML algorithms of each node. For example, this applies when a specific node receives support from other nodes (for training, adaptation, etc.) and vice versa. At this level, model exchange between network nodes is not required.

[0277] Cat 2) Collaborative AI / ML tasks can be performed among multiple nodes. This level requires the exchange of AI / ML model commands or network nodes.

[0278] AI / ML models can be classified into one-side models and two-side models depending on whether training and / or inference are performed on a single node or jointly / sequentially on multiple nodes.

[0279] A one-side model can refer to an AI / ML model where inference is performed entirely by a single node (e.g., a terminal or a network). Here, the training of the AI / ML model can also be performed entirely by a single node. The training and inference of the AI / ML model may be performed by the same node, or they may be performed by different nodes.

[0280] A two-side model can refer to an AI / ML model in which joint inference is performed across multiple nodes (e.g., terminals and networks). Joint inference means that inference is performed collaboratively across multiple nodes; for example, the first part of the inference may be performed by the first node, and the remainder by the second node. Two-side models can be classified into various types as follows, depending on the training method of the AI / ML model.

[0281] - First type: An AI / ML model can be trained on a single node. In this case, joint training can be performed. The trained model can then be distributed to other nodes / entities.

[0282] - Second type: Joint training of AI / ML models can be performed on multiple nodes / entities (e.g., networks and terminals). Joint training can mean that model generation (e.g., CSI generation) and model reconstruction (CSI compression by sub-use cases) are trained in the same loop for forward activation and backward gradient. In this type, joint training can include both simultaneous training (i.e., model generation training and model reconstruction training are performed simultaneously) and sequential training (i.e., model reconstruction training is performed after model generation training).

[0283] - Third Type: Separate training of AI / ML models can be performed at multiple nodes (e.g., networks and terminals). Separate training may mean that training starts sequentially at one node and continues at another node. In this case, if the first node performs the AI / ML model first and shares the training data with the second node, the second node can perform the AI / ML model using the shared training data. For example, training for the CSI generation part may be performed by the terminal, while CSI reconstruction may be performed by the network.

[0284] CSI-free NW-side monitoring for CSI compression

[0285] Figure 17 is a diagram illustrating an inference method during AI / ML-based CSI compression, and Figure 18 is a diagram illustrating a method for performing AI / ML-based CSI prediction.

[0286] Due to the advancement of computational processing technology and AI (artificial intelligence) / ML (machine learning) technologies as described above, the nodes and terminals constituting wireless communication networks are becoming more intelligent and sophisticated. In particular, due to the intelligence of the network, it is expected that various network decision parameter values ​​can be rapidly optimized, derived, and applied according to various network environment parameters (e.g., distribution / location of base stations, distribution / location / material of buildings / furniture, location / movement direction / speed of terminals, weather information, etc.). Here, these various network decision parameter values ​​may include the transmit / receive power of each base station, the transmit power of each terminal, the precoders and beams of base stations and terminals, time / frequency resource allocation for each terminal, and the duplex method of each base station.

[0287] In the following sections, we describe in detail a method for dynamically performing CSI reporting on various CSI-RS resources using a single AI / ML model when wireless communication services are supported based on AI / ML, in order to support advanced network / base station-based dynamic systems. In particular, the proposed method can be effectively applied even in situations where there are constraints on the terminal's memory and computation costs.

[0288] In the Rel-18 AI / ML study item, discussions were held regarding CSI compression based on AI / ML models to improve CSI accuracy and reduce feedback overhead in existing NR systems. As illustrated in Fig. 17, CSI compression can be operated / performed based on two-sided models. For example, a terminal can generate compressed channel information by performing inference based on a CSI generation submodel using channel information or channel state information measured based on CSI-RS as input data, and can quantize the compressed channel information and feed it back to the base station. The base station can restore CSI (e.g., channel information measured by the terminal based on CSI-RS) by performing inference based on a CSI reconstruction submodel using information obtained by de-quantizing feedback bits (e.g., quantized compressed channel information) received from the terminal as input data.

[0289] AI / ML models have a structural characteristic in which input and output dimensions are fixed. Therefore, multiple configuration-specific models may be required to perform CSI reporting for various CSI-RS resources, and such multiple configuration-specific models can cause storage issues for terminals, base stations, and networks. NR CSI reporting measurement / reporting can be performed based on CSI-RS(s) of configured N-ports ({1,2,4,8,16,24,32,48,64,128}, N_max=128) and M-subbands ({1,...,19}, M_max=19). In this case, a total of 10*19=190 configuration-specific models may be required to perform CSI reporting for all CSI-RS resources. Moreover, considering the massive MIMO (e.g., N_max=256, 512, etc.) and wide bandwidth environments of 6G, including Upper mid-band (e.g., 7-24GHz) and Sub-Thz (e.g., 90-300GHz), it is expected that even more models will be needed.

[0290] In the following, we consider a network-side performance monitoring method for AI / ML (Artificial Intelligence / Machine Learning)-based CSI compression and propose a monitoring method in which the UE does not need to report ground-truth (target) CSI to the NW.

[0291] In Rel-18, three major use cases regarding NR air interfaces via AI / ML were discussed as study items: CSI feedback enhancement, beam management, and positioning accuracy enhancement. Regarding CSI feedback enhancement, various methods were discussed to reduce overhead and improve accuracy in CSI reporting using AI / ML models. For the upcoming Rel-19, the study item description (SID) shown in Table 5 was agreed upon in RAN#102 as an additional study item.

[0292] CSI feedback enhancement[RAN1]:For CSI compression (two-sided model), further study ways to:-> Improve trade-off between performance and complexity / overhead- e.g., considering extending the spatial / frequency compression to spatial / temporal / frequency compression, cell / site specific models, CSI compression plus prediction (compared to Rel-18 non-AI / ML based approach), etc.-- Alleviate / resolve issues related to inter-vendor training collaboration.- while addressing other aspects requiring further study / conclusion as captured in the conclusions section of the TR 38.843.-> For CSI prediction (one-sided model), further study performance gain over Rel-18 non-AI / ML based approach and associated complexity, while addressing other aspects requiring further study / conclusion as captured in the conclusions section of the TR 38.843 (e.g., cell / site specific model could be considered to improve performance gain).

[0293] Referring to FIG. 17, CSI compression is operated based on a two-sided model (in terms of inference) in which AI / ML models are provided at the terminal and the base station, respectively, and operate as a pair. For convenience, the terminal-side model can be defined as an AI-encoder and the base station-side model as an AI-decoder. As illustrated in FIG. 17, the terminal applies channel information (e.g., raw channel matrix or precoder-type channel information vector (e.g., eigenvector)) measured / estimated (or, if necessary, pre-processing may be performed) as input to the AI-encoder and can generate / obtain output information from the AI-encoder. The terminal can quantize the output information and feed it back to the base station as AI-CSI. In this case, the base station can generate / acquire the output CSI of the AI-decoder by applying the de-quantized information of the feedbacked AI-CSI as input to the AI-decoder. Since the channel information to be actually sent is compressed and fed back through an AI / ML-based model through this process, the feedback overhead can be reduced, so this is called CSI compression.

[0294] Referring to FIG. 18, a UE-side model is illustrated that performs model inference in the sense that the CSI prediction is equipped with an AI / ML model only on the terminal-side. For example, the CSI prediction may be an AI / ML-based channel estimation method that applies a plurality of historical measurements as input to the AI / ML (or AL / ML model) and estimates at least one future CSI (or at least one time point) as the output of the AI / ML model.

[0295] As such, regarding CSI feedback enhancement, two main cases—CSI compression and CSI prediction—are discussed, and CSI compression will be explained in detail below.

[0296] Referring to FIG. 19 (a) and (b), channel measurement information (V, raw data) is input to an AI / ML encoder (ENC), and output information (z) or latent information (z) obtained from the AI / ML encoder (ENC) can be quantized and fed back to a network / base station. The network / base station can dequantize the quantized latent information (z) and input it to an AI / ML decoder (DEC), and obtain a reconstructed CSI (or output information; V_hat) from the AI / ML decoder (DENC).

[0297] Performance monitoring of CSI compression can be performed using a UE-side SGCS estimation model for intermediate Key Performance Indicator (KPI) estimation as shown in FIG. 19 (a), or an NW-side SGCS estimation model for intermediate KPI estimation as shown in FIG. 19 (b).

[0298] Specifically, to monitor the performance of CSI compression using a two-sided model through an intermediate KPI (e.g., Squared Generalized Cosine Similarity; SGCS), it is essentially necessary to compare the ground-truth value of the target CSI (e.g., V in Fig. 19) with the reconstructed CSI (e.g., V_hat in Fig. 19). However, since V exists on the UE-side while the reconstructed CSI (V_hat) exists on the network-side (NW-side), the most naive approach to consider is to send information about the target CSI (V) to the network-side (NW-side) or to inform the UE-side of the reconstructed CSI (V_hat). Both methods may require additional signaling overhead for monitoring the model's performance. In this regard, various methods to reduce signaling overhead have been proposed, as shown in Table 6.

[0299] Further study following monitoring options in Rel-19, including the necessity and feasibility,-> NW-side monitoring, considering overhead, latency, complexity, monitoring accuracy, UE capability- Based on the target CSI reported by the UE via legacy eT2 codebook or eT2-like high-resolution codebook (Case 1)- SRS-based monitoring-> UE-side monitoring, considering overhead, latency, complexity, monitoring accuracy, UE capability- Based on the output of the CSI reconstruction model at the UE (Case 2-1)-- * Note: CSI reconstruction model at the UE-side can be the same as the actual CSI reconstruction model used at the NW-side, a reference model provided by NW, or a proxy model developed by the UE side.- Via direct estimation of intermediate KPI (e.g., SGCS) without reconstructing a target CSI (Case 2-2)- Via estimation of monitoring output other than intermediate KPI without reconstructing a target CSI- Based on precoded RS (e.g., CSI-RS, DMRS) transmitted from NW based on the output of the CSI reconstruction model- Based on the output of the CSI reconstruction model indicated by the NW via legacy eT2 codebook or eT2-like high-resolution codebookRegarding monitoring metrics:- Monitoring accuracy also includes generalization considerations, if applicable.- Complexity also includes LCM complexity, if applicable.- Monitoring overhead, latency, complexity, and accuracy analysis may have to consider using at N>1 CSI feedback occasions.- Testability of UE reported metricsDiscussion may include the following aspects:- Consideration of Options 1-5 and their sub-options for alleviating / resolving the issues related to inter-vendor training collaboration- Temporal domain aspects of CSI compression- How the above monitoring approaches or combination of them may help identifying the cause (e.g., NW side, UE side, data drift) of the performance degradation* Note: for UE-side monitoring, the final reported monitoring output, if specified, may be different, eg, be further derived based on the output of the above approaches.* Note: implementation-based monitoring solutions can be considered in assessing the necessity of the above monitoring approaches.

[0300] Conventional network-side monitoring methods can operate by having the UE report target CSIs (V) based on high-resolution codebooks, such as the legacy eT2 codebook. This approach may have two problems (mutual trade-offs).

[0301] Since the performance gains of AI / ML-based CSI compression are based on non-linearity and expressiveness in latent space that are not captured in codebook-based reporting, errors in intermediate KPI calculations caused by quantization errors may occur when reporting target CSI (V) or ground-truth (e.g., channel measurement information measured by the UE) based on codebooks, which can lead to monitoring results (or monitoring outputs) that differ from the actual model performance.

[0302] - If a codebook with higher resolution is used to reduce the above quantization error, the signaling overhead for reporting the target CSI (V) for monitoring or ground-truth (e.g., channel measurement information measured by the UE) may increase.

[0303] Therefore, the following describes in detail a method for reporting target CSI (V) or ground-truth (V) based on latent information, which is a representation in latent space, rather than the codebook-based method described above.

[0304] 1. Proposal 1

[0305] In Proposal 1, for performance monitoring of CSI compression, an NW-side SGCS estimation model for intermediate KPI estimation can be used, as shown in Fig. 19 (b). For example, in relation to Proposal 1, it can be assumed that there exists an estimation model on the network side (NW-side) that receives CSI feedback (z in Fig. 19 (b)) as input and outputs intermediate KPIs such as SGCS or monitoring output information as output.

[0306] For example, the performance of the UE-side SGCS estimator (Fig. 19 (a)) may be sufficiently good. For example, the input to the SGCS estimator may be latent information within the model rather than the target CSI (V) or ground-truth (V), and the output may be the estimated SGCS. For example, it may be possible to estimate intermediate KPIs directly from information originating from the target CSI (V) or ground-truth (V) without the need for the reconstructed CSI (V_hat). The latent information may generally be a value different from z, but it can be considered essentially the same as z in that it is latent information originating from the target CSI (V). Therefore, it may also be considered to estimate the SGCS directly from z.

[0307] For example, if a sample different from the expected distribution is given as input to the network-side (NW-side) CSI reconstruction part, it can be expected that the model performance will be poor. Here, "different from the expected distribution" may mean a data sample that has not been experienced during the training process of the CSI reconstruction part. Furthermore, since information about V_hat exists on the network-side (NW-side), it will be possible to calculate intermediate KPIs or monitoring outputs without a separate report on ground-truth by utilizing V_hat information together with z. For example, in the case of Fig. 19 (a), SGCS is estimated directly from somewhere between V and z, but the proposed method can estimate values ​​such as SGCS using V_hat as well as z, and in this case, at least a similar level of estimation accuracy to that of using V and V_hat can be expected. The proposed network-side (NW-side) SGCS estimator can be trained using actual SGCS as labels and z (and / or, V_hat) as training data. For V_hat, if a sample different from the distribution expected as the output of the network-side (NW-side) CSI reconstruction part is given, poor model performance can be expected (e.g., when using a UE-side SGCS estimator). In contrast, it can be expected that there will be no issues with model performance when using a network-side (NW-side) SGCS estimator.

[0308] For example, if the z values ​​z1, z2, z3, ... for time slot t=1, 2, 3, ... change rapidly beyond a certain level within a certain window, NW can observe the z values ​​and infer that the current performance of CSI compression has changed significantly.

[0309] A key approach considering these points is to further improve the accuracy of performance monitoring by having the UE report z to the NW with a finer granularity compared to the quantization granularity used for inference. Meanwhile, the aforementioned z may be used for inference as well as monitoring. Regarding operations related to this, the NW-side operations and the UE-side operations will be explained separately.

[0310] (1) NW operation

[0311] For example, the NW can increase the CSI feedback payload through a separate (though not strictly necessary) CSI reporting setting for monitoring (simultaneous inference possible). In other words, the NW can set the payload size of the latent information for monitoring (or the payload size of the CSI feedback) through the CSI reporting setting, or set the payload size of the latent information for monitoring (or the payload size of the CSI feedback) through a separate parameter.

[0312] Generally, it is difficult to distinguish a UE's CSI reporting by its purpose (e.g., whether for inference or monitoring); however, as described above, the purpose / use can be distinguished through the CSI reporting settings. For instance, even if the CSI reporting setting on the UE is configured for monitoring purposes, the UE has the characteristic of being able to report CSI feedback (latent vector, information) to the Network without reporting the target CSI; furthermore, the reported CSI feedback can be utilized by the Network for both monitoring and inference. In other words, model performance monitoring is possible using only the CSI feedback itself, rather than the target CSI.

[0313] For example, the NW may set a monitoring opportunity or a monitoring resource / report (whereby CSI inference may also be performed on such a resource) to the UE (or transmit the monitoring resource / report setting to the UE). Meanwhile, instructions / settings regarding the payload size of the CSI report for monitoring may be provided together with the monitoring resource / report setting or provided in advance before the monitoring resource / report setting. Alternatively, the monitoring resource / report setting may include settings for the time-domain behavior of the CSI report in the UE, or, if necessary, the settings for the time-domain behavior of the CSI report in the UE may be signaled separately. For example, regarding the time-domain behavior of the CSI report in the UE, the monitoring resource / report setting may be set / instructed to periodically report z with a relatively high granularity (e.g., granularity finer than the granularity of z according to inference) once out of a total of N CSI feedback reports. Alternatively, regarding the time-domain behavior of CSI reporting in the UE, the monitoring resource / reporting setting may be configured so that the granularity of z becomes relatively high only at a specific point in time triggered by the NW (e.g., a specific point in time triggered through the monitoring setting). Alternatively, the time-domain behavior may not be configured through the monitoring resource / reporting setting as described above, but may be configured in the UE through separate signaling. The configuration of the time-domain behavior may set a monitoring period, or, in the case of non-periodical, set NW-triggered monitoring reporting opportunities.

[0314] Subsequently, the NW may receive a z (fineer version of z; CSI feedback information) from the UE that has a relatively finer granularity than the z for CSI inference (e.g., CSI feedback information) based on the monitoring settings and / or the indicators. In this case, the network may input the received z into an NW-side SGCS estimator to obtain a performance / monitoring result (or monitoring output information), and may transmit setting information to a terminal that instructs the model to be enabled, disabled, fallback, or switched based on the monitoring output information.

[0315] (2) UE operation

[0316] The UE can adjust the particle size (or quantized particle size or reported particle size) for z for monitoring based on the CSI reporting settings received from the NW (or the CSI reporting settings including the monitoring resource / reporting settings) and report the particle size-adjusted z to the NW.

[0317] Generally, it is difficult to distinguish a UE's CSI reporting by its purpose (e.g., whether for inference or monitoring); however, as described above, the purpose / use can be distinguished through the CSI reporting settings. For instance, even if the CSI reporting setting on the UE is configured for monitoring purposes, the UE has the characteristic of being able to report CSI feedback (latent vector, information) to the Network without reporting the target CSI; furthermore, the reported CSI feedback can be utilized by the Network for both monitoring and inference. In other words, model performance monitoring is possible using only the CSI feedback itself, rather than the target CSI.

[0318] The reporting granularity of z for monitoring may be set to a relatively higher or finer granularity only at specific points among the CSI reporting points (e.g., points for CSI reporting based on CSI inference). The specific points may be pre-set by the NW or set through the monitoring settings. For example, the UE may periodically change / adjust the reporting granularity to a relatively higher / fine level once out of a total of N CSI reporting points (e.g., points for CSI reporting based on CSI inference) and transmit CSI feedback information to the NW. Alternatively, as described above, z may be reported at a relatively higher / fine granularity level during NW-triggered monitoring reporting opportunities. In other words, z may be reported at a reporting point different from the reporting granularity of other CSI reports at one reporting point out of a total of N CSI reporting points, or at a reporting point triggered by the NW in relation to monitoring.

[0319] After reporting CSI feedback information for such monitoring (e.g., a report of z), the UE can receive configuration information of a model (e.g., an AI / ML encoder) from the NW and can perform activation, deactivation, fallback, or model switching operations of the model according to the instructions in the configuration information.

[0320] Referring to FIG. 19 (c), the UE can report / transmit to the base station not only information about z but also additional information as feedback information for monitoring. Here, the additional information serves as an (additional) input to a network-side (NW-side) SGCS estimator, and information corresponding to the input of the SGCS estimator in FIG. 19 (a) can be derived from the additional information (additional information and z). The additional information is characterized by having a smaller payload size compared to z and can be used to estimate SGCS alone (or together with z). The configuration method for the additional information is explained by distinguishing between NW operation and UE operation.

[0321] (1) NW operation

[0322] NW can be configured to increase the CSI feedback payload and report additional information together through a separate (though not strictly necessary) CSI reporting setting for monitoring (simultaneous inference possible). For example, NW can be configured to specify the payload size (or reporting granularity) of the CSI report for monitoring through the CSI reporting setting, and to report additional information as CSI feedback information along with the aforementioned latent information (z). Alternatively, this can be specified / configured through a separate setting parameter rather than the CSI measurement reporting setting.

[0323] For example, the NW may set a monitoring opportunity or a monitoring resource / report (whereby CSI inference may also be performed on such resources) to the UE (or transmit the monitoring resource / report setting to the UE). Meanwhile, instructions / settings regarding the payload size of the CSI report for monitoring may be provided together with the monitoring resource / report setting or provided in advance prior to the monitoring resource / report setting. The monitoring resource / report setting may include settings for the time-domain behavior of the CSI report in the UE. For example, regarding the time-domain behavior of the CSI report in the UE, the monitoring resource / report setting may be configured to periodically report z with a relatively high granularity (e.g., granularity finer than the granularity of z according to inference) once out of a total of N CSI feedback reports. Alternatively, regarding the time-domain behavior of the CSI report in the UE, the monitoring resource / report setting may be configured to have the granularity of z relatively high only at a specific point in time triggered by the NW (e.g., a specific point in time triggered through the monitoring setting). The above time-domain behavior can be configured through the monitoring resource / reporting settings or configured in the UE via separate signaling. The configuration of the time-domain behavior can set a monitoring cycle, or, in the case of non-periodical monitoring, set NW-triggered monitoring reporting opportunities.

[0324] Subsequently, the NW may receive z (CSI feedback) and additional information from the UE, input it into a network-side (NW-side) SGCS estimator to obtain a monitoring output, and transmit configuration information to the UE regarding whether to enable, disable, fallback, or perform switching on the model based on the output. For example, the NW may input the feedback z (CSI feedback) and additional information into the SGCS estimator to obtain monitoring output information, and input the feedback z (CSI feedback) into an AL / ML decoder to obtain a reconstructed CSI. And / or, the feedback z (CSI feedback) input into the SGCS estimator or AL / ML decoder may be inversely quantized.

[0325] (2) UE operation

[0326] The UE can adjust the particle size (or quantized particle size or reporting particle size) for z for monitoring based on the CSI reporting settings received from the NW (e.g., CSI reporting settings including the monitoring resource / reporting settings) and report the particle size-adjusted z and the additional information together to the NW.

[0327] The reporting granularity of z for monitoring may be set to a relatively higher or finer granularity only at specific points among the CSI reporting points (e.g., points for CSI reporting based on CSI inference). The specific points may be pre-set by the NW or set through the monitoring settings. For example, the UE may periodically change / adjust the reporting granularity to a relatively higher / fine level once out of a total of N CSI reporting points (e.g., points for CSI reporting based on CSI inference) and transmit CSI feedback information to the NW. Alternatively, as described above, z may be reported at a relatively higher / fine granularity level during NW-triggered monitoring reporting opportunities. In other words, z may be reported at a reporting point different from the reporting granularity of other CSI reports at one reporting point out of a total of N CSI reporting points, or at a reporting point triggered by the NW in relation to monitoring.

[0328] After reporting CSI feedback information for such monitoring (e.g., z and additional information), the UE can receive configuration information of a model (e.g., AI / ML encoder) from the NW and can perform activation, deactivation, fallback, or model switching operations of the model according to instructions in the configuration information.

[0329] The above-described network-side (NW-side) intermediate KPI (e.g., SGCS) estimator can be implemented as an AI / ML model and can also be defined as a network-side (NW-side) ground-truth (GT) CSI-free monitoring model.

[0330] In Proposal 1, for an intermediate KPI (e.g., SGSC) estimator for monitoring to operate sufficiently without GT-CSI, it may be necessary to finer the quantization granularity because there are limitations with a relatively coarse z. Therefore, the UE may perform an action to increase the payload size. For example, the UE may determine the quantization granularity and / or quantization method for z and report this to the NW. If the UE determines that monitoring of CSI compression is necessary, it may adjust the quantization granularity / method and report the adjusted value to the network side (NW-side). Alternatively, if the NW sets / instructs the UE to monitor (e.g., when the monitoring reporting time is specified by monitoring resources / reporting settings), the UE may adjust the quantization granularity / method and report the adjusted value to the network side (NW-side). Meanwhile, as described above, this case may apply when there is no setting / instruction from the NW side regarding the granularity / payload size for monitoring CSI reporting.

[0331] Alternatively, in Proposal ②, the UE may need to increase the total payload size for CSI feedback to include additional information along with z. To this end, the UE may determine the quantization granularity and / or quantization method for the additional information and report this to the NW. For example, if the UE determines that monitoring of CSI compression is necessary, it may adjust the payload size and report the adjusted value to the network side (NW-side). Or, if the NW sets up / instructs the UE to monitor, the UE may adjust the payload size and report the adjusted value to the network side (NW-side). In this case, the UE may also report to the network side (NW-side) which part of the total payload corresponds to z and which part corresponds to the additional information. Meanwhile, as described above, this case may apply when there is no setting / instruction from the NW side regarding the granularity / payload size for monitoring CSI reporting.

[0332] Below, the method for configuring the proposed additional information is explained in detail.

[0333] The additional information described above may basically be information regarding how different a given sample (data point) is from the data distribution learned by the model, with respect to the target CSI (V) and the latent z corresponding to the CSI feedback. For example, it may refer to the discrepancy between the input / output of the encoder model and the data distribution experienced by the encoder model through learning. Meanwhile, the additional information may be configured in a manner that performs residual coding or conditional coding using information about the target CSI (V) and information about the latent z. For example, the additional information may be configured in a data-driven manner according to at least one of the two cases below.

[0334] - Case 1: For example, the UE can first configure the signal flow as follows. The output of a part (which may be a specific layer) of the CSI generation part (encoder model) is fed as input to a UE-side nominal / reference SGCS estimator, and the label for the output of the SGCS estimator is the SGCS obtained by comparing V and V_hat (ground-truth or equivalent to GT), thereby allowing the encoder model and the UE-side nominal / reference SGCS estimator to be trained together. To obtain V_hat on the UE-side, the UE-side nominal / reference CSI reconstruction part can be utilized. Alternatively, V_hat may be received from the network side (NW-side). A dataset representing the parameters / model or inputs / outputs for the trained UE-side nominal / reference SGCS estimator can be transferred / delivered to the NW.

[0335] - Case 2: When there is a V corresponding to GT CSI on the network side (NW-side), the NW can train a network-side (NW-side) nominal / reference SGCS estimator by utilizing the network-side (NW-side) nominal / reference encoder (CSI generation part) model. After training is completed, a dataset representing the parameters / model or input / output of the network-side (NW-side) nominal / reference SGCS estimator can be transferred / delivered to the UE.

[0336] Below, the training method for the network-side (NW-side) GT CSI-free monitoring model is explained in detail.

[0337] The training method for a monitoring model based on the training location may be as follows.

[0338] (1) Network-side training method

[0339] The UE may report GT CSI to the NW for the training of the above-mentioned monitoring model. Reporting of the GT CSI may occur only during the training phase. For example, GT CSI may not be reported for inference or monitoring.

[0340] (2) UE-side training method

[0341] - 1. Pre-training process: In order to train the above GT CSI-free monitoring model on the UE-side, the CSI reconstruction part (hereinafter referred to as the decoder) and / or the above monitoring model located on the network side (NW-side) may be transferred / delivered to the UE-side. At this time, the model transferred / delivered by the NW to the UE-side may be the actual model or a proxy / reference model. Prior to training, model parameters and / or a training dataset may be transferred from the NW to the UE for training.

[0342] - 2. Training process: A monitoring model that has been delivered to or is already present on the UE-side can be trained. At this time, labels can be obtained from the decoder model and the UE-side CSI generation part (hereinafter the encoder) that are transferred / delivered from the network side (NW-side) in the “pre-training” stage.

[0343] - 3. After training: After training is completed, the monitoring model can be transferred / delivered to the network side.

[0344] The method for configuring the training dataset for the aforementioned learning (when training on the network side) may be as follows.

[0345] In Proposal 1, since z values ​​reflecting multiple quantization methods / graininess are transmitted to the NW for latent information (or z) corresponding to CSI feedback, the training dataset may include information regarding the said graininess / method. For example, the said graininess / method information may vary at the dataset level, or the said graininess / method information may vary at the data sample level. Of course, the constructed training dataset may be transmitted to the network side prior to network-side training.

[0346] Performance monitoring methods for network-side (NW-side) GT CSI-free monitoring models may be as follows. Monitoring methods may be distinguished based on the monitoring location.

[0347] (1) Network-side monitoring method

[0348] The UE can report the GT CSI to the NW, and monitoring (or performance evaluation) of the proposed monitoring model (e.g., SGCS estimator or SGCS estimation model) can be performed by calculating the actual intermediate KPI through the GT CSI and comparing it with the estimated intermediate KPI of the monitoring model.

[0349] (2) UE-side monitoring method

[0350] A proxy / reference model for a network-side (NW-side) GT CSI-free monitoring model may exist on the UE-side. In this case, the terminal can perform monitoring on the said proxy / reference model, update the monitoring model based on the monitoring results, and transfer / delivery the updated model to the network-side (NW-side).

[0351] (3) UE-assisted monitoring method

[0352] A proxy / reference model for a network-side (NW-side) GT CSI-free monitoring model may exist on the UE-side. In this case, the terminal may calculate a monitoring output for the proxy / reference model and report it to the NW. Here, the monitoring output may be an estimation accuracy for SGCS.

[0353] As such, the proposed method enables performance monitoring with relatively low signaling overhead compared to ground-truth target CSI reporting. Furthermore, relatively accurate performance monitoring can be expected compared to codebook-based GT CSI reporting.

[0354] FIG. 20 is a diagram illustrating a method for a UE to transmit output information for monitoring an AI model to a base station.

[0355] As described above, the UE performs CSI inference based on an AI model (or AL / ML) and can report / transmit output information, which is the result of the AI ​​model's CSI inference, to the base station as feedback information. Here, the output information may be latent information representing the CSI in latent space. For example, the UE may input the measured CSI into the CSI compression model, which is the AI ​​model, and obtain latent information representing the CSI in latent space from the CSI compression model (or calculate latent information representing the CSI in latent space using the CSI compression model). In this case, the UE may report the latent information, which is the output information of the CSI compression model, to the base station as feedback information for channel measurement.

[0356] Meanwhile, as described above, the UE may provide / report to the base station potential information, which is the output information of the AI ​​model or CSI compression model, rather than the raw target CSI(V) or ground-truth CSI measured through CSI-RS resources, etc., as information for evaluating the performance of the AI ​​model or CSI compression model. Below, a method for the UE to provide the output information of the AI ​​model or CSI compression model (hereinafter referred to as the AI ​​model) to the base station as information for evaluating the performance of the AI ​​model or CSI compression model will be explained in detail.

[0357] Referring to FIG. 20, the UE may receive CSI reporting setting information from a base station (S201). The CSI reporting setting information may be basic setting information for CSI measurement and reporting. Additionally, the CSI reporting setting information may further include setting information for performance monitoring of the AI ​​model. For example, the CSI reporting setting information may include setting information regarding the payload size (or quantization granularity) of the output information related to performance monitoring of the AI ​​model. And / or, the CSI reporting setting information may include information for directing / setting CSI reporting settings related to performance monitoring of the AI ​​model. And / or, the CSI reporting setting information may further include information for directing time-domain reporting behavior.

[0358] The UE can measure the CSI based on the CSI reporting setting information (S203). The UE inputs the measured CSI into the AI ​​model and, accordingly, obtains output information of the AI ​​model (e.g., latent information, which is a representation of the measured CSI in a latent space). Here, the output information of the AI ​​model may be feedback information for reporting the measured CSI or monitoring information for monitoring the performance of the AI ​​model.

[0359] Next, the UE can transmit output information of an AI model with the CSI as input to a base station (S205). At this time, the payload size of the output information of the AI ​​model may be determined based on whether the output information of the AI ​​model is related to performance monitoring of the AI ​​model. For example, if the output information of the AI ​​model is output information for performance monitoring of the AI ​​model, the payload size of the output information of the AI ​​model may be determined as a first payload size. Conversely, if the output information of the AI ​​model is output information for CSI inference based on the measured CSI (e.g., output information intended to provide feedback information related to the measured CSI), the payload size of the output information of the AI ​​model may be determined as a second payload size different from the first payload size. As described above, the first payload size may be larger than the second payload size, and accordingly, the output information of the AI ​​model having the first payload size may have a finer granularity / quantization granularity than the output information of the AI ​​model having the second payload size. This is to improve the accuracy of the performance evaluation of the AI ​​model at the base station without the direct transmission of raw channel state information measured by the UE by transmitting monitoring latent information with a finer granularity than inference latent information. For example, the UE does not necessarily report the raw target CSI (V) or ground-truth CSI itself directly for performance monitoring, but may transmit / report the latent information with a relatively larger payload size (or finer granularity) than the inference latent information. Meanwhile, as illustrated in FIG. 17, the output information may be transmitted to the base station after undergoing a quantization process. In this case, the output information may be quantized to have a higher granularity.

[0360] Alternatively, the UE may transmit output information of the AI ​​model based on the CSI reporting setting information multiple times. For example, the UE may transmit output information of the AI ​​model a total of N (where N is an integer) times according to the same CSI reporting setting information, and at least one of the N transmissions may be a transmission of output information for monitoring the performance of the AI ​​model. For example, output information for monitoring with a relatively large payload size or finer quantization granularity may be transmitted in only one of the total N reporting opportunities, and output information for CSI inference may be transmitted in the remaining reporting opportunities. For example, the UE may be configured so that output information for monitoring the AI ​​model is transmitted in the last reporting opportunity among the N reporting opportunities. Alternatively, the base station may instruct / configure (e.g., specify / configure N and m) to transmit output information for monitoring the AI ​​model at the m-th reporting opportunity (where m is an integer) among N reporting opportunities, and may transmit output information for monitoring the AI ​​model (e.g., output information having a payload size larger than the output information for CSI inference) at the m-th reporting opportunity instructed / configured for each of the N reporting opportunities. Alternatively, it may be set to a specific reporting time point that is triggered non-periodically by the base station (e.g., a reporting time point for monitoring the AI ​​model), and may transmit / report the output information having the first payload size to the base station at such a specific reporting time point.

[0361] For example, the above CSI report setting information may include information for designating / setting a reporting opportunity related to performance monitoring of the AI ​​model among a plurality of reporting opportunities. For example, the above CSI report setting information may further include information regarding the monitoring period and a specific reporting time point at which monitoring is performed, as information for setting the time-domain behavior of the CSI report. Accordingly, the UE can transmit output information of the AI ​​model by applying a different payload size and / or a different quantization granularity at a monitoring reporting time point that is distinct from a general CSI inference report. Furthermore, the base station may transmit a control signal (DCI or MAC-CE) to the UE that triggers or activates at least one of the plurality of CSI report settings included in the above CSI report setting information for performance monitoring of the AI ​​model, and the UE may generate and transmit monitoring output information according to the corresponding CSI report setting in response to the control signal.

[0362] Alternatively, as described above, the UE may provide / transmit to the base station latent information, which is the output information of the AI ​​model rather than raw CSI information measured from the CSI measurement resource, as information for monitoring the performance of said AI model. For example, both the output information for monitoring the performance of said AI model and the output information for said CSI inference may be said latent information. For example, the UE transmits latent information generated from the measured CSI to the base station, and the base station may monitor the performance of said AI model through a network-side monitoring model while simultaneously or separately performing CSI inference using said latent information. In this case, the latent information for performance monitoring may have the same or a similar structure as the latent information for CSI inference, but may be reported with a larger payload size or finer quantization granularity.

[0363] Additionally, if the output information of the AI ​​model is output information for performance monitoring of the AI ​​model, the output information of the AI ​​model may be transmitted together with additional information (or supplementary information). The additional information may be information indicating the difference in data distribution between the training CSI included in the training data of the AI ​​model (e.g., the training CSI most similar to the measured CSI among the training CSIs included in the training data) and the CSI currently measured by the UE. In other words, the additional information may be discrepancy information indicating how different the current CSI sample is from the data distribution experienced by the AI ​​model during the training process. The base station or network-side monitoring model can calculate the performance monitoring results of the AI ​​model more accurately by using the additional information along with the potential information as input. Alternatively, the additional information may have a payload size smaller than the potential information and may be configured to be transmitted combined with the potential information.

[0364] For example, the additional information may be information indicating how much a currently measured or generated CSI-related sample deviates from the training data distribution used in the training process of an AI / ML model. Here, the “deviation from the data distribution” may include a difference at the level of a single data point, a difference at the level of a window composed of multiple samples, or a difference in the feature space. Additionally, the additional information may be generated based on the target CSI (V), latent information (z), the output value of the intermediate layer of the encoder model, the reconstructed CSI (V_hat), or a combination thereof. For example, the UE or network may pre-set reference information representing the training data distribution. The reference information may include at least one of the mean, variance, covariance, moment, quantile, centroid, codebook entry, probability density function, cumulative distribution function, histogram, or parameters of a neural network-based density estimator obtained from the training dataset. In this case, the above additional information may consist of a numerical value representing the difference between the current sample or current sample set and the above reference information.

[0365] Alternatively, the data distribution difference may be calculated using a distance-based method. For example, the UE may calculate the Euclidean distance, Manhattan distance, cosine distance, or Mahalanobis distance between the feature vector of the current sample and the representative vector of the training data distribution. For instance, the Mahalanobis distance can be used as a form of additional information, as it reflects the covariance of the training data and indicates how far the current sample deviates from the typical region of the training distribution. This additional information may be the distance value itself, or it may be an index, level, or flag resulting from a threshold comparison of the distance value. Alternatively, the data distribution difference may be calculated as the divergence between distributions or the distance metric between distributions. For example, the UE or network may generate an empirical distribution from the recent K sample windows containing the current sample and compare it with the training data distribution to calculate at least one of Kullback-Leibler divergence, Jensen-Shannon divergence, Wasserstein distance, total variation distance, Kolmogorov-Smirnov statistic, maximum mean discrepancy (MMD), energy distance, or Bhattacharyya distance. In this case, the additional information may be the divergence or distance value itself, or a vector composed of multiple divergence / distance values.

[0366] In this way, when the UE reports output information for the monitoring to the base station, the base station inputs the output information into the SGCS estimator and can transmit configuration information or control information for the AI ​​model (e.g., CSI compression model) determined based on the output value of the SGCS estimator to the UE. Here, the configuration information or control information may include instruction information or parameter sets for activating, deactivating, falling back, or performing model switching of the AI ​​model.

[0367] In this way, the UE inputs the measured CSI into an AI model based on the CSI reporting configuration information to generate output information, and can transmit this output information by applying different payload sizes and quantization granularities depending on whether it is for performance monitoring or CSI inference. Furthermore, by configuring some of the multiple reporting opportunities for monitoring or by activating specific CSI reporting settings for monitoring based on base station triggers, the performance of the AI ​​model can be monitored more accurately without excessively increasing signaling overhead. Moreover, if necessary, the accuracy of the network-side monitoring model can be further improved by transmitting additional information indicating the difference in data distribution between the training CSI and the measured CSI, along with latent information.

[0368] FIG. 21 is a diagram illustrating how a base station receives output information of an AI model from a UE.

[0369] Referring to FIG. 21, the base station may transmit CSI report setting information to the UE (S211). The CSI report setting information may be information set for CSI reporting in which monitoring and inference may be performed together. For example, the CSI report setting information may include information for setting the payload size of latent information (z) or CSI feedback information corresponding to the output information of the AI ​​model. For example, the base station may set the payload size of the latent information for monitoring or the payload size of the CSI feedback through the CSI report setting for monitoring, and may also set this through separate parameters if necessary. In this case, the CSI report setting information may be composed of information instructing that a relatively larger payload size or a finer report granularity / quantization granularity is applied compared to the inference report when the output information of the AI ​​model is related to the performance monitoring of the AI ​​model. For example, the above CSI report setting information may include setting information regarding monitoring opportunities (or monitoring resources / reports), time-domain behavior of CSI report / output information for monitoring, and / or the payload size of output information for monitoring (e.g., a payload size value, or a value representing the difference relative to the payload size of output information for inference). Meanwhile, the above CSI report setting may include information regarding CSI resource information, etc., as in existing settings.

[0370] For example, as described above, the monitoring resource / reporting setting included in the CSI reporting setting information may include the time-domain behavior of CSI reporting in the UE. For example, the monitoring resource / reporting setting (or the CSI reporting setting information) may further include setting information that causes reporting to occur periodically at a relatively high granularity (e.g., granularity finer than the granularity of z according to inference) once among N CSI feedback reports according to the CSI reporting setting information. Additionally, the CSI reporting setting information may be configured so that the granularity of z becomes relatively high only at a specific point in time triggered by the base station, or it may be implemented by setting a monitoring period or an NW-triggered monitoring reporting opportunity. Accordingly, the CSI reporting setting information may go beyond simple CSI measurement reporting settings and further include information that applies different payload sizes depending on whether the output information of the AI ​​model is for performance monitoring, and information that sets / instructs the time or opportunity for such reporting to be performed.

[0371] Next, the base station may receive output information of an AI model from the UE, which takes the measured CSI based on the CSI reporting setting information as input (S213). The output information of the AI ​​model may correspond to latent information (z), and in the case of monitoring, it may be received in the form of a finer version of z having a relatively finer granularity than z for CSI inference. For example, based on the monitoring setting and / or the indicator, the base station may receive AI model output information from the UE having a payload size or reporting granularity different from the AI ​​model output information used for inference reporting.

[0372] For example, since information regarding the reconstructed CSI (V_hat) may exist on the base station / NW side, the base station can utilize the z received from the UE together with the V_hat on the base station side to calculate intermediate KPIs or performance monitoring values ​​for the AI ​​model or CSI compression model in the UE without a separate report on ground-truth. For example, if the output information of the AI ​​model is related to performance monitoring, the base station can input the z received from the UE into a network-side SGCS estimator to obtain performance / monitoring results or monitoring output information.

[0373] Next, the base station may transmit configuration information (e.g., parameter set, etc.) to the UE, which instructs the activation, deactivation, fallback, or switching of the model based on the monitoring output information (S215). Accordingly, the base station may set the payload size of the AI ​​model output information for performance monitoring to be distinct from that for inference through the CSI report configuration information, and receive the output information of the AI ​​model from the UE according to the configured method to utilize for network-side performance monitoring. In this case, the base station can effectively evaluate the performance of the AI ​​model using output information reported with a higher granularity, even if raw channel measurement information (or ground-truth) measured by the UE is not provided by the UE. Meanwhile, the SGCS estimator may be included on the base station side as an estimator that is pre-trained to evaluate the performance of the AI ​​model based on output information (z) reported with a higher granularity as described above.

[0374] Thus, the proposed invention enables a network-side estimator to calculate intermediate KPIs or performance monitoring outputs with significant accuracy based on latent information of the CSI compression model reported or transmitted by the UE, particularly latent information expressed with higher or finer granularity, even without separately reporting raw target CSI or ground-truth CSI. And / or, the proposed invention has the effect of enabling performance monitoring that is more consistent with actual model performance through estimation based on latent information, which is the output information of the CSI compression model, while reducing the quantization error and signaling overhead associated with codebook-based target CSI or ground-truth CSI reporting.

[0375] Example of a communication system to which the invention is applied

[0376] Although not limited thereto, the various descriptions, functions, procedures, proposals, methods, and / or flowcharts of the invention disclosed in this document may be applied to various fields requiring wireless communication / connection (e.g., 5G) between devices.

[0377] Examples are provided in more detail below with reference to the drawings. In the following drawings and descriptions, the same reference numerals may represent the same or corresponding hardware blocks, software blocks, or function blocks unless otherwise described.

[0378] FIG. 22 illustrates a communication system to which the present invention is applied.

[0379] Referring to FIG. 22, the communication system (1) to which the present invention applies includes a wireless device, a base station, and a network. Here, the wireless device refers to a device that performs communication using wireless access technology (e.g., 5G NR (New RAT), LTE (Long Term Evolution)) and may be referred to as a communication / wireless / 5G device. Although not limited thereto, the wireless device may include a robot (100a), a vehicle (100b-1, 100b-2), an XR (eXtended Reality) device (100c), a hand-held device (100d), a home appliance (100e), an IoT (Internet of Thing) device (100f), and an AI device / server (400). For example, the vehicle may include a vehicle equipped with wireless communication functions, an autonomous vehicle, a vehicle capable of performing inter-vehicle communication, etc. Here, the vehicle may include an Unmanned Aerial Vehicle (UAV) (e.g., a drone). XR devices include AR (Augmented Reality) / VR (Virtual Reality) / MR (Mixed Reality) devices and can be implemented in the form of HMDs (Head-Mounted Devices), HUDs (Head-Up Displays) equipped in vehicles, televisions, smartphones, computers, wearable devices, home appliances, digital signage, vehicles, robots, etc. Portable devices may include smartphones, smartpads, wearable devices (e.g., smartwatches, smart glasses), computers (e.g., laptops, etc.). Home appliances may include TVs, refrigerators, washing machines, etc. IoT devices may include sensors, smart meters, etc. For example, base stations and networks may be implemented as wireless devices, and a specific wireless device (200a) may operate as a base station / network node to other wireless devices.

[0380] Wireless devices (100a to 100f) can be connected to a network (300) through a base station (200). Artificial Intelligence (AI) technology may be applied to the wireless devices (100a to 100f), and the wireless devices (100a to 100f) can be connected to an AI server (400) through the network (300). The network (300) can be configured using a 3G network, a 4G (e.g., LTE) network, or a 5G (e.g., NR) network. The wireless devices (100a to 100f) may communicate with each other through the base station (200) / network (300), but they may also communicate directly (e.g., sidelink communication) without going through the base station / network. For example, vehicles (100b-1, 100b-2) can communicate directly (e.g., V2V (Vehicle to Vehicle) / V2X (Vehicle to everything) communication). Also, IoT devices (e.g., sensors) can communicate directly with other IoT devices (e.g., sensors) or other wireless devices (100a to 100f).

[0381] Wireless communication / connection (150a, 150b, 150c) can be established between wireless devices (100a~100f) / base station (200) and base station (200) / base station (200). Here, wireless communication / connection can be achieved through various wireless access technologies (e.g., 5G NR), such as uplink / downlink communication (150a), sidelink communication (150b) (or D2D communication), and inter-base station communication (150c) (e.g., relay, IAB (Integrated Access Backhaul)). Through wireless communication / connection (150a, 150b, 150c), wireless devices and base stations / wireless devices, and base stations and base stations can transmit / receive wireless signals to / from each other. For example, wireless communication / connection (150a, 150b, 150c) can transmit / receive signals through various physical channels. To this end, based on various proposals of the present invention, at least some of the following may be performed: various configuration information setting processes for transmitting / receiving wireless signals, various signal processing processes (e.g., channel encoding / decoding, modulation / demodulation, resource mapping / demapping, etc.), resource allocation processes, etc.

[0382] Example of a wireless device to which the present invention is applied

[0383] FIG. 23 illustrates a wireless device that can be applied to the present invention.

[0384] Referring to FIG. 23, the first wireless device (100) and the second wireless device (200) can transmit and receive wireless signals through various wireless access technologies (e.g., LTE, NR). Here, {the first wireless device (100), the second wireless device (200)} may correspond to {wireless device (100x), base station (200)} and / or {wireless device (100x), wireless device (100x)} of FIG. 22.

[0385] The first wireless device (100) includes one or more processors (102) and one or more memories (104), and may additionally include one or more transceivers (106) and / or one or more antennas (108). The processor (102) controls the memory (104) and / or transceivers (106) and may be configured to implement the descriptions, functions, procedures, proposals, methods and / or flowcharts of operation disclosed in this document. For example, the processor (102) may process information within the memory (104) to generate a first information / signal and then transmit a wireless signal containing the first information / signal through the transceiver (106). Additionally, the processor (102) may receive a wireless signal containing a second information / signal through the transceiver (106) and then store information obtained from the signal processing of the second information / signal in the memory (104). The memory (104) may be connected to the processor (102) and may store various information related to the operation of the processor (102). For example, the memory (104) may store software code containing instructions for performing some or all of the processes controlled by the processor (102) or for performing the descriptions, functions, procedures, proposals, methods, and / or operation sequence diagrams disclosed in this document. Here, the processor (102) and the memory (104) may be part of a communication modem / circuit / chipset designed to implement wireless communication technology (e.g., LTE, NR). The transceiver (106) may be connected to the processor (102) and may transmit and / or receive wireless signals through one or more antennas (108). The transceiver (106) may include a transmitter and / or receiver. The transceiver (106) may be combined with an RF (Radio Frequency) unit. In the present invention, the wireless device may refer to a communication modem / circuit / chipset.

[0386] Specifically, the first wireless device or UE (100) may include a processor (102) connected to a transceiver (106) and a memory (104). The memory (104) may include at least one program capable of performing operations related to the embodiments described in the section described with reference to FIGS. 17 through 21 in the section “CSI-free NW-side monitoring for CSI compression”. The operations include controlling the RF transceiver to receive CSI (channel state information) reporting setting information, measuring the CSI based on the CSI reporting setting information, and transmitting output information of an AI (Artificial Intelligence) model with the CSI as input, and the payload size of the output information of the AI ​​model may be determined based on whether the output information of the AI ​​model is related to performance monitoring of the AI ​​model.

[0387] Alternatively, a processing device may be configured including a processor (102) and a memory (104) that control the UE. In this case, the processing device may include at least one processor; and at least one memory connected to the at least one processor and storing instructions that perform operations when executed by the at least one processor. The method includes receiving CSI (channel state information) reporting setting information, measuring the CSI based on the CSI reporting setting information, and transmitting output information of an AI (Artificial Intelligence) model that takes the CSI as input, wherein the payload size of the output information of the AI ​​model may be determined based on whether the output information of the AI ​​model is related to performance monitoring of the AI ​​model.

[0388] Alternatively, at least one non-transient computer-readable medium may be configured, in which a program / instruction for performing the above-described operations is stored.

[0389] The second wireless device (200) includes one or more processors (202) and one or more memories (204), and may additionally include one or more transceivers (206) and / or one or more antennas (208). The processor (202) controls the memory (204) and / or transceivers (206) and may be configured to implement the descriptions, functions, procedures, proposals, methods and / or operation sequences disclosed in this document. For example, the processor (202) may process information within the memory (204) to generate a third information / signal and then transmit a wireless signal containing the third information / signal through the transceiver (206). Additionally, the processor (202) may receive a wireless signal containing a fourth information / signal through the transceiver (206) and then store information obtained from the signal processing of the fourth information / signal in the memory (204). The memory (204) may be connected to the processor (202) and may store various information related to the operation of the processor (202). For example, the memory (204) may store software code containing instructions for performing some or all of the processes controlled by the processor (202) or for performing the descriptions, functions, procedures, proposals, methods, and / or operation sequence diagrams disclosed in this document. Here, the processor (202) and the memory (204) may be part of a communication modem / circuit / chip designed to implement wireless communication technology (e.g., LTE, NR). The transceiver (206) may be connected to the processor (202) and may transmit and / or receive wireless signals through one or more antennas (208). The transceiver (206) may include a transmitter and / or receiver. The transceiver (206) may be interchangeable with an RF unit. In the present invention, the wireless device may refer to a communication modem / circuit / chip.

[0390] Specifically, the second wireless device or base station (200) may include a processor (202) and a memory (204) connected to a transceiver or RF transceiver (206). The memory (204) may include at least one program capable of performing operations related to the embodiments described in the section “CSI-free NW-side monitoring for CSI compression” with reference to FIGS. 17 through 21. The operations include controlling the RF transceiver to transmit CSI (channel state information) reporting setting information and receiving output information of an AI (Artificial Intelligence) model from a UE (User Equipment) that takes a measured CSI based on the CSI reporting setting information as input, and the CSI reporting setting information may include information for setting the payload size of the AI ​​model output information differently based on whether the output information of the AI ​​model is related to performance monitoring of the AI ​​model.

[0391] Hereinafter, hardware elements of the wireless device (100, 200) will be described in more detail. Although not limited thereto, one or more protocol layers may be implemented by one or more processors (102, 202). For example, one or more processors (102, 202) may implement one or more layers (e.g., functional layers such as PHY, MAC, RLC, PDCP, RRC, SDAP). One or more processors (102, 202) may generate one or more Protocol Data Units (PDUs) and / or Service Data Units (SDUs) according to the descriptions, functions, procedures, proposals, methods, and / or flowcharts of operation disclosed in this document. One or more processors (102, 202) may generate messages, control information, data, or information according to the descriptions, functions, procedures, proposals, methods, and / or flowcharts of operation disclosed in this document. One or more processors (102, 202) may generate a signal (e.g., baseband signal) containing a PDU, SDU, message, control information, data, or information according to the functions, procedures, proposals, and / or methods disclosed in this document and provide it to one or more transceivers (106, 206). One or more processors (102, 202) may receive a signal (e.g., baseband signal) from one or more transceivers (106, 206) and may obtain a PDU, SDU, message, control information, data, or information according to the descriptions, functions, procedures, proposals, methods, and / or flowcharts disclosed in this document.

[0392] One or more processors (102, 202) may be referred to as a controller, microcontroller, microprocessor, or microcomputer. One or more processors (102, 202) may be implemented by hardware, firmware, software, or a combination thereof. For example, one or more Application Specific Integrated Circuits (ASICs), one or more Digital Signal Processors (DSPs), one or more Digital Signal Processing Devices (DSPDs), one or more Programmable Logic Devices (PLDs), or one or more Field Programmable Gate Arrays (FPGAs) may be included in one or more processors (102, 202). The descriptions, functions, procedures, proposals, methods, and / or flowcharts disclosed in this document may be implemented using firmware or software, and the firmware or software may be implemented to include modules, procedures, functions, etc. Firmware or software configured to perform the descriptions, functions, procedures, proposals, methods, and / or operation sequences disclosed in this document may be contained in one or more processors (102, 202) or stored in one or more memories (104, 204) and driven by one or more processors (102, 202). The descriptions, functions, procedures, proposals, methods, and / or operation sequences disclosed in this document may be implemented using firmware or software in the form of code, instructions, and / or sets of instructions.

[0393] One or more memories (104, 204) may be connected to one or more processors (102, 202) and may store various forms of data, signals, messages, information, programs, codes, instructions, and / or commands. One or more memories (104, 204) may be composed of ROM, RAM, EPROM, flash memory, hard drive, registers, cache memory, computer read storage media, and / or combinations thereof. One or more memories (104, 204) may be located inside and / or outside of one or more processors (102, 202). Additionally, one or more memories (104, 204) may be connected to one or more processors (102, 202) through various technologies such as wired or wireless connections.

[0394] One or more transceivers (106, 206) may transmit user data, control information, wireless signals / channels, etc., as mentioned in the methods and / or operation flowcharts, etc., of this document to one or more other devices. One or more transceivers (106, 206) may receive user data, control information, wireless signals / channels, etc., as mentioned in the descriptions, functions, procedures, proposals, methods and / or operation flowcharts, etc., disclosed in this document from one or more other devices. For example, one or more transceivers (106, 206) may be connected to one or more processors (102, 202) and may transmit and receive wireless signals. For example, one or more processors (102, 202) may control one or more transceivers (106, 206) to transmit user data, control information, or wireless signals to one or more other devices. Additionally, one or more processors (102, 202) may control one or more transceivers (106, 206) to receive user data, control information, or wireless signals from one or more other devices. Additionally, one or more transceivers (106, 206) may be connected to one or more antennas (108, 208), and one or more transceivers (106, 206) may be configured to transmit and receive user data, control information, wireless signals / channels, etc., as described in the descriptions, functions, procedures, proposals, methods, and / or flowcharts of operation disclosed in this document through one or more antennas (108, 208). In this document, one or more antennas may be multiple physical antennas or multiple logical antennas (e.g., antenna ports). One or more transceivers (106, 206) can convert the received wireless signal / channel, etc. from an RF band signal to a baseband signal in order to process the received user data, control information, wireless signal / channel, etc. using one or more processors (102, 202).One or more transceivers (106, 206) can convert user data, control information, wireless signals / channels, etc. processed using one or more processors (102, 202) from baseband signals to RF band signals. To this end, one or more transceivers (106, 206) may include (analog) oscillators and / or filters.

[0395] Examples of wireless device applications to which the present invention is applied

[0396] FIG. 24 illustrates another example of a wireless device to which the present invention applies. The wireless device may be implemented in various forms depending on the use-example / service (see FIG. 22).

[0397] Referring to FIG. 24, the wireless device (100, 200) corresponds to the wireless device (100, 200) of FIG. 23 and may be composed of various elements, components, units / parts, and / or modules. For example, the wireless device (100, 200) may include a communication unit (110), a control unit (120), a memory unit (130), and additional elements (140). The communication unit may include a communication circuit (112) and transceiver(s) (114). For example, the communication circuit (112) may include one or more processors (102, 202) and / or one or more memories (104, 204) of FIG. 24. For example, the transceiver(s) (114) may include one or more transceivers (106, 206) and / or one or more antennas (108, 208) of FIG. 23. The control unit (120) is electrically connected to the communication unit (110), the memory unit (130), and additional elements (140) and controls the general operation of the wireless device. For example, the control unit (120) may control the electrical / mechanical operation of the wireless device based on a program / code / command / information stored in the memory unit (130). Additionally, the control unit (120) may transmit information stored in the memory unit (130) to an external (e.g., another communication device) via a wireless / wired interface through the communication unit (110), or store information received from an external (e.g., another communication device) via a wireless / wired interface through the communication unit (110) in the memory unit (130).

[0398] The additional element (140) can be configured in various ways depending on the type of wireless device. For example, the additional element (140) may include at least one of a power unit / battery, an input / output unit (I / O unit), a driving unit, and a computing unit. Although not limited thereto, the wireless device may be implemented in the form of a robot (Fig. 22, 100a), a vehicle (Fig. 22, 100b-1, 100b-2), an XR device (Fig. 22, 100c), a portable device (Fig. 22, 100d), a home appliance (Fig. 22, 100e), an IoT device (Fig. 22, 100f), a digital broadcasting terminal, a hologram device, a public safety device, an MTC device, a medical device, a fintech device (or financial device), a security device, a climate / environment device, an AI server / device (Fig. 22, 400), a base station (Fig. 22, 200), a network node, etc. Wireless devices can be used in a movable or fixed location depending on the use—e.g., service.

[0399] In FIG. 24, various elements, components, units / parts, and / or modules within the wireless device (100, 200) may be entirely interconnected via a wired interface, or at least partially connected via a communication unit (110). For example, within the wireless device (100, 200), the control unit (120) and the communication unit (110) may be connected via a wire, and the control unit (120) and the first unit (e.g., 130, 140) may be connected wirelessly via the communication unit (110). Additionally, each element, component, unit / part, and / or module within the wireless device (100, 200) may include one or more additional elements. For example, the control unit (120) may be composed of one or more sets of processors. For example, the control unit (120) may be composed of a set of a communication control processor, an application processor, an Electronic Control Unit (ECU), a graphics processing processor, a memory control processor, etc. As another example, the memory unit (130) may be composed of RAM (Random Access Memory), DRAM (Dynamic RAM), ROM (Read Only Memory), flash memory, volatile memory, non-volatile memory and / or a combination thereof.

[0400] Examples of vehicles or autonomous vehicles to which the present invention is applied

[0401] FIG. 25 illustrates a vehicle or autonomous vehicle to which the present invention applies. The vehicle or autonomous vehicle may be implemented as a mobile robot, a vehicle, a train, an aerial vehicle (AV), a ship, etc.

[0402] Referring to FIG. 25, a vehicle or autonomous vehicle (100) may include an antenna unit (108), a communication unit (110), a control unit (120), a driving unit (140a), a power supply unit (140b), a sensor unit (140c), and an autonomous driving unit (140d). The antenna unit (108) may be configured as part of the communication unit (110). Blocks 110 / 130 / 140a to 140d each correspond to blocks 110 / 130 / 140 of FIG. 24.

[0403] The communication unit (110) can transmit and receive signals (e.g., data, control signals, etc.) with external devices such as other vehicles, base stations (e.g., base stations, roadside base stations (Roadside units), etc.), and servers. The control unit (120) can perform various operations by controlling elements of the vehicle or autonomous vehicle (100). The control unit (120) may include an Electronic Control Unit (ECU). The driving unit (140a) can drive the vehicle or autonomous vehicle (100) on the ground. The driving unit (140a) may include an engine, motor, power train, wheels, brakes, steering device, etc. The power supply unit (140b) supplies power to the vehicle or autonomous vehicle (100) and may include wired / wireless charging circuits, batteries, etc. The sensor unit (140c) can obtain vehicle status, surrounding environment information, user information, etc. The sensor unit (140c) may include an IMU (inertial measurement unit) sensor, a collision sensor, a wheel sensor, a speed sensor, an inclination sensor, a weight detection sensor, a heading sensor, a position module, a vehicle forward / reverse sensor, a battery sensor, a fuel sensor, a tire sensor, a steering sensor, a temperature sensor, a humidity sensor, an ultrasonic sensor, an illuminance sensor, a pedal position sensor, etc. The autonomous driving unit (140d) may implement technologies such as maintaining the driving lane, technologies for automatically adjusting speed such as adaptive cruise control, technologies for automatically driving along a predetermined path, and technologies for automatically setting a path and driving when a destination is set.

[0404] For example, the communication unit (110) can receive map data, traffic information data, etc. from an external server. The autonomous driving unit (140d) can generate an autonomous driving path and a driving plan based on the acquired data. The control unit (120) can control the drive unit (140a) so that the vehicle or the autonomous vehicle (100) moves along the autonomous driving path according to the driving plan (e.g., speed / direction control). During autonomous driving, the communication unit (110) can acquire the latest traffic information data from an external server non-periodically and can acquire surrounding traffic information data from surrounding vehicles. Additionally, during autonomous driving, the sensor unit (140c) can acquire vehicle status and surrounding environment information. The autonomous driving unit (140d) can update the autonomous driving path and the driving plan based on the newly acquired data / information. The communication unit (110) can transmit information regarding the vehicle location, autonomous driving path, driving plan, etc. to an external server. An external server can predict traffic information data in advance using AI technology, etc., based on information collected from vehicles or autonomous vehicles, and can provide the predicted traffic information data to vehicles or autonomous vehicles.

[0405] Here, the wireless communication technology implemented in the wireless device (XXX, YYY) of this specification may include LTE, NR, and 6G, as well as Narrowband Internet of Things for low-power communication. For example, NB-IoT technology may be an example of LPWAN (Low Power Wide Area Network) technology and may be implemented according to standards such as LTE Cat NB1 and / or LTE Cat NB2, but is not limited to the names mentioned above. Additionally, or generally, the wireless communication technology implemented in the wireless device (XXX, YYY) of this specification may perform communication based on LTE-M technology. For example, LTE-M technology may be an example of LPWAN technology and may be referred to by various names such as eMTC (enhanced Machine Type Communication). For example, LTE-M technology may be implemented in at least one of various standards such as 1) LTE CAT 0, 2) LTE Cat M1, 3) LTE Cat M2, 4) LTE non-BL (non-Bandwidth Limited), 5) LTE-MTC, 6) LTE Machine Type Communication, and / or 7) LTE M, and is not limited to the names mentioned above. Additionally or generally, wireless communication technology implemented in the wireless device (XXX, YYY) of this specification may include at least one of ZigBee, Bluetooth, and Low Power Wide Area Network (LPWAN) with consideration for low-power communication, and is not limited to the names mentioned above. As an example, ZigBee technology can create personal area networks (PANs) related to small / low-power digital communication based on various standards such as IEEE 802.15.4, and may be referred to by various names.

[0406] The embodiments described above are combinations of the components and features of the present invention in a specific form. Each component or feature should be considered optional unless otherwise explicitly stated. Each component or feature may be implemented in a form not combined with other components or features. Additionally, it is possible to construct embodiments of the present invention by combining some components and / or features. The order of operations described in the embodiments of the present invention may be changed. Some components or features of one embodiment may be included in another embodiment, or may be replaced with corresponding components or features of another embodiment. It is obvious that embodiments may be constructed by combining claims that do not have an explicit citation relationship in the claims, or that new claims may be included by amendment after filing.

[0407] In this document, embodiments of the present invention are described primarily with a focus on the signal transmission and reception relationship between a terminal and a base station. This transmission and reception relationship is extended in the same or similar manner to signal transmission and reception between a terminal and a relay or between a base station and a relay. Specific operations described in this document as being performed by a base station may, in some cases, be performed by an upper node. That is, it is self-evident that various operations performed for communication with a terminal in a network consisting of multiple network nodes including a base station may be performed by the base station or other network nodes other than the base station. The base station may be replaced by terms such as fixed station, Node B, eNode B (eNB), and access point. Additionally, the terminal may be replaced by terms such as User Equipment (UE), Mobile Station (MS), and Mobile Subscriber Station (MSS).

[0408] Embodiments according to the present invention may be implemented by various means, for example, hardware, firmware, software, or a combination thereof. In the case of implementation by hardware, one embodiment of the present invention may be implemented by one or more ASICs (application specific integrated circuits), DSPs (digital signal processors), DSPDs (digital signal processing devices), PLDs (programmable logic devices), FPGAs (field programmable gate arrays), processors, controllers, microcontrollers, microprocessors, etc.

[0409] In the case of implementation by firmware or software, an embodiment of the present invention may be implemented in the form of a module, procedure, function, etc., that performs the functions or operations described above. Software code may be stored in a memory unit and executed by a processor. The memory unit may be located inside or outside the processor and may exchange data with the processor by various means already known.

[0410] It is obvious to those skilled in the art that the present invention may be embodied in other specific forms without departing from the features of the invention. Accordingly, the foregoing detailed description should not be interpreted restrictively in all respects but should be considered exemplary. The scope of the invention shall be determined by a reasonable interpretation of the appended claims, and all modifications within the equivalent scope of the invention are included within the scope of the invention.

[0411] The embodiments of the present invention as described above can be applied to various mobile communication systems.

Claims

1. In a method using UE (User Equipment), Step of receiving CSI (channel state information) report configuration information; A step of measuring CSI based on the above CSI report setting information; and It includes the step of transmitting output information of an AI (Artificial Intelligence) model that takes the above CSI as input, and A method in which the payload size of the output information of the above AI model is determined based on whether the output information of the above AI model is related to performance monitoring of the above AI model.

2. In Paragraph 1, Based on the fact that the output information of the above AI model is output information for performance monitoring of the above AI model, the payload size of the output information of the above AI model is determined to be a first payload size, and A method in which, based on the fact that the output information of the AI ​​model is output information for CSI inference based on the CSI, the payload size of the output information of the AI ​​model is determined to be a second payload size different from the first payload size.

3. In Paragraph 2, A method in which the first payload size is larger than the second payload size.

4. In Paragraph 2, A method in which the output information of the AI ​​model having the first payload size has a finer quantization grain than the output information of the AI ​​model having the second payload size.

5. In Paragraph 2, A method in which output information for performance monitoring of the above AI model and output information for CSI inference are latent information.

6. In Paragraph 1, The above UE transmits the output information of the AI ​​model based on the above CSI report setting information N times, and A method in which one of the above N transmissions is a transmission of output information for monitoring the performance of the AI ​​model.

7. In Paragraph 1, A method comprising the above CSI reporting setting information including information for setting at least one reporting opportunity among a plurality of reporting opportunities as a reporting opportunity related to performance monitoring of the AI ​​model.

8. In Paragraph 1, A method further comprising the step of receiving a control signal from a base station to trigger or activate at least one CSI report setting among a plurality of CSI report settings included in the CSI report setting information for performance monitoring of the AI ​​model.

9. In Paragraph 1, A method in which, based on the fact that the output information of the AI ​​model is output information for monitoring the performance of the AI ​​model, the output information of the AI ​​model is transmitted along with additional information indicating the difference in data distribution between the training CSI included as training data of the AI ​​model and the CSI.

10. In at least one non-transient computer-readable recording medium, Includes instructions that perform operations when executed by at least one processor, The above operations are, Receive CSI (channel state information) report configuration information; Measure CSI based on the above CSI report setting information; and It includes transmitting output information of an AI (Artificial Intelligence) model that takes the above CSI as input, and At least one non-transient computer-readable recording medium, wherein the payload size of the output information of the AI ​​model is determined based on whether the output information of the AI ​​model is related to performance monitoring of the AI ​​model.

11. Regarding UE (User Equipment), RF (Radio Frequency) transceiver; and It includes a processor connected to the above RF transceiver, and The processor controls the RF transceiver to receive CSI (channel state information) reporting setting information, measures CSI based on the CSI reporting setting information, and transmits output information of an AI (Artificial Intelligence) model that takes the CSI as input. A UE whose payload size of the output information of the above AI model is determined based on whether the output information of the above AI model is related to performance monitoring of the above AI model.

12. In Paragraph 11, Based on the fact that the output information of the above AI model is output information for performance monitoring of the above AI model, the payload size of the output information of the above AI model is determined to be a first payload size, and A UE in which, based on the fact that the output information of the AI ​​model is output information for CSI inference based on the CSI, the payload size of the output information of the AI ​​model is determined to be a second payload size different from the first payload size.

13. In a processing device for controlling UE (User Equipment), At least one processor; and It includes at least one memory that stores instructions connected to the above at least one processor and performing operations when executed by the at least one processor, The above operations cause the UE to: Receive CSI (channel state information) report configuration information; Measure CSI based on the above CSI report setting information; and It includes transmitting output information of an AI (Artificial Intelligence) model that takes the above CSI as input, and A processing device in which the payload size of the output information of the above AI model is determined based on whether the output information of the above AI model is related to performance monitoring of the above AI model.

14. In the method using a base station, Step of transmitting CSI (channel state information) report configuration information; and The method includes the step of receiving output information of an AI (Artificial Intelligence) model from a UE (User Equipment) that takes a measured CSI based on the CSI report setting information as input. A method comprising the above CSI report setting information including information for setting the payload size of the output information of the AI ​​model differently based on whether the output information of the AI ​​model is related to performance monitoring of the AI ​​model.

15. Regarding base stations, RF (Radio Frequency) transceiver; and It includes a processor connected to the above RF transceiver, and The above processor controls the RF transceiver to transmit CSI (channel state information) reporting setting information and receives output information of an AI (Artificial Intelligence) model from a UE (User Equipment) that takes the measured CSI based on the CSI reporting setting information as input. A base station comprising information for setting the payload size of the output information of the AI ​​model differently based on whether the output information of the AI ​​model is related to performance monitoring of the AI ​​model, wherein the above CSI report setting information includes information for setting the payload size of the output information of the AI ​​model differently.

Images