Apparatus and method for efficiently operating ai / ML model in wireless communication system

Abstract

According to various embodiments of the present disclosure, an operation method for a first node in a wireless communication system is provided, the method comprising the steps of: receiving at least one synchronization signal from a second node; receiving control information from the second node; transmitting first communication environment data to the second node; receiving, from the second node, model information related to a first secondary artificial intelligence / machine learning (AI / ML) model based on a first sub-feature set related to the first communication environment data; transmitting, to the second node, second communication environment data changed from the first communication environment data; and receiving, from the second node, model update information for a second secondary AI / ML model, which is based on a second sub-feature set related to the second communication environment data and is changed from the first secondary AI / ML model.

Description

Device and method for efficiently operating AI / ML models in wireless communication systems

[0001] The present disclosure relates to a device and method for efficiently operating an AI / ML (artificial intelligence / machine learning) model in a wireless communication system. Specifically, the present disclosure relates to a device and method for dynamically changing an AI / ML model and feature set based on context in a wireless communication system and selecting and transmitting only features appropriate for the context, thereby minimizing overall signaling and resource overhead and efficiently updating the AI / ML model.

[0002] Wireless communication systems are widely deployed to provide diverse communication services, such as voice and data, and attempts to integrate AI into these systems are rapidly increasing. Current AI integration approaches can be broadly categorized into communications for AI (C4AI), which advances communication technologies to support artificial intelligence (AI), and AI for communications (AI4C), which utilizes AI to improve communication performance. In the AI4C space, there are attempts to improve design efficiency by replacing channel encoders / decoders with end-to-end autoencoders. In the C4AI space, federated learning, a distributed learning technique, shares only model weights and gradients with servers without sharing raw device data, protecting privacy while updating a common prediction model. Split inference is also being used to distribute the load across devices, network edges, and cloud servers.

[0003] To solve the above-described problems, the present disclosure provides a device and method for efficiently operating an AI / ML model in a wireless communication system.

[0004] The present disclosure provides a device and method for dynamically changing an AI / ML model and feature set based on context in a wireless communication system and selecting and transmitting only features appropriate for the context, thereby minimizing overall signaling and resource overhead and updating an AI / ML model in an efficient manner.

[0005] The technical problems to be achieved in the present disclosure are not limited to the technical problems mentioned above, and other technical problems not mentioned can be clearly understood by a person having ordinary skill in the technical field to which the present disclosure belongs from the description below.

[0006] According to various embodiments of the present disclosure, a method of operating a first node in a wireless communication system is provided, the method comprising: receiving at least one synchronization signal from a second node; receiving control information from the second node; transmitting first communication environment data to the second node; receiving model information related to a first secondary AI / ML (artificial intelligence / machine learning) model based on a first sub-feature set related to the first communication environment data from the second node; transmitting second communication environment data changed from the first communication environment data to the second node; and receiving model update information from the second node to a second secondary AI / ML model changed from the first secondary AI / ML model based on a second sub-feature set related to the second communication environment data.

[0007] According to various embodiments of the present disclosure, a method for operating a second node in a wireless communication system comprises the steps of: transmitting at least one synchronization signal to a plurality of first nodes within the coverage of the second node; transmitting control information to the plurality of first nodes; transmitting first communication environment data received from the plurality of first nodes to a third node; receiving model information related to a first secondary AI / ML (artificial intelligence / machine learning) model based on a first sub-feature set related to the first communication environment data from the third node; transmitting the model information related to the first secondary AI / ML model to the plurality of first nodes; transmitting second communication environment data received from the plurality of first nodes and changed from the first communication environment data to the third node; A method is provided, comprising: receiving model update information related to a second secondary AI / ML model that is changed from the first secondary AI / ML model based on a second sub-characteristic set related to the second communication environment data from the third node; and transmitting the model update information from the first secondary AI / ML model to the second secondary AI / ML model to the plurality of first nodes.

[0008] According to various embodiments of the present disclosure, in a wireless communication system, a first node is provided, comprising: a transceiver; at least one processor; and at least one memory operably connectable to the at least one processor and storing instructions that, when executed by the at least one processor, perform operations, wherein the operations include all steps of a method of operating the first node according to various embodiments of the present disclosure.

[0009] According to various embodiments of the present disclosure, in a wireless communication system, a second node is provided, comprising: a transceiver; at least one processor; and at least one memory operably connectable to the at least one processor and storing instructions that, when executed by the at least one processor, perform operations, wherein the operations include all steps of a method of operating the second node according to various embodiments of the present disclosure.

[0010] According to various embodiments of the present disclosure, a control device for controlling a first node in a wireless communication system is provided, comprising: at least one processor; and at least one memory operably connected to the at least one processor, wherein the at least one memory stores instructions for performing operations based on being executed by the at least one processor, wherein the operations include all steps of an operating method of the first node according to various embodiments of the present disclosure.

[0011] According to various embodiments of the present disclosure, a control device for controlling a second node in a wireless communication system is provided, comprising: at least one processor; and at least one memory operably connected to the at least one processor, wherein the at least one memory stores instructions for performing operations based on being executed by the at least one processor, wherein the operations include all steps of an operating method of the second node according to various embodiments of the present disclosure.

[0012] According to various embodiments of the present disclosure, one or more non-transitory computer-readable media storing one or more instructions, wherein the one or more instructions, based on being executed by one or more processors, perform operations, the operations comprising all steps of a method of operating a first node according to various embodiments of the present disclosure, are provided.

[0013] According to various embodiments of the present disclosure, there is provided one or more non-transitory computer-readable media storing one or more instructions, wherein the one or more instructions, when executed by one or more processors, perform operations, the operations comprising all steps of a method of operating a second node according to various embodiments of the present disclosure.

[0014] To solve the above-described problems, the present disclosure can provide a device and method for efficiently operating an AI / ML model in a wireless communication system.

[0015] The present disclosure provides a device and method for dynamically changing an AI / ML model and feature set based on context in a wireless communication system and selecting and transmitting only features appropriate for the context, thereby minimizing overall signaling and resource overhead and updating an AI / ML model in an efficient manner.

[0016] The accompanying drawings are intended to aid in understanding the present disclosure and, together with detailed descriptions, may provide embodiments of the present disclosure. However, the technical features of the present disclosure are not limited to specific drawings, and the features disclosed in each drawing may be combined with each other to form new embodiments. Reference numerals in each drawing may indicate structural elements.

[0017] Figure 1 is a diagram illustrating an example of physical channels and general signal transmission used in a 3GPP system.

[0018] Figure 2 is a diagram illustrating the system structure of a New Generation Radio Access Network (NG-RAN).

[0019] Figure 3 is a diagram illustrating the functional division between NG-RAN and 5GC.

[0020] Figure 4 is a diagram illustrating an example of a 5G usage scenario.

[0021] Figure 5 is a diagram illustrating an example of a communication structure that can be provided in a 6G system.

[0022] Figure 6 is a schematic diagram illustrating an example of a perceptron structure.

[0023] Figure 7 is a schematic diagram illustrating an example of a multilayer perceptron structure.

[0024] Figure 8 is a schematic diagram illustrating an example of a deep neural network.

[0025] Figure 9 is a schematic diagram illustrating an example of a convolutional neural network.

[0026] Figure 10 is a schematic diagram illustrating an example of a filter operation in a convolutional neural network.

[0027] Figure 11 is a schematic diagram illustrating an example of a neural network structure in which a recurrent loop exists.

[0028] Figure 12 is a diagram schematically illustrating an example of the operating structure of a recurrent neural network.

[0029] Figure 13 is a diagram illustrating an example of the electromagnetic spectrum.

[0030] Figure 14 is a diagram illustrating an example of a THz communication application.

[0031] Fig. 15 is a diagram illustrating an example of an electronic component-based THz wireless communication transmitter and receiver.

[0032] FIG. 16 is a diagram illustrating an example of a method for generating a THz signal based on an optical element.

[0033] Fig. 17 is a diagram illustrating an example of an optical element-based THz wireless communication transceiver.

[0034] Fig. 18 is a diagram illustrating the structure of a photon source-based transmitter.

[0035] Figure 19 is a drawing showing the structure of an optical modulator.

[0036] FIG. 20 is a diagram illustrating an example of a functional framework for RAN Intelligence in a system applicable to the present disclosure.

[0037] FIG. 21 is a diagram illustrating an example of model training at OAM and model inference at NG-RAN in a system applicable to the present disclosure.

[0038] FIG. 22 is a diagram illustrating an example of model training and model inference at NG-RAN in a system applicable to the present disclosure.

[0039] FIG. 23 is a diagram illustrating an example of model training and model inference both located in RAN nodes in a system applicable to the present disclosure.

[0040] FIG. 24 is a diagram illustrating an example of the process of creating and deploying an AI / ML model in a system applicable to the present disclosure.

[0041] FIG. 25 is a diagram illustrating an example of the operation and prediction process of an AI / ML model in a system applicable to the present disclosure.

[0042] FIG. 26 is a diagram illustrating an example of a deployment process of a primary AI / ML model in a system applicable to the present disclosure.

[0043] Figure 27 is a diagram illustrating an example of a process for deploying a secondary AI / ML model in a system applicable to the present disclosure.

[0044] FIG. 28 is a diagram illustrating an example of the operation and model update process of a secondary AI / ML model in a system applicable to the present disclosure.

[0045] FIG. 29 is a diagram illustrating an example of a functional framework for RAN Intelligence in TR 37.817 in a system applicable to the present disclosure.

[0046] FIG. 30 is a diagram illustrating an example of the operation process of the first node in a system applicable to the present disclosure.

[0047] FIG. 31 is a diagram illustrating an example of the operation process of a second node in a system applicable to the present disclosure.

[0048] FIG. 32 illustrates a communication system (1) applicable to various embodiments of the present disclosure.

[0049] FIG. 33 illustrates a wireless device that can be applied to various embodiments of the present disclosure.

[0050] FIG. 34 illustrates another example of a wireless device that can be applied to various embodiments of the present disclosure.

[0051] Figure 35 illustrates a signal processing circuit for a transmission signal.

[0052] FIG. 36 illustrates another example of a wireless device applicable to various embodiments of the present disclosure.

[0053] FIG. 37 illustrates a mobile device applicable to various embodiments of the present disclosure.

[0054] FIG. 38 illustrates a vehicle or autonomous vehicle applicable to various embodiments of the present disclosure.

[0055] FIG. 39 illustrates a vehicle applicable to various embodiments of the present disclosure.

[0056] FIG. 40 illustrates an XR device applicable to various embodiments of the present disclosure.

[0057] FIG. 41 illustrates a robot applicable to various embodiments of the present disclosure.

[0058] FIG. 42 illustrates an AI device applicable to various embodiments of the present disclosure.

[0059] In various embodiments of the present disclosure, “A or B” may mean “only A,” “only B,” or “both A and B.” In other words, in various embodiments of the present disclosure, “A or B” may be interpreted as “A and / or B.” For example, in various embodiments of the present disclosure, “A, B or C” may mean “only A,” “only B,” “only C,” or “any combination of A, B and C.”

[0060] In various embodiments of the present disclosure, a slash ( / ) or a comma may mean "and / or." For example, "A / B" may mean "A and / or B." Accordingly, "A / B" may mean "only A," "only B," or "both A and B." For example, "A, B, C" may mean "A, B, or C."

[0061] In various embodiments of the present disclosure, “at least one of A and B” may mean “only A,” “only B,” or “both A and B.” Furthermore, in various embodiments of the present disclosure, the expressions “at least one of A or B” or “at least one of A and / or B” may be interpreted as equivalent to “at least one of A and B.”

[0062] Additionally, in various embodiments of the present disclosure, “at least one of A, B and C” can mean “only A,” “only B,” “only C,” or “any combination of A, B and C.” Additionally, “at least one of A, B or C” or “at least one of A, B and / or C” can mean “at least one of A, B and C.”

[0063] Additionally, parentheses used in various embodiments of the present disclosure may mean "for example." Specifically, when indicated as "control information (PDCCH)", "PDCCH" may be proposed as an example of "control information." In other words, "control information" in various embodiments of the present disclosure is not limited to "PDCCH", and "PDDCH" may be proposed as an example of "control information." Furthermore, even when indicated as "control information (i.e., PDCCH)", "PDCCH" may be proposed as an example of "control information."

[0064] Technical features individually described in a single drawing in various embodiments of the present disclosure may be implemented individually or simultaneously.

[0065] The following technologies can be used in various wireless access systems, such as CDMA, FDMA, TDMA, OFDMA, and SC-FDMA. 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) / EDGE (Enhanced Data Rates for GSM Evolution). OFDMA can be implemented using wireless technologies such as IEEE 802.11 (Wi-Fi), IEEE 802.16 (WiMAX), IEEE 802-20, and E-UTRA (Evolved UTRA). UTRA is a part of UMTS (Universal Mobile Telecommunications System). 3GPP (3rd Generation Partnership Project) LTE (Long Term Evolution) is a part of E-UMTS (Evolved UMTS) that uses E-UTRA, and LTE-A (Advanced) / LTE-A pro is an evolved version of 3GPP LTE. 3GPP NR (New Radio or New Radio Access Technology) is an evolved version of 3GPP LTE / LTE-A / LTE-A pro. 3GPP 6G may be an evolved version of 3GPP NR.

[0066] For clarity, the description is based on 3GPP communication systems (e.g., LTE, NR, etc.), but the technical spirit of the present disclosure is not limited thereto. LTE refers to technology after 3GPP TS 36.xxx Release 8. Specifically, LTE technology after 3GPP TS 36.xxx Release 10 is referred to as LTE-A, and LTE technology after 3GPP TS 36.xxx Release 13 is referred to as LTE-A pro. 3GPP NR refers to technology after TS 38.xxx Release 15. 3GPP 6G may refer to technology after TS Release 17 and / or Release 18. “xxx” refers to a standard document detail number. LTE / NR / 6G may be collectively referred to as a 3GPP system. For background technology, terms, abbreviations, etc. used in the description of the present disclosure, reference may be made to matters described in standard documents published prior to the present disclosure. For example, reference may be made to the following documents.

[0067] 3GPP LTE

[0068] - 36.211: Physical channels and modulation

[0069] - 36.212: Multiplexing and channel coding

[0070] - 36.213: Physical layer procedures

[0071] - 36.300: Overall description

[0072] - 36.331: Radio Resource Control (RRC)

[0073] 3GPP NR

[0074] - 38.211: Physical channels and modulation

[0075] - 38.212: Multiplexing and channel coding

[0076] - 38.213: Physical layer procedures for control

[0077] - 38.214: Physical layer procedures for data

[0078] - 38.300: NR and NG-RAN Overall Description

[0079] - 38.331: Radio Resource Control (RRC) protocol specification

[0080] Physical Channel and Frame Structure

[0081] Physical channels and general signal transmission

[0082] Figure 1 is a diagram illustrating an example of physical channels and general signal transmission used in a 3GPP system.

[0083] In a wireless communication system, a terminal receives information from a base station via the downlink (DL) and transmits it to the base station via the uplink (UL). The information transmitted and received between the base station and the terminal includes data and various control information, and various physical channels exist depending on the type and purpose of the information being transmitted and received.

[0084] When a terminal is powered on or enters a new cell, it performs an initial cell search operation, such as synchronizing with the base station (S11). To this end, the terminal receives a Primary Synchronization Signal (PSS) and a Secondary Synchronization Signal (SSS) from the base station to synchronize with the base station and obtain information such as a cell ID. Afterwards, the terminal can receive a Physical Broadcast Channel (PBCH) from the base station to obtain broadcast information within the cell. Meanwhile, the terminal can receive a Downlink Reference Signal (DL RS) during the initial cell search phase to check the downlink channel status.

[0085] A terminal that has completed initial cell search can obtain more specific system information by receiving a physical downlink control channel (PDCCH) and a physical downlink shared channel (PDSCH) based on information contained in the PDCCH (S12).

[0086] Meanwhile, when accessing a base station for the first time or when there are no radio resources for signal transmission, the terminal may perform a random access procedure (RACH) for the base station (S13 to S16). To this end, the terminal may transmit a specific sequence as a preamble via a physical random access channel (PRACH) (S13 and S15) and receive a response message (RAR (Random Access Response) message) to the preamble via a PDCCH and a corresponding PDSCH. In the case of a contention-based RACH, a contention resolution procedure may additionally be performed (S16).

[0087] The terminal that has performed the procedure described above can then perform PDCCH / PDSCH reception (S17) and physical uplink shared channel (PUSCH) / physical uplink control channel (PUCCH) transmission (S18) as general uplink / downlink signal transmission procedures. In particular, the terminal can receive downlink control information (DCI) through the PDCCH. Here, the DCI includes control information such as resource allocation information for the terminal, and different formats can be applied depending on the purpose of use.

[0088] Meanwhile, the control information that the terminal transmits to the base station via the uplink or that the terminal receives from the base station may include downlink / uplink ACK / NACK signals, CQI (Channel Quality Indicator), PMI (Precoding Matrix Index), RI (Rank Indicator), etc. The terminal may transmit the above-described control information such as CQI / PMI / RI via PUSCH and / or PUCCH.

[0089] Structure of uplink and downlink channels

[0090] Downlink channel structure

[0091] The base station transmits a related signal to the terminal through a downlink channel described below, and the terminal receives the related signal from the base station through a downlink channel described below.

[0092] (1) Physical Downlink Shared Channel (PDSCH)

[0093] PDSCH carries downlink data (e.g., DL-shared channel transport block, DL-SCH TB) and applies modulation methods such as Quadrature Phase Shift Keying (QPSK), 16 Quadrature Amplitude Modulation (QAM), 64 QAM, and 256 QAM. Codewords are generated by encoding the TBs. PDSCH can carry multiple codewords. Scrambling and modulation mapping are performed for each codeword, and modulation symbols generated from each codeword are mapped to one or more layers (Layer mapping). Each layer is mapped to resources along with a Demodulation Reference Signal (DMRS), generated as an OFDM symbol signal, and transmitted through the corresponding antenna port.

[0094] (2) Physical downlink control channel (PDCCH)

[0095] The PDCCH carries downlink control information (DCI) and employs modulation methods such as QPSK. A PDCCH consists of 1, 2, 4, 8, or 16 Control Channel Elements (CCEs), depending on the Aggregation Level (AL). Each CCE is comprised of six Resource Element Groups (REGs). Each REG is defined by one OFDM symbol and one (P)RB.

[0096] The UE obtains DCI transmitted via the PDCCH by performing decoding (also known as blind decoding) on ​​a set of PDCCH candidates. The set of PDCCH candidates decoded by the UE is defined as a PDCCH search space set. The search space set may be a common search space or a UE-specific search space. The UE can obtain DCI by monitoring PDCCH candidates within one or more search space sets established by the MIB or higher layer signaling.

[0097] Uplink channel structure

[0098] The terminal transmits a related signal to the base station through the uplink channel described below, and the base station receives the related signal from the terminal through the uplink channel described below.

[0099] (1) Physical Uplink Shared Channel (PUSCH)

[0100] PUSCH carries uplink data (e.g., UL-shared channel transport block, UL-SCH TB) and / or uplink control information (UCI), and is transmitted based on a CP-OFDM (Cyclic Prefix - Orthogonal Frequency Division Multiplexing) waveform, a DFT-s-OFDM (Discrete Fourier Transform - spread - Orthogonal Frequency Division Multiplexing) waveform, etc. When the PUSCH is transmitted based on a DFT-s-OFDM waveform, the UE transmits the PUSCH by applying transform precoding. For example, when transform precoding is disabled (e.g., transform precoding is disabled), the UE transmits the PUSCH based on the CP-OFDM waveform, and when transform precoding is enabled (e.g., transform precoding is enabled), the UE can transmit the PUSCH based on the CP-OFDM waveform or the DFT-s-OFDM waveform. PUSCH transmissions can be dynamically scheduled by UL grants in DCI, or semi-statically scheduled (configured grant) based on higher layer (e.g., RRC) signaling (and / or Layer 1 (L1) signaling (e.g., PDCCH)). PUSCH transmissions can be performed in a codebook-based or non-codebook-based manner.

[0101] (2) Physical Uplink Control Channel (PUCCH)

[0102] PUCCH carries uplink control information, HARQ-ACK and / or scheduling request (SR), and can be divided into multiple PUCCHs depending on the PUCCH transmission length.

[0103] Below, we describe new radio access technology (new RAT, NR).

[0104] As more and more communication devices demand greater communication capacity, the need for improved mobile broadband communication compared to existing radio access technology (RAT) is emerging. Furthermore, massive Machine Type Communications (MTC), which connects numerous devices and objects to provide various services anytime, anywhere, is also a key issue to be considered in next-generation communication. Furthermore, communication system design that considers reliability and latency-sensitive services / terminals is being discussed. The introduction of next-generation radio access technologies that take into account enhanced mobile broadband communication, massive MTC, and URLLC (Ultra-Reliable and Low Latency Communication) is being discussed, and in various embodiments of the present disclosure, these technologies are conveniently referred to as new RAT or NR.

[0105] Figure 2 is a diagram illustrating the system structure of a New Generation Radio Access Network (NG-RAN).

[0106] Referring to FIG. 2, the NG-RAN may include a gNB and / or an eNB that provides user plane and control plane protocol termination to the UE. FIG. 1 illustrates a case where only a gNB is included. The gNB and eNB are connected to each other via an Xn interface. The gNB and eNB are connected to the 5th generation core network (5G Core Network: 5GC) via the NG interface. More specifically, the gNB is connected to the access and mobility management function (AMF) via the NG-C interface, and the gNB is connected to the user plane function (UPF) via the NG-U interface.

[0107] Figure 3 is a diagram illustrating the functional division between NG-RAN and 5GC.

[0108] Referring to FIG. 3, the gNB can provide functions such as inter-cell radio resource management (Inter Cell RRM), radio bearer management (RB control), connection mobility control (Connection Mobility Control), radio admission control (Radio Admission Control), measurement configuration and provision, and dynamic resource allocation. The AMF can provide functions such as NAS security and idle state mobility processing. The UPF can provide functions such as mobility anchoring and PDU processing. The SMF (Session Management Function) can provide functions such as terminal IP address allocation and PDU session control.

[0109] Figure 4 is a diagram illustrating an example of a 5G usage scenario.

[0110] The 5G usage scenario illustrated in FIG. 4 is merely exemplary, and the technical features of various embodiments of the present disclosure can also be applied to other 5G usage scenarios not illustrated in FIG. 4.

[0111] Referring to Figure 4, the three key requirement areas for 5G include (1) enhanced mobile broadband (eMBB), (2) massive machine type communication (mMTC), and (3) ultra-reliable and low latency communications (URLLC). Some use cases may require optimization across multiple areas, while others may focus on just one key performance indicator (KPI). 5G supports these diverse use cases in a flexible and reliable manner.

[0112] eMBB focuses on improving data speeds, latency, user density, and overall capacity and coverage of mobile broadband connections. It targets throughputs of around 10 Gbps. eMBB significantly exceeds basic mobile internet access, enabling rich interactive experiences, media and entertainment applications in the cloud, and augmented reality. Data is a key driver of 5G, and for the first time, dedicated voice services may not be available in the 5G era. In 5G, voice is expected to be handled as an application, simply using the data connection provided by the communication system. The increased traffic volume is primarily due to the increasing content size and the growing number of applications that require high data rates. Streaming services (audio and video), interactive video, and mobile internet connectivity will become more prevalent as more devices connect to the internet. Many of these applications require always-on connectivity to push real-time information and notifications to users. Cloud storage and applications are rapidly growing on mobile communication platforms, and this can be applied to both work and entertainment. Cloud storage is a particular use case driving the growth of uplink data rates. 5G is also used for remote work in the cloud, requiring significantly lower end-to-end latency to maintain a superior user experience when tactile interfaces are used. In entertainment, for example, cloud gaming and video streaming are other key factors driving the demand for mobile broadband. Entertainment is essential on smartphones and tablets, regardless of location, including in highly mobile environments like trains, cars, and airplanes. Another use case is augmented reality and information retrieval for entertainment, where augmented reality requires extremely low latency and instantaneous data volumes.

[0113] mMTC is designed to enable communication between a large number of low-cost, battery-powered devices, supporting applications such as smart metering, logistics, field, and body sensors. mMTC targets a battery life of approximately 10 years and / or a population of approximately 1 million devices per square kilometer. mMTC enables seamless connectivity of embedded sensors across all sectors and is one of the most anticipated 5G use cases. The number of IoT devices is projected to reach 20.4 billion by 2020. Industrial IoT is one area where 5G will play a key role, enabling smart cities, asset tracking, smart utilities, agriculture, and security infrastructure.

[0114] URLLC is ideal for vehicle communications, industrial control, factory automation, remote surgery, smart grids, and public safety applications by enabling devices and machines to communicate with high reliability, very low latency, and high availability. URLLC targets latency on the order of 1 ms. URLLC encompasses new services that will transform industries through ultra-reliable, low-latency links, such as remote control of critical infrastructure and autonomous vehicles. This level of reliability and latency is essential for smart grid control, industrial automation, robotics, and drone control and coordination.

[0115] Next, we will look more specifically at a number of usage examples included within the triangle in Fig. 4.

[0116] 5G can complement fiber-to-the-home (FTTH) and cable-based broadband (or DOCSIS) by delivering streams rated at hundreds of megabits per second to gigabits per second. These high speeds may be required to deliver TV at resolutions beyond 4K (6K, 8K, and beyond), as well as virtual reality (VR) and augmented reality (AR). VR and AR applications include near-immersive sports events. Certain applications may require specialized network configurations. For example, for VR gaming, a gaming company may need to integrate its core servers with a network operator's edge network servers to minimize latency.

[0117] Automotive is expected to be a significant new driver for 5G, with numerous use cases for in-vehicle mobile communications. For example, passenger entertainment demands both high capacity and high mobile broadband, as future users will consistently expect high-quality connectivity regardless of their location and speed. Another automotive application is augmented reality dashboards. An AR dashboard allows drivers to identify objects in the dark on top of what they see through the windshield. The AR dashboard overlays information to inform the driver about the distance and movement of objects. In the future, wireless modules will enable vehicle-to-vehicle communication, information exchange between vehicles and supporting infrastructure, and information exchange between vehicles and other connected devices (e.g., devices accompanying pedestrians). Safety systems can guide drivers to safer driving behaviors, reducing the risk of accidents. The next step will be remotely controlled or autonomous vehicles, which require highly reliable and fast communication between different autonomous vehicles and / or between vehicles and infrastructure. In the future, autonomous vehicles will perform all driving tasks, leaving drivers to focus solely on traffic anomalies that the vehicle itself cannot detect. The technological requirements for autonomous vehicles will require ultra-low latency and ultra-high-speed reliability, increasing traffic safety to levels unattainable by humans.

[0118] Smart cities and smart homes, often referred to as smart societies, will be embedded with dense wireless sensor networks. A distributed network of intelligent sensors will identify conditions for cost- and energy-efficient maintenance of cities or homes. Similar setups can be implemented for individual homes. Temperature sensors, window and heating controllers, burglar alarms, and appliances will all be wirelessly connected. Many of these sensors typically require low data rates, low power, and low cost. However, for example, real-time HD video may be required from certain types of devices for surveillance purposes.

[0119] The consumption and distribution of energy, including heat and gas, are becoming increasingly decentralized, requiring automated control of distributed sensor networks. Smart grids interconnect these sensors using digital information and communication technologies to collect and act on information. This information can include the behavior of suppliers and consumers, enabling smart grids to improve efficiency, reliability, economic efficiency, sustainable production, and the automated distribution of fuels like electricity. Smart grids can also be viewed as another low-latency sensor network.

[0120] The health sector has numerous applications that can benefit from mobile communications. Telecommunications systems can support telemedicine, which provides clinical care in remote locations. This can help reduce distance barriers and improve access to health services that are otherwise unavailable in remote rural areas. It can also be used to save lives in critical care and emergency situations. Mobile-based wireless sensor networks can provide remote monitoring and sensors for parameters such as heart rate and blood pressure.

[0121] Wireless and mobile communications are becoming increasingly important in industrial applications. Wiring is expensive to install and maintain. Therefore, the potential to replace cables with reconfigurable wireless links presents an attractive opportunity for many industries. However, achieving this requires wireless connections to operate with similar latency, reliability, and capacity to cables, while simplifying their management. Low latency and extremely low error rates are new requirements for 5G connectivity.

[0122] Logistics and freight tracking are important use cases for mobile communications, enabling the tracking of inventory and packages anywhere using location-based information systems. Logistics and freight tracking typically require low data rates but may require wide-range and reliable location information.

[0123] Below, examples of next-generation communications (e.g., 6G) that can be applied to various embodiments of the present disclosure will be described.

[0124] 6G system in general

[0125] The 6G (wireless communication) system aims to achieve (i) very high data rates per device, (ii) a very large number of connected devices, (iii) global connectivity, (iv) very low latency, (v) low energy consumption for battery-free IoT devices, (vi) ultra-reliable connectivity, and (vii) connected intelligence with machine learning capabilities. The vision of the 6G system can be divided into four aspects: intelligent connectivity, deep connectivity, holographic connectivity, and ubiquitous connectivity, and the 6G system can satisfy the requirements as shown in Table 1 below. In other words, Table 1 is a table showing an example of the requirements of a 6G system.

[0126] Per device peak data rate1TbpsE2E latency1msMaximum spectral efficiency100bps / HzMobility supportUp to 1000km / hrSatellite integrationFullyAIFullyAutonomous vehicleFullyXRFullyHaptic CommunicationFully

[0127] 6G systems can have key factors such as enhanced mobile broadband (eMBB), ultra-reliable low latency communications (URLLC), massive machine-type communication (mMTC), AI integrated communication, tactile internet, high throughput, high network capacity, high energy efficiency, low backhaul and access network congestion, and enhanced data security.

[0128] Figure 5 is a diagram illustrating an example of a communication structure that can be provided in a 6G system.

[0129] 6G systems are expected to have 50 times the simultaneous wireless connectivity of 5G systems. URLLC, a key feature of 5G, will become even more crucial in 6G communications by providing end-to-end latency of less than 1 ms. 6G systems will have significantly higher volumetric spectral efficiency, compared to the commonly used area spectral efficiency. 6G systems can offer extremely long battery life and advanced battery technologies for energy harvesting, eliminating the need for separate charging for mobile devices in 6G systems. New network characteristics in 6G may include:

[0130] - Satellite integrated network: 6G is expected to integrate with satellites to provide a global mobile network. The integration of terrestrial, satellite, and airborne networks into a single wireless communications system is crucial for 6G.

[0131] Connected Intelligence: Unlike previous generations of wireless communication systems, 6G is revolutionary, upgrading the wireless evolution from "connected objects" to "connected intelligence." AI can be applied at every stage of the communication process (or at every signal processing step, as described below).

[0132] - Seamless integration of wireless information and energy transfer: 6G wireless networks will transfer power to charge the batteries of devices such as smartphones and sensors. Therefore, wireless information and energy transfer (WIET) will be integrated.

[0133] - Ubiquitous super 3D connectivity: Access to networks and core network functions of drones and very low Earth orbit satellites will create super 3D connectivity in 6G ubiquitous.

[0134] Some general requirements for the new network characteristics of 6G, such as the above, may be as follows:

[0135] - Small cell networks: The concept of small cell networks was introduced to improve received signal quality in cellular systems by increasing throughput, energy efficiency, and spectral efficiency. Consequently, small cell networks are essential for 5G and beyond-5G (5GB) communication systems. Accordingly, 6G communication systems also adopt the characteristics of small cell networks.

[0136] Ultra-dense heterogeneous networks: Ultra-dense heterogeneous networks will be another key feature of 6G communication systems. Multi-tier networks comprised of heterogeneous networks improve overall QoS and reduce costs.

[0137] High-capacity backhaul: Backhaul connections are characterized by high-capacity backhaul networks to support high-volume traffic. High-speed fiber optics and free-space optics (FSO) systems may be potential solutions to this problem.

[0138] - Radar technology integrated with mobile technology: High-precision localization (or location-based services) through communications is a key feature of 6G wireless communication systems. Therefore, radar systems will be integrated with 6G networks.

[0139] - Softwarization and virtualization: Softwarization and virtualization are two critical features that form the foundation of the design process for 5GB networks to ensure flexibility, reconfigurability, and programmability. Furthermore, billions of devices can be shared on a shared physical infrastructure.

[0140] Core implementation technology of 6G systems

[0141] Artificial Intelligence

[0142] The most crucial and newly introduced technology for 6G systems is AI. 4G systems did not involve AI. 5G systems will support partial or very limited AI. However, 6G systems will fully support AI for automation. Advances in machine learning will create more intelligent networks for real-time communications in 6G. Incorporating AI into communications can streamline and improve real-time data transmission. AI can use numerous analyses to determine how complex target tasks should be performed. In other words, AI can increase efficiency and reduce processing delays.

[0143] Time-consuming tasks such as handover, network selection, and resource scheduling can be performed instantly using AI. AI can also play a crucial role in machine-to-machine (M2M), machine-to-human, and human-to-machine communications. Furthermore, AI can facilitate rapid communication in brain-computer interfaces (BCIs). 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.

[0144] Recent attempts to integrate AI into wireless communication systems have focused on the application layer, network layer, and especially deep learning in wireless resource management and allocation. However, this research is increasingly evolving to the MAC layer and physical layer, with attempts to combine deep learning with wireless transmission, particularly at the physical layer. AI-based physical layer transmission refers to the application of AI-based signal processing and communication mechanisms, rather than traditional communication frameworks, in the fundamental signal processing and communication mechanisms. For example, this may include deep learning-based channel coding and decoding, deep learning-based signal estimation and detection, deep learning-based MIMO mechanisms, and AI-based resource scheduling and allocation.

[0145] Machine learning can be used for channel estimation and channel tracking, as well as for power allocation and interference cancellation in the physical layer of the downlink (DL). Furthermore, machine learning can be used for antenna selection, power control, and symbol detection in MIMO systems.

[0146] However, the application of DNN for transmission at the physical layer may have the following problems.

[0147] Deep learning-based AI algorithms require a large amount of training data to optimize training parameters. However, due to limitations in obtaining training data from specific channel environments, a large amount of training data is used offline. This means that static training on training data in specific channel environments can lead to conflicts with the dynamic characteristics and diversity of the wireless channel.

[0148] Furthermore, current deep learning primarily targets real-world signals. However, signals at the physical layer of wireless communications are complex signals. Further research is needed on neural networks that detect complex-domain signals to match the characteristics of wireless communication signals.

[0149] Below, we will look at machine learning in more detail.

[0150] Machine learning refers to a series of operations that train machines to perform tasks that humans can or cannot perform. Machine learning requires data and a learning model. Data learning methods in machine learning can be broadly categorized into three types: supervised learning, unsupervised learning, and reinforcement learning.

[0151] Neural network training aims to minimize output errors. It involves repeatedly inputting training data into a neural network, calculating the neural network output and target error for the training data, and backpropagating the neural network error from the output layer to the input layer to update the weights of each node in the neural network to reduce the error.

[0152] Supervised learning uses labeled training data, while unsupervised learning may not have labeled training data. For example, in the case of supervised learning for data classification, the training data may be data in which each training data category is labeled. Labeled training data is input to a neural network, and the error can be calculated by comparing the output (categories) of the neural network with the training data labels. The calculated error is backpropagated through the neural network in the backward direction (i.e., from the output layer to the input layer), and the connection weights of each node in each layer of the neural network can be updated through backpropagation. The amount of change in the connection weights of each updated node can be determined by the learning rate. The neural network's calculation of the input data and the backpropagation of the error can constitute a learning cycle (epoch). The learning rate can be applied differently depending on the number of iterations of the neural network's learning cycle. For example, in the early stages of training a neural network, a high learning rate can be used to quickly allow the network to reach a certain level of performance, thereby improving efficiency. In the later stages of training, a low learning rate can be used to improve accuracy.

[0153] Learning methods may vary depending on the characteristics of the data. For example, if the goal is to accurately predict data transmitted by a transmitter in a communication system, supervised learning is preferable to unsupervised learning or reinforcement learning.

[0154] The learning model corresponds to the human brain, and the most basic linear model can be thought of, but the machine learning paradigm that uses highly complex neural network structures, such as artificial neural networks, as learning models is called deep learning.

[0155] The neural network cores used in learning methods are mainly divided into deep neural networks (DNN), convolutional deep neural networks (CNN), and recurrent boltzmann machines (RNN).

[0156] An artificial neural network is an example of a network of multiple perceptrons.

[0157] Figure 6 is a schematic diagram illustrating an example of a perceptron structure.

[0158] Referring to Fig. 6, when an input vector x=(x1,x2,...,xd) is input, the entire process of multiplying each component by a weight (W1,W2,...,Wd), adding up all the results, and then applying the activation function σ(·) is called a perceptron. A large-scale artificial neural network structure can extend the simplified perceptron structure illustrated in Fig. 6 to apply the input vector to perceptrons of different dimensions. For convenience of explanation, input values ​​or output values ​​are called nodes.

[0159] Meanwhile, the perceptron structure illustrated in Fig. 6 can be explained as consisting of a total of three layers based on input and output values. An artificial neural network in which there are H perceptrons of (d+1) dimensions between the 1st layer and the 2nd layer, and K perceptrons of (H+1) dimensions between the 2nd layer and the 3rd layer can be expressed as in Fig. 7.

[0160] Figure 7 is a schematic diagram illustrating an example of a multilayer perceptron structure.

[0161] The layer where the input vector is located is called the input layer, the layer where the final output value is located is called the output layer, and all layers located between the input layer and the output layer are called hidden layers. The example in Fig. 7 shows three layers, but when counting the number of layers in an actual artificial neural network, the input layer is excluded, so it can be viewed as a total of two layers. An artificial neural network is composed of perceptrons, which are basic blocks, connected in two dimensions.

[0162] The aforementioned input, hidden, and output layers can be applied jointly not only to multilayer perceptrons but also to various artificial neural network structures, such as CNNs and RNNs, which will be described later. The greater the number of hidden layers, the deeper the artificial neural network. The machine learning paradigm that uses sufficiently deep artificial neural networks as learning models is called deep learning. Furthermore, the artificial neural network used for deep learning is called a deep neural network (DNN).

[0163] Figure 8 is a schematic diagram illustrating an example of a deep neural network.

[0164] The deep neural network illustrated in Figure 8 is a multilayer perceptron consisting of eight hidden layers and eight output layers. The multilayer perceptron structure is referred to as a fully connected neural network. In a fully connected neural network, there is no connection between nodes located in the same layer, and there is a connection only between nodes located in adjacent layers. DNN has a fully connected neural network structure and is composed of a combination of multiple hidden layers and activation functions, and can be usefully applied to identify correlation characteristics between inputs and outputs. Here, the correlation characteristic can mean the joint probability of inputs and outputs.

[0165] Meanwhile, depending on how multiple perceptrons are connected to each other, various artificial neural network structures different from the aforementioned DNN can be formed.

[0166] Figure 9 is a schematic diagram illustrating an example of a convolutional neural network.

[0167] In DNN, nodes within a single layer are arranged vertically in a one-dimensional manner. However, Fig. 9 can assume a case where nodes are arranged two-dimensionally, with w nodes in width and h nodes in height (the convolutional neural network structure of Fig. 9). In this case, since a weight is added to each connection in the connection process from one input node to the hidden layer, a total of hΥw weights must be considered. Since there are hΥw nodes in the input layer, a total of h2w2 weights are required between two adjacent layers.

[0168] The convolutional neural network of Fig. 9 has a problem in that the number of weights increases exponentially according to the number of connections. Therefore, instead of considering the connections of all modes between adjacent layers, it assumes that there are small filters, and performs weighted sum and activation function operations on the overlapping portions of the filters, as in Fig. 10.

[0169] Figure 10 is a schematic diagram illustrating an example of a filter operation in a convolutional neural network.

[0170] Each filter has a weight corresponding to the number of its size, and weight learning can be performed so that a specific feature on the image can be extracted as a factor and output. In Fig. 10, a filter of size 3Y3 is applied to the upper left 3Y3 region of the input layer, and the output value resulting from performing weighted sum and activation function operations on the corresponding node is stored in z22.

[0171] The above filter performs weighted sum and activation function operations while moving at a certain horizontal and vertical interval while scanning the input layer, and places the output value at the current filter position. This operation method is similar to the convolution operation for images in the field of computer vision, so a deep neural network with this structure is called a convolutional neural network (CNN), and the hidden layer generated as a result of the convolution operation is called a convolutional layer. In addition, a neural network with multiple convolutional layers is called a deep convolutional neural network (DCNN).

[0172] In the convolutional layer, the number of weights can be reduced by calculating a weighted sum that includes only the nodes located in the area covered by the filter, starting from the node where the current filter is located. This allows a single filter to focus on features within a local area. Accordingly, CNNs can be effectively applied to image data processing where physical distance in a two-dimensional area is an important criterion for judgment. Meanwhile, CNNs can apply multiple filters immediately before the convolutional layer, and can generate multiple output results through the convolution operation of each filter.

[0173] Meanwhile, depending on the data properties, there may be data for which sequence characteristics are important. Considering the length variability and chronological relationship of such sequence data, a structure that applies a method of inputting one element of the data sequence at each timestep and inputting the output vector (hidden vector) of the hidden layer output at a specific timestep together with the immediately following element in the sequence is called a recurrent neural network structure.

[0174] Figure 11 is a schematic diagram illustrating an example of a neural network structure in which a recurrent loop exists.

[0175] Referring to Figure 11, a recurrent neural network (RNN) is a structure that inputs elements (x1(t), x2(t), ,..., xd(t)) of a data sequence at a time point t into a fully connected neural network, and then inputs the hidden vectors (z1(t-1), z2(t-1),..., zH(t-1)) of the immediately preceding time point t-1 together and applies a weighted sum and activation function. The reason for transmitting the hidden vector to the next time point in this way is because the information in the input vectors of the preceding time points is considered to be accumulated in the hidden vector of the current time point.

[0176] Figure 12 is a diagram schematically illustrating an example of the operating structure of a recurrent neural network.

[0177] Referring to Figure 12, the recurrent neural network operates in a predetermined order of time for the input data sequence.

[0178] When the input vector (x1(t), x2(t), ,..., xd(t)) at time point 1 is input to the recurrent neural network, the hidden vector (z1(1), z2(1),..., zH(1)) is input together with the input vector (x1(2), x2(2),..., xd(2)) at time point 2, and the vector (z1(2), z2(2),..., zH(2)) of the hidden layer is determined through a weighted sum and an activation function. This process is repeatedly performed until time points 2, 3, ,,, T.

[0179] Meanwhile, when multiple hidden layers are placed within a recurrent neural network, it is called a deep recurrent neural network (DRNN). Recurrent neural networks are designed to be useful for processing sequence data (e.g., natural language processing).

[0180] It is a neural network core used in a learning manner, and includes various deep learning techniques such as DNN, CNN, RNN, Restricted Boltzmann Machine (RBM), Deep Belief Network (DBN), and Deep Q-Network, and can be applied to fields such as computer vision, speech recognition, natural language processing, and speech / signal processing.

[0181] Recent attempts to integrate AI into wireless communication systems have focused on the application layer, network layer, and especially deep learning in wireless resource management and allocation. However, this research is increasingly evolving to the MAC layer and physical layer, with attempts to combine deep learning with wireless transmission, particularly at the physical layer. AI-based physical layer transmission refers to the application of AI-based signal processing and communication mechanisms, rather than traditional communication frameworks, in the fundamental signal processing and communication mechanisms. For example, this may include deep learning-based channel coding and decoding, deep learning-based signal estimation and detection, deep learning-based MIMO mechanisms, and AI-based resource scheduling and allocation.

[0182] THz (Terahertz) communication

[0183] Data rates can be increased by increasing bandwidth. This can be achieved by utilizing sub-THz communications with wide bandwidths and applying advanced massive MIMO technology. THz waves, also known as sub-millimeter waves, typically refer to the frequency range between 0.1 THz and 10 THz, with corresponding wavelengths ranging from 0.03 mm to 3 mm. The 100 GHz to 300 GHz band (sub-THz band) is considered a key part of the THz band for cellular communications. Adding the sub-THz band to the mmWave band will increase the capacity of 6G cellular communications. Among the defined THz bands, 300 GHz to 3 THz lies in the far infrared (IR) frequency band. While part of the optical band, the 300 GHz to 3 THz band lies at the boundary of the optical band, immediately following the RF band. Therefore, this 300 GHz to 3 THz band exhibits similarities to RF.

[0184] Figure 13 is a diagram illustrating an example of the electromagnetic spectrum.

[0185] Key characteristics of THz communications include (i) the widely available bandwidth to support very high data rates and (ii) the high path loss that occurs at high frequencies (requiring highly directional antennas). The narrow beamwidths generated by highly directional antennas reduce interference. The small wavelength of THz signals allows for a significantly larger number of antenna elements to be integrated into devices and base stations operating in this band. This enables the use of advanced adaptive array technologies to overcome range limitations.

[0186] Optical wireless technology

[0187] OWC technology is designed for 6G communications, in addition to RF-based communications for all possible device-to-access networks. These networks connect to network-to-backhaul / fronthaul networks. OWC technology has already been used in 4G communication systems, but it will be used more widely to meet the demands of 6G communication systems. OWC technologies such as light fidelity, visible light communication, optical camera communication, and wideband-based FSO communication are already well-known. Communications based on optical wireless technology can provide very high data rates, low latency, and secure communications. LiDAR can also be used for ultra-high-resolution 4D mapping in 6G communications based on wideband.

[0188] FSO backhaul network

[0189] The transmitter and receiver characteristics of an FSO system are similar to those of a fiber-optic network. Therefore, data transmission in an FSO system is similar to that of a fiber-optic system. Therefore, FSO can be a promising technology for providing backhaul connectivity in 6G systems, in conjunction with fiber-optic networks. Using FSO, ultra-long-distance communications are possible, even over distances exceeding 10,000 km. FSO supports high-capacity backhaul connectivity for remote and non-remote areas, such as the ocean, space, underwater, and isolated islands. FSO also supports cellular base station (BS) connections.

[0190] Massive MIMO technology

[0191] One of the key technologies for improving spectral efficiency is the application of MIMO technology. As MIMO technology improves, spectral efficiency also improves. Therefore, massive MIMO technology will be crucial in 6G systems. Because MIMO technology utilizes multiple paths, multiplexing technology must be considered to ensure that data signals can be transmitted along more than one path, as well as beam generation and operation technologies suitable for the THz band.

[0192] Blockchain

[0193] Blockchain will become a crucial technology for managing massive amounts of data in future communication systems. Blockchain is a form of distributed ledger technology. A distributed ledger is a database distributed across numerous nodes or computing devices. Each node replicates and stores an identical copy of the ledger. Blockchains are managed by a peer-to-peer network and can exist without being managed by a central authority or server. Data on a blockchain is collected and organized into blocks. Blocks are linked together and protected using cryptography. Blockchain perfectly complements large-scale IoT with its inherently enhanced interoperability, security, privacy, reliability, and scalability. Therefore, blockchain technology offers several features, such as interoperability between devices, traceability of large amounts of data, autonomous interaction with other IoT systems, and the massive connectivity stability of 6G communication systems.

[0194] 3D networking

[0195] 6G systems integrate terrestrial and airborne networks to support vertically expanded user communications. 3D BS will be provided via low-orbit satellites and UAVs. Adding a new dimension in altitude and associated degrees of freedom, 3D connections differ significantly from existing 2D networks.

[0196] Quantum communication

[0197] Unsupervised reinforcement learning holds promise in the context of 6G networks. Supervised learning approaches cannot label the massive amounts of data generated by 6G networks. Unsupervised learning does not require labeling. Therefore, this technology can be used to autonomously build representations of complex networks. Combining reinforcement learning and unsupervised learning allows for truly autonomous network operation.

[0198] drone

[0199] Unmanned Aerial Vehicles (UAVs), or drones, will be a key element in 6G wireless communications. In most cases, high-speed wireless connections will be provided using UAV technology. BS entities are installed on UAVs to provide cellular connectivity. UAVs offer specific capabilities not found in fixed BS infrastructure, such as easy deployment, robust line-of-sight links, and controlled mobility. During emergencies such as natural disasters, deploying terrestrial communication infrastructure is not economically feasible, and sometimes, volatile environments make it impossible to provide services. UAVs can easily handle these situations. UAVs will become a new paradigm in wireless communications. This technology facilitates three fundamental requirements for wireless networks: enhanced mobile broadband (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 important technologies for 6G communications.

[0200] Cell-free Communication

[0201] Tight integration of multiple frequencies and heterogeneous communication technologies is crucial in 6G systems. As a result, users will be able to seamlessly move from one network to another without requiring any manual configuration on their devices. The best network will be automatically selected from available communication technologies. This will break the limitations of the cell concept in wireless communications. Currently, user movement from one cell to another in dense networks results in excessive handovers, resulting in handover failures, handover delays, data loss, and a ping-pong effect. 6G cell-free communications will overcome all of these challenges and provide better QoS. Cell-free communications will be achieved through multi-connectivity and multi-tier hybrid technologies, as well as heterogeneous radios on devices.

[0202] Integration of wireless information and energy transmission

[0203] WIET uses the same fields and waves as wireless communication systems. Specifically, sensors and smartphones will be charged using wireless power transfer during communication. WIET is a promising technology for extending the life of battery-powered wireless systems. Therefore, battery-less devices will be supported by 6G communications.

[0204] Integration of sensing and communication

[0205] Autonomous wireless networks are capable of continuously sensing dynamically changing environmental conditions and exchanging information between different nodes. In 6G, sensing will be tightly integrated with communications to support autonomous systems.

[0206] Integration of Access Backhaul Networks

[0207] In 6G, the density of access networks will be enormous. Each access network will be connected to backhaul connections, such as fiber optics and FSO networks. To cope with the enormous number of access networks, there will be tight integration between access and backhaul networks.

[0208] Holographic beam forming

[0209] Beamforming is a signal processing procedure that adjusts an antenna array to transmit a wireless signal in a specific direction. It is a subset of smart antennas or advanced antenna systems. Beamforming technology offers several advantages, including high signal-to-noise ratio, interference avoidance and rejection, and high network efficiency. Holographic beamforming (HBF) is a novel beamforming method that differs significantly from MIMO systems because it uses software-defined antennas. HBF will be a highly effective approach for efficient and flexible signal transmission and reception in multi-antenna communication devices in 6G.

[0210] Big data analysis

[0211] Big data analytics is a complex process for analyzing diverse, large-scale data sets, or "big data." This process uncovers hidden data, unknown correlations, and customer trends, ensuring complete data management. Big data is collected from various sources, such as video, social networks, images, and sensors. This technology is widely used to process massive amounts of data in 6G systems.

[0212] Large Intelligent Surface (LIS)

[0213] THz-band signals have strong linearity, which can create many shadow areas due to obstacles. LIS technology, which enables expanded communication coverage, enhanced communication stability, and additional value-added services by installing LIS near these shadow areas, is becoming increasingly important. LIS is an artificial surface made of electromagnetic materials that can alter the propagation of incoming and outgoing radio waves. While LIS can be viewed as an extension of massive MIMO, it differs from massive MIMO in its array structure and operating mechanism. Furthermore, LIS operates as a reconfigurable reflector with passive elements, passively reflecting signals without using active RF chains, which offers the advantage of low power consumption. Furthermore, because each passive reflector in LIS must independently adjust the phase shift of the incoming signal, this can be advantageous for wireless communication channels. By appropriately adjusting the phase shift via the LIS controller, the reflected signal can be collected at the target receiver to boost the received signal power.

[0214] Terahertz (THz) wireless communications in general

[0215] THz wireless communication uses THz waves with a frequency of approximately 0.1 to 10 THz (1 THz = 1012 Hz), and can refer to terahertz (THz) band wireless communication using a very high carrier frequency of 100 GHz or higher. THz waves are located between the RF (Radio Frequency) / millimeter (mm) and infrared bands, and (i) compared to visible light / infrared light, they penetrate non-metallic / non-polarizable materials well, and compared to RF / millimeter waves, they have a shorter wavelength, so they have high linearity and can focus beams. In addition, since the photon energy of THz waves is only a few meV, they have the characteristic of being harmless to the human body. The frequency bands expected to be used for THz wireless communication may be the D-band (110 GHz to 170 GHz) or H-band (220 GHz to 325 GHz), which have low propagation loss due to molecular absorption in the air. Discussions on standardization of THz wireless communication are being centered around the IEEE 802.15 THz working group in addition to 3GPP, and standard documents issued by the IEEE 802.15 Task Group (TG3d, TG3e) may specify or supplement the contents described in various embodiments of the present disclosure. THz wireless communication can be applied to wireless cognition, sensing, imaging, wireless communication, THz navigation, etc.

[0216] Figure 14 is a diagram illustrating an example of a THz communication application.

[0217] As illustrated in Figure 14, THz wireless communication scenarios can be categorized into macro networks, micro networks, and nanoscale networks. In macro networks, THz wireless communication can be applied to vehicle-to-vehicle and backhaul / fronthaul connections. In micro networks, THz wireless communication can be applied to fixed point-to-point or multi-point connections, such as indoor small cells, wireless connections in data centers, and near-field communications, such as kiosk downloads.

[0218] Table 2 below shows examples of technologies that can be used in THz waves.

[0219] Transceivers DeviceAvailable immature: UTC-PD, RTD and SBDModulation and CodingLow order modulation techniques (OOK, QPSK), LDPC, Reed Soloman, Hamming, Polar, TurboAntennaOmni and Directional, phased array with low number of antenna elementsBandwidth69GHz (or 23 GHz) at 300GHzChannel modelsPartiallyData rate100GbpsOutdoor deploymentNoFree space lossHighCoverageLowRadio Measurements300GHz indoorDevice sizeFew micrometers

[0220] THz wireless communications can be categorized based on the methods used to generate and receive THz waves. THz generation methods can be categorized as either optical or electronic-based.

[0221] Fig. 15 is a diagram illustrating an example of an electronic component-based THz wireless communication transmitter and receiver.

[0222] Methods for generating THz using electronic components include a method using semiconductor components such as a resonant tunneling diode (RTD), a method using a local oscillator and a multiplier, a MMIC (Monolithic Microwave Integrated Circuits) method using an integrated circuit based on a compound semiconductor HEMT (High Electron Mobility Transistor), and a method using a Si-CMOS-based integrated circuit. In the case of Fig. 15, a multiplier (doubler, tripler, multiplier) is applied to increase the frequency, and it passes through a subharmonic mixer and is radiated by an antenna. Since the THz band forms a high frequency, a multiplier is essential. Here, the multiplier is a circuit that has an output frequency that is N times that of the input, and matches it to the desired harmonic frequency and filters out all remaining frequencies. In addition, beamforming can be implemented by applying an array antenna or the like to the antenna of Fig. 15. In Fig. 15, IF represents intermediate frequency, tripler and multiplexer represent multipliers, PA represents power amplifier, LNA represents low noise amplifier, and PLL represents phase-locked loop.

[0223] FIG. 16 is a diagram illustrating an example of a method for generating a THz signal based on an optical element.

[0224] Fig. 17 is a diagram illustrating an example of an optical element-based THz wireless communication transceiver.

[0225] Optical component-based THz wireless communication technology refers to a method of generating and modulating THz signals using optical components. Optical component-based THz signal generation technology generates an ultra-high-speed optical signal using a laser and an optical modulator, and converts it into a THz signal using an ultra-high-speed photodetector. Compared to technologies that use only electronic components, this technology can easily increase the frequency, generate high-power signals, and obtain flat response characteristics over a wide frequency band. As illustrated in Figure 16, optical component-based THz signal generation requires a laser diode, a wideband optical modulator, and an ultra-high-speed photodetector. In the case of Figure 16, the light signals of two lasers with different wavelengths are combined to generate a THz signal corresponding to the wavelength difference between the lasers. In Fig. 16, an optical coupler refers to a semiconductor device that transmits an electrical signal using optical waves to provide electrical isolation and coupling between circuits or systems, and a UTC-PD (Uni-Travelling Carrier Photo-Detector) is a type of photodetector that uses electrons as active carriers and reduces the travel time of electrons with bandgap grading. The UTC-PD is capable of detecting light at 150 GHz or higher. In Fig. 17, an EDFA (Erbium-Doped Fiber Amplifier) ​​represents an erbium-doped fiber amplifier, a PD (Photo Detector) represents a semiconductor device that can convert an optical signal into an electrical signal, an OSA represents an optical module (Optical Sub Assembly) that modularizes various optical communication functions (photoelectric conversion, electro-optical conversion, etc.) into a single component, and a DSO represents a digital storage oscilloscope.

[0226] The structure of a photoelectric converter (or photoelectric converter) is described with reference to FIGS. 18 and 19.

[0227] Fig. 18 is a diagram illustrating the structure of a photon source-based transmitter.

[0228] Figure 19 is a drawing showing the structure of an optical modulator.

[0229] In general, the phase of a signal can be changed by passing the optical source of a laser through an optical wave guide. At this time, data is loaded by changing the electrical characteristics through a microwave contact, etc. Therefore, the optical modulator output is formed as a modulated waveform. An opto-electrical modulator (O / E converter) can generate THz pulses by optical rectification operation by a nonlinear crystal, photoelectric conversion by a photoconductive antenna, emission from a bunch of relativistic electrons, etc. Terahertz pulses generated in the above manner can have a length in units of femtoseconds to picoseconds. An optical / electronic converter (O / E converter) performs down conversion by utilizing the non-linearity of the device.

[0230] Considering the THz spectrum usage, it is likely that THz systems will use multiple contiguous gigahertz bands for fixed or mobile service purposes. Based on the outdoor scenario criteria, the available bandwidth can be classified based on the oxygen attenuation of 10^2 dB / km in the spectrum up to 1 THz. Accordingly, a framework in which the available bandwidth is composed of multiple band chunks can be considered. As an example of the above framework, if the THz pulse length for one carrier is set to 50 ps, ​​the bandwidth (BW) becomes approximately 20 GHz.

[0231] Effective down-conversion from the infrared band (IR band) to the terahertz band (THz band) depends on how to utilize the nonlinearity of the optical / electrical converter (O / E converter). In other words, to down-convert to the desired terahertz band (THz band), it is necessary to design an optical / electrical converter (O / E converter) with the most ideal non-linearity for transferring to the corresponding terahertz band (THz band). If an optical / electrical converter (O / E converter) that is not suitable for the target frequency band is used, errors are likely to occur in the amplitude and phase of the corresponding pulse.

[0232] In a single-carrier system, a terahertz transmission and reception system can be implemented using a single optical-to-electrical converter. Depending on the channel environment, in a multi-carrier system, the number of optical-to-electrical converters may be equal to the number of carriers. This phenomenon will be particularly noticeable in a multi-carrier system that utilizes multiple broadbands according to the aforementioned spectrum usage plan. In this regard, a frame structure for the multi-carrier system may be considered. A signal down-converted using an optical-to-electrical converter may be transmitted in a specific resource region (e.g., a specific frame). The frequency region of the specific resource region may include multiple chunks. Each chunk may be composed of at least one component carrier (CC).

[0233] Hereinafter, various embodiments of the present disclosure will be described in more detail.

[0234] The present disclosure provides a device and method for efficiently operating an AI / ML model in a wireless communication system. Specifically, the present disclosure provides a device and method for performing adaptive operation of an optimal path algorithm according to network scale when searching for a path for entangled routing between any two source and destination nodes connected via a multi-hop path in a wireless communication system.

[0235] Background to various embodiments of the present disclosure

[0236] 3GPP TR 37.817 V0.3.0 (2021-08)

[0237] 4 General Framework

[0238] Editor Note: high level principles for RAN intelligence enabled by AI, the functional framework (e.g. the AI functionality and the input / output of the component for AI enabled optimization)

[0239] Editor Note: FFS if the study assumes single vendor environment, e.g., if the model deployment / update procedure is proprietary.

[0240] 4.1 High-level Principles

[0241] The following high level principles should be applied for AI-enabled RAN intelligence:

[0242] - The detailed AI / ML algorithms and models for use cases are implementation specific and out of RAN3 scope.

[0243] - The study focuses on AI / ML functionality and corresponding types of inputs / outputs.

[0244] - The input / output and the location of the Model Training and Model Inference function should be studied case by case.

[0245] - The study focuses on the analysis of data needed at the Model Training function from Data Collection, while the aspects of how the Model Training function uses inputs to train a model are out of RAN3 scope.

[0246] - The study focuses on the analysis of data needed at the Model Inference function from Data Collection, while the aspects of how the Model Inference function uses inputs to derive outputs are out of RAN3 scope.

[0247] - Where AI / ML functionality resides within the current RAN architecture, depends on deployment and on the specific use cases.

[0248] - The Model Training and Model Inference functions should be able to request, if needed, specific information to be used to train or execute the AI / ML algorithm and to avoid reception of unnecessary information. The nature of such information depends on the use case and on the AI / ML algorithm.

[0249] - The Model Inference function should signal the outputs of the model only to nodes that have explicitly requested them (eg via subscription), or nodes that are subject to actions based on the output from Model Inference.

[0250] - An AI / ML model used in a Model Inference function has to be initially trained, validated and tested before deployment.

[0251] - NG-RAN is prioritized; EN-DC is included in the scope. FFS on whether MR-DC should be down-prioritized.

[0252] - A general framework and workflow for AI / ML optimization should be defined and captured in the TR. The generalized workflow should not prevent to “think beyond” the workflow if the use case requires so.

[0253] - User data privacy and anonymization should be respected during AI / ML operation.

[0254] 4 General Framework

[0255] Editor's Note: High-level principles and functional framework for AI-enabled RAN intelligence (e.g., inputs / outputs of components for AI functionality and AI-enabled optimization).

[0256] Editor's Note: If the study assumes a single-vendor environment (e.g., model deployment / update procedures are proprietary), it is FFS.

[0257] 4.1 Higher Principles

[0258] AI-enabled RAN intelligence must adhere to the following high-level principles:

[0259] - Detailed AI / ML algorithms and models for use cases are implementation-dependent and outside the scope of RAN3.

[0260] - This study focuses on AI / ML functions and their input / output types.

[0261] - The input / output and location of model learning and model inference functions should be studied on a case-by-case basis.

[0262] - While this study focuses on the data analysis required for the model training function of data collection, aspects of how the model training function uses inputs to train a model are outside the scope of RAN3.

[0263] - While this study focuses on the data analysis required for the model inference function of data collection, the aspect of the model inference function using inputs to derive outputs is beyond the scope of RAN3.

[0264] - Where AI / ML capabilities exist within current RAN architectures depends on deployment and specific use cases.

[0265] Model training and model inference functions should be able to request specific information, if necessary, to train or execute AI / ML algorithms and prevent the receipt of unnecessary information. The nature of this information varies depending on the use case and AI / ML algorithm.

[0266] - The model inference function should signal the model's output only to nodes that explicitly request it (e.g., via a subscription) or that are the target of an action based on the output of the model inference.

[0267] - AI / ML models used in model inference functions must be initially trained, validated, and tested before deployment.

[0268] - NG-RAN is given priority. EN-DC is included in the scope. This is an FFS regarding whether MR-DC should be given lower priority.

[0269] A general framework and workflow for AI / ML optimization should be defined and captured in the TR. Generalized workflows should not prevent "thinking beyond" the workflow when required by use cases.

[0270] - User data privacy and anonymization must be respected during AI / ML work.

[0271] FIG. 20 is a diagram illustrating an example of a functional framework for RAN Intelligence in a system applicable to the present disclosure.

[0272] 4.2 Functional Framework

[0273] Editor's Note: Data Preparation aspects may be further refined

[0274] Editor Note: FFS whether and how to signal metrics (eg, accuracy, uncertainty, etc.) and validity time together with or as part of the inference output.

[0275] Editor Note: FFS on whether model testing / generating of model performance metrics is performed in Model Inference.

[0276] This section introduces the common terminologies related to the functional framework for RAN intelligence illustrated in Figure 20. For the functions and data / information flows shown in the Figure 20, whether there is any standardization impact and what is the standardization impact are discussed in clause 5.

[0277] - Data Collection is a function that provides input data to Model training and Model inference functions. AI / ML algorithm specific data preparation (e.g., data pre-processing and cleaning, formatting, and transformation) is not carried out in the Data Collection function.

[0278] Examples of input data may include measurements from UEs or different network entities, feedback from Actor, output from an AI / ML model.

[0279] o Training Data: Data needed as input for the AI / ML Model Training function.

[0280] o Inference Data: Data needed as input for the AI / ML Model Inference function.

[0281] - Model Training is a function that performs the ML model training, validation, and testing. The Model training function is also responsible for data preparation (e.g. data pre-processing and cleaning, formatting, and transformation) based on Training Data delivered by a Data Collection function, if required.

[0282] o (FFS) Model Deployment / Update: Deploy or update an AI / ML model to Model Inference function.

[0283] - Model Inference is a function that provides AI / ML model inference output (e.g. predictions or decisions). The Model inference function is also responsible for data preparation (e.g. data pre-processing and cleaning, formatting, and transformation) based on Inference Data delivered by a Data Collection function, if required.

[0284] o Output: The inference output of the AI / ML model produced by a Model Inference function.

[0285] - Actor is a function that receives the output from the Model inference function and triggers or performs corresponding actions. The Actor may trigger actions directed to other entities or to itself.

[0286] o Feedback: Information that may be needed to derive training or inference data or performance feedback.

[0287] 4.2 Functional Framework

[0288] Editor's note: The data preparation aspect could be further improved.

[0289] Editor's Note: FFS indicates whether and how metrics (e.g., accuracy, uncertainty, etc.) and validation times are signaled along with or as part of the inference output.

[0290] Editor's Note: FFS regarding whether model testing / model performance metrics generation is performed in model inference.

[0291] This section introduces general terminology related to the functional framework for RAN intelligence depicted in Figure 20. Whether and what the standardization implications are for the functions and data / information flows depicted in Figure 20 are discussed in Section 5.

[0292] Data collection is the function that provides input data to model training and model inference functions. Data preparation for AI / ML algorithms (e.g., data preprocessing and cleaning, formatting, and transformation) is not performed in the data collection function.

[0293] Examples of input data may include measurements from UEs or various network entities, feedback from actors, and output from AI / ML models.

[0294] o Training data: Data required as input for AI / ML model training functions.

[0295] o Inference data: Data required as input for the AI / ML model inference function.

[0296] Model Training is the function that performs ML model training, validation, and testing. If necessary, the model training function also handles data preparation (e.g., data preprocessing and cleaning, formatting, and transformation) based on training data delivered via the data collection function.

[0297] o (FFS) Model Deployment / Update: Deploy or update AI / ML models to the model inference function.

[0298] Model inference is the function that provides AI / ML model inference output (e.g., predictions or decisions). If necessary, the model inference function also handles data preparation (e.g., data preprocessing and cleaning, formatting, and transformation) based on the inference data delivered by the data collection function.

[0299] o Output: The inference output of the AI / ML model generated by the model inference function.

[0300] An actor is a function that receives the output of a model inference function and triggers or performs a corresponding action. An actor can trigger actions on other objects or on itself.

[0301] o 피드백: 훈련이나 추론 데이터 또는 성과 피드백을 도출하는 데 필요할 수 있는 정보이다.

[0302] 5 Use Cases and Solutions for Artificial Intelligence in RAN

[0303] 5.1 Network Energy Saving

[0304] 5.1.1 Use case description

[0305] To meet the 5G network requirements of key performance and the demands of the unprecedented growth of the mobile subscribers, millions of base stations (BSs) are being deployed. Such rapid growth brings the issues of high energy consumption, CO2 emissions and operation expenditures (OPEX). Therefore, energy saving is an important use case which may involve different layers of the network, with mechanisms operating at different time scales.

[0306] Cell activation / deactivation is an energy saving scheme in the spatial domain that exploits traffic offloading in a layered structure to reduce the energy consumption of the whole radio access network (RAN). When the expected traffic volume is lower than a fixed threshold, the cells may be switched off, and the served Ues may be offloaded to a new target cell.

[0307] Efficient energy consumption can also be achieved by other means such as reduction of load, coverage modification, or other RAN configuration adjustments. The optimal EE decision depends on many factors including the load situation at different nodes, RAN nodes capabilities, KPI / QoS requirements, number of active Ues and UE mobility, cell utilization, etc.

[0308] However, the identification of actions aimed at energy efficiency improvements is not a trivial task. Wrong switch-off of the cells may seriously deteriorate the network performance since the remaining active cells need to serve the additional traffic. Wrong traffic offload actions may lead to a deterioration of Energy Efficiency instead of an improvement. The current energy-saving schemes are vulnerable to potential issues listed as follows:

[0309] - Inaccurate cell load prediction. Currently, energy-saving decisions rely on current traffic load without considering future traffic load.

[0310] - Conflicting targets between system performance and energy efficiency. Maximizing the system’s key performance indicator (KPI) is usually done at the expense of energy efficiency. Similarly, the most energy efficient solution may impact system performance. Thus, there is a need to balance and manage the trade-off between the two.

[0311] - Conventional energy-saving related parameters adjustment. Energy-saving related parameters configuration is set by traditional operation, e.g., based on different thresholds of cell load for cell switch on / off which is somewhat a rigid mechanism since it is difficult to set a reasonable threshold.

[0312] - Actions that may produce a local (e.g. limited to a single RAN node) improvement of Energy Efficiency, while producing an overall (e.g. involving multiple RAN nodes) deterioration of Energy Efficiency.

[0313] To deal with issues listed above, ML techniques could be utilized to leverage on the data collected in the RAN network. ML algorithms may predict the energy efficiency and load state of the next period, which can be used to make better decisions on cell activation / deactivation for ES. Based on the predicted load, the system may dynamically configure the energy-saving strategy (e.g. the switch-off timing and granularity, offloading actions) to keep a balance between system performance and energy efficiency and to reduce the energy consumption.

[0314] 5.1.2 Solutions and standard impacts

[0315] Editor Note: Capture the solutions for the use case, including potential standard impacts on existing Nodes, functions, and interfaces

[0316] 5.1 네트워크 에너지 절약

[0317] 5.1.1 사용 사례 설명

[0318] To meet 5G network requirements for core performance and the unprecedented growth in mobile subscribers, millions of base stations (BSs) are being deployed. This rapid growth brings with it high energy consumption, CO2 emissions, and operating expenses (OPEX). Therefore, energy conservation is a critical use case that can encompass various layers of the network, along with mechanisms that operate over various time scales.

[0319] Cell activation / deactivation is a spatial energy-saving method that utilizes hierarchical traffic offloading to reduce energy consumption across the entire Radio Access Network (RAN). When the expected traffic volume falls below a fixed threshold, a cell can be deactivated, and the UE receiving the service can be offloaded to a new target cell.

[0320] Efficient energy consumption can also be achieved through other means, such as load reduction, coverage modification, and other RAN configuration adjustments. The optimal EE decision depends on several factors, including the load conditions of various nodes, RAN node capabilities, KPI / QoS requirements, the number of active UEs and their mobility, and cell utilization.

[0321] However, identifying measures aimed at improving energy efficiency is no simple task. Incorrectly shutting down cells can severely degrade network performance, forcing the remaining active cells to handle additional traffic. Incorrect traffic offloading measures can actually worsen energy efficiency, rather than improve it. Current energy-saving plans are vulnerable to the following potential problems:

[0322] - Inaccurate cell load predictions. Current energy-saving decisions rely on current traffic loads without considering future traffic loads.

[0323] There is a conflict between the goals of system performance and energy efficiency. Maximizing a system's key performance indicators (KPIs) typically comes at the expense of energy efficiency. Similarly, the most energy-efficient solution can impact system performance. Therefore, a balance between the two is necessary, and management is required to maintain this balance.

[0324] - Adjusting existing energy-saving parameters. Energy-saving parameters are traditionally set based on various cell load thresholds, such as those for turning cell switches on and off. This is a rather rigid mechanism, as it is difficult to set reasonable thresholds.

[0325] - Work that can produce localized (e.g., limited to a single RAN node) improvements in energy efficiency while simultaneously producing overall (e.g., involving multiple RAN nodes) degradations in energy efficiency.

[0326] To address the challenges listed above, ML techniques can be leveraged to leverage data collected from RAN networks. ML algorithms can predict energy efficiency and load conditions for the next period, which can be used to make better decisions about activating and deactivating cells in ESs. Based on the predicted load, the system can dynamically configure energy-saving strategies (e.g., power-off timing and segmentation, offloading tasks) to balance system performance and energy efficiency and reduce energy consumption.

[0327] 5.1.2 Solution and Standard Impact

[0328] Editor's Note: Captures solutions for use cases, including potential standards impacts on existing nodes, features, and interfaces.

[0329] 도 21은 본 개시에 적용 가능한 시스템에서 OAM의 모델 훈련, NG-RAN의 모델 추론(Model Training at OAM, Model Inference at NG-RAN)의 일례를 도시한 도면이다.

[0330] 5.1.2.1 Model Training at OAM and Model Inference at NG-RAN

[0331] In this solution, NG-RAN predicts energy saving decisions by AI / ML model trained from OAM.

[0332] Step 0: NG-RAN node 1 is assumed to have a AI / ML model trained by OAM, NG-RAN node 2 is assumed to have a AI / ML model trained by OAM optionally.

[0333] Step 1: NG-RAN node 2 sends the required input data to NG-RAN node 1 for model inference of AI / ML-based network energy saving.

[0334] Step 2: UE sends UE measurement report to NG-RAN node 1. (FFS on if triggered)

[0335] Step 3: Based on local inputs of NG-RAN node 1 and received inputs from NG-RAN node 2, NG-RAN node 1 generates model inference output(s) (e.g. energy saving strategy, handover strategy, etc).

[0336] Step 4: NG-RAN node 1 selects the most appropriate target cell for each UE before it performs handover and goes to the predicted energy state.

[0337] Step 5: NG-RAN node 1 and NG-RAN node 2 provide feedback to OAM.

[0338] 5.1.2.1 Model Training in OAM and Model Inference in NG-RAN

[0339] In this solution, NG-RAN predicts energy-saving decisions using AI / ML models trained in OAM.

[0340] Step 0: Assume that NG-RAN node 1 has an AI / ML model trained by OAM, and NG-RAN node 2 optionally has an AI / ML model trained by OAM.

[0341] Step 1: NG-RAN node 2 sends the input data required for AI / ML-based network energy saving model inference to NG-RAN node 1.

[0342] Step 2: The UE sends the UE measurement report to NG-RAN Node 1 (FFS turned on if triggered).

[0343] Step 3: Based on the local input of NG-RAN node 1 and the input received from NG-RAN node 2, NG-RAN node 1 generates model inference outputs (e.g., energy saving strategy, handover strategy, etc.).

[0344] Step 4: NG-RAN node 1 selects the most suitable target cell for each UE and moves to the predicted energy state before performing the handover.

[0345] Step 5: NG-RAN Node 1 and NG-RAN Node 2 provide feedback to OAM.

[0346] 도 22은 본 개시에 적용 가능한 시스템에서 NG-RAN의 모델 훈련 및 모델 추론 (Model Training and Model Inference at NG-RAN)의 일례를 도시한 도면이다.

[0347] 5.1.2.2 Model Training and Model Inference at NG-RAN

[0348] In this solution, NG-RAN is responsible for model training and generates energy saving decisions.

[0349] Editor’s Notes: FFS on data collection.

[0350] Step 1: NG-RAN node 1 trains AI / ML model for AI / ML-based energy saving based on collected data. NG-RAN node 2 is assumed to have AI / ML model for AI / ML-based energy saving optionally, which can also generate predicted results / actions.

[0351] Step 2: NG-RAN node 2 sends the required input data to NG-RAN node 1 for model inference of AI / ML-based network energy saving.

[0352] Step 3: UE sends UE measurement report to NG-RAN node 1. (FFS on if triggered)

[0353] Step 4: Based on local inputs of NG-RAN node 1 and received inputs from NG-RAN node 2, NG-RAN node 1 generates model inference output (e.g. energy saving strategy, handover strategy, etc).

[0354] Step 5: NG-RAN node 1 selects the most appropriate target cell for each UE before it performs handover and goes to the predicted energy state.

[0355] Step 6: NG-RAN node 2 provides feedback to NG-RAN node 1.

[0356] 5.1.2.3 Input of AI / ML-based Network Energy Saving

[0357] To predict the optimized network energy saving decisions, NG-RAN may need following information as input data for AI / ML-based network energy saving:

[0358] - Current / Predicted resource status of ES-Cell and its neighbor nodes

[0359] - Current / Predicted energy information of ES-Cell and its neighbor nodes

[0360] - UE measurement report (e.g. UE RSRP, RSRQ, SINR measurement, etc)

[0361] If existing UE measurements are needed by a gNB for AI / ML-based network energy saving, RAN3 shall reuse the existing framework (including MDT and RRM measurements). FFS on whether new UE measurements are needed.

[0362] Editor’s Note: FFS other input information required for AI / ML-based network energy saving. FFS energy information is exact energy consumption value or energy efficiency gain.

[0363] 5.1.2.4 Output of AI / ML-based Network Energy Saving

[0364] AI / ML-based network energy saving model can generate following information as output:

[0365] - Energy saving strategy

[0366] - Handover strategy, including recommended candidate cells for taking over the traffic

[0367] - Predicted energy information

[0368] Editor's Note: FFS other output information expected from AI / ML-based network energy saving. FFS detailed granularity and action of energy saving strategy. FFS on accuracy of predicted energy saving decision.

[0369] 5.1.2.5 Feedback of AI / ML-based Network Energy Saving

[0370] To optimize the performance of AI / ML-based network energy saving model, following feedback can be considered to be collected from NG-RAN nodes:

[0371] - Load measurement

[0372] - Energy information

[0373] Editor's Note: FFS other feedback expected from AI / ML-based network energy saving.

[0374] 5.1.2.2 Model Training and Model Inference in NG-RAN

[0375] In this solution, NG-RAN is responsible for model training and making energy-saving decisions.

[0376] Editor's note: FFS on data collection.

[0377] Step 1: NG-RAN node 1 trains an AI / ML model for AI / ML-based energy savings based on the collected data. NG-RAN node 2 is optionally assumed to have an AI / ML model for AI / ML-based energy savings, which can also generate predicted results / actions.

[0378] Step 2: NG-RAN node 2 sends the input data required for AI / ML-based network energy-saving model inference to NG-RAN node 1.

[0379] Step 3: The UE sends the UE measurement report to NG-RAN Node 1 (FFS turned on if triggered).

[0380] Step 4: Based on the local input of NG-RAN node 1 and the input received from NG-RAN node 2, NG-RAN node 1 generates model inference outputs (e.g., energy saving strategy, handover strategy, etc.).

[0381] Step 5: NG-RAN node 1 selects the most suitable target cell for each UE and moves to the predicted energy state before performing the handover.

[0382] Step 6: NG-RAN node 2 provides feedback to NG-RAN node 1.

[0383] 5.1.2.3 AI / ML-based network energy saving input

[0384] To predict optimized network energy saving decisions, NG-RAN may require the following information as input data for AI / ML-based network energy saving:

[0385] - Current / expected resource status of ES-Cell and adjacent nodes

[0386] - Current / expected energy information of ES-Cell and surrounding nodes

[0387] - UE measurement reports (e.g. UE RSRP, RSRQ, SINR measurements, etc.)

[0388] If existing UE measurements are required at gNB for AI / ML-based network energy savings, RAN3 should reuse existing frameworks (including MDT and RRM measurements). This is an FFS decision on whether new UE measurements are required.

[0389] Editor's Note: FFS and other input information required for AI / ML-based network energy savings. FFS energy information is either accurate energy consumption values ​​or energy efficiency gains.

[0390] 5.1.2.4 AI / ML-based network energy-saving output

[0391] AI / ML-based network energy-saving models can produce the following information as output:

[0392] - Energy saving strategies

[0393] - Handover strategy, including recommended candidate cells to take over traffic

[0394] - Expected energy information

[0395] Editor's Note: Expected FFS output information from AI / ML-based network energy savings. FFS detailed breakdown and energy-saving strategy implementation. FFS on the accuracy of predicted energy-saving decisions.

[0396] 5.1.2.5 AI / ML-based Network Energy Saving Feedback

[0397] To optimize the performance of AI / ML-based network energy-saving models, one could consider collecting the following feedback from NG-RAN nodes:

[0398] - Load measurement

[0399] - Energy information

[0400] Editor's Note: Expected FFS and other feedback on AI / ML-based network energy savings.

[0401] 5.2 Load Balancing

[0402] 5.2.1 Use case description

[0403] The rapid traffic growth and multiple frequency bands utilized in a commercial network make it challenging to steer the traffic in a balanced distribution. To address the problem, load balancing had been proposed. The objective of load balancing is to distribute load evenly among cells and among areas of cells, or to transfer part of the traffic from congested cells or from congested areas of cells, or to offload users from one cell, cell area, carrier or RAT to improve network performance. This can be done by means of optimization of handover parameters and handover actions. The automation of such optimisation can provide high quality user experience, while simultaneously improving the system capacity and also to minimize human intervention in the network management and optimization tasks.

[0404] However, the optimization of the load balancing is not an easy task as follows:

[0405] - Currently the load balancing decisions relying on the current / past-state cell load status are insufficient. The traffic load and resource status of the network changes rapidly, especially in the scenarios with high-mobility and large number of connections, which may lead to ping-pong handover between different cells, cell overload and degradation of user service quality.

[0406] - It is difficult to guarantee the overall network and service performance when performing load balancing. For the load balancing, the UEs in the congested cell may be offloaded to the target cell, by means of handover procedure or adapting handover configuration. For example, if the UEs with time-varying traffic load are offloaded to the target cell, the target cell may be overloaded with new-arrival heavy traffic. It is difficult to determine whether the service performance after the offloading action meets the desired targets.

[0407] To deal with the above issues, solutions based on AI / ML model could be introduced to improve the load balancing performance. Based on collection of various measurements and feedbacks from UEs and network nodes, historical data, etc. ML model based solutions and predicted load could improve load balancing performance, in order to provide higher quality user experience and to improve the system capacity.

[0408] 5.2.2 Solutions and standard impacts

[0409] Editor Note: Capture the solutions for the use case, including potential standard impacts on existing Nodes, functions, and interfaces

[0410] The following solutions can be considered for supporting AI / ML-based load balancing:

[0411] - AI / ML Model Training is located in the OAM and AI / ML Model Inference is located in the gNB.

[0412] - AI / ML Model Training and AI / ML Model Inference are both located in the gNB.

[0413] In case of CU-DU split architecture, the following solutions are possible:

[0414] - AI / ML Model Training is located in the OAM and AI / ML Model Inference is located in the gNB-CU.

[0415] - AI / ML Model Training and Model Inference are both located in the gNB-CU.

[0416] Other possible locations of the AI / ML Model Training and AI / ML Model Inference are FFS.

[0417] To improve the load balancing decisions at a gNB (gNB-CU), a gNB can request load predictions from a neighbouring node. Details of the procedure are FFS.

[0418] If existing UE measurements are needed by a gNB for AI / ML-based load balancing, RAN3 shall reuse the existing framework (including MDT and RRM measurements). FFS on whether new UE measurements are needed.

[0419] 5.2 로드 밸런싱

[0420] 5.2.1 사용 사례 설명

[0421] The rapid growth of traffic and the diverse frequency bands used in commercial networks make it difficult to manage traffic with a balanced distribution. Load balancing has been proposed to address this issue. The goal of load balancing is to evenly distribute the load across cells and cell regions, redirect some of the traffic from congested cells or congested areas of a cell, or offload users from one cell, cell region, carrier, or RAT to the next, thereby improving network performance. This can be achieved by optimizing handover parameters and handover operations. Automating these optimizations can improve system capacity while providing a high-quality user experience and minimizing human intervention in network management and optimization.

[0422] However, optimizing load balancing is not an easy task, as follows:

[0423] Currently, load balancing decisions based on current and past cell load conditions are inadequate. Network traffic load and resource conditions change rapidly, especially in scenarios with high mobility and a large number of connections. This can lead to ping-pong handovers between different cells, cell overload, and degraded user service quality.

[0424] It's difficult to guarantee overall network and service performance when performing load balancing. For load balancing, UEs in congested cells can be offloaded to a target cell through handover procedures or handover configuration adjustments. For example, if UEs with time-varying traffic loads are offloaded to a target cell, the target cell may become overloaded due to the large volume of newly arriving traffic. It's difficult to determine whether service performance after the offloading process meets the desired goals.

[0425] To address the above issues, AI / ML model-based solutions can be introduced to improve load balancing performance. These solutions are based on various measurements and feedback collected from UEs and network nodes, as well as historical data. ML model-based solutions and predictive load can improve load balancing performance, provide a higher-quality user experience, and increase system capacity.

[0426] 5.2.2 Solution and Standard Impact

[0427] Editor's Note: Captures solutions for use cases, including potential standards impacts on existing nodes, features, and interfaces.

[0428] To support AI / ML-based load balancing, you can consider the following solutions:

[0429] - AI / ML model training is on OAM and AI / ML model inference is on gNB.

[0430] - Both AI / ML model training and AI / ML model inference are on gNB.

[0431] For the CU-DU partitioned architecture, the following solutions are possible:

[0432] - AI / ML model training is on OAM and AI / ML model inference is on gNB-CU.

[0433] - AI / ML model training and model inference are both on the gNB-CU.

[0434] Another possible location for AI / ML model training and AI / ML model inference is FFS.

[0435] To improve load balancing decisions in a gNB (gNB-CU), a gNB can request load predictions from neighboring nodes. The details of the procedure are described in FFS.

[0436] If existing UE measurements are required at the gNB for AI / ML-based load balancing, RAN3 should reuse existing frameworks (including MDT and RRM measurements). Whether new UE measurements are required is subject to FFS.

[0437] 5.3 Mobility Optimization

[0438] 5.3.1 Use case description

[0439] Mobility management is the scheme to guarantee the service-continuity during the mobility by minimizing the call drops, RLFs, unnecessary handovers, and ping-pong. For the future high-frequency network, as the coverage of a single node decreases, the frequency for UE to handover between nodes becomes high, especially for high-mobility UE. In addition, for the applications characterized with the stringent QoS requirements such as reliability, latency etc., the QoE is sensitive to the handover performance, so that mobility management should avoid unsuccessful handover and reduce the latency during handover procedure. However, for the conventional method, it is challengeable for trial-and-error-based scheme to achieve nearly zero-failure handover. The unsuccessful handover cases are the main reason for packet dropping or extra delay during the mobility period, which is unexpected for the packet-drop-intolerant and low-latency applications.In addition, the effectiveness of adjustment based on feedback may be weak due to randomness and inconstancy of transmission environment. Besides the baseline case of mobility, areas of optimization for mobility include dual connectivity, CHO, and DAPS, which each have additional aspects to handle in the optimization of mobiltity.

[0440] Mobility aspects of SON that can be enhanced by the use of AI / ML include

[0441] - Reduction of the probability of unintended events

[0442] - UE Location / Mobility / Performance prediction

[0443] - Traffic Steering

[0444] Reduction of the probability of unintended events associated with mobility.

[0445] Examples of such unintended events are:

[0446] - Intra-system Too Late Handover: A radio link failure (RLF) occurs after the UE has stayed for a long period of time in the cell; the UE attempts to re-establish the radio link connection in a different cell.

[0447] - Intra-system Too Early Handover: An RLF occurs shortly after a successful handover from a source cell to a target cell or a handover failure occurs during the handover procedure; the UE attempts to re-establish the radio link connection in the source cell.

[0448] - Intra-system Handover to Wrong Cell: An RLF occurs shortly after a successful handover from a source cell to a target cell or a handover failure occurs during the handover procedure; the UE attempts to re-establish the radio link connection in a cell other than the source cell and the target cell.

[0449] RAN Intelligence could observe multiple HO events with associated parameters, use this information to train its ML model and try to identify sets of parameters that lead to successful Hos and sets of parameters that lead to unintended events.

[0450] UE Location / Mobility / Performance Prediction

[0451] Predicting UE’s location is a key part for mobility optimisation, as many RRM actions related to mobility (e.g. selecting handover target cells) can benefit from the predicted UE location / trajectory. UE mobility prediction is also one key factor in the optimization of early data forwarding particularly for CHO. UE Performance prediction when the UE is served by certain cells is a key factor in determining which is the best mobility target for maximisation of efficiency and performance.

[0452] Traffic Steering

[0453] Efficient resource handling can be achieved adjusting handover trigger points and selecting optimal combination of Pcell / PSCell / Scells to serve a user.

[0454] Existing traffic steering can also be improved by providing a RAN node with information related to mobility or dual connectivity.

[0455] For example, before initiating a handover, the source gNB, could use feedbacks on UE performance collected for successful handovers occurred in the past and received from neighboring gNBs.

[0456] Similarly, for the case of dual connectivity, before triggering the addition of a secondary gNB or triggering SN change, an eNB could use information (feedbacks) received in the past from the gNB for successfully completed SN Addition or SN Change procedures.

[0457] In the two reported examples, the source RAN node of a mobility event, or the RAN node acting as Master Node (a eNB for EN-DC, a gNB for NR-DC) can use feedbacks received from the other RAN node, as input to an AI / ML function supporting traffic related decisions (e.g. selection of target cell in case of mobility, selection of a PSCell / Scell(s) in the other case), so that future decisions can be optimized.

[0458] 5.3 모빌리티 최적화

[0459] 5.3.1 사용 사례 설명

[0460] Mobility Management (MSM) ensures service continuity during mobility by minimizing dropped calls, RLFs, unnecessary handovers, and ping-pong. In future high-frequency networks, as the coverage of a single node decreases, the frequency of UE handovers between nodes will increase, especially for highly mobile UEs. Furthermore, for applications with stringent QoS requirements, such as reliability and latency, QoE is sensitive to handover performance. Therefore, mobility management must prevent failed handovers and reduce latency during the handover process. However, existing methods struggle to achieve near-zero handovers using trial-and-error methods. Failed handovers are a major cause of packet drop or additional delay during mobility, which is unpredictable for applications that tolerate packet drop and require low latency. Furthermore, the randomness and inconsistency of the transmission environment can hinder the effectiveness of feedback-based adjustments. Beyond the basic mobility case, optimization areas for mobility include dual connectivity, CHO, and DAPS, each of which presents additional aspects that must be addressed in mobility optimization.

[0461] Some mobility aspects of SON that can be improved using AI / ML include:

[0462] - Reduces the likelihood of unintended events occurring

[0463] - UE location / mobility / performance prediction

[0464] - Traffic coordination

[0465] Reduces the likelihood of unintended incidents related to mobility.

[0466] Examples of unintended events include:

[0467] - Intra-system Too Late Handover: Radio Link Failure (RLF) occurs after the UE has stayed in a cell for an extended period of time. The UE attempts to reestablish a radio link connection in another cell.

[0468] - Premature handover within the system: An RLF occurs immediately after a successful handover from a source cell to a target cell or after a handover failure occurs during the handover procedure. The UE attempts to re-establish the radio link connection with the source cell.

[0469] - Intra-system handover to a wrong cell: An RLF occurs immediately after a successful handover from a source cell to a target cell, or a handover failure occurs during the handover procedure. The UE attempts to re-establish a radio link connection in a cell other than the source or target cell.

[0470] RAN Intelligence can observe multiple HO events with related parameters, use this information to train ML models, and attempt to identify sets of parameters that lead to successful HOs and those that lead to unintended events.

[0471] UE location / mobility / performance prediction

[0472] Predicting UE location is a key component of mobility optimization, as many mobility-related RRM tasks (e.g., handover target cell selection) benefit from predicted UE location / trajectory. Predicting UE mobility is also a key element in optimizing initial data delivery, particularly for CHO. When a UE is served by a specific cell, predicting UE performance is key to determining the optimal mobility target for maximizing efficiency and performance.

[0473] Traffic steering

[0474] Efficient resource handling can be achieved by adjusting the handover trigger point and selecting the optimal Pcell / PSCell / Scell ​​combination to provide services to users.

[0475] It may also improve existing traffic steering by providing information related to mobility or dual connectivity to RAN nodes.

[0476] For example, the source gNB may use feedback on UE performance collected from past successful handovers and received from neighboring gNBs before initiating the handover.

[0477] Similarly, in case of dual connectivity, before triggering secondary gNB addition or SN change, the eNB may use information (feedback) previously received from the gNB for successfully completed SN addition or SN change procedures.

[0478] In the two reported examples, the RAN node acting as the source or master node of the mobility event (eNB for EN-DC, gNB for NR-DC) can use feedback received from other RAN nodes as input. AI / ML capabilities that support traffic-related decisions (e.g., target cell selection for mobility, PSCell / Scell ​​selection in other cases) can then be used to optimize future decisions.

[0479] 5.3.2 Solutions and standard impacts

[0480] Editor Note: Capture the solutions for the use case, including potential standard impacts on existing Nodes, functions, and interfaces

[0481] Considering the locations of AI / ML Model Training and AI / ML Model Inference for mobility solution, following two options are considered:

[0482] - The AI / ML Model Training function is deployed in OAM, while the Model Inference function resides within the RAN node

[0483] - Both the AI / ML Model Training function and the AI / ML Model Inference function reside within the RAN node

[0484] Furthermore, for CU-DU split scenario, following option is possible:

[0485] - AI / ML Model Training is located in CU-CP or OAM, and AI / ML Model Inference function is located in CU-CP

[0486] 5.3.2.1 AI / ML Model Training in OAM and AI / ML Model Inference in NG-RAN node

[0487] Step 1: The RAN is assumed to have in use a trained AI / ML model for inference

[0488] Step 2. Model Inference. Required measurements are leveraged into Model Inference to output the prediction, eg UE trajectory prediction, target cell prediction, target NG-RAN node prediction, etc.

[0489] Step 3. According to the prediction, recommended actions or configuration are executed for Mobility Optimization.

[0490] 5.3.2 Solution and Standard Impact

[0491] Editor's Note: Captures solutions for use cases, including potential standards impacts on existing nodes, features, and interfaces.

[0492] When considering where to place AI / ML model training and AI / ML model inference for mobility solutions, two options are considered:

[0493] - AI / ML model training functions are deployed in OAM, and model inference functions are within the RAN node.

[0494] - Both AI / ML model training and AI / ML model inference functions reside within the RAN node.

[0495] Additionally, for the CU-DU split scenario, the following options are possible:

[0496] - AI / ML model training is located in CU-CP or OAM, and AI / ML model inference function is located in CU-CP.

[0497] 5.3.2.1 AI / ML Model Training in OAM and AI / ML Model Inference in NG-RAN Nodes

[0498] Step 1: RAN assumes that it is using a trained AI / ML model for inference.

[0499] Step 2. Model Inference. The required measurements are used in model inference to produce predictions. These include UE trajectory prediction, target cell prediction, and target NG-RAN node prediction.

[0500] Step 3. Based on the prediction, recommended actions or configurations for Mobility Optimization are executed.

[0501]

[0502] FIG. 23 is a diagram illustrating an example of model training and model inference both located in RAN nodes in a system applicable to the present disclosure.

[0503] 5.3.2.2 AI / ML Model Training and AI / ML Model Inference in NG-RAN node

[0504] Step 1. NG-RAN node1 configures the measurement information on the UE side and sends configuration message to UE including configuration information.

[0505] Step 2. UE collects the indicated measurement, eg, UE measurements related to RSRP, RSRQ, SINR of serving cell and neighboring cells.

[0506] Step 3. UE sends measurement report message to NG-RAN node1 including the required measurement.

[0507] Step 4. Model training. Required measurements are leveraged to train ML model for mobility optimization.

[0508] Step 5. NG-RAN node1 obtains the measurement report as inference data for real-time UE mobility optimization.

[0509] Step 6. Model Inference. Required measurements are leveraged into Model Inference to output the prediction, including eg, UE trajectory prediction, target cell prediction, target NG-RAN node prediction, etc.

[0510] Step 7. According to the prediction, recommended actions are executed for Mobility Optimization. NG-RAN node1 may send the predicted mobility optimization solution to NG-RAN node2.

[0511] 5.3.2.2 AI / ML Model Training and AI / ML Model Inference on NG-RAN Nodes

[0512] Step 1. NG-RAN node1 configures measurement information on the UE side and sends a configuration message containing the configuration information to the UE.

[0513] Step 2. The UE collects indicated measurements, for example, UE measurements related to RSRP, RSRQ, and SINR of the serving cell and neighboring cells.

[0514] Step 3. The UE sends a measurement report message containing the required measurements to NG-RAN node1.

[0515] Step 4. Model training. The essential measurements are used to train an ML model for mobility optimization.

[0516] Step 5. NG-RAN node1 obtains a measurement report as inference data for real-time UE mobility optimization.

[0517] Step 6. Model Inference. The required measurements are used for model inference, which outputs predictions, including UE trajectory predictions, target cell predictions, and target NG-RAN node predictions.

[0518] Step 7. Based on the prediction, recommended actions for mobility optimization are implemented. NG-RAN node 1 can send the predicted mobility optimization solution to NG-RAN node 2.

[0519] 5.3.2.3 Input data

[0520] The following data is required as input data for mobility optimization.

[0521] Input Information from UE:

[0522] - FFS UE historical location information from MDT, eg, Latitude, longitude, altitude, cell ID

[0523] - Radio measurements related to serving cell and neighboring cells associated with UE location information, eg, RSRP, RSRQ, SINR

[0524] - UE historical serving cells and their locations

[0525] - Moving velocity

[0526] - FFS predicted traffic

[0527] Input Information from the neighbouring RAN nodes:

[0528] - UE’s successful handover information in the past and received from neighboring RAN nodes

[0529] - UE’s history information from neighbor

[0530] - Position, resource status, FFS QoS parameters of historical HO-ed UE (e.g., loss rate, delay, etc.)

[0531] - Resource status and utilization prediction / estimation

[0532] - SON Reports of handovers that are successful, too-early, too-late, or handover to wrong (sub-optimal) cell

[0533] - FFS Information about the performance of handed over UEs

[0534] Input Information from the local node:

[0535] - UE trajectory prediction output (will be used by the RAN node internally)

[0536] - Local load prediction

[0537] If existing UE measurements are needed by a gNB for AI / ML-based network energy saving, RAN3 shall reuse the existing framework (including MDT and RRM measurements). FFS on whether new UE measurements are needed.

[0538] 5.3.2.4 Output data

[0539] - FFS UE trajectory prediction (Latitude, longitude, altitude of UE over a future period of time)

[0540] - Estimated arrival probability in CHO and relevant confidence interval

[0541] Predicted handover target node, candidate cells in CHO, may together with the confidence of the predication

[0542] 5.3.2.3 Input Data

[0543] The following data is required as input data for mobility optimization.

[0544] Input information from UE:

[0545] - FFS UE past location information of MDT (e.g. latitude, longitude, altitude, cell ID)

[0546] - Radio measurements related to serving cell and neighboring cells related to UE location information (e.g. RSRP, RSRQ, SINR)

[0547] - UE history serving cell and location

[0548] - Movement speed

[0549] - FFS predicted traffic

[0550] Input information from adjacent RAN nodes:

[0551] - Information on successful handovers of past UEs and information received from adjacent RAN nodes.

[0552] - UE history information from neighbors

[0553] - Location, resource status, and FFS QoS parameters (e.g., loss rate, delay, etc.) of previously HO-ed UEs.

[0554] - Resource status and utilization prediction / estimation

[0555] - SON reports on handovers that were successful, too early, too late, or to the wrong (suboptimal) cell.

[0556] - FFS information on the performance of the handed over UE

[0557] Input information for local node:

[0558] - UE trajectory prediction output (used internally in RAN nodes)

[0559] - Local load prediction

[0560] If existing UE measurements are required at gNB for AI / ML-based network energy savings, RAN3 should reuse existing frameworks (including MDT and RRM measurements). This is an FFS decision on whether new UE measurements are required.

[0561] 5.3.2.4 Output Data

[0562] - FFS UE trajectory prediction (latitude, longitude, and altitude of UE for future period)

[0563] - CHO's expected arrival probability and associated confidence intervals

[0564] The predicted handover target node, the candidate cell of CHO, may be present along with the confidence of the prediction.

[0565] 본 개시에서 사용되는 기호 / 약어 / 용어는 다음과 같다.

[0566] - NW: NetWork

[0567] - UE: User Equipment

[0568] - AI: Artificial Intelligence

[0569] - ML: Machine Learning

[0570] - RAN: Radio Access Network

[0571] - CSI: Channel State Information

[0572] - RS: Reference Signal

[0573] - RRM: Radio Resource Management

[0574] - PHY: PHYsical layer

[0575] - MAC: Medium Access Control

[0576] - RRC: Radio Resource Control

[0577] - MIB: Master Information Block

[0578] - SIB: System Information Block

[0579] - DCI: Downlink Control Information

[0580] - PLMN: Public Land Mobile Network

[0581] - ECGI: E-UTRA Cell Global Identifier

[0582] - LAC: Location Area Code

[0583] - RSRP: Reference Signal Received Power

[0584] - RSRQ: Reference Signal Received Quality)

[0585] - MCC / MNC: Mobile Country Code / Mobile Network Code

[0586] -IMSI: International Mobile Subscriber Identity

[0587] - NAS: Non-Access Stratum

[0588] - O&M: Operation and Maintenance

[0589] - DN: Data Network

[0590] - gNB: Next generation NodeB

[0591] Technical problems solved by this disclosure

[0592] FIG. 24 is a diagram illustrating an example of the process of creating and deploying an AI / ML model in a system applicable to the present disclosure.

[0593] AI is the most crucial and newly introduced technology for 6G systems. Research is currently underway to support partial or very limited AI in 5G systems. Beginning with data collection in RAN3, NR Rel-18 is exploring use cases for AI in RAN1 and RAN2 technologies and studying their impact on specifications. Going forward, AI is being applied not only to network technologies but also to improve technologies related to the PHY and MAC / RRC layers between terminals and base stations, such as CSI feedback enhancement, beam management, positioning, RS overhead reduction, and RRM mobility enhancement. It is expected that AI will be able to perform these time- and power-intensive tasks immediately.

[0594] As efforts to integrate AI into a wider range of communications technologies are increasing, a priority is placed on analyzing the impact of specifications on the scope and location of AI / ML model application in air interfaces such as RAN1 and RAN2. Considerable AI / ML model application scenarios can be broadly categorized into three categories:

[0595] Case 1. Performance improvement through AI / ML model implementation in NW and / or UE.

[0596] Case 2. Independent AI / ML model implementation in NW and / or UE, but performance improvement through input / output definition

[0597] Here, Case 1 refers to a situation where improvements are purely implementation-based, with no impact on specifications. This means that performance improvements can be achieved through independent implementation on the network or terminal. Case 2, for AI / ML models, algorithms are implemented independently on the terminal and / or network, but the standard defines inputs and outputs that can affect model operation.

[0598] The technology of the present disclosure focuses on the second of the two scenarios described above. It considers a situation where independent AI / ML models are implemented in the NW and / or UE, and input / output is exchanged and the AI / ML model is operated.

[0599] In the present disclosure, we consider a network (NW) that applies an AI / ML model that predicts UE mobility and traffic. The NW performs various network resource management functions, such as load balancing, channel selection, and power allocation, by predicting UE mobility and traffic through the AI / ML model. The operator deploys and operates the AI / ML model that predicts mobility and traffic to the UE, RAN, 5G Core, and DN according to the system design. For example, the DN collects mobility and traffic-related information from the UE, RAN, and 5G Core entities through RRC, NAS, EPS Session Management, or context-related messages.

[0600] Depending on each message type, the following content may include mobility and traffic-related information.

[0601] 1. RRC message

[0602] RRC-related messages include RRC Connection Request, RRC Reconfiguration, Tracking Area Update Request, and RRC Connection Re-establishment Request. UE movement and location information can be obtained through RRC messages. Information such as Cell ID, RSRP (Reference Signal Received Power), RSRQ (Reference Signal Received Quality), Timing Advance, Tracking Area ID, and LAC (Location Area Code) can be obtained. ECGI (E-UTRA Cell Global Identifier) ​​can be obtained through E-UTRAN Handover Request. Device-related information such as Network Type and Frequency Band can be obtained to indirectly identify movement patterns and traffic characteristics.

[0603] 2. NAS message

[0604] NAS-related messages include Attach Request, Attach Accept, Service Request, and PLMN (Public Land Mobile Network) Selection Request. Through these messages, location-related information such as MCC / MNC (Mobile Country Code / Mobile Network Code) and device-related information such as UE ID, UE Capability, User Location Information (ULI), and Temporary Mobile Subscriber Identity (TMSI) can be obtained. Furthermore, information necessary for traffic prediction, such as Requested Service, Service Type, and QoS (Quality of Service), can be collected. Policy Rule Information can be obtained through Create Session Request and Modify Bearer Request exchanged by MME or P-GW. Policy Rule Information specifies the network access rules for a specific service or application by the UE. It is also used to provide Quality of Service (QoS) to the UE or perform traffic management. Therefore, AI / ML models are used as training data and inference data for traffic prediction. IMSI (International Mobile Subscriber Identity) (Identity Request and Authentication, HSS) and IMEI (International Mobile Equipment Identity) that can be collected from HSS through Identity Request and Authentication messages are also necessary to predict mobility, traffic, or related correlations.

[0605] 3. EPS Session Management messages

[0606] EPS Session related messages are indirectly related to mobility because they are used when session information is updated to match a new location, and thus can be used to predict service usage patterns. The Activate Default EPS Bearer Context Request is a message sent from the UE to the MME to request the creation of an EPS Bearer for the purpose of bearer management. The Deactivate EPS Bearer Context Request is a message sent from the UE to the MME to request the deletion of an EPS Bearer during bearer management. The Modify EPS Bearer Context Request is a message sent from the UE to the MME to request the modification of an EPS Bearer during bearer management. The Bearer Resource Allocation Request is a message sent from the UE to the S-GW to request the allocation of resources for the EPS Bearer during session management. The Bearer Resource Modification Request is a message sent from the UE to the S-GW to request the modification of resources for the EPS Bearer during session management.

[0607] 4. Other Context messages

[0608] Context messages, although not directly defined in the 3GPP standard, contain mobility and traffic-related information that is delivered to the DN while managing the NW in the O&M (Operation and Maintenance) system. Messages include Fault Management and Performance Management, and can include NE (Network Element) ID, RANAP (Radio Access Network Application Part), Alarm Information, etc. They can also include information such as weather and temperature. Since information such as weather and temperature is also related to the UE's movement pattern, an AI / ML model is required to predict the UE's mobility.

[0609] Training data is collected from UE, RAN, and 5G Core over a certain period of time (training data collection period) and the collected information is transmitted to the entity that will deploy the AI / ML model. When deploying the AI / ML model to the DN, the information acquired through Messages 1 to 4 is transmitted to the DN. The DN stores data based on feature sets (Attributes) depending on the type of collected data and trains and deploys the Mobility / Traffic Prediction AI / ML model.

[0610] When there is an inference request, the entity operating the AI / ML Model collects inference data for a certain period of time (Inference data collection period), and the Mobility / Traffic Prediction AI / ML Model predicts mobility and traffic.

[0611] FIG. 25 is a diagram illustrating an example of the operation and prediction process of an AI / ML model in a system applicable to the present disclosure.

[0612] In Fig. 25, the procedure for operating and predicting after deploying the AI / ML model to the DN is shown. 0. In the Inference Request Stage, a prediction request event occurs in the deployed AI / ML Model. 5. In the Inference Data Collection Period Stage, the AI / ML Model collects inference data to be used as input during the inference data collection period. Then, preprocessing is performed so that it can be used as input for the model. 6. In the Inference Output Stage, when collection is complete, the Output (prediction result) is derived through inference. Since the cycle for acquiring data is different for each Message 1 to 4, the Inference Data Collection Period depends on the delay of the final acquired Message. In the example of Fig. 25, data for 25 types of the entire Feature Set (Attributes) must be collected to enable inference.

[0613] As the feature set of inference data required by AI / ML models grows, the number of messages to be collected also increases, leading to increased signaling overhead and resource consumption. To enable base stations to efficiently share AI / ML model information with terminals, a new sharing technique is required.

[0614] For an AI / ML model to perform predictions, inference data must be input to the model, aligned with the feature set used in the training phase. This data has the following characteristics:

[0615] Characteristic 1. The larger the feature set size, the longer it takes to collect inference data and the more overhead there is in data transmission.

[0616] Characteristic 2. A small feature set degrades AI / ML model performance.

[0617] Considering features 1 and 2, an AI / ML model can be created by statically determining an optimally sized feature set based on only specific features. However, the optimal feature set constantly changes depending on the context. Therefore, using a fixed feature set (or applying and operating an AI / ML model trained with a fixed feature set) leads to inefficiencies.

[0618] NW needs to detect contextual changes and dynamically adjust the optimal feature set within that context to operate AI / ML models. Methods for deriving feature sets based on contextual changes and exchanging information about these feature sets are not yet defined, and this needs to be defined in a standard.

[0619] Composition of various embodiments of the present disclosure

[0620] The technology of this disclosure proposes a method of operating separately the AI / ML model for selecting a feature set (attributes) and the operational AI / ML model. It dynamically modifies the operational AI / ML model according to a context change request and defines a new message to transmit the changes to other entities. By optimizing the input feature set required for prediction, the time required for inference data collection and message overhead are minimized. Through this disclosure, the AI / ML model and feature set are dynamically modified according to the context, and only the features appropriate for the context are selected and transmitted, thereby minimizing the overall signaling and resource overhead and enabling the AI / ML model to be updated in an efficient manner.

[0621] Networks (NWs) are affected by various environmental factors. For example, NW UE mobility and traffic are affected by time of day. Mobility and traffic patterns can change significantly depending on day or night. Furthermore, mobility and traffic patterns also differ depending on the type of device or the service it supports, such as whether it's a healthcare service device, a user's mobile phone, or a device for factory automation. As these contexts change, the performance of the learning model applied to the NW can also change. While it's possible to operate an AI / ML model trained for all contextual situations, this increases the size of the feature set. Therefore, new learning is necessary to adapt to the changing cell environment depending on the context, and a method for selecting a new feature set is proposed.

[0622] FIG. 26 is a diagram illustrating an example of a deployment process of a primary AI / ML model in a system applicable to the present disclosure.

[0623] Figure 26 illustrates the development process of the first AI / ML model. After collecting training data for a certain period (training data collection period), the first AI / ML model (feature selection model) is deployed based on the collected full feature set (full attributes). The first AI / ML model (feature selection model) reduces the feature set within a range that maintains the performance tolerance threshold (the first AI / ML model utilizes the full feature set) that the first AI / ML model can achieve in the current context. This is called the sub feature set (sub attributes). Then, the AI / ML model is retrained based on the sub feature set. The AI / ML model trained using the sub feature set is called the second AI / ML model (operation model). If the DN decides to operate the AI / ML model, the DN stores the first AI / ML model and deploys and operates the second AI / ML model. The secondary AI / ML model can reduce the inference data collection period because the feature sets required for inference are small, and the overhead of transmitted messages is also reduced.

[0624] Figure 27 is a diagram illustrating an example of a process for deploying a secondary AI / ML model in a system applicable to the present disclosure.

[0625] Figure 27 illustrates the process of deploying a secondary AI / ML model. For example, assuming a performance tolerance threshold of 10%, the secondary AI / ML model deployment process is as follows. DN creates a primary AI / ML model using the full feature set (25 full attributes including Time, UE ID, Cell ID, …, Policy Rule). The primary AI / ML model is used to evaluate its performance in the current context. Assuming a performance of 97%, a sub-feature set is derived within 87% accuracy, which is within the performance tolerance threshold range. DN creates, deploys, and operates a secondary AI / ML model based on the sub-feature set.

[0626] FIG. 28 is a diagram illustrating an example of the operation and model update process of a secondary AI / ML model in a system applicable to the present disclosure.

[0627] 7. In the Context Change step, if a change in the surrounding Context is detected, DN creates a new secondary AI / ML Model based on the primary AI / ML Model. The procedure for creating the secondary AI / ML Model is the same as the procedure described in Fig. 27. However, since the Context has changed, a change occurs in the Sub Feature set. 8. In the Model Update step, DN sends information about the Sub Feature set required by the new secondary AI / ML Model to other entities that collect information for mobility and traffic prediction through the Model Update Message. The 8. Model Update Message includes information about the Sub Feature set (Attributes) and may additionally include information about the 7. Context Change Event occurrence cycle. In addition, it may include the following information.

[0628] - Data or performance evaluation values ​​that affect the performance evaluation of AI / ML models

[0629] - AI / ML model type (eg, DNN, RNN, etc.)

[0630] - AI / ML model coefficient quantization level (eg, 8bit, 16bit, etc.)

[0631] - Same use case / procedure / processing block

[0632] - AI / ML related capabilities / versions

[0633] - Valid area information, e.g., valid PLMN, cell group area, cell only, etc.

[0634] The present disclosure relates to a dynamic information exchange procedure that separates a feature selection NN from a NN that performs the actual task (e.g., mobility / traffic prediction) to reduce overhead and latency in data movement between network entities, thereby collecting only data corresponding to features selected by the selection NN. In other words, it may relate to life cycle management (LCM) for AI / ML models.

[0635] The prediction NN model for performing the actual task has only selected features as input.

[0636] To further reduce overhead and latency incurred during the creation / deployment process of a prediction NN model, models that only use selected features as input can be learned in advance (e.g., pre-trained) and stored (retained) by the network entity during the pre-training process.

[0637] The number of elements of the feature set in the present disclosure may be referred to by other terms such as feature size or input size for the NN model.

[0638] FIG. 29 is a diagram illustrating an example of a functional framework for RAN Intelligence in TR 37.817 in a system applicable to the present disclosure.

[0639] The technology of the present disclosure assumes the following:

[0640] - It can follow the framework of 3GPP TR 37.817.

[0641] - The AI / ML model in the technology of the present disclosure performs model training in NW.

[0642] - The terminal and / or base station performs model inference using the trained AI / ML model.

[0643] - Based on the results of recognizing the context change, NW performs a secondary AI / ML model update.

[0644] Next, we describe an example of the operation of a terminal initially entering a cell using the technology of the present disclosure. First, a terminal can operate as follows when the first message is defined as one of the SIBs.

[0645] 1. A terminal that initially enters a cell receives basic system information (MIB, SIB1) broadcast from the base station.

[0646] 2. A terminal supporting AI / ML determines that the base station supports AI / ML through system information.

[0647] This can be determined by whether it is a base station transmitting SIB-x (the first message proposed in the technology of this disclosure, hereinafter referred to as SIB15) that transmits AI / ML model information.

[0648] 3. The terminal receives SIB15. At this time, one of the following two methods can be used depending on the base station policy.

[0649] 3-1. When SIB15 is broadcast, SIB15 is received based on scheduling information for the corresponding SIB.

[0650] 3-2. If SIB15 is not broadcast, SIB15 can be received through an on-demand request.

[0651] 4. Receive information about each AI / ML model through SIB15 reception. For terminals that initially enter the network, scheduling information for the Model Update Message is obtained.

[0652] Alternatively, the technology of the present disclosure can be defined so that a terminal receives AI / ML model information (Model Update) without using an SIB. That is, it can be included in a short message or new DCI rather than an SIB, and a terminal that has received AI / ML model information can check whether an update has occurred through the Model Update-related information field in the short message or new DCI. This means that, unlike the previously described embodiment, it can also operate as a separate procedure from the conventional SIB.

[0653] The AI / ML model proposed in the present disclosure includes defining valid area information for each group. This means that valid area information can be defined for each group for AI / ML models affected by context changes. A terminal receiving a message containing this information can determine whether a specific group is within the valid area for each cell movement. If it determines that it has left the valid area, it can update only the AI / ML model for that group. This can be done without an update instruction from the base station.

[0654] Effects of various embodiments of the present disclosure

[0655] The expected effects of various embodiments of the present disclosure are as follows.

[0656] Various embodiments of the present disclosure enable efficient transmission and reception of AI / ML-related information exchanged between a terminal and a network in a wireless communication environment that operates AI / ML models for various purposes to optimize wireless resources. In particular, by dynamically modifying the AI / ML model and feature set according to the context and selecting and transmitting only features appropriate for the context, the overall signaling and resource overhead is minimized, and the AI / ML model can be updated efficiently.

[0657] Characteristic configurations of various embodiments of the present disclosure are as follows.

[0658] (1) In the method of applying and operating AI / ML Model to NW

[0659] A method for updating an AI / ML model and transmitting and receiving messages containing updated AI / ML model information according to changes in NW operating hours, number of UEs served, types of services supported, types of devices, etc.

[0660] (2) NW is a system configured to operate the AI / ML model by dividing it into the first model that uses all the information required for prediction and the second model that uses some information of the performance tolerance threshold.

[0661] (3) When changing the second model, the AI / ML Update message transmitted and received includes at least one of the following: information required for prediction by the second model (Sub Feature Set), data or performance evaluation values ​​affecting the performance evaluation of the AI / ML model, context change and Model Update message scheduling information, AI / ML model type (e.g., DNN, RNN, etc.), AI / ML model coefficient quantization level (e.g., 8 bit, 16 bit, etc.), same use case / procedure / processing block, AI / ML-related capability / version, valid area information, e.g., valid PLMN, cell group area, cell only information.

[0662] [Description of the first node claim]

[0663] The embodiments described below are specifically described with reference to FIG. 30 in terms of the operation of the first node. The methods described below are distinguished for convenience of explanation, and it is understood that some components of one method may be substituted for or combined with some components of another method, as long as they are not mutually exclusive.

[0664] FIG. 30 is a diagram illustrating an example of the operation process of the first node in a system applicable to the present disclosure.

[0665] According to various embodiments of the present disclosure, a method performed by a first node in a communication system is provided.

[0666] According to various embodiments of the present disclosure, each of the first node and the second node may correspond to either a terminal or a base station in a wireless communication system.

[0667] The embodiment of FIG. 30 may further include, before step S3001, one or more of the following steps: a step in which the first node receives one or more synchronization signals from the second node; a step in which the first node receives system information from the second node; a step in which the first node receives configuration information from the second node; and a step in which the first node receives control information from the second node.

[0668] The embodiment of FIG. 30 may further include, before step S3001, one or more of the following steps: a step in which the first node transmits a random access preamble to the second node; a step in which the first node receives a random access response (RAR) from the second node; a step in which the first node transmits a random access message 3 to the second node; and a step in which the first node receives a contention resolution message from the second node. Message 3 is a first PUSCH transmission scheduled by the RAR together with an RAR UL grant.

[0669] In step S3001, the first node transmits first communication environment data to the second node.

[0670] In step S3002, the first node receives model information related to a first secondary AI / ML (artificial intelligence / machine learning) model based on a first sub-feature set related to the first communication environment data from the second node.

[0671] In step S3003, the first node transmits second communication environment data changed from the first communication environment data to the second node.

[0672] In step S3004, the first node receives model update information from the second node about a second secondary AI / ML model that is changed from the first secondary AI / ML model based on a second sub-characteristic set related to the second communication environment data.

[0673] According to various embodiments of the present disclosure, the first sub-feature set may be a feature set selected from the full feature set within a range in which the performance of the first primary AI / ML model based on the first communication environment data and the full feature set can be maintained within a predefined performance tolerance threshold.

[0674] According to various embodiments of the present disclosure, the second sub-feature set may be a feature set selected from the entire feature set within a range in which the performance of the second primary AI / ML model based on the second communication environment data and the entire feature set can be maintained within the performance tolerance threshold.

[0675] According to various embodiments of the present disclosure, the first communication environment data and the second communication environment data may include context information.

[0676] The context information may include one or more of the operating time associated with the second node, the number of user equipment (UE) to which the second node provides service, the type of service supported by the second node, and the type of device supported by the second node.

[0677] According to various embodiments of the present disclosure, the model update information may include information on the occurrence cycle of a change event of the context information.

[0678] According to various embodiments of the present disclosure, the model update information may include one or more of data affecting performance evaluation of the second secondary AI / ML model, a model type of the second secondary AI / ML model, an AI / ML model coefficient quantization level of the second secondary AI / ML model, a public land mobile network (PLMN) associated with the second secondary AI / ML model, and a cell group area associated with the second secondary AI / ML model.

[0679] According to various embodiments of the present disclosure, the embodiment of FIG. 30 may further include a step of receiving scheduling information for the model update information from the second node.

[0680] According to various embodiments of the present disclosure, the model update information can be received based on the scheduling information.

[0681] According to various embodiments of the present disclosure, the first primary AI / ML model, the second primary AI / ML model, the first secondary AI / ML model, and the second secondary AI / ML model may be configured to predict information related to one or more of mobility or traffic volume of the plurality of user equipments within the coverage of the second node.

[0682] According to various embodiments of the present disclosure, a first node is provided in a communication system. The first node includes a transceiver and at least one processor, wherein the at least one processor may be configured to perform the operating method of the first node according to FIG. 30.

[0683] According to various embodiments of the present disclosure, a device for controlling a first node in a communication system is provided. The device includes at least one processor and at least one memory operably connected to the at least one processor. The at least one memory may be configured to store instructions for performing an operating method of the first node according to FIG. 30 based on instructions executed by the at least one processor.

[0684] According to various embodiments of the present disclosure, one or more non-transitory computer-readable media (CRM) storing one or more instructions are provided. The one or more instructions, when executed by one or more processors, perform operations, and the operations may include the operating method of the first node according to FIG. 30.

[0685] [Description of the second node claim]

[0686] The embodiments described below are specifically described with reference to FIG. 31 in terms of the operation of the second node. The methods described below are distinguished for convenience of explanation, and it is understood that some components of one method may be substituted for or combined with some components of another method, as long as they are not mutually exclusive.

[0687] FIG. 31 is a diagram illustrating an example of the operation process of a second node in a system applicable to the present disclosure.

[0688] According to various embodiments of the present disclosure, a method performed by a second node in a communication system is provided.

[0689] According to various embodiments of the present disclosure, each of the first node, the second node, and the third node may correspond to one of a terminal or a base station in a wireless communication system.

[0690] The embodiment of FIG. 31 may further include, before step S3101, one or more of the following steps: a step in which the second node transmits one or more synchronization signals to the first node; a step in which the second node transmits system information to the first node; a step in which the second node transmits configuration information to the first node; and a step in which the second node transmits control information to the first node.

[0691] The embodiment of FIG. 31 may further include, before step S3101, one or more of the following steps: a step in which the second node receives a random access preamble from the first node; a step in which the second node transmits a random access response (RAR) to the first node; a step in which the second node receives a random access message 3 from the first node; and a step in which the second node transmits a contention resolution message to the first node. Message 3 is a first PUSCH transmission scheduled by RAR together with an RAR UL grant.

[0692] In step S3101, the second node transmits the first communication environment data received from the plurality of first nodes to the third node.

[0693] In step S3102, the second node receives model information related to a first secondary AI / ML (artificial intelligence / machine learning) model based on a first sub-feature set related to the first communication environment data from the third node.

[0694] In step S3103, the second node transmits the model information related to the first secondary AI / ML model to the plurality of first nodes.

[0695] In step S3104, the second node transmits second communication environment data received from the plurality of first nodes and changed from the first communication environment data to the third node.

[0696] In step S3105, the second node receives model update information related to a second secondary AI / ML model that is changed from the first secondary AI / ML model based on a second sub-characteristic set related to the second communication environment data from the third node.

[0697] In step S3106, the second node transmits model update information from the first secondary AI / ML model to the second secondary AI / ML model to the plurality of first nodes.

[0698]

[0699] According to various embodiments of the present disclosure, the first sub-feature set may be a feature set selected from the full feature set within a range in which the performance of the first primary AI / ML model based on the first communication environment data and the full feature set can be maintained within a predefined performance tolerance threshold.

[0700] According to various embodiments of the present disclosure, the second sub-feature set may be a feature set selected from the entire feature set within a range in which the performance of the second primary AI / ML model based on the second communication environment data and the entire feature set can be maintained within the performance tolerance threshold.

[0701] According to various embodiments of the present disclosure, the first communication environment data and the second communication environment data may include context information.

[0702] The context information may include one or more of the following: an operating time associated with the second node, the number of the plurality of first nodes for which the second node provides a service, a type of service supported by the second node, and a type of device supported by the second node.

[0703] According to various embodiments of the present disclosure, the model update information may include information on the occurrence cycle of a change event of the context information.

[0704] According to various embodiments of the present disclosure, the model update information may include one or more of data affecting performance evaluation of the second secondary AI / ML model, a model type of the second secondary AI / ML model, an AI / ML model coefficient quantization level of the second secondary AI / ML model, a public land mobile network (PLMN) associated with the second secondary AI / ML model, and a cell group area associated with the second secondary AI / ML model.

[0705] According to various embodiments of the present disclosure, the embodiment of FIG. 31 may further include a step of transmitting scheduling information for the model update information to the plurality of first nodes within the coverage of the second node.

[0706] According to various embodiments of the present disclosure, the model update information may be transmitted based on the scheduling information.

[0707] According to various embodiments of the present disclosure, the first primary AI / ML model, the second primary AI / ML model, the first secondary AI / ML model, and the second secondary AI / ML model may be configured to predict information related to one or more of mobility or traffic volume of the plurality of first nodes within the coverage of the second node.

[0708] According to various embodiments of the present disclosure, a second node is provided in a communication system. The second node includes a transceiver and at least one processor, wherein the at least one processor may be configured to perform the operating method of the second node according to FIG. 31.

[0709] According to various embodiments of the present disclosure, a device for controlling a second node in a communication system is provided. The device includes at least one processor and at least one memory operably connected to the at least one processor. The at least one memory may be configured to store instructions for performing an operating method of the second node according to FIG. 31 based on instructions executed by the at least one processor.

[0710] According to various embodiments of the present disclosure, one or more non-transitory computer-readable media (CRM) storing one or more instructions are provided. The one or more instructions, when executed by one or more processors, perform operations, and the operations may include the operating method of a second node according to FIG. 31.

[0711] Communication system applicable to the present disclosure

[0712] FIG. 32 illustrates a communication system (1) applicable to various embodiments of the present disclosure.

[0713] Referring to FIG. 32, a communication system (1) applicable to various embodiments of the present disclosure includes a wireless device, a base station, and a network. Here, the wireless device refers to a device that performs communication using a wireless access technology (e.g., 5G NR (New RAT), LTE (Long Term Evolution), 6G wireless communication), and may be referred to as a communication / wireless / 5G device / 6G 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 Things) device (100f), and an AI device / server (400). For example, the vehicle may include a vehicle equipped with a wireless communication function, an autonomous vehicle, a vehicle capable of performing vehicle-to-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 may be implemented in the form of a Head-Mounted Device (HMD), a Head-Up Display (HUD) installed in a vehicle, a television, a smartphone, a computer, a wearable device, a home appliance, digital signage, a vehicle, a robot, etc. Mobile devices may include a smartphone, a smart pad, a wearable device (e.g., a smart watch, smart glasses), a computer (e.g., a laptop, etc.), etc. Home appliances may include a TV, a refrigerator, a washing machine, etc. IoT devices may include a sensor, a smart meter, etc. For example, a base station and a network may also be implemented as a wireless device, and a specific wireless device (200a) may act as a base station / network node to other wireless devices.

[0714] Wireless devices (100a to 100f) can be connected to a network (300) via a base station (200). Artificial Intelligence (AI) technology can be applied to the wireless devices (100a to 100f), and the wireless devices (100a to 100f) can be connected to an AI server (400) via the network (300). The network (300) can be configured using a 3G network, a 4G (e.g., LTE) network, a 5G (e.g., NR) network, or a 6G network. The wireless devices (100a to 100f) can communicate with each other via the base station (200) / network (300), but can 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). In addition, IoT devices (e.g., sensors) can communicate directly with other IoT devices (e.g., sensors) or other wireless devices (100a to 100f).

[0715] Wireless communication / connection (150a, 150b, 150c) can be established between wireless devices (100a~100f) / base stations (200), and base stations (200) / base stations (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 communication between base stations (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 each other. For example, wireless communication / connection (150a, 150b, 150c) can transmit / receive signals through various physical channels. To this end, at least some of 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.), and resource allocation processes can be performed based on various proposals of various embodiments of the present disclosure.

[0716] Meanwhile, NR supports multiple numerologies (or subcarrier spacing (SCS)) to support various 5G services. For example, an SCS of 15 kHz supports a wide area in traditional cellular bands; an SCS of 30 kHz / 60 kHz supports dense urban areas, lower latency, and wider carrier bandwidth; and an SCS of 60 kHz or higher supports a bandwidth greater than 24.25 GHz to overcome phase noise.

[0717] The NR frequency band can be defined by two types of frequency ranges (FR1, FR2). The numerical values ​​of the frequency ranges can be changed, and for example, the frequency ranges of the two types (FR1, FR2) can be as shown in Table 3 below. For convenience of explanation, among the frequency ranges used in the NR system, FR1 can mean the "sub 6 GHz range", and FR2 can mean the "above 6 GHz range" and can be called millimeter wave (mmW).

[0718] Frequency Range designationCorresponding frequency rangeSubcarrier SpacingFR1450MHz-6000MHz15, 30, 60kHzFR224250MHz-52600MHz60, 120, 240kHz

[0719] As described above, the numerical value of the frequency range of the NR system can be changed. For example, FR1 may include a band from 410 MHz to 7125 MHz, as shown in Table 4 below. That is, FR1 may include a frequency band above 6 GHz (or 5850, 5900, 5925 MHz, etc.). For example, the frequency band above 6 GHz (or 5850, 5900, 5925 MHz, etc.) included within FR1 may include an unlicensed band. The unlicensed band may be used for various purposes, such as for vehicular communications (e.g., autonomous driving).

[0720] Frequency Range designationCorresponding frequency rangeSubcarrier SpacingFR141MHz-7125MHz15, 30, 60kHzFR224250MHz-52600MHz60, 120, 240kHz

[0721] According to various embodiments of the present disclosure, the communication system (1) can support terahertz (THz) wireless communication. THz wireless communication is a wireless communication using THz waves having a frequency of approximately 0.1 to 10 THz (1 THz = 1012 Hz), and may refer to terahertz (THz) band wireless communication using a very high carrier frequency of 100 GHz or higher. The frequency band expected to be used for THz wireless communication may be a D-band (110 GHz to 170 GHz) or H-band (220 GHz to 325 GHz) band where propagation loss due to absorption of molecules in the air is small.

[0722] Wireless devices applicable to the present disclosure

[0723] Below, examples of wireless devices to which various embodiments of the present disclosure are applied are described.

[0724] FIG. 33 illustrates a wireless device that can be applied to various embodiments of the present disclosure.

[0725] Referring to FIG. 33, the first wireless device (100) and the second wireless device (200) can transmit and receive wireless signals via various wireless access technologies (e.g., LTE, NR). Here, {the first wireless device (100), the second wireless device (200)} can correspond to {the wireless device (100x), the base station (200)} and / or {the wireless device (100x), the wireless device (100x)} of FIG. 32.

[0726] A first wireless device (100) includes one or more processors (102) and one or more memories (104), and may further include one or more transceivers (106) and / or one or more antennas (108). The processor (102) controls the memories (104) and / or the transceivers (106), and may be configured to implement the descriptions, functions, procedures, proposals, methods, and / or operational flowcharts disclosed in this document. For example, the processor (102) may process information in the memory (104) to generate first information / signal, and then transmit a wireless signal including the first information / signal via the transceiver (106). In addition, the processor (102) may receive a wireless signal including second information / signal via the transceiver (106), and then store information obtained from 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 perform some or all of the processes controlled by the processor (102), or may store software code including commands for performing the descriptions, functions, procedures, proposals, methods, and / or operation flowcharts disclosed in this document. Here, the processor (102) and the memory (104) may be part of a communication modem / circuit / chip designed to implement a 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 via one or more antennas (108). The transceiver (106) may include a transmitter and / or a receiver. The transceiver (106) may be used interchangeably with an RF (Radio Frequency) unit. In various embodiments of the present disclosure, a wireless device may mean a communication modem / circuit / chip.

[0727] The second wireless device (200) includes one or more processors (202), one or more memories (204), and may further include one or more transceivers (206) and / or one or more antennas (208). The processor (202) controls the memories (204) and / or the transceivers (206), and may be configured to implement the descriptions, functions, procedures, proposals, methods, and / or operational flowcharts disclosed in this document. For example, the processor (202) may process information in the memory (204) to generate third information / signals, and then transmit a wireless signal including the third information / signals via the transceivers (206). Furthermore, the processor (202) may receive a wireless signal including fourth information / signals via the transceivers (206), and then store information obtained from signal processing of the fourth information / signals 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 perform some or all of the processes controlled by the processor (202), or may store software code including commands for performing the descriptions, functions, procedures, proposals, methods, and / or operation flowcharts 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 via one or more antennas (208). The transceiver (206) may include a transmitter and / or a receiver. The transceiver (206) may be used interchangeably with an RF unit. In various embodiments of the present disclosure, a wireless device may also mean a communication modem / circuit / chip.

[0728] Hereinafter, the 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 one or more Service Data Units (SDUs) according to the descriptions, functions, procedures, proposals, methods, and / or operation flowcharts 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 operation flowcharts disclosed in this document. One or more processors (102, 202) can generate signals (e.g., baseband signals) including PDUs, SDUs, messages, control information, data or information according to the functions, procedures, proposals and / or methods disclosed herein, and provide the signals to one or more transceivers (106, 206). One or more processors (102, 202) can receive signals (e.g., baseband signals) from one or more transceivers (106, 206) and obtain PDUs, SDUs, messages, control information, data or information according to the descriptions, functions, procedures, proposals, methods and / or operational flowcharts disclosed herein.

[0729] One or more processors (102, 202) may be referred to as a controller, a microcontroller, a microprocessor, or a 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 operational 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. The descriptions, functions, procedures, suggestions, methods and / or operation flowcharts disclosed in this document may be implemented using firmware or software configured to perform one or more processors (102, 202) or stored in one or more memories (104, 204) and executed by one or more processors (102, 202). The descriptions, functions, procedures, suggestions, methods and / or operation flowcharts disclosed in this document may be implemented using firmware or software in the form of codes, instructions and / or sets of instructions.

[0730] One or more memories (104, 204) may be coupled to one or more processors (102, 202) and may store various forms of data, signals, messages, information, programs, codes, instructions, and / or commands. The one or more memories (104, 204) may be configured as ROM, RAM, EPROM, flash memory, hard drives, registers, cache memory, computer-readable storage media, and / or combinations thereof. The one or more memories (104, 204) may be located internally and / or externally to the one or more processors (102, 202). Additionally, the one or more memories (104, 204) may be coupled to the one or more processors (102, 202) via various technologies, such as wired or wireless connections.

[0731] One or more transceivers (106, 206) can transmit user data, control information, wireless signals / channels, etc., as mentioned in the methods and / or flowcharts of this document, to one or more other devices. One or more transceivers (106, 206) can receive user data, control information, wireless signals / channels, etc., as mentioned in the descriptions, functions, procedures, proposals, methods and / or flowcharts of this document, from one or more other devices. For example, one or more transceivers (106, 206) can be connected to one or more processors (102, 202) and can transmit and receive wireless signals. For example, one or more processors (102, 202) can 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 coupled 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, or the like, as referred to in the descriptions, functions, procedures, proposals, methods, and / or operational flowcharts disclosed herein, via 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 received user data, control information, wireless signals / channels, etc. from RF band signals to baseband signals in order to process the received user data, control information, wireless signals / channels, etc. using one or more processors (102, 202).One or more transceivers (106, 206) may convert user data, control information, wireless signals / channels, etc. processed by one or more processors (102, 202) from baseband signals to RF band signals. For this purpose, one or more transceivers (106, 206) may include an (analog) oscillator and / or filter.

[0732] FIG. 34 illustrates another example of a wireless device that can be applied to various embodiments of the present disclosure.

[0733] According to FIG. 34, the wireless device may include at least one processor (102, 202), at least one memory (104, 204), at least one transceiver (106, 206), and one or more antennas (108, 208).

[0734] The difference between the example of the wireless device described in FIG. 33 and the example of the wireless device in FIG. 34 is that in FIG. 33, the processor (102, 202) and the memory (104, 204) are separated, but in the example of FIG. 34, the memory (104, 204) is included in the processor (102, 202).

[0735] Here, the specific description of the processor (102, 202), memory (104, 204), transceiver (106, 206), and one or more antennas (108, 208) is as described above, so in order to avoid unnecessary repetition of description, the description of the repeated description is omitted.

[0736] Below, examples of signal processing circuits to which various embodiments of the present disclosure are applied are described.

[0737] Figure 35 illustrates a signal processing circuit for a transmission signal.

[0738] Referring to FIG. 35, the signal processing circuit (1000) may include a scrambler (1010), a modulator (1020), a layer mapper (1030), a precoder (1040), a resource mapper (1050), and a signal generator (1060). Although not limited thereto, the operations / functions of FIG. 35 may be performed in the processor (102, 202) and / or the transceiver (106, 206) of FIG. 33. The hardware elements of FIG. 35 may be implemented in the processor (102, 202) and / or the transceiver (106, 206) of FIG. 33. For example, blocks 1010 to 1060 may be implemented in the processor (102, 202) of FIG. 33. Additionally, blocks 1010 to 1050 may be implemented in the processor (102, 202) of FIG. 33, and block 1060 may be implemented in the transceiver (106, 206) of FIG. 33.

[0739] The codeword can be converted into a wireless signal through the signal processing circuit (1000) of FIG. 35. Here, the codeword is an encoded bit sequence of an information block. The information block may include a transport block (e.g., an UL-SCH transport block, a DL-SCH transport block). The wireless signal may be transmitted through various physical channels (e.g., a PUSCH or a PDSCH).

[0740] Specifically, the codeword can be converted into a bit sequence scrambled by a scrambler (1010). The scramble sequence used for scrambling is generated based on an initialization value, and the initialization value may include ID information of the wireless device, etc. The scrambled bit sequence can be modulated into a modulation symbol sequence by a modulator (1020). The modulation method may include pi / 2-BPSK (pi / 2-Binary Phase Shift Keying), m-PSK (m-Phase Shift Keying), m-QAM (m-Quadrature Amplitude Modulation), etc. The complex modulation symbol sequence can be mapped to one or more transmission layers by a layer mapper (1030). The modulation symbols of each transmission layer can be mapped to the corresponding antenna port(s) by a precoder (1040) (precoding). The output z of the precoder (1040) can be obtained by multiplying the output y of the layer mapper (1030) by a precoding matrix W of N*M. Here, N is the number of antenna ports, and M is the number of transmission layers. Here, the precoder (1040) can perform precoding after performing transform precoding (e.g., DFT transform) on complex modulation symbols. In addition, the precoder (1040) can perform precoding without performing transform precoding.

[0741] The resource mapper (1050) can map modulation symbols of each antenna port to time-frequency resources. The time-frequency resources can include multiple symbols (e.g., CP-OFDMA symbols, DFT-s-OFDMA symbols) in the time domain and multiple subcarriers in the frequency domain. The signal generator (1060) generates a wireless signal from the mapped modulation symbols, and the generated wireless signal can be transmitted to another device through each antenna. To this end, the signal generator (1060) can include an Inverse Fast Fourier Transform (IFFT) module, a Cyclic Prefix (CP) inserter, a Digital-to-Analog Converter (DAC), a frequency uplink converter, etc.

[0742] The signal processing process for receiving signals in a wireless device can be configured in reverse order of the signal processing process (1010 to 1060) of FIG. 35. For example, a wireless device (e.g., 100, 200 of FIG. 33) can receive wireless signals from the outside through an antenna port / transceiver. The received wireless signals can be converted into baseband signals through a signal restorer. For this purpose, the signal restorer can include a frequency downlink converter, an analog-to-digital converter (ADC), a CP remover, and a fast Fourier transform (FFT) module. Thereafter, the baseband signal can be restored to a codeword through a resource demapper process, a postcoding process, a demodulation process, and a descrambling process. The codewords can be restored to the original information blocks through decoding. Accordingly, a signal processing circuit (not shown) for a received signal may include a signal restorer, a resource de-mapper, a postcoder, a demodulator, a de-scrambler, and a decoder.

[0743] Below, examples of wireless device utilization to which various embodiments of the present disclosure are applied are described.

[0744] Figure 36 illustrates another example of a wireless device applicable to various embodiments of the present disclosure. The wireless device may be implemented in various forms depending on the use case / service (see Figure 32).

[0745] Referring to FIG. 36, the wireless device (100, 200) corresponds to the wireless device (100, 200) of FIG. 33 and may be composed of various elements, components, units / units, 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 an additional element (140). The communication unit may include a communication circuit (112) and a 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. 33. For example, the transceiver(s) (114) may include one or more transceivers (106, 206) and / or one or more antennas (108, 208) of FIG. 33. The control unit (120) is electrically connected to the communication unit (110), the memory unit (130), and the additional elements (140) and controls the overall operation of the wireless device. For example, the control unit (120) may control the electrical / mechanical operation of the wireless device based on the program / code / command / information stored in the memory unit (130). In addition, the control unit (120) may transmit information stored in the memory unit (130) to an external device (e.g., another communication device) via a wireless / wired interface through the communication unit (110), or store information received from an external device (e.g., another communication device) via a wireless / wired interface in the memory unit (130).

[0746] The additional element (140) may be configured in various ways depending on the type of the 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. 32, 100a), a vehicle (Fig. 32, 100b-1, 100b-2), an XR device (Fig. 32, 100c), a portable device (Fig. 32, 100d), a home appliance (Fig. 32, 100e), an IoT device (Fig. 32, 100f), a digital broadcasting terminal, a hologram device, a public safety device, an MTC device, a medical device, a fintech device (or a financial device), a security device, a climate / environmental device, an AI server / device (Fig. 32, 400), a base station (Fig. 32, 200), a network node, etc. Wireless devices may be mobile or stationary depending on the use / service.

[0747] In FIG. 36, 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 some may be wirelessly 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 wired, and the control unit (120) and a first unit (e.g., 130, 140) may be wirelessly connected via the communication unit (110). In addition, each element, component, unit / part, and / or module within the wireless device (100, 200) may further include one or more elements. For example, the control unit (120) may be composed of a set of one or more 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.

[0748] Below, the implementation example of Fig. 36 is described in more detail with reference to the drawings.

[0749] Figure 37 illustrates a mobile device applicable to various embodiments of the present disclosure. The mobile device may include a smartphone, a smart pad, a wearable device (e.g., a smartwatch, smartglasses), or a portable computer (e.g., a laptop, etc.). The mobile device may be referred to as a Mobile Station (MS), a User Terminal (UT), a Mobile Subscriber Station (MSS), a Subscriber Station (SS), an Advanced Mobile Station (AMS), or a Wireless Terminal (WT).

[0750] Referring to FIG. 37, the portable device (100) may include an antenna unit (108), a communication unit (110), a control unit (120), a memory unit (130), a power supply unit (140a), an interface unit (140b), and an input / output unit (140c). The antenna unit (108) may be configured as a part of the communication unit (110). Blocks 110 to 130 / 140a to 140c correspond to blocks 110 to 130 / 140 of FIG. 36, respectively.

[0751] The communication unit (110) can transmit and receive signals (e.g., data, control signals, etc.) with other wireless devices and base stations. The control unit (120) can control components of the mobile device (100) to perform various operations. The control unit (120) can include an AP (Application Processor). The memory unit (130) can store data / parameters / programs / codes / commands required for operating the mobile device (100). In addition, the memory unit (130) can store input / output data / information, etc. The power supply unit (140a) supplies power to the mobile device (100) and can include a wired / wireless charging circuit, a battery, etc. The interface unit (140b) can support connection between the mobile device (100) and other external devices. The interface unit (140b) can include various ports (e.g., audio input / output ports, video input / output ports) for connection with external devices. The input / output unit (140c) can input or output video information / signals, audio information / signals, data, and / or information input from a user. The input / output unit (140c) may include a camera, a microphone, a user input unit, a display unit (140d), a speaker, and / or a haptic module.

[0752] For example, in the case of data communication, the input / output unit (140c) obtains information / signals (e.g., touch, text, voice, image, video) input by the user, and the obtained information / signals can be stored in the memory unit (130). The communication unit (110) converts the information / signals stored in the memory into wireless signals, and can directly transmit the converted wireless signals to other wireless devices or to a base station. In addition, the communication unit (110) can receive wireless signals from other wireless devices or base stations, and then restore the received wireless signals to the original information / signals. The restored information / signals can be stored in the memory unit (130) and then output in various forms (e.g., text, voice, image, video, haptic) through the input / output unit (140c).

[0753] FIG. 38 illustrates a vehicle or autonomous vehicle applicable to various embodiments of the present disclosure.

[0754] Vehicles or autonomous vehicles can be implemented as mobile robots, cars, trains, manned or unmanned aerial vehicles (AVs), ships, etc.

[0755] Referring to FIG. 38, 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 a part of the communication unit (110). Blocks 110 / 130 / 140a to 140d correspond to blocks 110 / 130 / 140 of FIG. 36, respectively.

[0756] 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, road side units, etc.), and servers. The control unit (120) can control elements of the vehicle or autonomous vehicle (100) to perform various operations. The control unit (120) can include an ECU (Electronic Control Unit). The drive unit (140a) can drive the vehicle or autonomous vehicle (100) on the ground. The drive unit (140a) can include an engine, a motor, a power train, wheels, brakes, a steering device, etc. The power supply unit (140b) supplies power to the vehicle or autonomous vehicle (100) and can include a wired / wireless charging circuit, a battery, 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 incline sensor, a weight detection sensor, a heading sensor, a position module, a vehicle forward / backward 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 a technology for maintaining a driving lane, a technology for automatically controlling speed such as adaptive cruise control, a technology for automatically driving along a set path, a technology for automatically setting a path and driving when a destination is set, etc.

[0757] 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 route and driving plan based on the acquired data. The control unit (120) can control the drive unit (140a) so that the vehicle or autonomous vehicle (100) moves along the autonomous driving route according to the driving plan (e.g., speed / direction control). During autonomous driving, the communication unit (110) can irregularly / periodically acquire the latest traffic information data from an external server and can acquire surrounding traffic information data from surrounding vehicles. In addition, 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 route and driving plan based on newly acquired data / information. The communication unit (110) can transmit information regarding the vehicle location, autonomous driving route, driving plan, etc. to the external server. External servers can predict traffic information data in advance using AI technology or other technologies based on information collected from vehicles or autonomous vehicles, and provide the predicted traffic information data to the vehicles or autonomous vehicles.

[0758] Figure 39 illustrates a vehicle applicable to various embodiments of the present disclosure. The vehicle may also be implemented as a means of transportation, a train, an aircraft, a ship, or the like.

[0759] Referring to FIG. 39, the vehicle (100) may include a communication unit (110), a control unit (120), a memory unit (130), an input / output unit (140a), and a position measurement unit (140b). Here, blocks 110 to 130 / 140a to 140b correspond to blocks 110 to 130 / 140 of FIG. 36, respectively.

[0760] The communication unit (110) can transmit and receive signals (e.g., data, control signals, etc.) with other vehicles or external devices such as base stations. The control unit (120) can control components of the vehicle (100) to perform various operations. The memory unit (130) can store data / parameters / programs / codes / commands that support various functions of the vehicle (100). The input / output unit (140a) can output AR / VR objects based on information in the memory unit (130). The input / output unit (140a) can include a HUD. The position measurement unit (140b) can obtain position information of the vehicle (100). The position information can include absolute position information of the vehicle (100), position information within a driving line, acceleration information, position information with respect to surrounding vehicles, etc. The position measurement unit (140b) can include GPS and various sensors.

[0761] For example, the communication unit (110) of the vehicle (100) can receive map information, traffic information, etc. from an external server and store them in the memory unit (130). The location measurement unit (140b) can obtain vehicle location information through GPS and various sensors and store the information in the memory unit (130). The control unit (120) can create a virtual object based on the map information, traffic information, and vehicle location information, and the input / output unit (140a) can display the created virtual object on the vehicle window (1410, 1420). In addition, the control unit (120) can determine whether the vehicle (100) is being driven normally within the driving line based on the vehicle location information. If the vehicle (100) abnormally deviates from the driving line, the control unit (120) can display a warning on the vehicle window through the input / output unit (140a). Additionally, the control unit (120) can broadcast a warning message regarding driving abnormalities to surrounding vehicles through the communication unit (110). Depending on the situation, the control unit (120) can transmit vehicle location information and information regarding driving / vehicle abnormalities to relevant authorities through the communication unit (110).

[0762] Figure 40 illustrates an XR device applicable to various embodiments of the present disclosure. The XR device may be implemented as an HMD, a head-up display (HUD) installed in a vehicle, a television, a smartphone, a computer, a wearable device, a home appliance, digital signage, a vehicle, a robot, and the like.

[0763] Referring to FIG. 40, the XR device (100a) may include a communication unit (110), a control unit (120), a memory unit (130), an input / output unit (140a), a sensor unit (140b), and a power supply unit (140c). Here, blocks 110 to 130 / 140a to 140c correspond to blocks 110 to 130 / 140 of FIG. 36, respectively.

[0764] The communication unit (110) can transmit and receive signals (e.g., media data, control signals, etc.) with external devices such as other wireless devices, portable devices, or media servers. The media data can include videos, images, sounds, etc. The control unit (120) can control components of the XR device (100a) to perform various operations. For example, the control unit (120) can be configured to control and / or perform procedures such as video / image acquisition, (video / image) encoding, metadata generation and processing, etc. The memory unit (130) can store data / parameters / programs / codes / commands required for driving the XR device (100a) / generating XR objects. The input / output unit (140a) can obtain control information, data, etc. from the outside, and output the generated XR objects. The input / output unit (140a) can include a camera, a microphone, a user input unit, a display unit, a speaker, and / or a haptic module, etc. The sensor unit (140b) can obtain the XR device status, surrounding environment information, user information, etc. The sensor unit (140b) may include a proximity sensor, an illuminance sensor, an acceleration sensor, a magnetic sensor, a gyro sensor, an inertial sensor, an RGB sensor, an IR sensor, a fingerprint recognition sensor, an ultrasonic sensor, a light sensor, a microphone, and / or a radar. The power supply unit (140c) supplies power to the XR device (100a) and may include a wired / wireless charging circuit, a battery, etc.

[0765] For example, the memory unit (130) of the XR device (100a) may include information (e.g., data, etc.) required for creating an XR object (e.g., AR / VR / MR object). The input / output unit (140a) may obtain a command to operate the XR device (100a) from the user, and the control unit (120) may operate the XR device (100a) according to the user's operating command. For example, when a user attempts to watch a movie, news, etc. through the XR device (100a), the control unit (120) may transmit content request information to another device (e.g., a mobile device (100b)) or a media server through the communication unit (130). The communication unit (130) may download / stream content such as movies and news from another device (e.g., a mobile device (100b)) or a media server to the memory unit (130). The control unit (120) controls and / or performs procedures such as video / image acquisition, (video / image) encoding, and metadata generation / processing for content, and can generate / output an XR object based on information about surrounding space or real objects acquired through the input / output unit (140a) / sensor unit (140b).

[0766] In addition, the XR device (100a) is wirelessly connected to the mobile device (100b) through the communication unit (110), and the operation of the XR device (100a) can be controlled by the mobile device (100b). For example, the mobile device (100b) can act as a controller for the XR device (100a). To this end, the XR device (100a) can obtain three-dimensional position information of the mobile device (100b), and then generate and output an XR object corresponding to the mobile device (100b).

[0767] Figure 41 illustrates robots applicable to various embodiments of the present disclosure. Robots may be classified into industrial, medical, household, military, and other categories depending on their intended use or field.

[0768] Referring to FIG. 41, the robot (100) may include a communication unit (110), a control unit (120), a memory unit (130), an input / output unit (140a), a sensor unit (140b), and a driving unit (140c). Here, blocks 110 to 130 / 140a to 140c correspond to blocks 110 to 130 / 140 of FIG. 36, respectively.

[0769] The communication unit (110) can transmit and receive signals (e.g., driving information, control signals, etc.) with external devices such as other wireless devices, other robots, or control servers. The control unit (120) can control components of the robot (100) to perform various operations. The memory unit (130) can store data / parameters / programs / codes / commands that support various functions of the robot (100). The input / output unit (140a) can obtain information from the outside of the robot (100) and output information to the outside of the robot (100). The input / output unit (140a) can include a camera, a microphone, a user input unit, a display unit, a speaker, and / or a haptic module. The sensor unit (140b) can obtain internal information of the robot (100), surrounding environment information, user information, etc. The sensor unit (140b) may include a proximity sensor, an illuminance sensor, an acceleration sensor, a magnetic sensor, a gyro sensor, an inertial sensor, an IR sensor, a fingerprint recognition sensor, an ultrasonic sensor, a light sensor, a microphone, a radar, etc. The driving unit (140c) may perform various physical operations such as moving the robot joints. In addition, the driving unit (140c) may enable the robot (100) to drive on the ground or fly in the air. The driving unit (140c) may include an actuator, a motor, wheels, brakes, propellers, etc.

[0770] FIG. 42 illustrates an AI device applicable to various embodiments of the present disclosure.

[0771] AI devices can be implemented as fixed or mobile devices, such as TVs, projectors, smartphones, PCs, laptops, digital broadcasting terminals, tablet PCs, wearable devices, set-top boxes (STBs), radios, washing machines, refrigerators, digital signage, robots, and vehicles.

[0772] Referring to FIG. 42, the AI ​​device (100) may include a communication unit (110), a control unit (120), a memory unit (130), an input / output unit (140a / 140b), a learning processor unit (140c), and a sensor unit (140d). Blocks 110 to 130 / 140a to 140d correspond to blocks 110 to 130 / 140 of FIG. 36, respectively.

[0773] The communication unit (110) can transmit and receive wired and wireless signals (e.g., sensor information, user input, learning models, control signals, etc.) with external devices such as other AI devices (e.g., FIG. W1, 100x, 200, 400) or AI servers (200) using wired and wireless communication technology. To this end, the communication unit (110) can transmit information within the memory unit (130) to the external device or transfer a signal received from the external device to the memory unit (130).

[0774] The control unit (120) may determine at least one executable operation of the AI ​​device (100) based on information determined or generated using a data analysis algorithm or a machine learning algorithm. In addition, the control unit (120) may control components of the AI ​​device (100) to perform the determined operation. For example, the control unit (120) may request, search, receive, or utilize data from the learning processor unit (140c) or the memory unit (130), and may control components of the AI ​​device (100) to perform at least one executable operation, a predicted operation, or an operation determined to be desirable. In addition, the control unit (120) may collect history information including the operation contents of the AI ​​device (100) or user feedback on the operation, and store the collected history information in the memory unit (130) or the learning processor unit (140c), or transmit the collected history information to an external device such as an AI server (FIG. W1, 400). The collected history information may be used to update a learning model.

[0775] The memory unit (130) can store data that supports various functions of the AI ​​device (100). For example, the memory unit (130) can store data obtained from the input unit (140a), data obtained from the communication unit (110), output data of the learning processor unit (140c), and data obtained from the sensing unit (140). In addition, the memory unit (130) can store control information and / or software codes necessary for the operation / execution of the control unit (120).

[0776] The input unit (140a) can obtain various types of data from the outside of the AI ​​device (100). For example, the input unit (120) can obtain learning data for model learning, input data to which the learning model will be applied, etc. The input unit (140a) may include a camera, a microphone, and / or a user input unit. The output unit (140b) may generate output related to sight, hearing, or touch. The output unit (140b) may include a display unit, a speaker, and / or a haptic module, etc. The sensing unit (140) can obtain at least one of internal information of the AI ​​device (100), information about the surrounding environment of the AI ​​device (100), and user information using various sensors. The sensing unit (140) may include a proximity sensor, an illuminance sensor, an acceleration sensor, a magnetic sensor, a gyro sensor, an inertial sensor, an RGB sensor, an IR sensor, a fingerprint recognition sensor, an ultrasonic sensor, a light sensor, a microphone, and / or a radar, etc.

[0777] The learning processor unit (140c) can train a model composed of an artificial neural network using learning data. The learning processor unit (140c) can perform AI processing together with the learning processor unit of the AI ​​server (Figure W1, 400). The learning processor unit (140c) can process information received from an external device via the communication unit (110) and / or information stored in the memory unit (130). In addition, the output value of the learning processor unit (140c) can be transmitted to an external device via the communication unit (110) and / or stored in the memory unit (130).

[0778] The claims described in the various embodiments of the present disclosure may be combined in various ways. For example, the technical features of the method claims of the various embodiments of the present disclosure may be combined and implemented as a device, and the technical features of the device claims of the various embodiments of the present disclosure may be combined and implemented as a method. Furthermore, the technical features of the method claims of the various embodiments of the present disclosure may be combined and implemented as a device, and the technical features of the method claims of the various embodiments of the present disclosure may be combined and implemented as a method.

Claims

1. In a method of operating a first node in a wireless communication system, A step of receiving at least one synchronization signal from a second node; A step of receiving a control signal (control information) from the second node; A step of transmitting first communication environment data to the second node; A step of receiving model information related to a first secondary AI / ML (artificial intelligence / machine learning) model based on a first sub-feature set related to the first communication environment data from the second node; A step of transmitting second communication environment data changed from the first communication environment data to the second node; A step of receiving model update information from the second node to a second secondary AI / ML model that is changed from the first secondary AI / ML model based on a second sub-characteristic set related to the second communication environment data, method.

2. In paragraph 1, The first sub-feature set is a feature set selected from the full feature set within a range in which the performance of the first primary AI / ML model based on the first communication environment data and the full feature set can be maintained within a predefined performance tolerance threshold. The second sub-feature set is a feature set selected from the entire feature set within a range where the performance of the second primary AI / ML model based on the second communication environment data and the entire feature set can be maintained within the performance tolerance threshold. method.

3. In paragraph 1, The first communication environment data and the second communication environment data include context information, The context information includes at least one of an operating time associated with the second node, the number of user equipment (UE) to which the second node provides a service, a type of service supported by the second node, and a type of device supported by the second node. method.

4. In paragraph 3, The above model update information includes information on the occurrence cycle of a change event of the above context information. method.

5. In paragraph 2, The model update information includes at least one of data affecting performance evaluation of the second secondary AI / ML model, a model type of the second secondary AI / ML model, an AI / ML model coefficient quantization level of the second secondary AI / ML model, a public land mobile network (PLMN) related to the second secondary AI / ML model, and a cell group area related to the second secondary AI / ML model. method.

6. In paragraph 1, Further comprising a step of receiving scheduling information for the model update information from the second node, The above model update information is received based on the above scheduling information. method.

7. In paragraph 2, The first primary AI / ML model, the second primary AI / ML model, the first secondary AI / ML model, and the second secondary AI / ML model are for predicting information related to one or more of mobility or traffic volume of the plurality of user devices within the coverage of the second node. method.

8. In a method of operating a second node in a wireless communication system, A step of transmitting at least one synchronization signal to a plurality of first nodes within the coverage of the second node; A step of transmitting a control signal (control information) to the plurality of first nodes; A step of transmitting first communication environment data received from the plurality of first nodes to a third node; A step of receiving model information related to a first secondary AI / ML (artificial intelligence / machine learning) model based on a first sub-feature set related to the first communication environment data from the third node; A step of transmitting the model information related to the first secondary AI / ML model to the plurality of first nodes; A step of transmitting second communication environment data received from the plurality of first nodes and changed from the first communication environment data to the third node; A step of receiving model update information related to a second secondary AI / ML model changed from the first secondary AI / ML model based on a second sub-characteristic set related to the second communication environment data from the third node; A step of transmitting model update information from the first secondary AI / ML model to the second secondary AI / ML model to the plurality of first nodes, method.

9. In paragraph 8, The first sub-feature set is a feature set selected from the full feature set within a range in which the performance of the first primary AI / ML model based on the first communication environment data and the full feature set can be maintained within a predefined performance tolerance threshold. The second sub-feature set is a feature set selected from the entire feature set within a range where the performance of the second primary AI / ML model based on the second communication environment data and the entire feature set can be maintained within the performance tolerance threshold. method.

10. In paragraph 8, The first communication environment data and the second communication environment data include context information, The context information includes at least one of an operating time associated with the second node, the number of the plurality of first nodes for which the second node provides a service, a type of service supported by the second node, and a type of device supported by the second node. method.

11. In paragraph 10, The above model update information includes information on the occurrence cycle of a change event of the above context information. method.

12. In paragraph 9, The model update information includes at least one of data affecting performance evaluation of the second secondary AI / ML model, a model type of the second secondary AI / ML model, an AI / ML model coefficient quantization level of the second secondary AI / ML model, a public land mobile network (PLMN) related to the second secondary AI / ML model, and a cell group area related to the second secondary AI / ML model. method.

13. In paragraph 8, Further comprising a step of transmitting scheduling information for the model update information to the plurality of first nodes within the coverage of the second node, The above model update information is transmitted based on the above scheduling information. method.

14. In paragraph 9, The first primary AI / ML model, the second primary AI / ML model, the first secondary AI / ML model, and the second secondary AI / ML model are for predicting information related to one or more of mobility or traffic volume of the plurality of first nodes within the coverage of the second node. method.

15. In a first node in a wireless communication system, Transmitter and receiver; at least one processor; and At least one memory operably connectable to said at least one processor and storing instructions that, when executed by said at least one processor, perform operations; The above actions are, Comprising all steps of the method according to one of claims 1 to 7, Node 1.

16. In a second node in a wireless communication system, Transmitter and receiver; at least one processor; and At least one memory operably connectable to said at least one processor and storing instructions that, when executed by said at least one processor, perform operations; The above actions are, Comprising all steps of a method according to one of claims 8 to 14, Second node.

17. In a control device for controlling a first node in a wireless communication system, at least one processor; and comprising at least one memory operably connected to at least one of the processors; The at least one memory stores instructions for performing operations based on being executed by the at least one processor, The above actions are, Comprising all steps of the method according to one of claims 1 to 7, controller.

18. In a control device for controlling a second node in a wireless communication system, at least one processor; and comprising at least one memory operably connected to at least one of the processors; The at least one memory stores instructions for performing operations based on being executed by the at least one processor, The above actions are, Comprising all steps of a method according to one of claims 8 to 14, controller.

19. In one or more non-transitory computer-readable media storing one or more instructions, The one or more instructions perform operations based on being executed by one or more processors, The above actions are, Comprising all steps of the method according to one of claims 1 to 7, Computer readable medium.

20. In one or more non-transitory computer-readable media storing one or more instructions, The one or more instructions perform operations based on being executed by one or more processors, The above actions are, Comprising all steps of a method according to one of claims 8 to 14, Computer readable medium.

Images