Method and apparatus for supporting ai and ML operations in wireless communication system

Abstract

A method and apparatus for supporting AI and ML operations in a wireless communication system are provided. The wireless device detects a fault in operation with the network; and transmitting a message including (i) information about the fault and (ii) information about predictive information related to the fault, where the predictive information is derived from an artificial intelligence (AI) and / or machine learning (ML) model of the wireless device.

Description

Technical Field

[0001] The present disclosure relates to methods and apparatus for supporting AI and ML operations in wireless communication systems. Background Art

[0002] The 3rd Generation Partnership Project (3GPP) Long Term Evolution (LTE) is a technology that enables high-speed packet communications. Many solutions have been proposed for LTE, including those aimed at reducing user and provider costs, improving service quality, and expanding and improving coverage and system capacity. As high-level requirements, 3GPP LTE requires reduced cost per bit, increased service availability, flexible use of frequency bands, a simple structure, open interfaces, and appropriate power consumption of terminals.

[0003] The International Telecommunication Union (ITU) and 3GPP have begun developing requirements and specifications for New Radio (NR) systems. 3GPP must identify and develop the technical components necessary for the successful standardization of new RATs that will meet both immediate market needs and the longer-term requirements outlined by the ITU Radiocommunication Sector (ITU-R) International Mobile Telecommunications (IMT)-2020 process. Furthermore, NR should be able to use any spectrum band available for wireless communications in the more distant future, at least up to 100 GHz.

[0004] The goal of NR is to be a single technology framework that addresses all use cases, requirements, and deployment scenarios, including enhanced mobile broadband (eMBB), massive machine-type communications (mMTC), ultra-reliable and low-latency communications (URLLC), etc. NR should be inherently forward-compatible. Summary of the Invention

[0005] Technical issues

[0006] A UE supporting AI / ML operation may derive measurement results based on an artificial intelligence (AI) / machine learning (ML)-based method or a non-AI / ML-based method. In the event that a UE experiences a connectivity failure, the UE may report the available measurement results to the network as part of a failure report. Upon receiving the failure report, the network may not clearly determine which of the following is the possible cause of the failure:

[0007] - Case a) The expected failure is caused by inappropriate AI / ML operation of the UE, while the failure is not caused by network coverage issues or other inappropriate UE configurations.

[0008] - Case b) The expected failure is not caused by inappropriate AI / ML operation of the UE, but by network coverage issues or other inappropriate UE configurations.

[0009] This ambiguity from the network side arises primarily because existing fault reporting procedures are insufficient, particularly when the UE is configured or enabled to perform AI / ML-based prediction tasks for measurements and / or other 3GPP procedures. For example, existing fault reporting procedures do not indicate whether the UE is performing AI / ML-based CSI measurement reporting / RRM measurement reporting or traditional measurement reporting, and existing fault reporting procedures do not indicate whether a faulty connection may be caused by incorrectly predicted mobility based on incorrect AI / ML operation by the UE or by incorrect network decisions.

[0010] Even if the network can infer that the fault may be related to AI / ML operations, it is difficult to know how to correct the AI / ML-related operations because existing fault reports lack information related to the AI / ML operations performed by the UE at the time of the fault. As a result, the network may not be able to identify whether AI / ML-related settings need to be changed, and erroneous results in AI / ML operations may have been used as input for training or as interference results.

[0011] Therefore, research is needed on wireless communication systems to support AI and ML operations.

[0012] Technical Solution

[0013] In one aspect, a method performed by a wireless device in a wireless communication system is provided. The method includes the steps of: detecting a failure in operation with the network; and sending a message including (i) information about the failure and (ii) information about predicted information related to the failure, wherein the predicted information is derived by an artificial intelligence (AI) and / or machine learning (ML) model of the wireless device.

[0014] In another aspect, a device for implementing the above method is provided.

[0015] Beneficial effects

[0016] The present disclosure may have various beneficial effects.

[0017] According to some embodiments of the present disclosure, a wireless device may efficiently support AI / ML operations by reporting information related to the AI / ML operations.

[0018] For example, if the network can identify problems with the AI / ML model through fault reports, the network can manage the AI / ML model well from a model monitoring perspective by updating the parameters of the current AI / ML model or changing to a more suitable AI / ML model.

[0019] In other words, for example, the network can efficiently identify AI / ML problems in the UE and can efficiently manage the AI / ML models to be used in the UE (e.g., the UE can efficiently receive configurations of new AI / ML models).

[0020] According to some embodiments of the present disclosure, a wireless network system may provide efficient management of AI / ML operations of wireless devices by receiving information related to the AI / ML operations.

[0021] The advantageous effects that can be obtained by the specific embodiments of the present disclosure are not limited to the advantageous effects listed above. For example, there may be various technical effects that a person of ordinary skill in the relevant art can understand and / or deduce based on the present disclosure. Therefore, the specific effects of the present disclosure are not limited to those explicitly described herein, but can include various effects that can be understood or derived from the technical features of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 An example of a communication system to which an implementation of the present disclosure is applied is shown.

[0023] Figure 2 An example of a wireless device to which an implementation of the present disclosure is applied is shown.

[0024] Figure 3 An example of a wireless device to which an implementation of the present disclosure is applied is shown.

[0025] Figure 4 Another example of a wireless device to which implementations of the present disclosure are applied is shown.

[0026] Figure 5 An example of a UE to which an implementation of the present disclosure is applied is shown.

[0027] Figure 6 and Figure 7 An example of a protocol stack in a 3GPP-based wireless communication system to which implementations of the present disclosure are applied is shown.

[0028] Figure 8 The frame structure in a 3GPP-based wireless communication system to which the implementation of the present disclosure is applied is shown.

[0029] Figure 9 An example of data flow in a 3GPP NR system to which an implementation of the present disclosure is applied is shown.

[0030] Figure 10 An example of a functional framework for RAN intelligence is shown.

[0031] Figure 11Examples of AI / ML model training in OAM and AI / ML model inference in NG-RAN nodes are shown.

[0032] Figure 12 An example of model training and model inference, both located in RAN nodes, is shown.

[0033] Figure 13 and Figure 14 Examples of neurons and neural network architectures are shown.

[0034] Figure 15 An example of AI / ML inference is shown.

[0035] Figure 16 An example of an MLP DNN model is shown.

[0036] Figure 17 An example of a CNN model is shown.

[0037] Figure 18 An example of an RNN model is shown.

[0038] Figure 19 An example of reinforcement learning is shown.

[0039] Figure 20 An example of a method for supporting AI and ML operations in a wireless communication system is shown.

[0040] Figure 21 An example of a method for supporting AI and ML operations in a wireless communication system is shown.

[0041] Figure 22 An example of the time elapsed from receiving an AI / ML model configuration is shown. DETAILED DESCRIPTION

[0042] The following techniques, devices, and systems can be applied to various wireless multiple access systems. Examples of multiple access systems include code division multiple access (CDMA) systems, frequency division multiple access (FDMA) systems, time division multiple access (TDMA) systems, orthogonal frequency division multiple access (OFDMA) systems, single carrier frequency division multiple access (SC-FDMA) systems, and multi-carrier frequency division multiple access (MC-FDMA) systems. CDMA can be implemented using radio technologies such as Universal Terrestrial Radio Access (UTRA) or CDMA2000. TDMA can be implemented using radio technologies such as Global System for Mobile Communications (GSM), General Packet Radio Service (GPRS), or Enhanced Data Rates for GSM Evolution (EDGE). OFDMA can be implemented using radio technologies such as Institute of Electrical and Electronics Engineers (IEEE) 802.11 (Wi-Fi), IEEE 802.16 (WiMAX), IEEE 802.20, or Evolved UTRA (E-UTRA). UTRA is part of Universal Mobile Telecommunications System (UMTS). 3rd Generation Partnership Project (3GPP) Long Term Evolution (LTE) is part of Evolved UMTS (E-UMTS) using E-UTRA. 3GPP LTE adopts OFDMA in DL and SC-FDMA in UL. LTE-Advanced (LTE-A) is an evolved version of 3GPP LTE.

[0043] For ease of description, the implementation of the present disclosure will be primarily described with respect to a 3GPP-based wireless communication system. However, the technical features of the present disclosure are not limited thereto. For example, although the following detailed description is given based on a mobile communication system corresponding to a 3GPP-based wireless communication system, aspects of the present disclosure that are not limited to 3GPP-based wireless communication systems are applicable to other mobile communication systems.

[0044] For terms and techniques not specifically described in the terms and techniques adopted in the present disclosure, reference may be made to wireless communication standard documents issued prior to the present disclosure.

[0045] In the present disclosure, "A or B" may mean "only A", "only B", or "both A and B". In other words, in the present disclosure, "A or B" may be interpreted as "A and / or B". For example, in the present disclosure, "A, B, or C" may mean "only A", "only B", "only C", or "any combination of A, B, and C".

[0046] In the present disclosure, a slash ( / ) or a comma (,) may mean "and / or". For example, "A / B" may mean "A and / or B". Thus, "A / B" may mean "only A", "only B", or "both A and B". For example, "A, B, C" may mean "A, B, or C".

[0047] In the present disclosure, “at least one of A and B” may mean “only A”, “only B”, or “both A and B”. In addition, the expression “at least one of A or B” or “at least one of A and / or B” in the present disclosure may be interpreted as being the same as “at least one of A and B”.

[0048] In addition, in the present disclosure, “at least one of A, B, and C” may mean “only A,” “only B,” “only C,” or “any combination of A, B, and C.” In addition, “at least one of A, B, or C” or “at least one of A, B, and / or C” may mean “at least one of A, B, and C.”

[0049] In addition, the brackets used in the present disclosure may mean "for example". In detail, when "control information (PDCCH)" is shown, "PDCCH" may be proposed as an example of "control information". In other words, in the present disclosure, "control information" is not limited to "PDCCH", and "PDCCH" may be proposed as an example of "control information". In addition, even when "control information (i.e., PDCCH)" is shown, "PDCCH" may be proposed as an example of "control information".

[0050] The technical features described separately in one figure in this disclosure can be implemented separately or simultaneously.

[0051] Although not limited thereto, the various descriptions, functions, processes, suggestions, methods and / or operational flowcharts of the present disclosure disclosed herein may be applied to various fields requiring wireless communication and / or connectivity between devices (e.g., 5G).

[0052] Hereinafter, the present disclosure will be described in more detail with reference to the accompanying drawings. Unless otherwise specified, the same reference numerals in the following drawings and / or descriptions may refer to the same and / or corresponding hardware blocks, software blocks and / or functional blocks.

[0053] Figure 1 An example of a communication system to which an implementation of the present disclosure is applied is shown.

[0054] exist Figure 1 The 5G usage scenarios shown in the present disclosure are only exemplary, and the technical features of the present disclosure can be applied to Figure 1 Other 5G usage scenarios shown in .

[0055] The three main demand categories for 5G include: (1) enhanced mobile broadband (eMBB) category, (2) massive machine type communication (mMTC) category, and (3) ultra-reliable and low-latency communication (URLLC) category.

[0056] Some use cases may require multiple categories for optimization, while others may focus on just one key performance indicator (KPI). 5G supports such a variety of use cases using a flexible and reliable approach.

[0057] eMBB goes far beyond basic mobile internet access and covers rich two-way work and media and entertainment applications in the cloud and augmented reality. Data is one of the core driving forces of 5G, and for the first time in the 5G era, dedicated voice services may not be provided. In 5G, voice is expected to be simply processed as an application using the data connection provided by the communication system. The main reasons for the increase in service capacity are the increase in content size and the increase in the number of applications requiring high data transmission rates. As more and more devices connect to the internet, streaming services (audio and video), conversational video, and mobile internet access will become more widely used. Many of these applications require always-on connectivity to push real-time information and alerts to users. Cloud storage and applications are rapidly increasing in mobile communication platforms and can be applied to both work and entertainment. Cloud storage is a special use case that is accelerating the growth of uplink data transmission rates. 5G is also used for remote work in the cloud. When using tactile interfaces, 5G requires much lower end-to-end latency to maintain a good user experience. Entertainment, such as cloud gaming and video streaming, is another core element that is increasing the demand for mobile broadband capabilities. Entertainment is essential for smartphones and tablets anywhere, including in highly mobile environments such as trains, cars, and airplanes. Another use case is augmented reality for entertainment and information search. In this case, augmented reality requires very low latency and instantaneous data capacity.

[0058] Furthermore, one of the most anticipated 5G use cases involves the ability to seamlessly connect embedded sensors across all fields, known as mMTC. It is expected that the number of potential Internet of Things (IoT) devices will reach 204 billion by 2020. Industrial IoT is one of the key categories that will play a major role in enabling smart cities, asset tracking, smart utilities, agriculture, and security infrastructure through 5G.

[0059] URLLC, which includes remote control and ultra-reliable / available low-latency links over the primary infrastructure, will transform new industrial services (such as autonomous vehicles). This level of reliability and latency is necessary to control smart grids, automate industry, enable robotics, and control and coordinate drones.

[0060] 5G is a means of providing streams estimated to be hundreds of megabits per second to gigabits per second, and can supplement fiber to the home (FTTH) and cable-based broadband (or DOCSIS). Such fast speeds are needed to deliver TV with a resolution of 4K or more (6K, 8K and more), as well as virtual reality and augmented reality. Virtual reality (VR) and augmented reality (AR) applications include almost immersive sports games. Specific applications may require special network configurations. For example, for VR games, game companies need to merge core servers into the network operator's edge network servers to minimize latency.

[0061] Automobiles, along with their numerous use cases for mobile communications, are expected to be a significant new driver for 5G. For example, passenger entertainment will require high simultaneous capacity and mobile broadband with high mobility. This is because future users will continue to expect high-quality connectivity, regardless of their location and speed. Another use case in the automotive sector is augmented reality (AR) dashboards. AR dashboards allow drivers to identify objects in the dark, in addition to those visible through the front window, and display distance to and movement of objects by overlaying information spoken to the driver. In the future, wireless modules will enable communication between vehicles, information exchange between vehicles and supporting infrastructure, and information exchange between vehicles and other connected devices (e.g., devices accompanying pedestrians). Safety systems will guide alternative routes, enabling drivers to drive more safely and thus reducing the risk of accidents. The next stage will be remotely controlled or autonomous vehicles. This requires extremely high reliability and very fast communication between autonomous vehicles and between vehicles and infrastructure. In the future, autonomous vehicles will perform all driving activities, and drivers will only focus on unusual traffic events that the vehicle cannot identify. The technical requirements for autonomous vehicles require ultra-low latency and ultra-high reliability, increasing traffic safety to a level that cannot be achieved by humans.

[0062] Smart cities and smart homes / buildings, often referred to as smart societies, will be embedded in high-density wireless sensor networks. Distributed networks of smart sensors will identify conditions for cost- and energy-efficient maintenance in cities or homes. Similar configurations can be implemented for corresponding homes. All temperature sensors, window and heating controls, burglar alarms, and household appliances will be wirelessly connected. Many of these sensors are typically low in terms of data transmission rate, power, and cost. However, certain types of devices may require real-time HD video for monitoring.

[0063] The consumption and distribution of energy, including heat and gas, is becoming increasingly distributed, necessitating the automated control of distribution sensor networks. Smart grids collect information and use digital information and communication technologies to connect sensors to each other, thereby acting upon this information. Because this information can include the behavior of both supply companies and consumers, smart grids can improve the distribution of fuels such as electricity through methods that enhance efficiency, reliability, economic viability, sustainable production, and automation. Smart grids can also be considered another sensor network with low latency.

[0064] Mission-critical applications (e.g., e-health) are one of the 5G use cases. The health sector includes many applications that can benefit from mobile communications. Communication systems can support telemedicine, which provides clinical treatment in remote locations. Telemedicine can help reduce the barriers of distance and improve access to medical services that are not continuously available in remote rural areas. Telemedicine is also used to perform important treatments and save lives in emergency situations. Wireless sensor networks based on mobile communications can provide remote monitoring and sensors for parameters such as heart rate and blood pressure.

[0065] Wireless and mobile communications are becoming increasingly important in industrial applications. Cabling is expensive to install and maintain. Therefore, the potential to replace cables with reconfigurable radio links presents an attractive opportunity in many industrial sectors. However, to achieve this replacement, wireless connections must have similar latency, reliability, and capacity to cables, and their management must be simplified. When it comes to 5G connectivity, low latency and a very low probability of error are new requirements.

[0066] Logistics and freight tracking are important use cases for mobile communications, allowing inventory and packages to be tracked anywhere using location-based information systems. Logistics and freight tracking use cases typically require low data rates but require location information with wide range and reliability.

[0067] Reference Figure 1 , the communication system 1 includes wireless devices 100a to 100f, a base station (BS) 200, and a network 300. Figure 1 A 5G network is illustrated as an example of the network of the communication system 1 , but implementations of the present disclosure are not limited to the 5G system and may be applied to future communication systems other than the 5G system.

[0068] BS 200 and network 300 may be implemented as wireless devices, and certain wireless devices may operate as BSs / network nodes relative to other wireless devices.

[0069] Wireless devices 100a to 100f represent devices that perform communication using a radio access technology (RAT) (e.g., 5G New RAT (NR) or LTE) and may be referred to as communication / wireless / 5G devices. Wireless devices 100a to 100f may include, but are not limited to, a robot 100a, vehicles 100b-1 and 100b-2, an extended reality (XR) device 100c, a handheld device 100d, a home appliance 100e, an IoT device 100f, and an artificial intelligence (AI) device / server 400. For example, a vehicle may include a vehicle with wireless communication capabilities, an autonomous vehicle, and a vehicle capable of performing communication between vehicles. A vehicle may include an unmanned aerial vehicle (UAV) (e.g., a drone). XR devices may include AR / VR / mixed reality (MR) 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 device, a digital signage, a vehicle, a robot, and the like. Handheld devices can include smartphones, smart tablets, wearable devices (e.g., smart watches or smart glasses), and computers (e.g., laptops). Home appliances can include TVs, refrigerators, and washing machines. IoT devices can include sensors and smart meters.

[0070] In the present disclosure, wireless devices 100a to 100f may be referred to as user equipment (UE). For example, UE may include a cellular phone, a smartphone, a laptop computer, a digital broadcast terminal, a personal digital assistant (PDA), a portable multimedia player (PMP), a navigation system, a tablet-shaped personal computer (PC), a tablet PC, an ultrabook, a vehicle, a vehicle with an autonomous driving function, a connected car, a UAV, an AI module, a robot, an AR device, a VR device, an MR device, a hologram device, a public safety device, an MTC device, an IoT device, a medical device, a Fintech device (or a financial device), a security device, a weather / environmental device, a device related to 5G services, or a device related to the fourth industrial evolution field.

[0071] A UAV may be, for example, an aerial vehicle that is piloted by wireless control signals without a human on board.

[0072] VR devices may include, for example, devices for realizing objects or backgrounds in a virtual world. AR devices may include, for example, devices that realize this by connecting objects or backgrounds in a virtual world to objects or backgrounds in the real world. MR devices may include, for example, devices that realize this by merging objects or backgrounds in a virtual world into objects or backgrounds in the real world. Hologram devices may include, for example, devices for realizing 360-degree stereoscopic images by recording and reproducing stereoscopic information, which uses the interference phenomenon of light generated when two lasers meet, known as holographic imaging.

[0073] Public safety devices may include, for example, image relay devices or image devices wearable on a user's body.

[0074] MTC devices and IoT devices may be devices that do not require direct human intervention or manipulation, for example, and may include smart meters, vending machines, thermometers, smart light bulbs, door locks, or various sensors.

[0075] A medical device may be, for example, a device used for the purpose of diagnosing, treating, alleviating, curing, or preventing a disease. For example, a medical device may be a device used for the purpose of diagnosing, treating, alleviating, or correcting an injury or damage. For example, a medical device may be a device used for the purpose of inspecting, replacing, or modifying a structure or function. For example, a medical device may be a device used for the purpose of regulating pregnancy. For example, a medical device may include a device for treatment, a device for operation, a device for (in vitro) diagnosis, a hearing aid, or a device for surgery.

[0076] The safety device may be, for example, a device installed to prevent possible danger and maintain safety. For example, the safety device may be a camera, a closed-circuit TV (CCTV), a recorder, or a black box.

[0077] A Fintech device may be, for example, a device that can provide financial services such as mobile payments. For example, a Fintech device may include a payment device or a point of sale (POS) system.

[0078] Weather / environmental devices may include, for example, devices for monitoring or predicting weather / environmental conditions.

[0079] Wireless devices 100a to 100f can connect to network 300 via BS 200. AI technology can be applied to wireless devices 100a to 100f, and wireless devices 100a to 100f can connect to AI server 400 via network 300. Network 300 can be configured using 3G networks, 4G (e.g., LTE) networks, 5G (e.g., NR) networks, and beyond 5G networks. Although wireless devices 100a to 100f can communicate with each other via BS 200 / network 300, wireless devices 100a to 100f can perform direct communication (e.g., sidelink communication) with each other without going through BS 200 / network 300. For example, vehicles 100b-1 and 100b-2 can perform direct communication (e.g., vehicle-to-vehicle (V2V) / vehicle-to-everything (V2X) communication). IoT devices (e.g., sensors) can perform direct communication with other IoT devices (e.g., sensors) or other wireless devices 100a to 100f.

[0080] Wireless communications / connections 150a, 150b, and 150c may be established between wireless devices 100a to 100f and / or between wireless devices 100a to 100f and BS 200 and / or between BSs 200. Here, wireless communications / connections may be established via various RATs (e.g., 5G NR), such as uplink / downlink communication 150a, sidelink communication (or device-to-device (D2D) communication) 150b, and inter-base station communication 150c (e.g., relay, integrated access and backhaul (IAB)). Wireless devices 100a to 100f and BS 200 / wireless devices 100a to 100f may transmit / receive radio signals to / from each other via wireless communications / connections 150a, 150b, and 150c. For example, wireless communications / connections 150a, 150b, and 150c may transmit / receive signals via various physical channels. To this end, various configuration information configuration processes for sending / receiving radio signals, various signal processing processes (e.g., channel coding / decoding, modulation / demodulation, and resource mapping / demapping), and at least a portion of the resource allocation process can be performed based on the various proposals of the present disclosure.

[0081] Here, the radio communication technology implemented in the wireless device in the present disclosure may include narrowband Internet of Things (NB-IoT) technology for low-power communication as well as LTE, NR and 6G. For example, NB-IoT technology may be an example of low-power wide area network (LPWAN) technology, may be implemented in specifications such as LTE Cat NB1 and / or LTE Cat NB2, and may not be limited to the above names. Additionally and / or alternatively, the radio communication technology implemented in the wireless device in the present disclosure may communicate based on LTE-M technology. For example, LTE-M technology may be an example of LPWAN technology and may be referred to as various names such as enhanced machine type communication (eMTC). For example, LTE-M technology may be implemented in at least one of various specifications such as 1) LTE Cat 0, 2) LTE Cat M1, 3) LTE Cat M2, 4) LTE non-bandwidth limited (non-BL), 5) LTE-MTC, 6) LTE machine type communication, and / or 7) LTE M, and may not be limited to the above names. Additionally and / or alternatively, the radio communication technology implemented in the wireless device of the present disclosure may include at least one of ZigBee, Bluetooth, and / or LPWAN, which are considered low-power communication technologies, and may not be limited to the above names. For example, ZigBee technology can generate a personal area network (PAN) associated with small / low-power digital communication based on various specifications such as IEEE 802.15.4, and may be referred to by various names.

[0082] Figure 2An example of a wireless device to which an implementation of the present disclosure is applied is shown.

[0083] Reference Figure 2 , the first wireless device 100 and the second wireless device 200 can transmit / receive radio signals to / from an external device through various RATs (eg, LTE and NR). Figure 2 In the example, {first wireless device 100 and second wireless device 200} may correspond to the attached Figure 1 At least one of {wireless devices 100a to 100f and BS200}, {wireless devices 100a to 100f and wireless devices 100a to 100f} and / or {BS200 and BS200}.

[0084] The first wireless device 100 may include one or more processors 102 and one or more memories 104, and may also include one or more transceivers 106 and / or one or more antennas 108. The processor 102 may control the memory 104 and / or the transceiver 106 and may be configured to implement the descriptions, functions, processes, suggestions, methods, and / or operational flowcharts described in this disclosure. For example, the processor 102 may process information within the memory 104 to generate first information / signals, and then transmit a radio signal including the first information / signals through the transceiver 106. The processor 102 may receive a radio signal including second information / signals through the transceiver 106, and then store information obtained by processing the second information / signals in the memory 104. The memory 104 may be connected to the processor 102 and may store various information related to the operation of the processor 102. For example, the memory 104 may store software code including commands for executing part or all of the processes controlled by the processor 102 or for executing the descriptions, functions, processes, suggestions, methods, and / or operational flowcharts described in this disclosure. In this document, the processor 102 and the memory 104 may be part of a communication modem / circuit / chip designed to implement a RAT (e.g., LTE or NR). The transceiver 106 may be connected to the processor 102 and transmit and / or receive radio signals via one or more antennas 108. Each of the transceivers 106 may include a transmitter and / or a receiver. The transceiver 106 may be used interchangeably with a radio frequency (RF) unit. In this disclosure, the first wireless device 100 may represent a communication modem / circuit / chip.

[0085] The second wireless device 200 may include one or more processors 202 and one or more memories 204, and may also include one or more transceivers 206 and / or one or more antennas 208. The processor 202 may control the memory 204 and / or the transceiver 206 and may be configured to implement the descriptions, functions, processes, suggestions, methods, and / or operational flowcharts described in this disclosure. For example, the processor 202 may process information within the memory 204 to generate third information / signals, and then transmit a radio signal including the third information / signals through the transceiver 206. The processor 202 may receive a radio signal including fourth information / signals through the transceiver 106, and then store information obtained by processing 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 store software code including commands for executing part or all of the processes controlled by the processor 202 or for executing the descriptions, functions, processes, suggestions, methods, and / or operational flowcharts described in this disclosure. Herein, the processor 202 and the memory 204 may be part of a communication modem / circuit / chip designed to implement a RAT (e.g., LTE or NR). The transceiver 206 may be connected to the processor 202 and transmit and / or receive radio signals via one or more antennas 208. Each of the transceivers 206 may include a transmitter and / or a receiver. The transceiver 206 may be used interchangeably with an RF unit. In the present disclosure, the second wireless device 200 may represent a communication modem / circuit / chip.

[0086] In the following, the hardware elements of the wireless devices 100 and 200 will be described in more detail. One or more protocol layers may be implemented by, but not limited to, one or more processors 102 and 202. For example, one or more processors 102 and 202 may implement one or more layers (e.g., functional layers such as a physical (PHY) layer, a medium access control (MAC) layer, a radio link control (RLC) layer, a packet data convergence protocol (PDCP) layer, a radio resource control (RRC) layer, and a service data adaptation protocol (SDAP) layer). According to the descriptions, functions, processes, suggestions, methods, and / or operational flowcharts disclosed in the present disclosure, one or more processors 102 and 202 may generate one or more protocol data units (PDUs) and / or one or more service data units (SDUs). One or more processors 102 and 202 may generate messages, control information, data, or information according to the descriptions, functions, processes, suggestions, methods, and / or operational flowcharts disclosed in the present disclosure. The one or more processors 102 and 202 may generate a signal (e.g., a baseband signal) including a PDU, an SDU, a message, control information, data, or information according to the description, functions, procedures, suggestions, methods, and / or operational flowcharts disclosed in the present disclosure, and provide the generated signal to the one or more transceivers 106 and 206. The one or more processors 102 and 202 may receive a signal (e.g., a baseband signal) from the one or more transceivers 106 and 206, and obtain the PDU, SDU, message, control information, data, or information according to the description, functions, procedures, suggestions, methods, and / or operational flowcharts disclosed in the present disclosure.

[0087] One or more processors 102 and 202 may be referred to as controllers, microcontrollers, microprocessors, or microcomputers. One or more processors 102 and 202 may be implemented by hardware, firmware, software, or a combination thereof. As an 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 and 202. The descriptions, functions, processes, suggestions, methods, and / or operational flowcharts disclosed in this disclosure may be implemented using firmware or software, and the firmware or software may be configured to include modules, processes, or functions. Firmware or software configured to execute the descriptions, functions, processes, suggestions, methods, and / or operational flowcharts disclosed in this disclosure may be included in one or more processors 102 and 202, or stored in one or more memories 104 and 204, thereby being driven by one or more processors 102 and 202. The descriptions, functions, processes, suggestions, methods and / or operational flowcharts disclosed in this disclosure may be implemented using firmware or software in the form of codes, commands and / or command sets.

[0088] One or more memories 104 and 204 can be connected to one or more processors 102 and 202 and store various types of data, signals, messages, information, programs, codes, instructions and / or commands. One or more memories 104 and 204 can be configured by read-only memory (ROM), random access memory (RAM), electrically erasable programmable read-only memory (EPROM), flash memory, hard drive, registers, cache memory, computer-readable storage media and / or combinations thereof. One or more memories 104 and 204 can be located internally and / or externally to one or more processors 102 and 202. One or more memories 104 and 204 can be connected to one or more processors 102 and 202 by various technologies such as wired or wireless connections.

[0089] One or more transceivers 106 and 206 can transmit user data, control information, and / or radio signals / channels mentioned in the descriptions, functions, processes, suggestions, methods, and / or operational flowcharts disclosed in this disclosure to one or more other devices. One or more transceivers 106 and 206 can receive user data, control information, and / or radio signals / channels mentioned in the descriptions, functions, processes, suggestions, methods, and / or operational flowcharts disclosed in this disclosure from one or more other devices. For example, one or more transceivers 106 and 206 can be connected to one or more processors 102 and 202 and transmit and receive radio signals. For example, one or more processors 102 and 202 can perform control so that one or more transceivers 106 and 206 can transmit user data, control information, or radio signals to one or more other devices. One or more processors 102 and 202 can perform control so that one or more transceivers 106 and 206 can receive user data, control information, or radio signals from one or more other devices.

[0090] One or more transceivers 106 and 206 may be connected to one or more antennas 108 and 208, and the one or more transceivers 106 and 206 may be configured to transmit and receive user data, control information, and / or radio signals / channels mentioned in the descriptions, functions, processes, suggestions, methods, and / or operational flowcharts disclosed in the present disclosure through the one or more antennas 108 and 208. In the present disclosure, the one or more antennas may be multiple physical antennas or multiple logical antennas (e.g., antenna ports).

[0091] One or more transceivers 106 and 206 may convert received radio signals / channels, etc., from RF band signals to baseband signals in order to process received user data, control information, radio signals / channels, etc. One or more transceivers 106 and 206 may convert user data, control information, radio signals / channels, etc., processed using one or more processors 102 and 202, from baseband signals to RF band signals. To this end, one or more transceivers 106 and 206 may include (analog) oscillators and / or filters. For example, under the control of processors 102 and 202, transceivers 106 and 206 may up-convert an OFDM baseband signal to a carrier frequency using their (analog) oscillators and / or filters, and transmit the up-converted OFDM signal at the carrier frequency. Transceivers 106 and 206 may receive an OFDM signal at a carrier frequency and, under the control of processors 102 and 202, down-convert the OFDM signal to an OFDM baseband signal using their (analog) oscillators and / or filters.

[0092] In implementations of the present disclosure, a UE may operate as a transmitting device in the uplink (UL) and as a receiving device in the downlink (DL). In implementations of the present disclosure, a base station (BS) may operate as a receiving device in the UL and as a transmitting device in the DL. Hereinafter, for ease of description, it is primarily assumed that the first wireless device 100 acts as a UE and the second wireless device 200 acts as a base station (BS). For example, the processor 102 connected to, installed on, or activated in the first wireless device 100 may be configured to perform UE behavior according to implementations of the present disclosure, or to control the transceiver 106 to perform UE behavior according to implementations of the present disclosure. The processor 202 connected to, installed on, or activated in the second wireless device 200 may be configured to perform BS behavior according to implementations of the present disclosure, or to control the transceiver 206 to perform BS behavior according to implementations of the present disclosure.

[0093] In this disclosure, a BS is also referred to as a Node B (NB), an eNodeB (eNB), or a gNB.

[0094] Figure 3 An example of a wireless device to which an implementation of the present disclosure is applied is shown.

[0095] The wireless device may be implemented in various forms depending on the use case / service (see Figure 1 ).

[0096] Reference Figure 3 , the wireless devices 100 and 200 may correspond to Figure 2 The wireless devices 100 and 200 may be configured by various elements, components, units / parts and / or modules. For example, each of the wireless devices 100 and 200 may include a communication unit 110, a control unit 120, a memory unit 130 and an additional component 140. The communication unit 110 may include a communication circuit 112 and a transceiver 114. For example, the communication circuit 112 may include Figure 2 One or more processors 102 and 202 and / or Figure 2 One or more memories 104 and 204. For example, the transceiver 114 may include Figure 2 One or more transceivers 106 and 206 and / or Figure 2The control unit 120 is electrically connected to the communication unit 110, the memory unit 130, and the additional components 140, and controls the overall operation of each of the wireless devices 100 and 200. For example, the control unit 120 can control the electrical / mechanical operation of each of the wireless devices 100 and 200 based on the program / code / command / information stored in the memory unit 130. The control unit 120 can transmit information stored in the memory unit 130 to the outside (e.g., other communication devices) through a wireless / wired interface via the communication unit 110, or store information received from the outside (e.g., other communication devices) through a wireless / wired interface in the memory unit 130 via the communication unit 110.

[0097] The additional component 140 may be configured differently depending on the type of the wireless devices 100 and 200. For example, the additional component 140 may include at least one of a power supply unit / battery, an input / output (I / O) unit (e.g., an audio I / O port, a video I / O port), a driving unit, and a computing unit. The wireless devices 100 and 200 may be configured in the form of, but not limited to, robots ( Figure 1 100a), vehicles ( Figure 1 100b-1 and 100b-2), XR devices ( Figure 1 100c), handheld device ( Figure 1 100d), household appliances ( Figure 1 100e), IoT devices ( Figure 1 100f), digital broadcast terminal, hologram device, public safety device, MTC device, medical device, Fintech device (or financial device), security device, climate / environmental device, AI server / device ( Figure 1 400), BSS( Figure 1 The wireless devices 100 and 200 may be implemented in the form of a mobile or fixed location, depending on the use case / service.

[0098] exist Figure 3In the wireless devices 100 and 200, the various elements, components, units / parts, and / or modules in their entirety may be connected to each other via a wired interface, or at least a portion thereof may be wirelessly connected via the communication unit 110. For example, in each of the wireless devices 100 and 200, the control unit 120 and the communication unit 110 may be connected via a wired interface, and the control unit 120 and the first unit (e.g., 130 and 140) may be wirelessly connected via the communication unit 110. Each element, component, unit / part, and / or module within the wireless devices 100 and 200 may also include one or more elements. For example, the control unit 120 may be configured by a group of one or more processors. As an example, the control unit 120 may be configured by a group of a communication control processor, an application processor (AP), an electronic control unit (ECU), a graphics processing unit, and a memory control processor. As another example, the memory unit 130 may be configured by RAM, DRAM, ROM, flash memory, volatile memory, non-volatile memory, and / or a combination thereof.

[0099] Figure 4 Another example of a wireless device to which implementations of the present disclosure are applied is shown.

[0100] Reference Figure 4 , the wireless devices 100 and 200 may correspond to Figure 2 The wireless devices 100 and 200 may be configured by various elements, components, units / portions and / or modules.

[0101] First wireless device 100 may include at least one transceiver, such as transceiver 106, and at least one processing chip, such as processing chip 101. Processing chip 101 may include at least one processor, such as processor 102, and at least one memory, such as memory 104. Memory 104 may be operably connected to processor 102. Memory 104 may store various types of information and / or instructions. Memory 104 may store software code 105 that, when executed by processor 102, implements instructions for performing the descriptions, functions, procedures, suggestions, methods, and / or operational flowcharts disclosed in this disclosure. For example, software code 105 may implement instructions for performing the descriptions, functions, procedures, suggestions, methods, and / or operational flowcharts disclosed in this disclosure, when executed by processor 102. For example, software code 105 may control processor 102 to execute one or more protocols. For example, software code 105 may control processor 102 to execute one or more layers of a radio interface protocol.

[0102] The second wireless device 200 may include at least one transceiver, such as transceiver 206, and at least one processing chip, such as processing chip 201. Processing chip 201 may include at least one processor, such as processor 202, and at least one memory, such as memory 204. Memory 204 may be operably connected to processor 202. Memory 204 may store various types of information and / or instructions. Memory 204 may store software code 205 that, when executed by processor 202, implements instructions for performing the descriptions, functions, procedures, suggestions, methods, and / or operational flowcharts disclosed in this disclosure. For example, software code 205 may implement instructions for performing the descriptions, functions, procedures, suggestions, methods, and / or operational flowcharts disclosed in this disclosure, when executed by processor 202. For example, software code 205 may control processor 202 to execute one or more protocols. For example, software code 205 may control processor 202 to execute one or more layers of a wireless interface protocol.

[0103] Figure 5 An example of a UE to which an implementation of the present disclosure is applied is shown.

[0104] Reference Figure 5 , UE 100 may correspond to the attached Figure 2 The first wireless device 100 and / or Figure 4 The first wireless device 100 is configured to:

[0105] UE 100 includes a processor 102 , memory 104 , a transceiver 106 , one or more antennas 108 , a power management module 110 , a battery 1112 , a display 114 , a keypad 116 , a subscriber identity module (SIM) card 118 , a speaker 120 , and a microphone 122 .

[0106] The processor 102 may be configured to implement the descriptions, functions, processes, suggestions, methods, and / or operational flowcharts disclosed in the present disclosure. The processor 102 may be configured to control one or more other components of the UE 100 to implement the descriptions, functions, processes, suggestions, methods, and / or operational flowcharts disclosed in the present disclosure. The layers of the radio interface protocol may be implemented in the processor 102. The processor 102 may include an ASIC, other chipsets, logic circuits, and / or data processing devices. The processor 102 may be an application processor. The processor 102 may include at least one of a digital signal processor (DSP), a central processing unit (CPU), a graphics processing unit (GPU), and a modem (modulator and demodulator). Examples of the processor 102 may be found at SNAPDRAGON MANUFACTURED TM series processors, Manufactured by EXYNOSTM series processors, A series processors manufactured by HELIO manufactured TM series processors, ATOM manufactured TM series processors or corresponding to the next generation processors.

[0107] The memory 104 is operatively coupled to the processor 102 and stores a variety of information to operate the processor 102. The memory 104 may include ROM, RAM, flash memory, a memory card, a storage medium, and / or other storage devices. When the embodiment is implemented in software, the techniques described herein may be implemented with modules (e.g., processes, functions, etc.) that execute the descriptions, functions, processes, suggestions, methods, and / or operational flow charts disclosed in this disclosure. The modules may be stored in the memory 104 and executed by the processor 102. The memory 104 may be implemented within the processor 102 or external to the processor 102, in which case the memory 104 may be communicatively coupled to the processor 102 via various means known in the art.

[0108] The transceiver 106 is operatively coupled to the processor 102 and transmits and / or receives radio signals. The transceiver 106 includes a transmitter and a receiver. The transceiver 106 may include baseband circuitry for processing radio frequency signals. The transceiver 106 controls one or more antennas 108 to transmit and / or receive radio signals.

[0109] The power management module 110 manages power to the processor 102 and / or the transceiver 106. The power management module 110 is powered by a battery 112.

[0110] The display 114 outputs the results processed by the processor 102. The keypad 116 receives input to be used by the processor 102. The keypad 116 may be displayed on the display 114.

[0111] The SIM card 118 is an integrated circuit designed to securely store an International Mobile Subscriber Identity (IMSI) number and its associated keys, which are used to identify and authenticate subscribers on mobile telephony devices such as mobile phones and computers. Contact information can also be stored on many SIM cards.

[0112] The speaker 120 outputs sound-related results processed by the processor 102. The microphone 122 receives sound-related input to be used by the processor 102.

[0113] Figure 6 and Figure 7 An example of a protocol stack in a 3GPP-based wireless communication system to which implementations of the present disclosure are applied is shown.

[0114] Specifically, Figure 6 illustrates an example of a radio interface user plane protocol stack between a UE and a BS, and Figure 7 An example of a radio interface control plane protocol stack between a UE and a BS is illustrated. The control plane refers to a path through which control messages for managing calls between the UE and the network are transmitted. The user plane refers to a path through which data generated in the application layer (for example, voice data or Internet packet data) is transmitted. Figure 6 , the user plane protocol stack can be divided into layer 1 (ie, PHY layer) and layer 2. Figure 7 , the control plane protocol stack can be divided into layer 1 (ie, PHY layer), layer 2, layer 3 (eg, RRC layer) and non-access stratum (NAS) layer. Layer 1, layer 2 and layer 3 are called access stratum (AS).

[0115] In 3GPP LTE systems, Layer 2 is divided into the following sublayers: MAC, RLC, and PDCP. In 3GPP NR systems, Layer 2 is divided into the following sublayers: MAC, RLC, PDCP, and SDAP. The PHY layer provides transport channels to the MAC sublayer, the MAC sublayer provides logical channels to the RLC sublayer, the RLC sublayer provides RLC channels to the PDCP sublayer, and the PDCP sublayer provides radio bearers to the SDAP sublayer. The SDAP sublayer provides Quality of Service (QoS) flows to the 5G core network.

[0116] In 3GPP NR systems, the main services and functions of the MAC sublayer include: mapping between logical channels and transport channels; multiplexing / demultiplexing MAC SDUs belonging to one or different logical channels onto / from transport blocks (TBs) delivered to / from the physical layer on transport channels; scheduling information reporting; error correction through hybrid automatic repeat request (HARQ) (one HARQ entity per cell in the case of carrier aggregation (CA); priority handling between UEs through dynamic scheduling; priority handling between logical channels of a UE through logical channel prioritization; and padding. A single MAC entity can support multiple parameter sets, transmission timings, and cells. Mapping restrictions in logical channel prioritization control which parameter set(s), cell, and transmission timing can be used by a logical channel.

[0117] MAC provides different types of data transfer services. To accommodate different types of data transfer services, multiple types of logical channels are defined, that is, each logical channel supports the transmission of a specific type of information. Each logical channel type is defined by the type of information transmitted. Logical channels are divided into two groups: control channels and traffic channels. Control channels are used only for the transmission of control plane information, and traffic channels are used only for the transmission of user plane information. The Broadcast Control Channel (BCCH) is a downlink logical channel used to broadcast system control information, the Paging Control Channel (PCCH) is a downlink logical channel that transmits paging information, system information change notifications, and indications of ongoing Public Warning Service (PWS) broadcasts, the Common Control Channel (CCCH) is a logical channel used to send control information between the UE and the network and is used by UEs that do not have an RRC connection with the network, and the Dedicated Control Channel (DCCH) is a point-to-point bidirectional logical channel that sends dedicated control information between the UE and the network and is used by UEs with an RRC connection. The Dedicated Traffic Channel (DTCH) is a point-to-point logical channel dedicated to one UE and is used to transmit user information. The DTCH can exist in both the uplink and downlink. In the downlink, the following connections exist between logical channels and transport channels: BCCH can be mapped to the broadcast channel (BCH); BCCH can be mapped to the downlink shared channel (DL-SCH); PCCH can be mapped to the paging channel (PCH); CCCH can be mapped to DL-SCH; DCCH can be mapped to DL-SCH; and DTCH can be mapped to DL-SCH. In the uplink, the following connections exist between logical channels and transport channels: CCCH can be mapped to the uplink shared channel (UL-SCH); DCCH can be mapped to UL-SCH; and DTCH can be mapped to UL-SCH.

[0118] The RLC sublayer supports three transmission modes: transparent mode (TM), unacknowledged mode (UM), and acknowledged mode (AM). The RLC configuration is per logical channel, without dependency on parameter sets and / or transmission duration. In 3GPP NR systems, the main services and functions of the RLC sublayer depend on the transmission mode and include: delivery of upper layer PDUs; sequence numbering independent of sequence numbering in PDCP (UM and AM); error correction through ARQ (AM only); segmentation (AM and UM) and re-segmentation (AM only) of RLC SDUs; reassembly of SDUs (AM and UM); duplicate detection (AM only); RLC SDU discard (AM and UM); RLC re-establishment; protocol error detection (AM).

[0119] In the 3GPP NR system, the main services and functions of the PDCP sublayer for the user plane include: sequence numbering; header compression and decompression using Robust Header Compression (RoHC); delivery of user data; reordering and duplicate detection; in-sequence delivery; PDCP PDU routing (in the case of split bearers); retransmission of PDCP SDUs; ciphering, deciphering, and integrity protection; PDCP SDU discard; PDCP re-establishment and data recovery for RLC AM; PDCP status reporting for RLC AM; PDCP PDU duplication and duplicate discard indication to lower layers. The main services and functions of the PDCP sublayer for the control plane include: sequence numbering; ciphering, deciphering, and integrity protection; delivery of control plane data; reordering and duplicate detection; in-sequence delivery; PDCP PDU duplication and duplicate discard indication to lower layers.

[0120] In 3GPP NR systems, the main services and functions of SDAP include: mapping between QoS flows and data radio bearers; marking QoS flow IDs (QFIs) in both DL and UL packets. A single protocol entity of SDAP is configured for each individual PDU session.

[0121] In the 3GPP NR system, the main services and functions of the RRC sublayer include: broadcast of system information related to AS and NAS; paging initiated by 5GC or NG-RAN; establishment, maintenance and release of RRC connections between UE and NG-RAN; security functions including key management; establishment, configuration, maintenance and release of signaling radio bearers (SRBs) and data radio bearers (DRBs); mobility functions (including handover and context transfer, UE cell selection and reselection and control of cell selection and reselection, and inter-RAT mobility); QoS management functions; UE measurement reporting and control of reporting; detection and recovery of radio link failures; and NAS message transmission from UE to NAS / from NAS to UE.

[0122] Figure 8 The frame structure in a 3GPP-based wireless communication system to which the implementation of the present disclosure is applied is shown.

[0123] Figure 8The frame structure shown in is merely exemplary, and the number of subframes, the number of time slots, and / or the number of symbols in a frame may vary. In a 3GPP-based wireless communication system, OFDM parameter sets (e.g., subcarrier spacing (SCS), transmission time interval (TTI) duration) may be configured differently between multiple cells aggregated for one UE. For example, if a UE is configured with different SCSs for cells aggregated for a cell, the (absolute time) duration of time resources (e.g., subframes, time slots, or TTIs) comprising the same number of symbols may be different among the aggregated cells. In this document, the symbols may include OFDM symbols (or CP-OFDM symbols), SC-FDMA symbols (or discrete Fourier transform-spread-OFDM (DFT-s-OFDM) symbols).

[0124] Reference Figure 8 , downlink and uplink transmissions are organized into frames. Each frame has T f = 10ms duration. Each frame is divided into two half-frames, where each half-frame has a duration of 5ms. Each half-frame includes 5 sub-frames, where the duration of each sub-frame is T sf is 1 ms. Each subframe is divided into slots, and the number of slots in a subframe depends on the subcarrier spacing. Each slot includes 14 or 12 OFDM symbols based on the cyclic prefix (CP). In normal CP, each slot includes 14 OFDM symbols, and in extended CP, each slot includes 12 OFDM symbols. The parameter set is based on an exponentially scalable subcarrier spacing Δf=2 u *15kHz.

[0125] Table 1 shows the subcarrier spacing Δf=2 u *N, the number of OFDM symbols per time slot of 15 kHz slot symb , the number of time slots per frame N frame,u slot , and the number of slots N per subframe for normal CP subframe,u slot .

[0126] [Table 1]

[0127] u <![CDATA[N slot symb ]]> <![CDATA[N frame,u slot ]]> <![CDATA[N subframe，u slot ]]> 0 14 10 1 1 14 20 2 2 14 40 4 3 14 80 8 4 14 160 16

[0128] Table 2 shows the subcarrier spacing Δf=2 u *N, the number of OFDM symbols per time slot of 15 kHz slot symb , the number of time slots per frame N frame,u slot , and the number of slots N per subframe for the extended CPsubframe,u slot .

[0129] [Table 2]

[0130] u <![CDATA[N slot symb ]]> <![CDATA[N frame，u slot ]]> <![CDATA[N subframe，u slot ]]> 2 12 40 4

[0131] A slot includes multiple symbols (e.g., 14 or 12 symbols) in the time domain. For each parameter set (e.g., subcarrier spacing) and carrier, a common resource block (CRB) N is allocated from the CRBs indicated by higher layer signaling (e.g., RRC signaling). start,u grid To begin, define N size,u grid,x *N RB sc subcarriers and N subframe,u symb OFDM symbol resource grid, where N size,u grid,x N is the number of resource blocks (RBs) in the resource grid, with the subscript x being DL for downlink and UL for uplink. RB sc is the number of subcarriers per RB. In 3GPP-based wireless communication systems, N RB sc Typically 12. For a given antenna port p, subcarrier spacing configuration u, and transmission direction (DL or UL), there is one resource grid. The carrier bandwidth N for subcarrier spacing configuration u is size,u grid Given by higher-layer parameters (e.g., RRC parameters). Each element in the resource grid for antenna port p and subcarrier spacing configuration u is called a resource element (RE), and one complex symbol can be mapped to each RE. Each RE in the resource grid is uniquely identified by an index k in the frequency domain and an index l representing the symbol position relative to a reference point in the time domain.

[0132] In 3GPP-based wireless communication systems, an RB is defined by 12 consecutive subcarriers in the frequency domain. In 3GPP NR systems, RBs are classified into CRBs and physical resource blocks (PRBs). CRBs are numbered upwards from 0 in the frequency domain for subcarrier spacing configuration u. The center of subcarrier 0 of CRB 0 for subcarrier spacing configuration u coincides with "point A", which is used as a common reference point for the resource block grid. In 3GPP NR systems, PRBs are defined within a bandwidth part (BWP) and are numbered from 0 to N. size BWP,i -1 numbering, where i is the number of the bandwidth part. Physical resource block n in bandwidth part i PRB With public resource block n CRB The relationship between them is as follows:PRB =n CRB +N size BWP,i , where N size BWP,i A BWP is a common resource block that begins with CRB 0. A BWP consists of multiple contiguous RBs. A carrier can include up to N (e.g., 5) BWPs. A UE can be configured with one or more BWPs on a given component carrier. Of the BWPs configured for a UE, only one can be active at a time. The active BWP defines the UE's operating bandwidth within the cell's operating bandwidth.

[0133] The NR frequency band can be defined as two types of frequency ranges, namely, FR1 and FR2. The numerical values of the frequency ranges can be changed. For example, the two types of frequency ranges (FR1 and FR2) can be shown in Table 3 below. For ease of explanation, in the frequency range used in the NR system, FR1 can represent "sub-6 GHz range", FR2 can represent "above 6 GHz range" and can be called millimeter wave (mmW).

[0134] [Table 3]

[0135] Frequency range specification Corresponding frequency range Subcarrier spacing FR1 450MHz-6000MHz 15, 30, 60kHz FR2 24250MHz-52600MHz 60, 120, 240kHz

[0136] As described above, the numerical value of the frequency range of the NR system can be changed. For example, FR1 can include a frequency band of 410 MHz to 7125 MHz, as shown in Table 4 below. That is, FR1 can include a frequency band of 6 GHz (or 5850, 5900, 5925 MHz, etc.) or larger. For example, the frequency band of 6 GHz (or 5850, 5900, 5925 MHz, etc.) or larger included in FR1 can include an unlicensed frequency band. The unlicensed frequency band can be used for various purposes, such as for communication in vehicles (e.g., autonomous driving).

[0137] [Table 4]

[0138] Frequency range specification Corresponding frequency range Subcarrier spacing FR1 410MHz-7125MHz 15, 30, 60kHz FR2 24250MHz-52600MHz 60, 120, 240kHz

[0139] In the present disclosure, the term "cell" may refer to a geographical area in which one or more nodes provide a communication system or to a radio resource. A "cell" as a geographical area may be understood as a coverage area within which a node can provide services using a carrier, and a "cell" as a radio resource (e.g., a time-frequency resource) is associated with a bandwidth, which is a frequency range configured by a carrier. A "cell" associated with a radio resource is defined by a combination of downlink resources and uplink resources (e.g., a combination of a DL component carrier (CC) and a UL CC). A cell may be configured only by downlink resources, or by downlink resources and uplink resources. Since the DL coverage (which is the range within which a node can send a valid signal) and the UL coverage (which is the range within which a node can receive a valid signal from a UE) depend on the carrier carrying the signal, the coverage of a node may be associated with the coverage of a "cell" of the radio resource used by the node. Therefore, the term "cell" may be used to sometimes refer to the service coverage of a node, to refer to a radio resource at other times, or to refer to the range within which a signal using a radio resource can reach with effective strength at other times.

[0140] In CA, two or more CCs are aggregated. The UE can receive or transmit on one or more CCs simultaneously depending on its capabilities. CA is supported for both contiguous CCs and non-contiguous CCs. When CA is configured, the UE has only one RRC connection with the network. During RRC connection establishment / reestablishment / handover, one serving cell provides NAS mobility information, and during RRC connection reestablishment / handover, one serving cell provides security input. This cell is called a primary cell (PCell). A PCell is a cell operating on the primary frequency, where the UE performs an initial connection establishment procedure or initiates a connection reestablishment procedure. Depending on the UE capabilities, a secondary cell (SCell) can be configured to form a set of serving cells together with the PCell. An SCell is a cell that provides additional radio resources on top of a special cell (PCell). Therefore, the set of configured serving cells for a UE always consists of one PCell and one or more SCells. For dual connectivity (DC) operation, the term "PCell" refers to the PCell of a primary cell group (MCG) or the primary SCell (PSCell) of a secondary cell group (SCG). SpCell supports PUCCH transmission and contention-based random access and is always activated. MCG is a set of serving cells associated with the master node, which includes SpCell (PCell) and optionally one or more SCells. For UEs configured with DC, SCG is a subset of serving cells associated with the secondary node, which includes PSCell and zero or more SCells. For UEs in RRC_CONNECTED that are not configured with CA / DC, there is only one serving cell consisting of PCell. For UEs in RRC_CONNECTED that are configured with CA / DC, the term "serving cell" is used to refer to the set of cells consisting of SpCell and all SCells. In DC, two MAC entities are configured in the UE: one for MCG and one for SCG.

[0141] Figure 9 An example of data flow in a 3GPP NR system to which an implementation of the present disclosure is applied is shown.

[0142] Reference Figure 9 "RB" stands for radio bearer, and "H" stands for header. Radio bearers are categorized into two groups: DRBs for user plane data and SRBs for control plane data. MAC PDUs are transmitted and received to and from external devices via the PHY layer using radio resources. MAC PDUs arrive at the PHY layer in the form of transport blocks.

[0143] In the PHY layer, the uplink transport channels UL-SCH and RACH are mapped to their physical channels PUSCH and PRACH, respectively, and the downlink transport channels DL-SCH, BCH and PCH are mapped to PDSCH, PBCH and PDSCH, respectively. In the PHY layer, uplink control information (UCI) is mapped to the physical PUCCH, and downlink control information (DCI) is mapped to the PDCCH. The MAC PDU associated with the UL-SCH is sent by the UE via the PUSCH based on the UL grant, and the MAC PDU associated with the DL-SCH is sent by the BS via the PDSCH based on the DL assignment.

[0144] In the following, technical features related to AI / ML are described.

[0145] To date, the application of AI / ML to wireless communications has been limited to implementation-based approaches on both the network and UE sides. Research on enhancements to NR and ENDC data collection (FS_NR_ENDC_data_collect) has examined the functional framework for RAN intelligence through further enhancements to data collection through use cases, examples, and more, and identified potential standardization impacts on current NG-RAN nodes and interfaces. Within SA WG2 AI / ML-related research, the network function NWDAF (Network Data Analysis Function) was introduced in Rel-15 and enhanced in Rel-16 and Rel-17.

[0146] In this study, we explored the benefits of enhancing the air interface with features that enable improved support for AI / ML-based algorithms to enhance performance and / or reduce complexity / overhead. The enhanced performance here depends on the use case under consideration and can be, for example, improved throughput, robustness, accuracy, or reliability.

[0147] By studying a few carefully selected use cases, evaluating their performance compared to traditional approaches, and the associated potential regulatory impact of solutions implementing them, this SI will serve as the foundation for future air interface use cases leveraging AI / ML technologies.

[0148] The goal is to consider enough use cases to enable the identification of common AI / ML frameworks, including functional requirements for AI / ML architectures, that can be used in subsequent projects. The research should also identify areas where AI / ML can improve the performance of air interface functions.

[0149] This study will be used to identify what is needed to adequately characterize and describe AI / ML models, thereby establishing relevant notations for discussion and subsequent evaluation. Identify and consider various levels of collaboration between gNB and UE.

[0150] An assessment of the achievable gains from applying AI / ML-based technologies for the considered use cases will be performed with the corresponding identification of KPIs, with the goal of better understanding the achievable gains and the associated complexity requirements.

[0151] Finally, the regulatory impact will be assessed to improve the overall understanding required to enable AI / ML technologies for air interfaces.

[0152] For research on AI / ML for air interfaces, the basic framework and principles for the FS_NR_ENDC_data_collect convention should be considered for possible applicability.

[0153] Study the 3GPP framework for AI / ML over the air interface corresponding to each target use case with regard to aspects such as performance, complexity, and potential specification impact.

[0154] Use cases to focus on:

[0155] 1>The initial set of use cases includes:

[0156] a) CSI feedback enhancements, e.g., reduced overhead, improved accuracy, prediction

[0157] b) Beam management, e.g., beam prediction in the time and / or spatial domain for overhead and latency reduction, beam selection accuracy improvement

[0158] c) Enhanced positioning accuracy for different scenarios, including, for example, scenarios with a large number of NLOS conditions

[0159] 2>Finally determine representative sub-use cases for each use case for characterization and baseline performance evaluation

[0160] a) AI / ML approaches for selected sub-use cases need to be diverse enough to support various requirements regarding gNB-UE collaboration levels

[0161] The selection of use cases for this study is intended solely for the development of a framework to apply AI / ML to the air interface for these and other use cases. The selection itself is not intended to provide any indication of the prospects for any future specification project.

[0162] AI / ML models, terms, and descriptions used to identify common and specific characteristics for the framework investigation:

[0163] 3> Characterize the limited stages and associated complexity of AI / ML related algorithms:

[0164] a) Model generation, e.g., model training (including input / output, pre-processing / post-processing, online / offline if applicable), model validation, model testing, if applicable

[0165] b) Inference operations, e.g., input / output, pre-processing / post-processing, if applicable

[0166] 4> Identify the various levels of collaboration between UE and gNB relevant to the selected use case, e.g.

[0167] a) No collaboration: AI / ML algorithms based solely on implementation, without information exchange [for comparison purposes]

[0168] b) Various levels of UE / gNB collaboration targeting standalone or joint ML operation.

[0169] 5> Characterize the lifecycle management of AI / ML models: for example, model training, model deployment, model inference, model monitoring, and model updates

[0170] 6>Datasets for training, validation, testing, and inference

[0171] 7> Identify common symbols and terminology for AI / ML-related functions, processes, and interfaces

[0172] 8> Comment: Consider work done for FS_NR_ENDC_data_collect where appropriate

[0173] For the use case considered:

[0174] - Evaluate the performance benefits of AI / ML-based algorithms for a final representative set of agreed use cases:

[0175] a) Statistical model-based approach for link and system level simulation.

[0176] i. Extensions to 3GPP evaluation methodologies should be considered as needed to better suit AI / ML-based technologies.

[0177] ii. Whether field data are optionally required to further estimate the performance and robustness in real-world environments should be discussed as part of the study.

[0178] iii. Common assumptions in dataset construction are required for training, validation, and testing for the chosen use case.

[0179] iv. Consider appropriate model training strategies, collaboration levels, and associated implications

[0180] v. Consider convention-based AI models for calibration

[0181] vi. The AI model description and training methodology used for evaluation should be reported for information and cross-checking purposes.

[0182] b) KPIs: Identify common KPIs and corresponding requirements for AI / ML operations. Identify use case-specific KPIs and benchmarks for selected use cases.

[0183] i. The performance, inference latency, and computational complexity of AI / ML-based algorithms should be compared with the performance, inference latency, and computational complexity of existing technology baselines.

[0184] ii. The overhead, power consumption (including computation), memory storage and hardware requirements (including for a given processing latency) and generalization capabilities associated with enabling the corresponding AI / ML solutions should be considered.

[0185] - Estimate potential normative impact, particularly for agreed use cases and common frameworks in the final representative set:

[0186] c) PHY layer,

[0187] i. Consider aspects related to, for example, potential specifications for AI model lifecycle management and dataset construction for training, validation, and testing for the chosen use case.

[0188] ii. Use case and collaboration level specific specification impacts, such as new signaling, means for training and validation data assistance, auxiliary information, measurements and feedback

[0189] d) Protocol-wise, for example (RAN2)-RAN2 will only start work after sufficient progress has been made in the use case studies in RAN1.

[0190] i. Consider aspects related to, for example, capability indication per RAN1 input, configuration and control processes (training / inference), and management of data and AI / ML models.

[0191] ii. Collaboration level specific specification impact per use case

[0192] e) Interoperability and testability aspects, e.g. (RAN4) - RAN4 will only start work after there is sufficient progress in use case studies in RAN1 and RAN2.

[0193] i. Requirements and test framework for validating AI / ML-based performance enhancements and ensuring that UEs and gNBs with AI / ML meet or exceed existing minimum requirements, where applicable

[0194] ii. Consider the need and implications of defining AI / ML processing capabilities

[0195] -Specific AI / ML models are not expected to be specified and depend on the implementation. User data privacy needs to be protected.

[0196] - Research on AI / ML for air interfaces is based on the current RAN architecture and no new interfaces should be introduced.

[0197] Figure 10 An example of a functional framework for RAN intelligence is shown.

[0198] Data collection is the function that provides input data to model training and model inference functions. AI / ML algorithm-specific data preparation (e.g., data preprocessing and cleaning, formatting, and transformation) is not performed in the data collection function.

[0199] Examples of input data may include measurements from the UE or different network entities, feedback from the actor, output from an AI / ML model.

[0200] >> Training data: Data required as input for training AI / ML models.

[0201] >>Inference data: Data required as input for the inference function of the AI / ML model.

[0202] Model training is the function that performs AI / ML model training, validation, and testing, and can generate model performance metrics as part of the model testing process. If required, the model training function is also responsible for data preparation (e.g., data preprocessing and cleaning, formatting, and transformation) based on the training data delivered by the data collection function.

[0203] >>Model deployment / update: Used to initially deploy trained, validated, and tested AI / ML models to model inference or to pass updated models to model inference.

[0204] Model inference is a function that provides AI / ML model inference output (e.g., predictions or decisions). The model inference function can provide model performance feedback to the model training function when applicable. If necessary, the model inference function is also responsible for data preparation (e.g., data preprocessing and cleaning, formatting, and transformation) based on the inference data delivered by the data collection function.

[0205] >>Output: The inference output of the AI / ML model produced by the model inference function.

[0206] >>>Note: The details of the inferred output are use case specific.

[0207] >>Model performance feedback: When available, it can be used to monitor the performance of AI / ML models.

[0208] >An actor is a function that receives output from the model inference function and triggers or performs corresponding actions. An actor can trigger actions on other entities or on itself.

[0209] >> Feedback: Information that may be needed to derive training data, inference data, or monitor the performance of AI / ML models and their impact on the network by updating KPIs and performance counters.

[0210] Hereinafter, technical features related to mobility optimization are described.

[0211] Mobility management is a solution that ensures service continuity during mobility by minimizing dropped calls, RLF, unnecessary handovers, and ping-pong. For future high-frequency networks, as the coverage of a single node decreases, the frequency of UE handovers between nodes becomes high, especially for highly mobile UEs. In addition, for applications characterized by strict QoS requirements such as reliability and latency, QoE is sensitive to handover performance, so that mobility management should avoid unsuccessful handovers and reduce the latency during the handover process. However, for conventional methods, it is difficult to achieve almost zero-failure handovers based on trial and error solutions. Unsuccessful handover situations are the main cause of packet loss or additional delays during mobility periods, which is undesirable for packet loss-intolerant and low-latency applications. In addition, due to the randomness and inconsistency of the transmission environment, the effectiveness of feedback-based adjustments may be weak. In addition to the baseline case of mobility, areas of optimization for mobility include dual connectivity, CHO, and DAPS, each of which has additional aspects to be addressed in the optimization of mobility.

[0212] Mobility aspects of SON that can be enhanced by using AI / ML include

[0213] -Reduction in the probability of unexpected events

[0214] -UE location / mobility / performance prediction

[0215] -Business Orientation

[0216] Reduce the probability of unexpected events associated with mobility.

[0217] Examples of such unexpected events are:

[0218] - Intra-system late handover: Radio link failure (RLF) occurs after the UE has stayed in a cell for a long period of time; the UE attempts to re-establish the radio link connection in a different cell.

[0219] - Intra-system premature handover: 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.

[0220] - Intra-system handover to wrong cell: 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.

[0221] -Successful switch: During a successful switch, there was a potential problem.

[0222] RAN intelligence can observe multiple HO events with associated parameters, use this information to train its ML model and try to identify parameter sets that lead to successful HO and those that lead to unexpected events.

[0223] UE location / mobility / performance prediction

[0224] Predicting the UE's location is a key component of mobility optimization, as many mobility-related RRM actions (e.g., selecting a handover target cell) can benefit from the predicted UE location / trajectory. UE mobility prediction is also a key factor in optimizing early data forwarding, particularly for CHO. Predicting the UE's performance when served by a specific cell is a key factor in determining the optimal mobility target for maximizing efficiency and performance.

[0225] Business Orientation

[0226] Efficient resource management can be achieved by adjusting the handover trigger point and selecting the optimal combination of Pcell / PSCell / Scell to serve users.

[0227] Existing traffic steering may also be improved by providing mobility or dual connectivity related information to the RAN node.

[0228] For example, before initiating a handover, the source gNB may use feedback on UE performance collected for successful handovers that occurred in the past and received from neighboring gNBs.

[0229] Similarly, for the dual connectivity case, before triggering the addition of a secondary gNB or triggering an SN change, the eNB can use the information (feedback) received from the gNB in the past to successfully complete the SN addition or SN change procedure.

[0230] In both reporting examples, the source RAN node of a mobility event or a RAN node acting as a master node (eNB for EN-DC, gNB for NR-DC) can use feedback received from another RAN node as input to AI / ML functions supporting service-related decisions (e.g., selection of a target cell in the case of mobility, selection of a PSCell / Scell in the other case) so that future decisions can be optimized.

[0231] Location for AI / ML model training and AI / ML model inference

[0232] Considering the location of AI / ML model training and AI / ML model inference for mobility solutions, consider the following two options:

[0233] -AI / ML model training functions are deployed in OAM, while model inference functions reside within RAN nodes

[0234] - Both AI / ML model training and AI / ML model inference functions reside within the RAN node

[0235] Additionally, for CU-DU split scenarios, the following options are possible:

[0236] -AI / ML model training is located in the CU-CP or OAM, and AI / ML model inference functions are located in the CU-CP.

[0237] It also allows gNB to continue model training based on the AI / ML model trained in OAM.

[0238] Figure 11 Examples of AI / ML model training in OAM and AI / ML model inference in NG-RAN nodes are shown.

[0239] Step 0. It is assumed that the NG-RAN node 2 optionally has an AI / ML model that can generate the required inputs (e.g., resource status and utilization prediction / assessment, etc.).

[0240] Step 1. The NG-RAN node configures measurement information on the UE side and sends a configuration message including the configuration information to the UE.

[0241] Step 2. The UE collects the indicated measurements (eg, UE measurement results related to RSRP, RSRQ, SINR of neighboring cells and serving cell).

[0242] Step 3. The UE sends a measurement report message including the required measurements to NG-RAN node 1.

[0243] Step 4. The NG-RAN node 1 sends input data for training to the OAM, where the input data for training includes required input information from the NG-RAN node 1 and measurements from the UE.

[0244] Step 5. The NG-RAN node 2 sends input data for training to the OAM, where the input data for training includes the required input information from the NG-RAN node 2. If the NG-RAN node 2 executes the AI / ML model, the input data for training may include the corresponding inference results from the NG-RAN node 2.

[0245] Step 6. Model training: Use the required measurements to train the AI / ML model optimized for UE mobility.

[0246] Step 7. OAM sends an AI / ML model deployment message to deploy the trained / updated AI / ML model to the NG-RAN node. The NG-RAN node can also continue model training based on the AI / ML model received from OAM.

[0247] Note: This step is outside the scope of RAN3 Rel-17.

[0248] Step 8. NG-RAN node 1 obtains the measurement report as inference data for UE mobility optimization.

[0249] Step 9. NG-RAN node 1 obtains input data for inference from NG-RAN node 2 for UE mobility optimization, where the input data for inference includes required input information from NG-RAN node 2. If NG-RAN node 2 executes an AI / ML model, the input data for inference may include a corresponding inference result from NG-RAN node 2.

[0250] Step 10. Model inference: Utilize the required measurements into model inference to output predictions (e.g., UE trajectory prediction, target cell prediction, target NG-RAN node prediction, etc.).

[0251] Step 11. NG-RAN1 sends model performance feedback to OAM (if applicable).

[0252] Note: This step is outside the scope of RAN3.

[0253] Step 12: Based on the prediction, recommended action or configuration, NG-RAN node 1, the target NG-RAN node (represented by NG-RAN node 2 in this step in the flowchart) and the UE perform a mobility optimization / handover procedure to hand over the UE from NG-RAN node 1 to the target NG-RAN node.

[0254] Step 13. NG-RAN node 1 sends feedback information to OAM.

[0255] Step 14. NG-RAN node 2 sends feedback information to OAM.

[0256] Figure 12 An example of model training and model inference, both located in RAN nodes, is shown.

[0257] Step 0. It is assumed that the NG-RAN node 2 optionally has an AI / ML model that can generate the required inputs (e.g., resource status and utilization prediction / assessment, etc.).

[0258] Step 1. NG-RAN node 1 configures measurement information on the UE side and sends a configuration message including the configuration information to the UE.

[0259] Step 2. The UE collects indicated measurements (eg, UE measurements related to RSRP, RSRQ, SINR of neighboring cells and serving cell).

[0260] Step 3. The UE sends a measurement report message including the required measurements to NG-RAN node 1.

[0261] Step 4. NG-RAN node 1 obtains input data for training from NG-RAN node 2, where the input data for training includes required input information from NG-RAN node 2. If NG-RAN node 2 executes the AI / ML model, the input data for training may include the corresponding inference results from NG-RAN node 2.

[0262] Step 5. Model training: Use the required measurements to train the AI / ML model for mobility optimization.

[0263] Step 6. NG-RAN node 1 obtains the measurement report as inference data for real-time UE mobility optimization.

[0264] Step 7. NG-RAN node 1 obtains input data for inference from NG-RAN node 2 for UE mobility optimization, where the input data for inference includes required input information from NG-RAN node 2. If NG-RAN node 2 executes an AI / ML model, the input data for inference may include a corresponding inference result from NG-RAN node 2.

[0265] Step 8. Model Inference: The required measurements are utilized in model inference to output predictions, including, for example, UE trajectory prediction, target cell prediction, target NG-RAN node prediction, etc.

[0266] Step 9: Based on the prediction, recommended action or configuration, NG-RAN node 1, the target NG-RAN node (represented by NG-RAN node 2 in this step in the flowchart) and the UE perform a mobility optimization / handover procedure to hand over the UE from NG-RAN node 1 to the target NG-RAN node.

[0267] Step 10. NG-RAN node 2 sends feedback information after the mobility optimization action to NG-RAN node 1.

[0268] For example, UE mobility information for training purposes is only sent to the gNB that requests such information or when triggered.

[0269] Inputs for AI / ML-based Mobility Optimization

[0270] The following data are required as input data for mobility optimization.

[0271] From UE:

[0272] -UE location information (e.g., coordinates, serving cell ID, mobile speed) interpreted by the gNB implementation when available.

[0273] - Radio measurements related to the serving cell and neighboring cells (e.g. RSRP, RSRQ, SINR) associated with the UE location information.

[0274] -UE mobility history information.

[0275] From neighboring RAN nodes:

[0276] - Historical information of UEs from neighbors

[0277] - Location, QoS parameters, and performance information (e.g., loss rate, delay, etc.) of UEs with historical HO

[0278] -Current / forecasted resource status

[0279] - Successful and unsuccessful handovers (including premature handovers, late handovers or handovers to wrong (suboptimal) cells) in the past by the UE based on the existing SON / RLF reporting mechanism.

[0280] From the local node:

[0281] -UE trajectory prediction

[0282] -Current / forecasted resource status

[0283] -Current / forecasted UE traffic

[0284] Output of AI / ML-based mobility optimization

[0285] AI / ML-based mobility optimization can generate the following information as output:

[0286] -UE trajectory prediction (UE latitude, longitude, altitude, cell ID in the future time period)

[0287] NOTE: Whether the UE trajectory prediction is an external output to the node hosting the model inference functionality should be discussed during the specification work phase.

[0288] -Estimated arrival probability in CHO and associated confidence intervals

[0289] - The predicted handover target node (candidate cell in CHO) may be together with the confidence of the prediction.

[0290] - Priority, handover execution timing, and predicted resource reservation time window for CHO.

[0291] -UE traffic prediction (to be used internally by RAN nodes and details depend on specification work phase)

[0292] - Model output validity time will be discussed during the R18 specification work for each inference output.

[0293] Feedback on AI / ML-based mobility optimization

[0294] The following data is required as feedback data for mobility optimization.

[0295] - QoS parameters (eg throughput, packet delay for handed-over UEs, etc.).

[0296] - Resource status information update from the target NG-RAN.

[0297] - Capability information from the target NG-RAN. Details of the capability information will be discussed during the specification work phase.

[0298] Standard impact

[0299] To improve mobility decisions at the gNB (gNB-CU), the gNB can request mobility feedback from neighboring nodes. The details of the procedure will be determined during the specification phase.

[0300] If the gNB requires existing UE measurements for AI / ML-based mobility optimization, RAN3 should reuse the existing framework (including MDT and RRM measurement results). Based on the use case description, whether new UE measurement results are required depends on the specification stage.

[0301] MDT process enhancements should be discussed during the specification phase.

[0302] Potential Xn interface impact:

[0303] - Predicted resource status information and performance information from candidate target NG-RAN nodes to source NG-RAN nodes

[0304] - A new signalling procedure or an existing procedure to retrieve input information via the Xn interface.

[0305] - A new signaling procedure or an existing procedure to retrieve feedback information via the Xn interface.

[0306] In the following, technical features related to AI and ML are described.

[0307] Artificial intelligence (AI) / machine learning (ML) is being used in a range of application areas across industrial sectors, achieving significant productivity gains. Specifically, in mobile communication systems, mobile devices (e.g., smartphones, smart vehicles, UAVs, mobile robots) are increasingly replacing conventional algorithms (e.g., speech recognition, machine translation, image recognition, video processing, user behavior prediction) with AI / ML models to enable applications such as enhanced photography, intelligent personal assistants, VR / AR, video games, video analysis, personalized shopping recommendations, autonomous driving / navigation, smart home appliances, mobile robots, mobile healthcare, and mobile finance.

[0308] Artificial intelligence (AI) is the science and engineering of building intelligent machines that can perform tasks like humans.

[0309] Deep Neural Networks

[0310] Figure 13 and Figure 14 Examples of neurons and neural network architectures are shown.

[0311] Within the field of ML, there is an area generally referred to as brain-inspired computing, which is a program that aims to simulate some aspects of our understanding of how the brain works. Since the main computing element of the human brain is believed to be 86 billion neurons, both sub-areas of brain-inspired computing are inspired by neuronal architectures, such as Figure 13 shown.

[0312] Compared to spike computing methods, more popular ML methods use "neural networks" as models. Neural networks (NNs) derive their inspiration from the concept that the computation of a neuron involves the weighted sum of input values. However, instead of simply outputting a weighted sum, NNs apply nonlinear functions to generate outputs only when the input exceeds a certain threshold, such as Figure 13 shown.

[0313] Figure 14 A schematic diagram of a computational neural network is shown. Neurons in the input layer receive values and propagate them to neurons in the middle layers of the network (also called "hidden layers"). The weighted sums from one or more hidden layers are ultimately propagated to the output layer, which presents the final output of the network.

[0314] A neural network with more than three layers (i.e., more than one hidden layer) is called a deep neural network (DNN). Compared to traditional shallow structured NN architectures, DNNs (also known as deep learning) have made amazing breakthroughs in many essential application areas since the 2010s because they can achieve human-level accuracy or even exceed human accuracy. Deep learning technology uses supervised and / or unsupervised strategies to automatically learn hierarchical representations in deep architectures for classification. In the case of a large number of hidden layers, the excellent performance of DNNs comes from their ability to extract high-level features from raw sensory data after using statistical learning through a large amount of data to obtain an effective representation of the input space. In recent years, due to the big data obtained from the real world, the rapidly increasing computing power and the continuously evolving algorithms, DNNs have become the most popular ML models for many AI applications.

[0315] Training and inference

[0316] Training is the process by which an AI / ML model learns to perform its given task, more specifically by optimizing the values of the weights in a DNN. A DNN is trained by feeding it a training set, typically correctly labeled training examples. For image classification, for example, the training set consists of correctly classified images. When training a network, the weights are typically updated using a hill-climbing optimization process called gradient descent. The gradient indicates how the weights should change to reduce the loss (the difference between the correct output and the output calculated by the DNN based on its current weights). The training process is repeated iteratively to continuously reduce the overall loss. A DNN with high accuracy is achieved until the loss falls below a predefined threshold.

[0317] There are many ways to train networks for different goals. Supervised learning, described above, uses labeled training examples to find the correct output for a task. Unsupervised learning uses unlabeled training examples to find structure or clusters in the data. Reinforcement learning can be used to output what actions the agent should take next to maximize the expected reward. Transfer learning uses a new training set to adjust previously trained weights (e.g., those in a global model) for faster or more accurate training of a personalized model.

[0318] Figure 15 An example of AI / ML inference is shown.

[0319] After training a DNN, it can perform its task by computing the output of the network using the weights determined during the training process, which is called inference. During model inference, input from the real world is passed through the DNN. Figure 15As shown, the network then outputs a prediction for the task. For example, the input could be the pixels of an image, the amplitude of audio wave samples, or a numerical representation of the state of some system or game. Correspondingly, the network's output could be the probability that an image contains a particular object, the probability that an audio sequence contains a particular word, or a bounding box in the image around an object, or a proposed action to take.

[0320] The performance of DNNs comes at the cost of high computational complexity. Therefore, more efficient computing engines (e.g., graphics processing units (GPUs) and network processing units (NPUs)) are typically used. Compared to inference, which only involves the feedforward process, training typically requires more computing and storage resources because it also involves the backward propagation process.

[0321] Widely used DNN models and algorithms

[0322] Figure 16 An example of an MLP DNN model is shown.

[0323] Many DNN models have been developed over the past two decades. Each of these models has a different "network architecture" in terms of the number of layers, layer types, layer shapes (i.e., filter sizes, number of channels and filters), and connections between layers. Figure 16 Three popular structures of DNN are presented: Multilayer Perceptron (MLP), Convolutional Neural Network (CNN) and Recurrent Neural Network (RNN). The Multilayer Perceptron (MLP) model is the most basic DNN, which consists of a series of fully connected layers. In a fully connected layer, all outputs are connected to all inputs, such as Figure 16 As shown in Figure 2, MLP requires a lot of storage and computation.

[0324] Figure 17 An example of a CNN model is shown.

[0325] One way to limit the number of weights that contribute to the output is to compute the output using only functions of a fixed-size input window. The very popular window-based DNN model uses convolution operations to construct the computation, hence the name Convolutional Neural Network (CNN). Figure 17 As shown in Figure 2, CNN consists of multiple convolutional layers. By applying various convolutional filters, CNN models can capture high-level representations of input data, making them popular in image classification and speech recognition tasks.

[0326] Figure 18 An example of an RNN model is shown.

[0327] Recurrent Neural Network (RNN) models are another type of DNN that uses sequential data feeds. The input to an RNN consists of the current input and the previous sample. Each neuron in an RNN has an internal memory that holds information from the calculations of the previous sample. Figure 18 As shown, the basic unit of RNN is called a cell, and in addition, each cell is composed of layers, and a series of cells enables sequential processing of RNN models. RNN models have been widely used in natural language processing tasks (e.g., language modeling, machine translation, question answering, word embedding, and document classification) on mobile devices.

[0328] Figure 19 An example of reinforcement learning is shown.

[0329] Deep reinforcement learning (DRL) is not another DNN model. It is composed of DNN and reinforcement learning. Figure 19 As shown in Figure 2, the goal of DRL is to create intelligent agents that can execute efficient policies to maximize rewards for long-term tasks with controllable actions. Typical applications of DRL are solving various scheduling problems such as decision-making problems in games and rate selection for video transmission.

[0330] Hereinafter, technical features related to connection failure are described. Reference may be made to a part of section 5.3.3.7 of 3GPP TS 38.331 v17.2.0.

[0331] The UE shall:

[0332] 1> If timer T300 expires, then:

[0333] 2> Reset MAC, release MAC configuration, and re-establish RLC for all established RBs;

[0334] 2> If the UE supports RRC connection establishment failure with temporary offset and T300 has expired for connEstFailCount consecutive times on the same cell, where connEstFailureControl is included in SIB1, then:

[0335] 3>For the period indicated by connEstFailOffsetValidity:

[0336] 4>When performing cell selection and reselection, connEstFailOffset is used for the parameter Qoffsettemp for the relevant cell;

[0337] - When performing cell selection, if no suitable or acceptable cell is found, it shall be UE implementation dependent whether to stop using the connEstFailOffset for the parameter Qoffsettemp during the connEstFailOffsetValidity for the concerned cell.

[0338] 2> If the UE supports multiple CEF reports, then:

[0339] 3> if the UE has connection establishment failure information or connection recovery failure information available in VarConnEstFailReport, and if the RPLMN is equal to the plmn-identity stored in VarConnEstFailReport; and

[0340] If the cell identity of the current cell is not equal to the cell identity in measResultFailedCell stored in VarConnEstFailReport, and if maxCEFReport-r17 has not been reached, then:

[0341] 4>Append VarConnEstFailReport as a new entry in VarConnEstFailReportList;

[0342] 2> if the UE has connection establishment failure information or connection recovery failure information available in VarConnEstFailReport, and if the RPLMN is not equal to the plmn-identity stored in VarConnEstFailReport; or

[0343] 2> If the cell ID of the current cell is not equal to the cell ID stored in measResultFailedCell in VarConnEstFailReport, then:

[0344] 3>Reset numberOfConnFail to 0;

[0345] 2> If the UE supports multiple CEF reports, and if the UE has connection establishment failure information or connection recovery failure information available in VarConnEstFailReportList, and if the RPLMN is not equal to the plmn-identity stored in any entry of VarConnEstFailReportList:

[0346] 3> Clear the contents included in VarConnEstFailReportList;

[0347] 2> Clear the content except numberOfConnFail (if any) included in VarConnEstFailReport;

[0348] 2>Store the following connection establishment failure information in VarConnEstFailReport by setting the fields as follows:

[0349] 3> Set plmn-Identity to the PLMN selected by the upper layer from the PLMNs in the plmn-IdentityInfoList included in SIB1;

[0350] 3> Based on the available SSB measurement results collected until the UE detects the connection establishment failure, measResultFailedCell is set to include the global cell identity, tracking area code, cell-level and SS / PBCH block-level RSRP, RSRQ and SS / PBCH block index of the failed cell;

[0351] 3> If available, set measResultNeighCells in order of descending sorting criteria for cell reselection to include neighbor cell measurements for up to the following number of neighbor cells: 6 intra-frequency and 3 inter-frequency neighbors per frequency, and 3 inter-RAT neighbors per frequency / frequency set per RAT, and according to the following:

[0352] 4> For each included neighbor cell, include the available optional fields;

[0353] - The UE includes the latest results of available measurements performed for cell reselection evaluation according to performance requirements.

[0354] 3>If available, set locationInfo as follows:

[0355] 4>If available, set commonLocationInfo to include detailed location information;

[0356] If available, set bt-LocationInfo to include Bluetooth measurements in order of decreasing RSSI for Bluetooth beacons.

[0357] If available, set wlan-LocationInfo to include WLAN measurements in order of decreasing RSSI for the WLAN APs.

[0358] If available, set sensor-LocationInfo to include sensor measurements as follows;

[0359] 5>If available, include sensor-MeasurementInformation;

[0360] 5> If available, include sensor-MotionInformation;

[0361] - Which location information related configuration the UE uses to make locationInfo available for inclusion in VarConnEstFailReport depends on the UE implementation.

[0362] 3> perRAInfoList is set to indicate the relevant information of the random access procedure performed;

[0363] 3> If numberOfConnFail is less than 8:

[0364] 4>Increment numberOfConnFail by 1;

[0365] 2> Notify the upper layer of the failure to establish the RRC connection, and the process ends at this point;

[0366] The UE may discard the connection establishment failure or connection recovery failure information, that is, release the UE variable VarConnEstFailReport 48 hours after detecting the last connection establishment failure.

[0367] The L2 U2N relay UE indicates to the upper layer (to trigger the release of the PC5 unicast link) or sends a notification message to the connected L2 U2N remote UE.

[0368] The following describes technical features related to SCGFailureInformation and MCGFailureInformation, and reference may be made to a portion of 3GPP TS 38.331 v17.2.0.

[0369] The SCGFailureInformation message is used to provide information about NR SCG failure detected by the UE.

[0370] -Signaling Radio Bearer: SRB1

[0371] -RLC-SAP:AM

[0372] -Logical channel: DCCH

[0373] - Direction: UE to network

[0374] SCGFailureInformation field description:

[0375] -measResultFreqList

[0376] This field contains the available results of measurements on the NR frequencies that the UE is configured to measure via measConfig.

[0377] -measResultSCG-Failure

[0378] This field contains the MeasResultSCG-Failure IE, which includes the available results of the measurements on the NR frequencies that the UE is configured to measure via the NR SCGRRCReconfiguration message.

[0379] -previousPSCellId

[0380] This field indicates the physical cell ID and carrier frequency of the cell that is the source PSCell of the last PSCell change.

[0381] -failedPSCellId

[0382] This field indicates the physical cell ID and carrier frequency of the cell where the SCG failure is detected or the target PSCell for failed PSCell change or failed PSCell addition.

[0383] -timeSCGFailure

[0384] This field is used to indicate the time that has passed since the last RRCReconfiguration with reconfigurationWithSync for the SCG until the SCG failure. Actual value = field value * 100ms. The maximum value of 1023 means 102.3s or longer.

[0385] The MCGFailureInformation message is used to provide information about NR MCG failure detected by the UE.

[0386] -Signaling Radio Bearer: SRB1

[0387] -RLC-SAP:AM

[0388] -Logical channel: DCCH

[0389] - Direction: UE to network

[0390] MCGFailureInformation field description:

[0391] -measResultFreqList

[0392] This field contains the available results of measurements on the NR frequencies that the UE is configured to measure via the measConfig associated with the MCG.

[0393] -measResultFreqListEUTRA

[0394] This field contains the available results of measurements on the E-UTRA frequencies that the UE is configured to measure via the measConfig associated with the MCG.

[0395] -measResultFreqListUTRA-FDD

[0396] This field contains the available results of measurements on the UTRA FDD frequencies that the UE is configured to measure via the measConfig associated with the MCG.

[0397] -measResultSCG

[0398] This field contains the MeasResultSCG-Failure IE, which includes the available measurement results on the NR frequencies that the UE is configured to measure through the measConfig associated with the SCG.

[0399] -measResultSCG-EUTRA

[0400] This field contains the EUTRAMeasResultSCG-FailureMRDC IE, which includes the available results of measurements on the E-UTRA frequencies that the UE is configured to measure via the E-UTRA RRCConnectionReconfiguration message.

[0401] In the following, reference may be made to technical features related to supporting SON / RLF reporting.

[0402] In the 3GPP specification, there are many methods to report fault information.

[0403] MCG / SCG fault information is used to report connection failures via SCG / MCG respectively. If a failure is detected in a CG, the fault information can be sent via the other CG. MCG / SCG fault information may include the following:

[0404] -Fault type: T310 expiration, random access problem, RLC maximum retransmission count, synchronous reconfiguration failure, etc.

[0405] - Measurement results of measurements on NR frequencies that the UE is configured to measure via measConfig

[0406] - Previous cell id indicating the carrier frequency and physical cell id of the source cell

[0407] - Faulty cell id indicating the physical cell id and carrier frequency of the cell where the fault was detected or faulty cell change / cell addition

[0408] -FailureTime indicating the time elapsed since the last execution of RRCReconfiguration with reconfigurationWithSync until the failure.

[0409] - Location information

[0410] Self-Organizing Networks (SON) and Minimized Drive Testing (MDT) are standardized mechanisms for collecting mobile network data using user devices in the network. For SON / MDT, the UE stores some information related to measurement results, connection failures, RLFs, mobility history, etc. The network requests this information via a UE Information Request message, and the UE then responds to the network with the stored information via a UE Information Response message. Although the corresponding information is not real-time, the UE can instead be notified of more detailed fault information. For example, the stored information may include the following information:

[0411] - The stored information can be used to notify connection establishment failure information, connection recovery failure information, RLF report

[0412] - The stored information may include the following information

[0413] >plmn logo

[0414] > The last serving cell measurement results including cell-level RSRP, RSRQ and available SINR of the source Pcell (in case of HO failure) or Pcell (in case of RLF) based on available SSB / CSI-RS measurements collected until the UE detects the failure

[0415] > RLM configuration information including the radio link monitoring configuration of the source Pcell (in case of HO failure) or Pcell (in case of RLF)

[0416] > Measurement results of neighboring cells including neighboring cell measurements

[0417] >> The measurement results include all available measurements of the best measured cell except the source Pcell (in case of HO failure) or Pcell (in case of RLF)

[0418] >>CHO Config includes conditional reconfiguration information (first trigger event, and the time between events if two events are configured)

[0419] >> The time elapsed between 1) CHO configuration and CHO execution (if CHO fails) and 2) CHO configuration and HO execution (if HO fails)

[0420] >CHO candidate cell list includes the global cell identity / physical cell identity / carrier frequency of each candidate cell for conditional handover in case of handover failure

[0421] >Finally switch the type, for example, cho, daps, etc.

[0422] >Connection fault type, for example, hof, rlf

[0423] >Fault Pcell Id sets the global cell identity / tracking area code / physical cell identity / carrier frequency where the radio link failure was detected

[0424] >Previous cell setting received the global cell identity / tracking area code of the Pcell of the last executed CHO message

[0425] >RLF causes, such as random access problems, beam failure recovery failures, etc.

[0426] > Location Information

[0427] Hereinafter, reference may be made to technical features related to supporting AI / ML model changes during mobility.

[0428] A valid question is whether mobility is supported for AI / ML operation over the air interface. If mobility is not supported for AI / ML over the air interface, AI / ML operation can be disabled immediately before a handover and then enabled after the handover to the target gNB is complete. For cooperation level z, model transfer may always be required when the serving gNB for a UE changes. Given the potentially large size of AI / ML models, transferring the model over the air interface at every handover clearly results in significant signaling overhead and consumes significant system capacity. It is desirable to implement a mechanism for AI / ML model change / reconfiguration during UE mobility.

[0429] Basically consider two scenarios:

[0430] 1. UE moves from one cell to another without AI / ML model change

[0431] 2. UE moves from one cell to another with AI / ML model changes

[0432] In the first scenario, the same AI / ML model is used in both the source gNB and the target gNB. There are two ways to continue using the AI / ML model in the target gNB. One way is for the UE to upload or download the AI / ML model whenever the anchor gNB or CN changes. The other way is for the source gNB to forward the AI / ML model or related information to the target gNB when a handover occurs.

[0433] In the second scenario, different AI / ML models are used in the source and target gNBs. There are generally two approaches to changing the AI / ML model. One approach is for the UE to upload or download the new AI / ML model when a handover to the target gNB occurs, similar to a full configuration. The other approach is to use incremental configuration to change only part or some parameters of the AI / ML model.

[0434] Mechanisms to support AI / ML changes during UE mobility are discussed.

[0435] Supporting AI / ML model changes during mobility for collaboration level z also requires consideration of model delivery options and model formats. Model delivery via RRC messages is compatible with current handover mechanisms and makes it easier to support AI / ML changes during UE mobility. Model delivery via NAS messages can also support AI / ML changes during UE mobility. However, if the AI / ML model is delivered via UP services, it is not possible to support AI / ML model changes during UE mobility using current handover mechanisms. If the AI / ML model is delivered in the format of an inter-operational image, it is difficult to support incremental configuration. If the model is delivered from one gNB to another, it is also a large burden on the Xn interface. If the AI / ML model is delivered in a format specified by 3GPP, model changes during mobility can be supported in a more signaling-efficient manner.

[0436] In addition, UEs supporting AI / ML operation may derive measurement results based on artificial intelligence (AI) / machine learning (ML)-based methods or based on non-AI / ML-based methods. In the event that a UE experiences a connection failure, the UE may report the available measurement results to the network as part of a failure report. Upon receiving the failure report, the network may not clearly determine which of the following is the possible cause of the failure:

[0437] - Case a) The expected failure is caused by inappropriate AI / ML operation of the UE, while the failure is not caused by network coverage issues or other inappropriate UE configurations.

[0438] - Case b) The expected failure is not caused by inappropriate AI / ML operation of the UE, but by network coverage issues or other inappropriate UE configurations.

[0439] This ambiguity from the network side arises primarily because existing fault reporting procedures are insufficient, particularly when the UE is configured or enabled to perform AI / ML-based prediction tasks for measurements and / or other 3GPP procedures. For example, existing fault reporting procedures do not indicate whether the UE is performing AI / ML-based CSI measurement reporting / RRM measurement reporting or traditional measurement reporting, and existing fault reporting procedures do not indicate whether a faulty connection may be caused by incorrectly predicted mobility based on incorrect AI / ML operation by the UE or by incorrect network decisions.

[0440] Even if the network can infer that the fault may be related to AI / ML operations, it is difficult to know how to correct the AI / ML-related operations because existing fault reports lack information related to the AI / ML operations performed by the UE at the time of the fault. As a result, the network may not be able to identify whether AI / ML-related settings need to be changed, and erroneous results in AI / ML operations may have been used as input for training or as interference results.

[0441] Therefore, research is needed to support AI and ML operations in wireless communication systems.

[0442] Hereinafter, a method for supporting AI and ML operations in a wireless communication system according to some embodiments of the present disclosure will be described with reference to the following drawings.

[0443] The following figures are created to illustrate specific embodiments of the present disclosure. The names of specific devices or the names of specific signals / messages / fields shown in the figures are provided by way of example, and therefore the technical features of the present disclosure are not limited to the specific names used in the following figures. In this document, a wireless device may be referred to as a user equipment (UE).

[0444] Figure 20 An example of a method for supporting AI and ML operations in a wireless communication system is shown.

[0445] Specifically, Figure 20 An example of a method performed by a wireless device in a wireless communication system is shown.

[0446] In step S2001, the wireless device may detect a failure in operation with the network.

[0447] For example, the operations may include predictive operations derived by AI and / or ML models.

[0448] For example, the operations may include conditional operation, conditional handover (CHO), conditional PScell change (CPC), cell and / or cell group (CG) addition, cell and / or CG activation, cell and / or CG deactivation and / or cell and / or CG release.

[0449] For example, failures in operation may include switching failure, radio link failure, beam failure, random access failure, configuration failure, connection failure, recovery failure, conditional reconfiguration failure, cell and / or CG addition failure, cell and / or CG activation failure and / or cell and / or CG deactivation failure.

[0450] In step S2002 , the wireless device may transmit a message including (i) information about a fault and (ii) information about prediction information related to the fault.

[0451] The prediction information may be derived by an artificial intelligence (AI) and / or machine learning (ML) model of the wireless device.

[0452] For example, the message also includes information about the operation in which the failure was detected.

[0453] For example, the message also includes information notifying the wireless device of the derivation of prediction information based on the AI and / or ML model.

[0454] For example, the message also includes information about the AI and / or ML model used to derive the predictive information.

[0455] For example, the message may also include information about the point in time at which the operation is performed.

[0456] According to some embodiments of the present disclosure, a wireless device may derive a predicted failure and a predicted time point at which the predicted failure occurs.

[0457] In this case, the information about the prediction information may include information about the predicted failure and information about the predicted time point.

[0458] For example, the information on the prediction information includes information on the time difference between the prediction time point and the time point at which the failure occurs.

[0459] For example, inputs to the AI and / or ML models may include stored measurements, location information, mobility information, user information, registration information, information about faults, stored messages, stored configurations, and / or UE capabilities.

[0460] That is, the wireless device may derive prediction information using information about stored measurement results, location information, mobility information, user information, registration information, information about failures, stored messages, stored configurations, and / or UE capabilities.

[0461] For example, the prediction information is the output of an AI and / or ML model. For example, the output of the AI and / or ML model may include a predicted failure at a future time point, a predicted measurement result at a future time point, a predicted time point at which the predicted failure will occur, a predicted operation for the wireless device, a predicted configuration for the wireless device, and / or a predicted location of the wireless device (e.g., a predicted cell, a predicted gNB, and / or a predicted tracking area).

[0462] For example, the input of the AI and / or ML model can be used to train the AI / ML model. That is, information about stored measurement results, location information, mobility information, user information, registration information and / or UE capabilities can be used to train the AI / ML model.

[0463] Additionally, the output of the AI and / or ML model can also be used to train the AI / ML model. Specifically, information regarding a predicted failure at a future point in time, predicted measurement results at a future point in time, a predicted time at which the predicted failure will occur, a predicted operation for the wireless device, a predicted configuration for the wireless device, and / or a predicted location of the wireless device (e.g., a predicted cell, a predicted gNB, and / or a predicted tracking area) can be used to train the AI / ML model.

[0464] According to some embodiments of the present disclosure, a wireless device may receive an AI and / or ML model configuration including information about the AI and / or ML model from a network. The wireless device may derive a predicted measurement result based on the AI and / or ML model. In this case, an operation may be triggered by the predicted measurement result.

[0465] For example, the message may further include information notifying that the operation is triggered by the predicted measurement result.

[0466] For example, the prediction information may include information about predicted measurement results.

[0467] The wireless device may include AI and / or ML (in other words, AI / ML) functionality. The AI / ML functionality is an entity used for AI / ML operations.

[0468] The wireless device may configure the AI / ML function using the AI / ML model. For example, the wireless device may receive a configuration for the AI / ML model from the network. The wireless device may apply the received configuration for the AI / ML model to the AI / ML function.

[0469] For example, the AI / ML function may be configured with a first AI / ML model. The AI / ML function may then derive prediction information (e.g., predicted measurement results) based on the first AI / ML model.

[0470] After receiving the configuration for the second AI / ML model, the wireless device may configure the AI / ML function using the second AI / ML model. The AI / ML function may then derive prediction information (e.g., predicted measurement results) based on the second AI / ML model.

[0471] For example, the entire AI / ML functionality will include several different components (e.g., data collection, model training, model inference, actors).

[0472] For example, the AI / ML model used in step S2001 may be described below.

[0473] Hereinafter, examples of AI / ML models used in the present disclosure are described.

[0474] An AI / ML model for CSI feedback enhancement is described.

[0475] The following set of goals have been identified for the dual-side CSI compression use case. First, to ensure that both the UE and network parts of the model are configured and applied according to their applicable scenarios and configurations. Second, to ensure that the models are correctly matched, ensuring that the CSI generation portion used at the UE corresponds to the CSI reconstruction portion employed at the gNB. Third, to allow for seamless operation, simultaneous activation (deactivation) and switching of the dual-side models is required.

[0476] Regarding the last point above, for the dual-side model CSI compression use case, the selection, activation (or deactivation), switching, and fallback of the AI / ML model or AI / ML functionality can be initiated by either the UE or the gNB. It is important to distinguish between the various scenarios and understand their applicability to both the UE-side model and the network-side model.

[0477] For data collection, model delivery / transfer, and function-to-entity mapping analysis, various scenarios are developed for both the bilateral CSI compression use case and the UE-side CSI prediction use case when the data generation and termination entities are different. For example, for:

[0478] 1>Model training:

[0479] 2> For the bilateral CSI compression use case, the training data can be generated by the UE or gNB, depending on the specific requirements, and the termination points for the training data can include the gNB, OAM, over-the-top (OTT) server, or UE.

[0480] 3> RAN2 identifies situations where the core network can be used for model training. However, this was not studied as it is beyond the scope of this working group.

[0481] 2> For the UE-side CSI prediction use case, the training data can be generated by the UE, and the termination point for the training data can include the UE or the UE-side OTT server.

[0482] 3>RAN2 identifies situations where OAM or core network can be used for UE-side model training. However, this was not studied as it is outside the scope of this working group.

[0483] RAN2 identified cases where the gNB can be used for UE-side model training. However, no conclusions were reached as this depends on RAN1 progress.

[0484] 1>Inference:

[0485] 2> For bilateral CSI compression use cases:

[0486] 3> For the network part of the two-sided model inference, the UE can generate the necessary input data, and the termination point for this input data is located in the gNB performing the inference process.

[0487] 3>For the UE part of the two-sided model inference, the input data is available internally at the UE that performs the inference process.

[0488] 2> For UE-side CSI prediction use case:

[0489] 3>For UE-side model inference, the input data is available internally at the UE where the inference process is performed.

[0490] 1> Management:

[0491] 2> For the bilateral CSI compression use case, the gNB performs model / functionality control (e.g., selection, activation (deactivation), switching, fallback, etc.).

[0492] 3>RAN2 identifies the situation where the UE performs control. However, no conclusion is drawn because this depends on the progress of RAN1.

[0493] 2> For UE-side CSI prediction use case:

[0494] 3> When monitoring resides within the UE, model / function control (e.g., selection, activation (deactivation), switching, fallback, etc.) can be performed by the UE.

[0495] 3> When monitoring resides within the gNB or UE, model / function control (e.g., selection, activation (deactivation), handover, fallback, etc.) can be performed by the gNB.

[0496] 2> Monitoring:

[0497] 3>UE monitors the performance of its UE-side model.

[0498] 3> For network-side monitoring of the UE-side model, if necessary, the UE can generate calculated performance metrics or data required for performance metric calculation, and the termination point of these is the gNB.

[0499] An AI / ML model for beam management is described.

[0500] For beam management, model or function selection, activation (deactivation), switching, and fallback can also be initiated by the UE or gNB. It is important to distinguish between the various cases and understand their applicability to UE-side models and network-side models.

[0501] For data collection, model delivery / transfer, and function-to-entity mapping analysis, various scenarios are explored when the data generating and terminating entities are different. For example, for:

[0502] 1>Model training:

[0503] 2> For the UE-side model, the training data can be generated by the UE, and the termination point for the training data can include the UE or the UE-side OTT server.

[0504] 3>RAN2 identifies situations where OAM or core network can be used for UE-side model training. However, this was not studied as it is outside the scope of this working group.

[0505] RAN2 identified cases where the gNB can be used for UE-side model training. However, no conclusions were reached as this depends on RAN1 progress.

[0506] 2> For the gNB-side model, the training data can be generated by the gNB or UE, and the termination point for the training data can include the gNB or OAM.

[0507] RAN2 identified scenarios where OTT servers and core networks can be used for gNB-side model training. However, this was not studied as it was outside the scope of this working group.

[0508] 1>Inference:

[0509] 2>For UE-side model inference, the input data is available internally at the UE where the inference process is performed.

[0510] 2> For network-side model inference, the UE can generate the necessary input data, and the termination point for this input data is located in the gNB that performs the inference process.

[0511] 1> Management:

[0512] 2> For the UE-side model, when monitoring resides within the UE, model / function control (e.g., selection, activation (deactivation), switching, fallback, etc.) can be performed by the UE.

[0513] 2> For the UE-side model, when monitoring resides within the gNB or UE, model / function control (e.g., selection, activation (deactivation), handover, fallback, etc.) can be performed by the gNB.

[0514] 2> Monitoring:

[0515] 3>UE monitors the performance of its UE-side model.

[0516] 3> For network-side monitoring of the UE-side model, if necessary, the UE can generate calculated performance metrics or data required for performance metric calculation, and the termination point for these is the gNB.

[0517] 3>For the network-side model, monitoring resides in the gNB.

[0518] An AI / ML model for enhanced positioning accuracy is described.

[0519] For positioning use cases, the selection, (or (de)activation,) switching, and fallback of models or functionalities can be initiated by the UE, gNB, or LMF. It is important to distinguish between the various cases and understand their applicability to UE-side models and network-side models.

[0520] For data collection, model delivery / transfer, and function-to-entity mapping analysis, various scenarios are explored when the data generating and terminating entities are different. For example, for:

[0521] 1>Model training:

[0522] 2> For the UE-side model, the training data can be generated by the UE, and the termination point for the training data can include the UE or the UE-side OTT server.

[0523] 3>RAN2 identifies situations where OAM or core network can be used for UE-side model training. However, this was not studied as it is outside the scope of this working group.

[0524] 3>RAN2 identified a case where LMF could be used for UE-side model training. However, no conclusion was reached as this depends on RAN1 progress.

[0525] 2> For the gNB-side model, the training data can be generated by the gNB, and the termination point for the training data can include the gNB or OAM.

[0526] RAN2 identified a scenario where LMF could be used for gNB-side model training. However, no conclusions were reached as this depends on RAN1 progress.

[0527] 2>For the LMF side model, LMF is the termination point for the training data.

[0528] 1>Inference:

[0529] 2>For UE-side model inference, the input data is available internally at the UE where the inference process is performed.

[0530] 2> For gNB-side model inference, the input data is available internally at the gNB. In this case, the UE can also generate the necessary input data, and the endpoint for this input data is within the gNB performing the inference process.

[0531] 2> For LMF side model inference, the UE or gNB can generate the necessary input data, and the termination point for this input data is located within the LMF where the inference process is performed.

[0532] 1> Management:

[0533] 2> For the UE-side model, when monitoring resides within the UE, model / function control (e.g., selection, activation (deactivation), switching, fallback, etc.) can be performed by the UE.

[0534] 2> For the gNB-side model, the gNB performs model / function control (e.g., selection, activation (deactivation), switching, fallback, etc.).

[0535] 2> When monitoring resides in the LMF or UE, model / function control (e.g., selection, activation (deactivation), switching, fallback, etc.) can be performed by the LMF.

[0536] 2> Monitoring:

[0537] 3>UE monitors the performance of its UE-side model.

[0538] 3> For monitoring on the gNB side, and if necessary, the calculated performance metrics or data required for performance metric calculation can be generated by at least the gNB.

[0539] 3> For monitoring on the LMF side, if necessary, the gNB or UE can generate the calculated performance metrics or the data required for performance metric calculation, and the termination point for these metrics is the LMF.

[0540] Hereinafter, examples related to the input and output of the AI / ML model of the present disclosure are described.

[0541] For example, the input and output of AI / ML models can be used to Figure 20 AI / ML models in .

[0542] For example, AI / ML models can be used for CSI compression.

[0543] The data content (ie, input data) for training of CSI compression may include (i) target CSI, (ii) CSI feedback, and (iii) a gradient for the CSI feedback.

[0544] The inferred data content for CSI compression may include CSI feedback.

[0545] The data content monitored for CSI compression may include (i) reconstructed CSI from the NW to the UE, (ii) calculated performance metrics, and (iii) target CSI.

[0546] For example, AI / ML models can be used for CSI prediction on the UE side.

[0547] The data content for training CSI prediction on the UE side may include target CSI in the observation and prediction windows.

[0548] The inferred data content for CSI prediction on the UE side may include predicted CSI feedback (AI / ML output).

[0549] The data content of monitoring for CSI prediction at the UE side may include (i) the ground truth corresponding to the predicted CSI (ie, target CSI) and (ii) the calculated performance metric / performance monitoring output.

[0550] For example, AI / ML models can be used for beam management.

[0551] The data content for training of beam management on the UE side and the network side may include L1-RSRP and / or beam ID.

[0552] The inferred data content for beam management on the UE side may include beam prediction results.

[0553] The inferred data content for beam management on the network side may include L1-RSRP and beam ID for set B (if needed).

[0554] The data content of monitoring for beam management on the UE side may include (i) event occurrence and / or calculated performance metrics (from UE to NW) and (ii) L1-RSRP and / or beam ID.

[0555] The data content monitored for beam management on the network side may include L1-RSRP and / or beam ID.

[0556] For example, AI / ML models can be used for positioning.

[0557] The data content for training on positioning may include (i) measurements (corresponding to model inputs): timing, power and / or phase information, (ii) labels: location coordinates as model outputs, and / or (iii) labels: intermediate positioning measurement results (timing information, LOS / NLOS indicators) as model outputs.

[0558] The inferred data content for positioning may include (i) position coordinates as model output, (ii) intermediate positioning measurements (timing information, LOS / NLOS indicators) as model output, and / or (iii) measurements (corresponding to model inputs): timing, power and / or phase information.

[0559] According to some embodiments of the present disclosure, a wireless device may communicate with at least one of a user device other than the wireless device, a network, or an autonomous vehicle.

[0560] In the following, some embodiments of a method of adding AI / ML information in a self-organizing network are described.

[0561] In the present disclosure, if a UE experiences a connection failure, the UE transmits information related to AI / ML operations that may be related to the connection failure.

[0562] For example, 1) The network configures the AI / ML configuration. 2) The UE derives the prediction results (e.g., measurement results rlf, bf, etc.). 3) The UE reports the prediction results to the network. 4-1) The network configures the UE to change some parameters or perform switching based on the prediction results, and the UE applies the configuration. 4-2) The UE autonomously performs conditional operations (e.g., CHO, CPAC). 5) The UE experiences a connection failure. 6) The UE constructs general fault-related information in the UE information response message or SCG / MCG fault information message for SON / MDT operations. At this time, AI / ML information related to the AI / ML model configuration and AI / ML prediction results is added to the corresponding message. 7) The UE reports the fault information to the network. 8) The network updates the AI / ML model / configuration.

[0563] Figure 21 An example of a method for supporting AI and ML operations in a wireless communication system is shown.

[0564] In step S2101, the network may configure the UE using the AI / ML model configuration.

[0565] 1> AI / ML model configuration can be transmitted via RRC message, NAS message or data packet (DRB).

[0566] >> In terms of radio bearers configured for AI / ML models, AI / ML specific SRBs / DRBs can be used.

[0567] 1> Each AI / ML model can be associated with a specific functional body, or a specific AI / ML model can be commonly used for a specific functional body.

[0568] 2> Functionality refers to use cases (e.g., beam prediction and RRM prediction).

[0569] 1>AI / ML model configuration can include complete model information, partial model information, or parameters related to each AI / ML model.

[0570] 1> AI / ML model configuration can include reporting conditions for each AI / ML model

[0571] 1>(Example 1) For example, an AI / ML model configuration may include the following:

[0572] 2>AI / ML Model A

[0573] 3>Functional body F_a

[0574] 3>Parameter P_a

[0575] 2>AI / ML Model B

[0576] 3>Functional body F_b

[0577] 3>Parameter P_b

[0578] 2>AI / ML Model C

[0579] 3>Parameter P_c

[0580] 2> In this example, based on AI / ML model configuration,

[0581] 3>UE can apply AI / ML model A to function F_a with parameters P_a.

[0582] 3>UE can apply AI / ML model B to function F_b with parameters P_b.

[0583] 3>UE can apply AI / ML model C to several functions with parameters P_c.

[0584] 1>(Example 2) For example, an AI / ML model configuration may include the following:

[0585] 2>Function F_1

[0586] 2>AI / ML Model A

[0587] 3>Parameter P_a

[0588] 2>AI / ML Model B

[0589] 3>Parameter P_b

[0590] 2> In this example, based on AI / ML model configuration,

[0591] 3>UE can apply AI / ML model A to function F_1 with parameters P_a.

[0592] 3>UE can apply AI / ML model B to function F_1 with parameters P_b.

[0593] In step S2102, in order to derive prediction results, the UE may be configured with more prediction model configurations.

[0594] 1> The prediction model configuration may include prediction model structure information,

[0595] 2> The network can configure the machine learning model to be used by the UE.

[0596] 3> The network can include machine learning types such as reinforcement learning, supervised learning, or unsupervised learning.

[0597] 3> The network can include machine learning models such as DNN, CNN, RNN and DRL.

[0598] 3> The configured ML model can be a pre-trained ML model that has been trained a priori by the network.

[0599] 4> Describe the configured ML model through model description information including model structure and parameters.

[0600] 4> For example, a neural network-based model may include an input layer, an output layer, and a hidden layer, wherein each layer includes one or more neurons (equivalent to nodes).

[0601] 5>Connect different layers based on the connections between neurons in different layers

[0602] 6> Each connection between two different neurons in two different layers can be directional (e.g., neuron A to neuron B, which means the output of neuron A is fed into neuron B)

[0603] 6> Each neuron can provide input to one or more connected neurons (1 to N connections).

[0604] 6> For a connection between two neurons (neuron A to neuron B), the output of one neuron (A) is scaled by the weight, and the other neuron takes the scaled output as its input.

[0605] 6> Each neuron can take input from one or more connected neurons (N to 1 connection) and combines the inputs from the connected neurons and produces an output based on an activation function.

[0606] 3>The configured ML model can be the ML model to be trained.

[0607] 4> The configured ML model is described by the model description information, which includes the model structure to be trained and the initial parameters.

[0608] 4> When the network configures the ML model to be trained, it is also possible to configure training parameters such as optimization objectives and optimization-related configuration parameters.

[0609] 3> The network may include machine learning input parameters for the machine learning model (e.g., UE location information, radio measurements related to serving and neighboring cells, UE mobility history).

[0610] 3> The network can include machine learning outputs (e.g., UE trajectory prediction, predicted target cell, predicted switching time, and UE traffic prediction).

[0611] 2>UE can perform machine learning model training, validation and testing, which can generate model performance metrics based on predictive model configuration.

[0612] 3>UE can use machine learning input parameters to perform model training.

[0613] 2>UE can use the configured ML model to perform ML tasks such as prediction of measurements.

[0614] 3>UE can derive machine learning output.

[0615] 3>UE can make inferences based on the output and use the output as feedback to the machine learning model.

[0616] 2> Regarding the results related to the machine learning output and the accuracy of the machine learning model, the UE can send feedback to the network.

[0617] 3> The network can update the machine learning model and parameters related to the machine learning model.

[0618] In step S2103 , the UE may derive / store prediction results for the functional body based on the AI / ML model configuration.

[0619] 1>UE can derive predicted measurement results

[0620] 1>UE can deduce and predict connection success / connection failure

[0621] 1>UE can deduce and predict radio link failure / beam failure

[0622] 1>UE can deduce and predict RACH failure

[0623] 1>UE can derive predicted location information

[0624] 1>UE can derive and predict mobility history

[0625] 1>UE can deduce and predict handover success / failure

[0626] 1> In each prediction, the UE can also derive the predicted time, location, cell, etc.

[0627] In step S2104, the UE may report the prediction results based on the reporting conditions configured by the AI / ML model.

[0628] 1> Prediction results can include measurement results

[0629] 1> The prediction result may include connection success / connection failure.

[0630] 1>Prediction results may include radio link failure / beam failure.

[0631] 1> The prediction result may include RACH failure.

[0632] 1> The prediction results can include location information

[0633] 1> Prediction results can include mobility history

[0634] 1> The prediction results can include switching success / failure

[0635] 1> The prediction results may include the predicted time, location, cell, etc.

[0636] In step S2105 , the UE may apply the new configuration or perform conditional operations.

[0637] For example, in step S2105-1, the network may configure the UE with a configuration based on the prediction result, and the UE may apply the new configuration.

[0638] 1> Configuration may include some parameter changes in the current cell (eg, RLM / BFD RS change, power control change, etc.).

[0639] 1> Configuration can include switching commands

[0640] 1> Configuration may include cell addition such as CA.

[0641] 1>Configuration may include CG additions such as SCG additions.

[0642] 1> Configuration can include conditional reconfiguration for CHO, CPAC, etc.

[0643] For another example, in step S2105-2, the UE may autonomously perform the conditional operation, that is, the UE may automatically apply the new configuration to the conditional operation.

[0644] 1>Conditional operations can be CHO and / or CPC.

[0645] 1>The conditional operation can be cell / CG addition.

[0646] 1>Conditional operation can be cell / CG activation (deactivation)

[0647] 1>The conditional operation may be cell / CG release.

[0648] In step S2106, the UE may detect a failure in UE operation.

[0649] 1>The fault may be a switching fault.

[0650] 1>The fault may be a radio link failure

[0651] 1>The fault may be a beam fault

[0652] 1>The fault may be a random access fault

[0653] 1>The fault may be a configuration fault

[0654] 1>The fault may be a connection failure

[0655] 1>The fault can be a recovery fault

[0656] 1> The fault may be a conditional reconfiguration (e.g., CHO, CPAC) fault.

[0657] 1>The fault may be a cell / CG addition failure.

[0658] 1>The failure may be a cell / CG activation (deactivation) failure.

[0659] 1> If the UE detects a failure, the UE may increase the number of (connection) failures associated with the AI / ML operation.

[0660] 2>Can count the number of AI / ML models

[0661] 2>Can count the number of AI / ML functions

[0662] 2>Can count the number of AI / ML configurations

[0663] In step S2107, the UE may send a report message including fault-related information.

[0664] 1> A report message can be reported immediately after a fault is detected (immediate information).

[0665] 1> Report message (stored information) can be reported after collecting fault history.

[0666] 1> Report messages can be delivered by network request or by UE decision.

[0667] 1> The report message may include the first information

[0668] 2> The first information may be related to a detected failure (eg, connection establishment failure information, connection recovery failure information, RLF report).

[0669] 2> For example, (R16 / 17 general fault information)

[0670] 3>Fault type: T310 expiration, random access problem, RLC maximum retransmission times, synchronous reconfiguration failure, etc.

[0671] 3>Measurement results on frequencies that the UE is configured to measure in idle / inactive state

[0672] 3>Measurement results of measurements on the frequencies that the UE is configured to measure via measConfig

[0673] 3>Previous cell id indicating the carrier frequency and physical cell id of the source cell

[0674] 3> Faulty cell identifier indicating the physical cell id and carrier frequency of the cell where the fault was detected or the faulty cell change / cell addition,

[0675] 3>Failure time indicating the time elapsed since the last execution of RRCReconfiguration with reconfigurationWithSync to the failure.

[0676] 3>plmn logo

[0677] 3> Measurement results of the last serving cell including cell-level RSRP, RSRQ and available SINR of the source Pcell (in case of HO failure) or Pcell (in case of RLF) based on available SSB / CSI-RS measurements collected until the UE detects the failure

[0678] 3> RLM configuration information including the radio link monitoring configuration of the source Pcell (in case of HO failure) or Pcell (in case of RLF)

[0679] 3> Measurement results of neighboring cells including neighboring cell measurements

[0680] 3> The measurement result includes all available measurement quantities of the best measurement cell except the source Pcell (in case of HO failure) or Pcell (in case of RLF)

[0681] 3>CHO Config includes conditional reconfiguration information (first trigger event, or the time between events if two events are configured)

[0682] 3> The time elapsed between 1) CHO configuration and CHO execution (if CHO fails) and 2) CHO configuration and HO execution (if HO fails)

[0683] 3>CHO candidate cell list includes the global cell ID / physical cell ID / carrier frequency of each candidate cell for conditional handover in case of handover failure

[0684] 3>Finally switch the type (for example, cho, daps, etc.)

[0685] 3>Connection failure type (for example, hof, rlf)

[0686] 3> Fault Pcell Id sets the global cell identity / tracking area code / physical cell identity / carrier frequency where the radio link failure was detected

[0687] 3> The previous cell sets the global cell identifier / tracking area code of the Pcell that received the last executed CHO message

[0688] 3>RLF reasons (e.g., random access problem, beam failure recovery failure, etc.).

[0689] 3> Location information

[0690] 1> The report message may include the second information

[0691] 2> The second information is related to AI / ML operations and may include the following:

[0692] 3> Information indicating whether the UE was performing AI / ML-based tasks around the time of the failure

[0693] 4> Whether the measurement results obtained near the failure time are AI / ML-based information.

[0694] 4> Whether the measurement results reported near the failure time are AI / ML-based information.

[0695] 4> Information that triggers fault mobility through AI / ML-based measurements or traditional measurements.

[0696] 3>AI / ML model information

[0697] 4> Indication that the AI / ML model is configured / activated / operated

[0698] 4> List of configured AI / ML models, if available

[0699] 4>AI / ML model ID (for example, Model A)

[0700] 4>AI / ML model capabilities (e.g., F_a and F_1)

[0701] 4> AI / ML model configuration associated with each prediction

[0702] 5> The model configuration may include bitmap type information to indicate the activated model.

[0703] 3>Time information related to prediction

[0704] 4>The time when the following items occur

[0705] 5> Receive AI / ML model configuration, derive prediction results, send prediction report, receive network configuration / command after the report, UE action based on network command or UE decision, fault occurs

[0706] 4> Time T1 between AI / ML model configuration and prediction

[0707] 5> If you update the AI / ML model based on the previous AI / ML model configuration, then

[0708] 6> Time T1_1 elapsed between AI / ML model configuration and AI / ML model update

[0709] 6> The time T1_2 elapsed between AI / ML model update and prediction

[0710] 4> Time T2 between prediction and prediction report

[0711] 4> Time T3 between prediction report and network configuration / command

[0712] 4> Time T4 between UE operation based on network command or UE decision and network configuration / command

[0713] 4> Time T5 between UE operation based on network command or UE decision and failure

[0714] 4> Elapsed time information may be included in combination (eg, T1+T2).

[0715] 3>Prediction result information

[0716] 4>Difference between predicted measurement results and actual measurement results

[0717] 4> Difference between predicted failure and actual failure (e.g. predicted RLF -> actual non-RLF)

[0718] 4> The difference between the predicted time and the actual time of the measurement result / fault-related prediction (e.g., predicted time t -> actual time t + T'). T' can be positive or negative.

[0719] 4> Separate predicted measurement results and actual measurement results

[0720] 4>Separate predicted failures and actual failures

[0721] 3>Cell Information

[0722] 4>Configure the PLMn ID, global cell ID, tracking area code, physical cell ID, and carrier frequency for AI / ML model configuration

[0723] 4> UE derives the predicted PLMn identity, global cell identity, tracking area code, physical cell identity and carrier frequency

[0724] 4> UE report prediction report PLMN identity, global cell identity, tracking area code, physical cell identity and carrier frequency

[0725] 4>UE receives the network command PLMn identity, global cell identity, tracking area code, physical cell identity and carrier frequency based on the prediction result

[0726] 4> UE PLMn identity, global cell identity, tracking area code, physical cell identity and carrier frequency based on network command or UE decision operation

[0727] 3>The number of failures associated with the AI / ML model

[0728] 2> The second information can be transmitted independently of the first information.

[0729] 2> The second information can be delivered in each content of the first information

[0730] In step S2108, the network may send the AI / ML model configuration.

[0731] 1>AI / ML model configuration may include updating a previous AI / ML model configuration with some parameters for a specific AI / ML model.

[0732] 1>AI / ML model configuration may include changing a previous AI / ML model configuration using a new AI / ML model.

[0733] 1>AI / ML model configuration can be used to activate (deactivate) specific AI / ML models

[0734] 1>AI / ML model configuration can be used to release AI / ML model operations.

[0735] Figure 22 An example of the time elapsed from receiving an AI / ML model configuration is shown.

[0736] Specifically, Figure 22 Time information related to the prediction described in step S2107 is illustrated.

[0737] T1 is the time that elapses between the configuration of the AI / ML model and the derivation of the prediction results.

[0738] T2 is the time that elapses between the derivation of the forecast and its reporting.

[0739] T3 is the time elapsed between the predicted report and the network configuration / command (that is, the time the UE receives the network configuration / command).

[0740] T4 is the time elapsed between network configuration / command and UE operation based on the network configuration / command. Otherwise, T4 is the time elapsed between network configuration / command and UE operation based on UE decision.

[0741] T5 is the time elapsed between the UE operation based on network configuration / command (or the UE operation based on UE decision) and the failure.

[0742] Time information (eg, T1, T2, T3, T4, and / or T5) may be included in the report message.

[0743] Figure 20 、 Figure 21 and Figure 22 Some of the specific steps shown in the examples may not be essential steps and may be omitted. Figure 20 、 Figure 21 and Figure 22 In addition to the steps shown, other steps may be added, and the order of the steps may be changed. Some of the above steps may have their own technical meanings.

[0744] Hereinafter, a device for supporting AI and ML operations in a wireless communication system according to some embodiments of the present disclosure will be described. In this article, the device may be Figure 2 、 Figure 3 and Figure 5 A wireless device (100 or 200) in.

[0745] For example, the wireless device may perform the above method. Detailed descriptions that overlap with the above content may be simplified or omitted.

[0746] Reference Figure 5 , the wireless device 100 may include a processor 102 , a memory 104 , and a transceiver 106 .

[0747] According to some embodiments of the present disclosure, the processor 102 may be configured to be operatively coupled with the memory 104 and the transceiver 106 .

[0748] The processor 102 may be configured to detect a fault in operation with the network. The processor 102 may be configured to control the transceiver 106 to transmit a message including (i) information about the fault and (ii) information about predicted information related to the fault. The predicted information may be derived by an artificial intelligence (AI) and / or machine learning (ML) model of the wireless device.

[0749] For example, the information regarding prediction information related to a fault may include information regarding an operation in which the fault was detected.

[0750] For example, the message may also include information notifying the wireless device of derivation of prediction information based on an AI and / or ML model.

[0751] For example, the message may also include information about the AI and / or ML model used to derive the predictive information.

[0752] For example, the processor 102 may be configured to derive a predicted failure and a predicted point in time at which the predicted failure will occur.

[0753] For example, the information about the prediction information may include information about the predicted failure and information about the predicted time point.

[0754] For example, the information about the prediction information may include information about a time difference between a prediction time point and a time point when a failure occurs.

[0755] For example, the message may also include information about the point in time at which the operation is performed.

[0756] For example, operations may include predictive operations derived from AI and / or ML models.

[0757] For example, the processor 102 may be configured to control the transceiver 106 to receive an AI and / or ML model configuration including information about the AI and / or ML model from the network. The processor 102 may be configured to derive a predicted measurement result based on the AI and / or ML model. The operation may be triggered by the predicted measurement result.

[0758] For example, the message may further include information notifying that the operation is triggered by the predicted measurement result.

[0759] For example, the prediction information may include information about predicted measurement results.

[0760] For example, the operations may include conditional operation, conditional handover (CHO), conditional PScell change (CPC), cell and / or cell group (CG) addition, cell and / or CG activation, cell and / or CG deactivation and / or cell and / or CG release.

[0761] For example, failures in operation may include switching failure, radio link failure, beam failure, random access failure, configuration failure, connection failure, recovery failure, conditional reconfiguration failure, cell and / or CG addition failure, cell and / or CG activation failure and / or cell and / or CG deactivation failure.

[0762] For example, the processor 102 may be configured to control the transceiver 106 to communicate with at least one of a user device other than a wireless device, a network, or an autonomous vehicle.

[0763] Hereinafter, a processor of a wireless device for supporting AI and ML operations in a wireless communication system according to some embodiments of the present disclosure will be described.

[0764] The processor may be configured to control the wireless device to detect a fault in operation with a network. The processor may be configured to control the wireless device to send a message including (i) information about the fault and (ii) information about prediction information related to the fault. The prediction information may be derived by an artificial intelligence (AI) and / or machine learning (ML) model of the wireless device.

[0765] For example, the information regarding prediction information related to a fault may include information regarding an operation in which the fault was detected.

[0766] For example, the message may also include information notifying the wireless device of derivation of prediction information based on an AI and / or ML model.

[0767] For example, the message may also include information about the AI and / or ML model used to derive the predictive information.

[0768] For example, the processor may be configured to control the wireless device to derive a predicted failure and a predicted time point at which the predicted failure occurs.

[0769] For example, the information about the prediction information may include information about the predicted failure and information about the predicted time point.

[0770] For example, the information about the prediction information may include information about a time difference between a prediction time point and a time point when a failure occurs.

[0771] For example, the message may also include information about the point in time at which the operation is performed.

[0772] For example, operations may include predictive operations derived from AI and / or ML models.

[0773] For example, the processor may be configured to control the wireless device to receive an AI and / or ML model configuration including information about the AI and / or ML model from the network. The processor may be configured to control the wireless device to derive a predicted measurement result based on the AI and / or ML model. The operation may be triggered by the predicted measurement result.

[0774] For example, the message may further include information notifying that the operation is triggered by the predicted measurement result.

[0775] For example, the prediction information may include information about predicted measurement results.

[0776] For example, the operations may include conditional operation, conditional handover (CHO), conditional PScell change (CPC), cell and / or cell group (CG) addition, cell and / or CG activation, cell and / or CG deactivation and / or cell and / or CG release.

[0777] For example, failures in operation may include switching failure, radio link failure, beam failure, random access failure, configuration failure, connection failure, recovery failure, conditional reconfiguration failure, cell and / or CG addition failure, cell and / or CG activation failure and / or cell and / or CG deactivation failure.

[0778] For example, the processor may be configured to control the wireless device to communicate with at least one of a user device other than the wireless device, a network, or an autonomous vehicle.

[0779] Hereinafter, a non-transitory computer-readable medium according to some embodiments of the present disclosure will be described, on which a plurality of instructions for supporting AI and ML operations in a wireless communication system are stored.

[0780] According to some embodiments of the present disclosure, the technical features of the present disclosure may be implemented directly in hardware, in software executed by a processor, or in a combination of the two. For example, a method performed by a wireless device in wireless communication may be implemented in hardware, software, firmware, or any combination thereof. For example, the software may reside in RAM memory, flash memory, ROM memory, EPROM memory, EEPROM memory, registers, a hard disk, a removable disk, a CD-ROM, or any other storage medium.

[0781] Some examples of storage media are coupled to a processor so that the processor can read information from the storage media. In another embodiment, the storage media can be integrated into the processor. The processor and storage media can reside in an ASIC. For another example, the processor and storage media can reside as discrete components.

[0782] Computer-readable media may include tangible and non-transitory computer-readable storage media.

[0783] For example, non-transitory computer-readable media may include random access memory (RAM) such as synchronous dynamic random access memory (SDRAM), read-only memory (ROM), non-volatile random access memory (NVRAM), electrically erasable programmable read-only memory (EEPROM), FLASH memory, magnetic or optical data storage media, or any other medium that can be used to store instructions or data structures. Non-transitory computer-readable media may also include combinations of the foregoing.

[0784] Furthermore, the methods described herein may be implemented at least in part by a computer-readable communication medium that carries or communicates code in the form of instructions or data structures and that can be accessed, read, and / or executed by a computer.

[0785] According to some embodiments of the present disclosure, a non-transitory computer-readable medium stores a plurality of instructions, and the stored plurality of instructions may be executed by a processor of a wireless device.

[0786] The stored plurality of instructions may cause the wireless device to detect a fault in operation with a network. The stored plurality of instructions may cause the wireless device to send a message including (i) information about the fault and (ii) information about prognostic information related to the fault. The prognostic information may be derived by an artificial intelligence (AI) and / or machine learning (ML) model of the wireless device.

[0787] For example, the information regarding prediction information related to a fault may include information regarding an operation in which the fault was detected.

[0788] For example, the message may also include information notifying the wireless device of derivation of prediction information based on an AI and / or ML model.

[0789] For example, the message may also include information about the AI and / or ML model used to derive the predictive information.

[0790] For example, the stored plurality of instructions may enable the wireless device to derive a predicted failure and a predicted point in time at which the predicted failure will occur.

[0791] For example, the information about the prediction information may include information about the predicted failure and information about the predicted time point.

[0792] For example, the information about the prediction information may include information about a time difference between a prediction time point and a time point when a failure occurs.

[0793] For example, the message may also include information about the point in time at which the operation is performed.

[0794] For example, operations may include predictive operations derived from AI and / or ML models.

[0795] For example, the stored instructions may cause the wireless device to receive an AI and / or ML model configuration including information about the AI and / or ML model from a network. The stored instructions may cause the wireless device to derive a predicted measurement result based on the AI and / or ML model. Operations may be triggered by the predicted measurement result.

[0796] For example, the message may further include information notifying that the operation is triggered by the predicted measurement result.

[0797] For example, the prediction information may include information about predicted measurement results.

[0798] For example, the operations may include conditional operation, conditional handover (CHO), conditional PScell change (CPC), cell and / or cell group (CG) addition, cell and / or CG activation, cell and / or CG deactivation and / or cell and / or CG release.

[0799] For example, failures in operation may include switching failure, radio link failure, beam failure, random access failure, configuration failure, connection failure, recovery failure, conditional reconfiguration failure, cell and / or CG addition failure, cell and / or CG activation failure and / or cell and / or CG deactivation failure.

[0800] According to some embodiments of the present disclosure, the stored plurality of instructions may enable the wireless device to communicate with at least one of a user device other than the wireless device, a network, or an autonomous vehicle.

[0801] Hereinafter, a method for supporting AI and ML operations in a wireless communication system performed by a base station (BS) according to some embodiments of the present disclosure will be described.

[0802] The BS may provide a configuration for an artificial intelligence (AI) and / or machine learning (ML) model to the wireless device. The BS may receive a message from the wireless device including (i) information about a fault detected in the operation of the wireless device and (ii) information about prediction information related to the fault.

[0803] Hereinafter, a base station (BS) for supporting AI and ML operations in a wireless communication system according to some embodiments of the present disclosure will be described.

[0804] The BS may include a transceiver, a memory, and a processor operatively coupled to the transceiver and the memory.

[0805] The processor may be configured to control the transceiver to provide a configuration for an artificial intelligence (AI) and / or machine learning (ML) model to the wireless device. The processor may be configured to control the transceiver to receive a message from the wireless device including (i) information about a fault detected in operation of the wireless device and (ii) information about prediction information related to the fault.

[0806] The present disclosure may have various beneficial effects.

[0807] According to some embodiments of the present disclosure, a wireless device may efficiently support AI / ML operations by reporting information related to the AI / ML operations.

[0808] For example, if the network can identify problems with the AI / ML model through fault reports, the network can manage the AI / ML model well from a model monitoring perspective by updating the parameters of the current AI / ML model or changing to a more suitable AI / ML model.

[0809] In other words, for example, the network can efficiently identify AI / ML problems in the UE and can efficiently manage the AI / ML models to be used in the UE (e.g., the UE can efficiently receive configurations of new AI / ML models).

[0810] According to some embodiments of the present disclosure, a wireless network system may provide efficient management of AI / ML operations of wireless devices by receiving information related to the AI / ML operations.

[0811] The advantageous effects that can be obtained by the specific embodiments of the present disclosure are not limited to the advantageous effects listed above. For example, there may be various technical effects that can be understood and / or derived from the present disclosure by a person of ordinary skill in the relevant art. Therefore, the specific effects of the present disclosure are not limited to those explicitly described herein, but may include various effects that can be understood or derived from the technical features of the present disclosure.

[0812] The claims in this disclosure may be combined in various ways. For example, the technical features in the method claims of this disclosure may be combined to be implemented or performed in a device, and the technical features in the device claims may be combined to be implemented or performed in a method. Furthermore, the technical features in the method claims and the device claims may be combined to be implemented or performed in a device. Furthermore, the technical features in the method claims and the device claims may be combined to be implemented or performed in a method. Other implementations are within the scope of the appended claims.

Claims

1. A method performed by a wireless device in a wireless communication system, the method comprising the following steps: Detecting faults in operations with the network; as well as Sending a message including (i) information about the fault and (ii) information about predicted information related to the fault, wherein the predicted information is derived by an artificial intelligence (AI) and / or machine learning (ML) model of the wireless device.

2. The method according to claim 1, in, The information regarding the prediction information related to the fault includes information regarding the operation in which the fault was detected.

3. The method according to claim 1, in, The message also includes information notifying the wireless device that the prediction information is derived based on the AI and / or ML model.

4. The method according to claim 1, in, The message also includes information about the AI and / or ML model used to derive the prediction information.

5. The method according to claim 1, wherein The method further comprises the following steps: A predicted failure and a predicted time point at which the predicted failure will occur are derived.

6. The method according to claim 5, in, The information about the prediction information includes information about the predicted failure and information about the predicted time point.

7. The method according to claim 5, in, The information about the prediction information includes information about a time difference between the prediction time point and a time point at which the fault occurs.

8. The method according to claim 1, in, The message also includes information about the point in time at which the operation is performed.

9. The method according to claim 1, in, The operations include predictive operations derived from the AI and / or ML models.

10. The method according to claim 1, wherein The method further comprises the following steps: receiving, from the network, an AI and / or ML model configuration including information about the AI and / or ML model; and deriving predicted measurement results based on the AI and / or ML models, The operation is triggered by the predicted measurement result.

11. The method according to claim 10, in, The message also includes information notifying that the operation is triggered by the predicted measurement result.

12. The method according to claim 10, in, The prediction information includes information about the predicted measurement result.

13. The method according to claim 1, in, The operations include conditional operation, conditional handover CHO, conditional PScell change CPC, cell and / or cell group CG addition, cell and / or CG activation, cell and / or CG deactivation and / or cell and / or CG release.

14. The method according to claim 1, in, The failures in the operation include switching failure, radio link failure, beam failure, random access failure, configuration failure, connection failure, recovery failure, conditional reconfiguration failure, cell and / or CG addition failure, cell and / or CG activation failure and / or cell and / or CG deactivation failure.

15. The method according to claim 1, in, The wireless device communicates with at least one of a user device other than the wireless device, a network, or an autonomous vehicle.

16. A wireless device in a wireless communication system, the wireless device comprising: transceiver; Memory; as well as at least one processor operatively coupled to the transceiver and the memory and adapted to: Detecting faults in operations with the network; as well as Sending a message including (i) information about the fault and (ii) information about prediction information related to the fault, wherein the prediction information is derived by an artificial intelligence (AI) and / or machine learning (ML) model of the wireless device.

17. The wireless device according to claim 16, in, The information about the prediction information related to the fault includes information about the operation in which the fault was detected.

18. The wireless device according to claim 16, in, The message also includes information notifying the wireless device that the prediction information is derived based on the AI and / or ML model.

19. The wireless device according to claim 16, in, The message also includes information about the AI and / or ML model used to derive the prediction information.

20. The wireless device of claim 16, wherein The at least one processor is further adapted to: A predicted failure and a predicted time point at which the predicted failure will occur are derived.

21. The wireless device according to claim 20, in, The information about the prediction information includes information about the predicted failure and information about the predicted time point.

22. The wireless device according to claim 20, in, The information about the prediction information includes information about a time difference between the prediction time point and a time point at which the fault occurs.

23. The wireless device according to claim 16, in, The message also includes information about the point in time at which the operation is performed.

24. The wireless device according to claim 16, in, The operations include predictive operations derived from the AI and / or ML models.

25. The wireless device of claim 16, wherein The at least one processor is further adapted to: receiving, from the network, an AI and / or ML model configuration including information about the AI and / or ML model; and deriving predicted measurement results based on the AI and / or ML models, The operation is triggered by the predicted measurement result.

26. The wireless device according to claim 25, in, The message also includes information notifying that the operation is triggered by the predicted measurement result.

27. The wireless device according to claim 25, in, The prediction information includes information about the predicted measurement result.

28. The wireless device according to claim 16, in, The operations include conditional operation, conditional handover CHO, conditional PScell change CPC, cell and / or cell group CG addition, cell and / or CG activation, cell and / or CG deactivation and / or cell and / or CG release.

29. The wireless device according to claim 16, in, The failures in the operation include switching failure, radio link failure, beam failure, random access failure, configuration failure, connection failure, recovery failure, conditional reconfiguration failure, cell and / or CG addition failure, cell and / or CG activation failure and / or cell and / or CG deactivation failure.

30. The wireless device of claim 16, in, The wireless device communicates with at least one of a user device other than the wireless device, a network, or an autonomous vehicle.

31. A processor for a wireless device in a wireless communication system, wherein: The processor is configured to control the wireless device to perform operations including: Detecting faults in the operation of the network; and sending a message including (i) information about the fault and (ii) information about prognostic information related to the fault, The prediction information is derived through an artificial intelligence (AI) and / or machine learning (ML) model of the wireless device.

32. A non-transitory computer-readable medium storing a plurality of instructions that, when executed by a processor of a wireless device, cause the wireless device to perform operations comprising: Detecting faults in operations with the network; as well as sending a message including (i) information about the fault and (ii) information about prognostic information related to the fault, The prediction information is derived by an artificial intelligence (AI) and / or machine learning (ML) model of the wireless device.

33. A method performed by a base station in a wireless communication system, the method comprising the steps of: Providing configuration for artificial intelligence (AI) and / or machine learning (ML) models to wireless devices; as well as A message is received from the wireless device including (i) information regarding a fault detected in operation of the wireless device and (ii) information regarding prognostic information related to the fault.

34. A base station in a wireless communication system, the base station comprising: transceiver; Memory; as well as a processor operatively coupled to the transceiver and the memory and adapted to: Providing configuration for artificial intelligence (AI) and / or machine learning (ML) models to wireless devices; and A message is received from the wireless device including (i) information regarding a fault detected in operation of the wireless device and (ii) information regarding prognostic information related to the fault.

Images