WO2025198238A1

WO2025198238A1 - Method and apparatus for monitoring inference function

Info

Publication number: WO2025198238A1
Application number: PCT/KR2025/003166
Authority: WO
Inventors: Myoungsoo Kim; Sunghoon Jung
Original assignee: LG Electronics Inc
Current assignee: LG Electronics Inc
Priority date: 2024-03-20
Filing date: 2025-03-11
Publication date: 2025-09-25
Anticipated expiration: 2026-09-20

Abstract

A method and apparatus for monitoring inference function is provided. The wireless device receives, from a network, a configuration related to at least one inference function, wherein the configuration includes information related to at least one validity condition. The wireless device activates a first inference function. The wireless device determines that monitoring results related to the first inference function does not satisfy the at least one validity condition. The wireless device transmits, the wireless device to the network, a monitoring report related to a second inference function which is not activated.

Description

METHOD AND APPARATUS FOR MONITORING INFERENCE FUNCTION

The present disclosure relates to a method and apparatus for monitoring inference function.

3rd generation partnership project (3GPP) long-term evolution (LTE) is a technology for enabling high-speed packet communications. Many schemes have been proposed for the LTE objective including those that aim to reduce user and provider costs, improve service quality, and expand and improve coverage and system capacity. The 3GPP LTE requires reduced cost per bit, increased service availability, flexible use of a frequency band, a simple structure, an open interface, and adequate power consumption of a terminal as an upper-level requirement.

Work has started in international telecommunication union (ITU) and 3GPP to develop requirements and specifications for new radio (NR) systems. 3GPP has to identify and develop the technology components needed for successfully standardizing the new RAT timely satisfying both the urgent market needs, and the more long-term requirements set forth by the ITU radio communication sector (ITU-R) international mobile telecommunications (IMT)-2020 process. Further, the NR should be able to use any spectrum band ranging at least up to 100 GHz that may be made available for wireless communications even in a more distant future.

The NR targets a single technical framework addressing all usage scenarios, requirements and deployment scenarios including enhanced mobile broadband (eMBB), massive machine-type-communications (mMTC), ultra-reliable and low latency communications (URLLC), etc. The NR shall be inherently forward compatible.

UE can evaluate and monitor the applicability and expected performance of an inactive model/functionality for purpose of activation/selection/switching of UE-side models/UE-part of two-sided models/functionalities. UE can evaluate applicability and performance based on additional conditions, input/output data distribution, monitoring conditions, and the inference accuracy.

However, if there is no restriction on the evaluation/monitoring of inactivate model/functionality, UE may consume more power to evaluate model performance. Additionally, UE may report the at least one inactive functionality/model's monitoring results even when UE does not need model switching, thereby leading to unnecessary signalling overhead.

Additionally, some models may need to be updated to reflect the current environment based on the latest data collection and training. Therefore, NW can configure a valid time for each model to ensure that the model is up to date. When expiring valid time, model management, e.g., (de)activation, switch, fallback, etc, can be necessary for accurate model operation.

Accordingly, when a model is no longer be valid, the NW may require knowing the applicability or performance of inactive models to activate a suitable model. If NW know the inactive model information after the expiration of the current active model, it may take time for the new model to resume operation due to the applicability/performance exchange of a specific model. This means that model operation efficiency may decrease.

Thus, studies for monitoring inference function are required.

In an aspect, a method comprises: receiving, by a wireless device from a network, a configuration related to at least one inference function, wherein the configuration includes information related to at least one validity condition; activating, by the wireless device, a first inference function; determining, by the wireless device, that monitoring results related to the first inference function does not satisfy the at least one validity condition; and transmitting, by the wireless device to the network, a monitoring report related to a second inference function which is not activated.

In another aspect, an apparatus for implementing the above method is provided.

The present disclosure can have various advantageous effects.

According to some embodiments of the present disclosure, a wireless device could efficiently monitor inference function.

For example, by monitoring the inactive model based on active model validity, it may reduce power consumption caused by model evaluation and report. The NW can use the evaluation information for inactive model for management only when it is necessary.

In other words, since the wireless device transmits a report about the inactive inference function only when the active inference function is invalid, the wireless device could save resources.

According to some embodiments of the present disclosure, the wireless communication system could provide an efficient solution for monitoring inference function.

Advantageous effects which can be obtained through specific embodiments of the present disclosure are not limited to the advantageous effects listed above. For example, there may be a variety of technical effects that a person having ordinary skill in the related art can understand and/or derive from the present disclosure. Accordingly, the specific effects of the present disclosure are not limited to those explicitly described herein, but may include various effects that may be understood or derived from the technical features of the present disclosure.

FIG. 1 shows an example of a communication system to which implementations of the present disclosure is applied.

FIG. 2 shows an example of wireless devices to which implementations of the present disclosure is applied.

FIG. 3 shows an example of a wireless device to which implementations of the present disclosure is applied.

FIG. 4 shows another example of wireless devices to which implementations of the present disclosure is applied.

FIG. 5 shows an example of UE to which implementations of the present disclosure is applied.

FIGS. 6 and 7 show an example of protocol stacks in a 3GPP based wireless communication system to which implementations of the present disclosure is applied.

FIG. 8 shows a frame structure in a 3GPP based wireless communication system to which implementations of the present disclosure is applied.

FIG. 9 shows a data flow example in the 3GPP NR system to which implementations of the present disclosure is applied.

FIG. 10 shows an example of a Functional Framework for RAN Intelligence.

FIG. 11 shows an example of an AI/ML Model Training in OAM and AI/ML Model Inference in NG-RAN node.

FIG. 12 shows an example of Model Training and Model Inference both located in RAN node.

FIGS. 13 and 14 show an example of an architecture of neuron and neural network.

FIG. 15 shows an example of an AI/ML inference.

FIG. 16 shows an example of an MLP DNN model.

FIG. 17 shows an example of a CNN model.

FIG. 18 shows an example of an RNN model.

FIG. 19 shows an example of Reinforcement learning.

FIG. 20 shows an example of network decision, network-initiated AI/ML management.

FIG. 21 shows an example of network decision, UE-initiated AI/ML management.

FIG. 22 shows an example of UE decision, event-triggered as configured by the network.

FIG. 23 shows an example of UE autonomous, decision reported to the network.

FIG. 24 shows an example of measurement reporting.

FIG. 25 shows an example of location measurement indication.

FIG. 26 shows an example of a method for monitoring inference function.

FIG. 27 shows an example of a method for inactive model monitoring based on active model validity.

FIG. 28 shows an example of a method for inactive model evaluation based on active model validity related information.

FIG. 29 shows an example of a method for validity of an active model based on valid information.

FIG. 30 shows an example of a method for inactive model monitoring based on active model validity.

The following techniques, apparatuses, and systems may be applied to a variety of wireless multiple access systems. Examples of the multiple access systems include a code division multiple access (CDMA) system, a frequency division multiple access (FDMA) system, a time division multiple access (TDMA) system, an orthogonal frequency division multiple access (OFDMA) system, a single carrier frequency division multiple access (SC-FDMA) system, and a multicarrier frequency division multiple access (MC-FDMA) system. CDMA may be embodied through radio technology such as universal terrestrial radio access (UTRA) or CDMA2000. TDMA may be embodied through radio technology such as global system for mobile communications (GSM), general packet radio service (GPRS), or enhanced data rates for GSM evolution (EDGE). OFDMA may be embodied through radio technology 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 a part of a universal mobile telecommunications system (UMTS). 3rd generation partnership project (3GPP) long term evolution (LTE) is a part of evolved UMTS (E-UMTS) using E-UTRA. 3GPP LTE employs OFDMA in DL and SC-FDMA in UL. LTE-advanced (LTE-A) is an evolved version of 3GPP LTE.

For convenience of description, implementations of the present disclosure are mainly described in regards 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 system are applicable to other mobile communication systems.

For terms and technologies which are not specifically described among the terms of and technologies employed in the present disclosure, the wireless communication standard documents published before the present disclosure may be referenced.

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

In the present disclosure, slash (/) or comma (,) may mean "and/or". For example, "A/B" may mean "A and/or B". Accordingly, "A/B" may mean "only A", "only B", or "both A and B". For example, "A, B, C" may mean "A, B or C".

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 same as "at least one of A and B".

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".

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

Technical features that are separately described in one drawing in the present disclosure may be implemented separately or simultaneously.

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

Hereinafter, the present disclosure will be described in more detail with reference to drawings. 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 unless otherwise indicated.

The 5G usage scenarios shown in FIG. 1 are only exemplary, and the technical features of the present disclosure can be applied to other 5G usage scenarios which are not shown in FIG. 1.

Three main requirement categories for 5G include (1) a category of enhanced mobile broadband (eMBB), (2) a category of massive machine type communication (mMTC), and (3) a category of ultra-reliable and low latency communications (URLLC).

Partial use cases may require a plurality of categories for optimization and other use cases may focus only upon one key performance indicator (KPI). 5G supports such various use cases using a flexible and reliable method.

eMBB far surpasses basic mobile Internet access and covers abundant bidirectional work and media and entertainment applications in cloud and augmented reality. Data is one of 5G core motive forces and, in a 5G era, a dedicated voice service may not be provided for the first time. In 5G, it is expected that voice will be simply processed as an application program using data connection provided by a communication system. Main causes for increased traffic volume are due to an increase in the size of content and an increase in the number of applications requiring high data transmission rate. A streaming service (of audio and video), conversational video, and mobile Internet access will be more widely used as more devices are connected to the Internet. These many application programs require connectivity of an always turned-on state in order to push real-time information and alarm for users. Cloud storage and applications are rapidly increasing in a mobile communication platform and may be applied to both work and entertainment. The cloud storage is a special use case which accelerates growth of uplink data transmission rate. 5G is also used for remote work of cloud. When a tactile interface is used, 5G demands much lower end-to-end latency to maintain user good experience. Entertainment, for example, cloud gaming and video streaming, is another core element which increases demand for mobile broadband capability. Entertainment is essential for a smartphone and a tablet in any place including high mobility environments such as a train, a vehicle, and an airplane. Other use cases are augmented reality for entertainment and information search. In this case, the augmented reality requires very low latency and instantaneous data volume.

In addition, one of the most expected 5G use cases relates a function capable of smoothly connecting embedded sensors in all fields, i.e., mMTC. It is expected that the number of potential Internet-of-things (IoT) devices will reach 204 hundred million up to the year of 2020. An industrial IoT is one of categories of performing a main role enabling a smart city, asset tracking, smart utility, agriculture, and security infrastructure through 5G.

URLLC includes a new service that will change industry through remote control of main infrastructure and an ultra-reliable/available low-latency link such as a self-driving vehicle. A level of reliability and latency is essential to control a smart grid, automatize industry, achieve robotics, and control and adjust a drone.

5G is a means of providing streaming evaluated as a few hundred megabits per second to gigabits per second and may complement fiber-to-the-home (FTTH) and cable-based broadband (or DOCSIS). Such fast speed is needed to deliver TV in 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. A specific application program may require a special network configuration. For example, for VR games, gaming companies need to incorporate a core server into an edge network server of a network operator in order to minimize latency.

Automotive is expected to be a new important motivated force in 5G together with many use cases for mobile communication for vehicles. For example, entertainment for passengers requires high simultaneous capacity and mobile broadband with high mobility. This is because future users continue to expect connection of high quality regardless of their locations and speeds. Another use case of an automotive field is an AR dashboard. The AR dashboard causes a driver to identify an object in the dark in addition to an object seen from a front window and displays a distance from the object and a movement of the object by overlapping information talking to the driver. In the future, a wireless module enables communication between vehicles, information exchange between a vehicle and supporting infrastructure, and information exchange between a vehicle and other connected devices (e.g., devices accompanied by a pedestrian). A safety system guides alternative courses of a behavior so that a driver may drive more safely drive, thereby lowering the danger of an accident. The next stage will be a remotely controlled or self-driven vehicle. This requires very high reliability and very fast communication between different self-driven vehicles and between a vehicle and infrastructure. In the future, a self-driven vehicle will perform all driving activities and a driver will focus only upon abnormal traffic that the vehicle cannot identify. Technical requirements of a self-driven vehicle demand ultra-low latency and ultra-high reliability so that traffic safety is increased to a level that cannot be achieved by human being.

A smart city and a smart home/building mentioned as a smart society will be embedded in a high-density wireless sensor network. A distributed network of an intelligent sensor will identify conditions for costs and energy-efficient maintenance of a city or a home. Similar configurations may be performed for respective households. All of temperature sensors, window and heating controllers, burglar alarms, and home appliances are wirelessly connected. Many of these sensors are typically low in data transmission rate, power, and cost. However, real-time HD video may be demanded by a specific type of device to perform monitoring.

Consumption and distribution of energy including heat or gas is distributed at a higher level so that automated control of the distribution sensor network is demanded. The smart grid collects information and connects the sensors to each other using digital information and communication technology so as to act according to the collected information. Since this information may include behaviors of a supply company and a consumer, the smart grid may improve distribution of fuels such as electricity by a method having efficiency, reliability, economic feasibility, production sustainability, and automation. The smart grid may also be regarded as another sensor network having low latency.

Mission critical application (e.g., e-health) is one of 5G use scenarios. A health part contains many application programs capable of enjoying benefit of mobile communication. A communication system may support remote treatment that provides clinical treatment in a faraway place. Remote treatment may aid in reducing a barrier against distance and improve access to medical services that cannot be continuously available in a faraway rural area. Remote treatment is also used to perform important treatment and save lives in an emergency situation. The wireless sensor network based on mobile communication may provide remote monitoring and sensors for parameters such as heart rate and blood pressure.

Wireless and mobile communication gradually becomes important in the field of an industrial application. Wiring is high in installation and maintenance cost. Therefore, a possibility of replacing a cable with reconstructible wireless links is an attractive opportunity in many industrial fields. However, in order to achieve this replacement, it is necessary for wireless connection to be established with latency, reliability, and capacity similar to those of the cable and management of wireless connection needs to be simplified. Low latency and a very low error probability are new requirements when connection to 5G is needed.

Logistics and freight tracking are important use cases for mobile communication that enables inventory and package tracking anywhere using a location-based information system. The use cases of logistics and freight typically demand low data rate but require location information with a wide range and reliability.

Referring to FIG. 1, the communication system 1 includes wireless devices 100a to 100f, base stations (BSs) 200, and a network 300. Although FIG. 1 illustrates a 5G network as an example of the network of the communication system 1, the implementations of the present disclosure are not limited to the 5G system, and can be applied to the future communication system beyond the 5G system.

The BSs 200 and the network 300 may be implemented as wireless devices and a specific wireless device may operate as a BS/network node with respect to other wireless devices.

The wireless devices 100a to 100f represent devices performing communication using radio access technology (RAT) (e.g., 5G new RAT (NR)) or LTE) and may be referred to as communication/radio/5G devices. The wireless devices 100a to 100f may include, without being limited to, a robot 100a, vehicles 100b-1 and 100b-2, an extended reality (XR) device 100c, a hand-held device 100d, a home appliance 100e, an IoT device 100f, and an artificial intelligence (AI) device/server 400. For example, the vehicles may include a vehicle having a wireless communication function, an autonomous driving vehicle, and a vehicle capable of performing communication between vehicles. The vehicles may include an unmanned aerial vehicle (UAV) (e.g., a drone). The XR device may include an AR/VR/Mixed Reality (MR) device and may be implemented in the form of a head-mounted device (HMD), a head-up display (HUD) mounted in a vehicle, a television, a smartphone, a computer, a wearable device, a home appliance device, a digital signage, a vehicle, a robot, etc. The hand-held device may include a smartphone, a smartpad, a wearable device (e.g., a smartwatch or a smartglasses), and a computer (e.g., a notebook). The home appliance may include a TV, a refrigerator, and a washing machine. The IoT device may include a sensor and a smartmeter.

In the present disclosure, the wireless devices 100a to 100f may be called user equipments (UEs). A UE may include, for example, 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 slate personal computer (PC), a tablet PC, an ultrabook, a vehicle, a vehicle having an autonomous traveling function, a connected car, an 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/environment device, a device related to a 5G service, or a device related to a fourth industrial revolution field.

The UAV may be, for example, an aircraft aviated by a wireless control signal without a human being onboard.

The VR device may include, for example, a device for implementing an object or a background of the virtual world. The AR device may include, for example, a device implemented by connecting an object or a background of the virtual world to an object or a background of the real world. The MR device may include, for example, a device implemented by merging an object or a background of the virtual world into an object or a background of the real world. The hologram device may include, for example, a device for implementing a stereoscopic image of 360 degrees by recording and reproducing stereoscopic information, using an interference phenomenon of light generated when two laser lights called holography meet.

The public safety device may include, for example, an image relay device or an image device that is wearable on the body of a user.

The MTC device and the IoT device may be, for example, devices that do not require direct human intervention or manipulation. For example, the MTC device and the IoT device may include smartmeters, vending machines, thermometers, smartbulbs, door locks, or various sensors.

The medical device may be, for example, a device used for the purpose of diagnosing, treating, relieving, curing, or preventing disease. For example, the medical device may be a device used for the purpose of diagnosing, treating, relieving, or correcting injury or impairment. For example, the medical device may be a device used for the purpose of inspecting, replacing, or modifying a structure or a function. For example, the medical device may be a device used for the purpose of adjusting pregnancy. For example, the 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 procedure.

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

The FinTech device may be, for example, a device capable of providing a financial service such as mobile payment. For example, the FinTech device may include a payment device or a point of sales (POS) system.

The weather/environment device may include, for example, a device for monitoring or predicting a weather/environment.

The wireless devices 100a to 100f may be connected to the network 300 via the BSs 200. An AI technology may be applied to the wireless devices 100a to 100f and the wireless devices 100a to 100f may be connected to the AI server 400 via the network 300. The network 300 may be configured using a 3G network, a 4G (e.g., LTE) network, a 5G (e.g., NR) network, and a beyond-5G network. Although the wireless devices 100a to 100f may communicate with each other through the BSs 200/network 300, the wireless devices 100a to 100f may perform direct communication (e.g., sidelink communication) with each other without passing through the BSs 200/network 300. For example, the vehicles 100b-1 and 100b-2 may perform direct communication (e.g., vehicle-to-vehicle (V2V)/vehicle-to-everything (V2X) communication). The IoT device (e.g., a sensor) may perform direct communication with other IoT devices (e.g., sensors) or other wireless devices 100a to 100f.

Wireless communication/connections 150a, 150b and 150c may be established between the wireless devices 100a to 100f and/or between wireless device 100a to 100f and BS 200 and/or between BSs 200. Herein, the wireless communication/connections may be established through various RATs (e.g., 5G NR) such as uplink/downlink communication 150a, sidelink communication (or device-to-device (D2D) communication) 150b, inter-base station communication 150c (e.g., relay, integrated access and backhaul (IAB)), etc. The wireless devices 100a to 100f and the BSs 200/the wireless devices 100a to 100f may transmit/receive radio signals to/from each other through the wireless communication/connections 150a, 150b and 150c. For example, the wireless communication/connections 150a, 150b and 150c may transmit/receive signals through various physical channels. To this end, at least a part of various configuration information configuring processes, various signal processing processes (e.g., channel encoding/decoding, modulation/demodulation, and resource mapping/de-mapping), and resource allocating processes, for transmitting/receiving radio signals, may be performed based on the various proposals of the present disclosure.

Here, the radio communication technologies implemented in the wireless devices 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-mentioned names. Additionally and/or alternatively, the radio communication technologies implemented in the wireless devices in the present disclosure may communicate based on LTE-M technology. For example, LTE-M technology may be an example of LPWAN technology and be called by various names such as enhanced machine type communication (eMTC). For example, LTE-M technology may be implemented in at least one of the 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-mentioned names. Additionally and/or alternatively, the radio communication technologies implemented in the wireless devices in the present disclosure may include at least one of ZigBee, Bluetooth, and/or LPWAN which take into account low-power communication, and may not be limited to the above-mentioned names. For example, ZigBee technology may generate personal area networks (PANs) associated with small/low-power digital communication based on various specifications such as IEEE 802.15.4 and may be called various names.

Referring to FIG. 2, a first wireless device 100 and a second wireless device 200 may transmit/receive radio signals to/from an external device through a variety of RATs (e.g., LTE and NR). In FIG. 2, {the first wireless device 100 and the second wireless device 200} may correspond to at least one of {the wireless device 100a to 100f and the BS 200}, {the wireless device 100a to 100f and the wireless device 100a to 100f} and/or {the BS 200 and the BS 200} of FIG. 1.

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

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

Hereinafter, hardware elements of the wireless devices 100 and 200 will be described more specifically. One or more protocol layers may be implemented by, without being limited to, one or more processors 102 and 202. For example, the one or more processors 102 and 202 may implement one or more layers (e.g., functional layers such as physical (PHY) layer, media access control (MAC) layer, radio link control (RLC) layer, packet data convergence protocol (PDCP) layer, radio resource control (RRC) layer, and service data adaptation protocol (SDAP) layer). The one or more processors 102 and 202 may generate one or more protocol data units (PDUs) and/or one or more service data unit (SDUs) according to the descriptions, functions, procedures, suggestions, methods and/or operational flowcharts disclosed in the present disclosure. The one or more processors 102 and 202 may generate messages, control information, data, or information according to the descriptions, functions, procedures, suggestions, methods and/or operational flowcharts disclosed in the present disclosure. The one or more processors 102 and 202 may generate signals (e.g., baseband signals) including PDUs, SDUs, messages, control information, data, or information according to the descriptions, functions, procedures, suggestions, methods and/or operational flowcharts disclosed in the present disclosure and provide the generated signals to the one or more transceivers 106 and 206. The one or more processors 102 and 202 may receive the signals (e.g., baseband signals) from the one or more transceivers 106 and 206 and acquire the PDUs, SDUs, messages, control information, data, or information according to the descriptions, functions, procedures, suggestions, methods and/or operational flowcharts disclosed in the present disclosure.

The one or more processors 102 and 202 may be referred to as controllers, microcontrollers, microprocessors, or microcomputers. The 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 the one or more processors 102 and 202. descriptions, functions, procedures, suggestions, methods and/or operational flowcharts disclosed in the present disclosure may be implemented using firmware or software and the firmware or software may be configured to include the modules, procedures, or functions. Firmware or software configured to perform the descriptions, functions, procedures, suggestions, methods and/or operational flowcharts disclosed in the present disclosure may be included in the one or more processors 102 and 202 or stored in the one or more memories 104 and 204 so as to be driven by the one or more processors 102 and 202. The descriptions, functions, procedures, suggestions, methods and/or operational flowcharts disclosed in the present disclosure may be implemented using firmware or software in the form of code, commands, and/or a set of commands.

The one or more memories 104 and 204 may be connected to the one or more processors 102 and 202 and store various types of data, signals, messages, information, programs, code, instructions, and/or commands. The one or more memories 104 and 204 may be configured by read-only memories (ROMs), random access memories (RAMs), electrically erasable programmable read-only memories (EPROMs), flash memories, hard drives, registers, cash memories, computer-readable storage media, and/or combinations thereof. The one or more memories 104 and 204 may be located at the interior and/or exterior of the one or more processors 102 and 202. The one or more memories 104 and 204 may be connected to the one or more processors 102 and 202 through various technologies such as wired or wireless connection.

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

The one or more transceivers 106 and 206 may be connected to the 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, procedures, 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 a plurality of physical antennas or a plurality of logical antennas (e.g., antenna ports).

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

In the implementations of the present disclosure, a UE may operate as a transmitting device in uplink (UL) and as a receiving device in downlink (DL). In the implementations of the present disclosure, a BS may operate as a receiving device in UL and as a transmitting device in DL. Hereinafter, for convenience of description, it is mainly assumed that the first wireless device 100 acts as the UE, and the second wireless device 200 acts as the BS. For example, the processor(s) 102 connected to, mounted on or launched in the first wireless device 100 may be configured to perform the UE behavior according to an implementation of the present disclosure or control the transceiver(s) 106 to perform the UE behavior according to an implementation of the present disclosure. The processor(s) 202 connected to, mounted on or launched in the second wireless device 200 may be configured to perform the BS behavior according to an implementation of the present disclosure or control the transceiver(s) 206 to perform the BS behavior according to an implementation of the present disclosure.

In the present disclosure, a BS is also referred to as a node B (NB), an eNode B (eNB), or a gNB.

The wireless device may be implemented in various forms according to a use-case/service (refer to FIG. 1).

Referring to FIG. 3, wireless devices 100 and 200 may correspond to the wireless devices 100 and 200 of FIG. 2 and may be configured by various elements, components, units/portions, 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 additional components 140. The communication unit 110 may include a communication circuit 112 and transceiver(s) 114. For example, the communication circuit 112 may include the one or more processors 102 and 202 of FIG. 2 and/or the one or more memories 104 and 204 of FIG. 2. For example, the transceiver(s) 114 may include the one or more transceivers 106 and 206 of FIG. 2 and/or the one or more antennas 108 and 208 of FIG. 2. The control unit 120 is electrically connected to the communication unit 110, the memory 130, and the additional components 140 and controls overall operation of each of the wireless devices 100 and 200. For example, the control unit 120 may control an electric/mechanical operation of each of the wireless devices 100 and 200 based on programs/code/commands/information stored in the memory unit 130. The control unit 120 may transmit the information stored in the memory unit 130 to the exterior (e.g., other communication devices) via the communication unit 110 through a wireless/wired interface or store, in the memory unit 130, information received through the wireless/wired interface from the exterior (e.g., other communication devices) via the communication unit 110.

The additional components 140 may be variously configured according to types of the wireless devices 100 and 200. For example, the additional components 140 may include at least one of a power unit/battery, input/output (I/O) unit (e.g., audio I/O port, video I/O port), a driving unit, and a computing unit. The wireless devices 100 and 200 may be implemented in the form of, without being limited to, the robot (100a of FIG. 1), the vehicles (100b-1 and 100b-2 of FIG. 1), the XR device (100c of FIG. 1), the hand-held device (100d of FIG. 1), the home appliance (100e of FIG. 1), the IoT device (100f of FIG. 1), a digital broadcast terminal, a hologram device, a public safety device, an MTC device, a medicine device, a FinTech device (or a finance device), a security device, a climate/environment device, the AI server/device (400 of FIG. 1), the BSs (200 of FIG. 1), a network node, etc. The wireless devices 100 and 200 may be used in a mobile or fixed place according to a use-example/service.

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

Referring to FIG. 4, wireless devices 100 and 200 may correspond to the wireless devices 100 and 200 of FIG. 2 and may be configured by various elements, components, units/portions, and/or modules.

The first wireless device 100 may include at least one transceiver, such as a transceiver 106, and at least one processing chip, such as a processing chip 101. The processing chip 101 may include at least one processor, such a processor 102, and at least one memory, such as a memory 104. The memory 104 may be operably connectable to the processor 102. The memory 104 may store various types of information and/or instructions. The memory 104 may store a software code 105 which implements instructions that, when executed by the processor 102, perform the descriptions, functions, procedures, suggestions, methods and/or operational flowcharts disclosed in the present disclosure. For example, the software code 105 may implement instructions that, when executed by the processor 102, perform the descriptions, functions, procedures, suggestions, methods and/or operational flowcharts disclosed in the present disclosure. For example, the software code 105 may control the processor 102 to perform one or more protocols. For example, the software code 105 may control the processor 102 may perform one or more layers of the radio interface protocol.

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

Referring to FIG. 5, a UE 100 may correspond to the first wireless device 100 of FIG. 2 and/or the first wireless device 100 of FIG. 4.

A UE 100 includes a processor 102, a 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 identification module (SIM) card 118, a speaker 120, and a microphone 122.

The processor 102 may be configured to implement the descriptions, functions, procedures, 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, procedures, suggestions, methods and/or operational flowcharts disclosed in the present disclosure. Layers of the radio interface protocol may be implemented in the processor 102. The processor 102 may include ASIC, other chipset, logic circuit and/or data processing device. 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), a modem (modulator and demodulator). An example of the processor 102 may be found in SNAPDRAGON^TM series of processors made by Qualcomm^®, EXYNOS^TM series of processors made by Samsung^®, A series of processors made by Apple^®, HELIO^TM series of processors made by MediaTek^®, ATOM^TM series of processors made by Intel^® or a corresponding next generation processor.

The memory 104 is operatively coupled with the processor 102 and stores a variety of information to operate the processor 102. The memory 104 may include ROM, RAM, flash memory, memory card, storage medium and/or other storage device. When the embodiments are implemented in software, the techniques described herein can be implemented with modules (e.g., procedures, functions, etc.) that perform the descriptions, functions, procedures, suggestions, methods and/or operational flowcharts disclosed in the present disclosure. The modules can be stored in the memory 104 and executed by the processor 102. The memory 104 can be implemented within the processor 102 or external to the processor 102 in which case those can be communicatively coupled to the processor 102 via various means as is known in the art.

The transceiver 106 is operatively coupled with the processor 102, and transmits and/or receives a radio signal. The transceiver 106 includes a transmitter and a receiver. The transceiver 106 may include baseband circuitry to process radio frequency signals. The transceiver 106 controls the one or more antennas 108 to transmit and/or receive a radio signal.

The power management module 110 manages power for the processor 102 and/or the transceiver 106. The battery 112 supplies power to the power management module 110.

The display 114 outputs results processed by the processor 102. The keypad 116 receives inputs to be used by the processor 102. The keypad 16 may be shown on the display 114.

The SIM card 118 is an　integrated circuit　that is intended to　securely store the　international mobile subscriber identity　(IMSI) number and its related　key, which are used to identify and authenticate subscribers on　mobile telephony　devices (such as　mobile phones　and　computers). It is also possible to store contact information on many SIM cards.

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

In particular, FIG. 6 illustrates an example of a radio interface user plane protocol stack between a UE and a BS and FIG. 7 illustrates an example of a radio interface control plane protocol stack between a UE and a BS. The control plane refers to a path through which control messages used to manage call by a UE and a network are transported. The user plane refers to a path through which data generated in an application layer, for example, voice data or Internet packet data are transported. Referring to FIG. 6, the user plane protocol stack may be divided into Layer 1 (i.e., a PHY layer) and Layer 2. Referring to FIG. 7, the control plane protocol stack may be divided into Layer 1 (i.e., a PHY layer), Layer 2, Layer 3 (e.g., an RRC layer), and a non-access stratum (NAS) layer. Layer 1, Layer 2 and Layer 3 are referred to as an access stratum (AS).

In the 3GPP LTE system, the Layer 2 is split into the following sublayers: MAC, RLC, and PDCP. In the 3GPP NR system, the Layer 2 is split into the following sublayers: MAC, RLC, PDCP and SDAP. The PHY layer offers to the MAC sublayer transport channels, the MAC sublayer offers to the RLC sublayer logical channels, the RLC sublayer offers to the PDCP sublayer RLC channels, the PDCP sublayer offers to the SDAP sublayer radio bearers. The SDAP sublayer offers to 5G core network quality of service (QoS) flows.

In the 3GPP NR system, the main services and functions of the MAC sublayer include: mapping between logical channels and transport channels; multiplexing/de-multiplexing of MAC SDUs belonging to one or different logical channels into/from transport blocks (TB) 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 case of carrier aggregation (CA)); priority handling between UEs by means of dynamic scheduling; priority handling between logical channels of one UE by means of logical channel prioritization; padding. A single MAC entity may support multiple numerologies, transmission timings and cells. Mapping restrictions in logical channel prioritization control which numerology(ies), cell(s), and transmission timing(s) a logical channel can use.

Different kinds of data transfer services are offered by MAC. To accommodate different kinds of data transfer services, multiple types of logical channels are defined, i.e., each supporting transfer of a particular type of information. Each logical channel type is defined by what type of information is transferred. Logical channels are classified into two groups: control channels and traffic channels. Control channels are used for the transfer of control plane information only, and traffic channels are used for the transfer of user plane information only. Broadcast control channel (BCCH) is a downlink logical channel for broadcasting system control information, paging control channel (PCCH) is a downlink logical channel that transfers paging information, system information change notifications and indications of ongoing public warning service (PWS) broadcasts, common control channel (CCCH) is a logical channel for transmitting control information between UEs and network and used for UEs having no RRC connection with the network, and dedicated control channel (DCCH) is a point-to-point bi-directional logical channel that transmits dedicated control information between a UE and the network and used by UEs having an RRC connection. Dedicated traffic channel (DTCH) is a point-to-point logical channel, dedicated to one UE, for the transfer of user information. A DTCH can exist in both uplink and downlink. In downlink, the following connections between logical channels and transport channels exist: BCCH can be mapped to broadcast channel (BCH); BCCH can be mapped to downlink shared channel (DL-SCH); PCCH can be mapped to 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 uplink, the following connections between logical channels and transport channels exist: CCCH can be mapped to uplink shared channel (UL-SCH); DCCH can be mapped to UL-SCH; and DTCH can be mapped to UL-SCH.

The RLC sublayer supports three transmission modes: transparent mode (TM), unacknowledged mode (UM), and acknowledged node (AM). The RLC configuration is per logical channel with no dependency on numerologies and/or transmission durations. In the 3GPP NR system, the main services and functions of the RLC sublayer depend on the transmission mode and include: transfer of upper layer PDUs; sequence numbering independent of the one 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 SDU (AM and UM); duplicate detection (AM only); RLC SDU discard (AM and UM); RLC re-establishment; protocol error detection (AM only).

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); transfer of user data; reordering and duplicate detection; in-order delivery; PDCP PDU routing (in 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; duplication of PDCP PDUs 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; transfer of control plane data; reordering and duplicate detection; in-order delivery; duplication of PDCP PDUs and duplicate discard indication to lower layers.

In the 3GPP NR system, the main services and functions of SDAP include: mapping between a QoS flow and a data radio bearer; marking QoS flow ID (QFI) in both DL and UL packets. A single protocol entity of SDAP is configured for each individual PDU session.

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 an RRC connection between the 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, inter-RAT mobility); QoS management functions; UE measurement reporting and control of the reporting; detection of and recovery from radio link failure; NAS message transfer to/from NAS from/to UE.

The frame structure shown in FIG. 8 is purely exemplary and the number of subframes, the number of slots, and/or the number of symbols in a frame may be variously changed. In the 3GPP based wireless communication system, OFDM numerologies (e.g., subcarrier spacing (SCS), transmission time interval (TTI) duration) may be differently configured between a plurality of cells aggregated for one UE. For example, if a UE is configured with different SCSs for cells aggregated for the cell, an (absolute time) duration of a time resource (e.g., a subframe, a slot, or a TTI) including the same number of symbols may be different among the aggregated cells. Herein, symbols may include OFDM symbols (or CP-OFDM symbols), SC-FDMA symbols (or discrete Fourier transform-spread-OFDM (DFT-s-OFDM) symbols).

Referring to FIG. 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 of the half-frames has 5ms duration. Each half-frame consists of 5 subframes, where the duration T_sf per subframe is 1ms. Each subframe is divided into slots and the number of slots in a subframe depends on a subcarrier spacing. Each slot includes 14 or 12 OFDM symbols based on a cyclic prefix (CP). In a normal CP, each slot includes 14 OFDM symbols and, in an extended CP, each slot includes 12 OFDM symbols. The numerology is based on exponentially scalable subcarrier spacing △f = 2^u*15 kHz.

Table 1 shows the number of OFDM symbols per slot N^slot _symb, the number of slots per frame N^frame,u _slot, and the number of slots per subframe N^subframe,u _slot for the normal CP, according to the subcarrier spacing △f = 2^u*15 kHz.

u	N ^slot _symb	N ^frame,u _slot	N ^subframe,u _slot
0	14	10	1
1	14	20	2
2	14	40	4
3	14	80	8
4	14	160	16

Table 2 shows the number of OFDM symbols per slot N^slot _symb, the number of slots per frame N^frame,u _slot, and the number of slots per subframe N^subframe,u _slot for the extended CP, according to the subcarrier spacing △f = 2^u*15 kHz.

u	N ^slot _symb	N ^frame,u _slot	N ^subframe,u _slot
2	12	40	4

A slot includes plural symbols (e.g., 14 or 12 symbols) in the time domain. For each numerology (e.g., subcarrier spacing) and carrier, a resource grid of N ^size,u _grid,x*N ^RB _sc subcarriers and N ^subframe,u _symb OFDM symbols is defined, starting at common resource block (CRB) N ^start,u _grid indicated by higher-layer signaling (e.g., RRC signaling), where N ^size,u _grid,x is the number of resource blocks (RBs) in the resource grid and the subscript x is DL for downlink and UL for uplink. N ^RB _sc is the number of subcarriers per RB. In the 3GPP based wireless communication system, N ^RB _sc is 12 generally. There is one resource grid for a given antenna port p, subcarrier spacing configuration u, and transmission direction (DL or UL). The carrier bandwidth N ^size,u _grid for subcarrier spacing configuration u is given by the higher-layer parameter (e.g., RRC parameter). Each element in the resource grid for the antenna port p and the subcarrier spacing configuration u is referred to as a resource element (RE) and one complex symbol may 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 a symbol location relative to a reference point in the time domain. In the 3GPP based wireless communication system, an RB is defined by 12 consecutive subcarriers in the frequency domain.

In the 3GPP NR system, RBs are classified into CRBs and physical resource blocks (PRBs). CRBs are numbered from 0 and upwards 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 serves as a common reference point for resource block grids. In the 3GPP NR system, PRBs are defined within a bandwidth part (BWP) and numbered from 0 to N ^size _BWP,i-1, where i is the number of the bandwidth part. The relation between the physical resource block n_PRB in the bandwidth part i and the common resource block n_CRB is as follows: n_PRB = n_CRB + N ^size _BWP,i, where N ^size _BWP,i is the common resource block where bandwidth part starts relative to CRB 0. The BWP includes a plurality of consecutive RBs. A carrier may include a maximum of N (e.g., 5) BWPs. A UE may be configured with one or more BWPs on a given component carrier. Only one BWP among BWPs configured to the UE can active at a time. The active BWP defines the UE's operating bandwidth within the cell's operating bandwidth.

The NR frequency band may be defined as two types of frequency range, i.e., FR1 and FR2. The numerical value of the frequency range may be changed. For example, the frequency ranges of the two types (FR1 and FR2) may be as shown in Table 3 below. For ease of explanation, in the frequency ranges used in the NR system, FR1 may mean "sub 6 GHz range", FR2 may mean "above 6 GHz range," and may be referred to as millimeter wave (mmW).

Frequency Range designation	Corresponding frequency range	Subcarrier Spacing
FR1	450MHz - 6000MHz	15, 30, 60kHz
FR2	24250MHz - 52600MHz	60, 120, 240kHz

As mentioned above, the numerical value of the frequency range of the NR system may be changed. For example, FR1 may include a frequency band of 410MHz to 7125MHz as shown in Table 4 below. That is, FR1 may include a frequency band of 6GHz (or 5850, 5900, 5925 MHz, etc.) or more. For example, a frequency band of 6 GHz (or 5850, 5900, 5925 MHz, etc.) or more included in FR1 may include an unlicensed band. Unlicensed bands may be used for a variety of purposes, for example for communication for vehicles (e.g., autonomous driving).

Frequency Range designation	Corresponding frequency range	Subcarrier Spacing
FR1	410MHz - 7125MHz	15, 30, 60kHz
FR2	24250MHz - 52600MHz	60, 120, 240kHz

In the present disclosure, the term "cell" may refer to a geographic area to which one or more nodes provide a communication system, or refer to radio resources. A "cell" as a geographic area may be understood as coverage within which a node can provide service using a carrier and a "cell" as radio resources (e.g., time-frequency resources) is associated with bandwidth which is a frequency range configured by the carrier. The "cell" associated with the radio resources is defined by a combination of downlink resources and uplink resources, for example, a combination of a DL component carrier (CC) and a UL CC. The cell may be configured by downlink resources only, or may be configured by downlink resources and uplink resources. Since DL coverage, which is a range within which the node is capable of transmitting a valid signal, and UL coverage, which is a range within which the node is capable of receiving the valid signal from the UE, depends upon a carrier carrying the signal, the coverage of the node may be associated with coverage of the "cell" of radio resources used by the node. Accordingly, the term "cell" may be used to represent service coverage of the node sometimes, radio resources at other times, or a range that signals using the radio resources can reach with valid strength at other times.

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

Referring to FIG. 9, "RB" denotes a radio bearer, and "H" denotes a header. Radio bearers are categorized into two groups: DRBs for user plane data and SRBs for control plane data. The MAC PDU is transmitted/received using radio resources through the PHY layer to/from an external device. The MAC PDU arrives to the PHY layer in the form of a transport block.

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 PUCCH, and downlink control information (DCI) is mapped to PDCCH. A MAC PDU related to UL-SCH is transmitted by a UE via a PUSCH based on an UL grant, and a MAC PDU related to DL-SCH is transmitted by a BS via a PDSCH based on a DL assignment.

Hereinafter, technical features related to AI/ML are described.

The application of AI/ML to wireless communications has been thus far limited to implementation-based approaches, both, at the network and the UE sides. A study on enhancement for data collection for NR and ENDC (FS_NR_ENDC_data_collect) has examined the functional framework for RAN intelligence enabled by further enhancement of data collection through use cases, examples etc. and identify the potential standardization impacts on current NG -RAN nodes and interfaces. In SA WG2 AI/ML related study, a network functionality NWDAF (Network Data Analytics Function) was introduced in Rel-15 and has been enhanced in Rel-16 and Rel-17.

In this study, we explore the benefits of augmenting the air-interface with features enabling improved support of AI/ML based algorithms for enhanced performance and/or reduced complexity/overhead. Enhanced performance here depends on the use cases under consideration and could be, e.g., improved throughput, robustness, accuracy or reliability, etc.

Through studying a few carefully selected use cases, assessing their performance in comparison with traditional methods and the associated potential specification impacts that enable their solutions, this SI will　lay the foundation for future air-interface use cases leveraging AI/ML techniques.

The goal is that sufficient use cases will be considered to enable the identification of a common AI/ML framework, including functional requirements of AI/ML architecture, which could be used in subsequent projects. The study should also identify areas where AI/ML could improve the performance of air-interface functions.

The study will serve identifying what is required for an adequate AI/ML model characterization and description establishing pertinent notation for discussions and subsequent evaluations. Various levels of collaboration between the gNB and UE are identified and considered.

Evaluations to exercise the attainable gains of AI/ML based techniques for the use cases under consideration will be carried out with the corresponding identification of KPIs with the goal to have a better understanding of the attainable gains and associated complexity requirements.

Finally, specification impact will be assessed in order to improve the overall understanding of what would be required to enable AI/ML techniques for the air-interface.

For the study on AI/ML for air-interface, the basic framework and principles agreed for FS _ NR _ ENDC _data_collect should be taken into consideration for possible applicability.

Study the 3GPP framework for AI/ML for air-interface corresponding to each target use case regarding aspects such as performance, complexity, and potential specification impact.

Use cases to focus on:

1> Initial set of use cases includes:

a) CSI feedback enhancement, e.g., overhead reduction, improved accuracy, prediction

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

c) Positioning accuracy enhancements for different scenarios including, e.g., those with heavy NLOS conditions

2> Finalize representative sub use cases for each use case for characterization and baseline performance evaluations

a) The AI/ML approaches for the selected sub use cases need to be diverse enough to support various requirements on the gNB-UE collaboration levels

- the selection of use cases for this study solely targets the formulation of a framework to apply AI/ML to the air-interface for these and other use cases. The selection itself does not intend to provide any indication of the prospects of any future normative project.

AI/ML model, terminology and description to identify common and specific characteristics for framework investigations:

3> Characterize the defining stages of AI/ML related algorithms and associated complexity:

a) Model generation, e.g., model training (including input/output, pre-/post-process, online/offline as applicable), model validation, model testing, as applicable

b) Inference operation, e.g., input/output, pre-/post-process, as applicable

4> Identify various levels of collaboration between UE and gNB pertinent to the selected use cases, e.g.,

a) No collaboration: implementation-based only AI/ML algorithms without information exchange [for comparison purposes]

b) Various levels of UE/gNB collaboration targeting at separate or joint ML operation.

5> Characterize lifecycle management of AI/ML model: e.g., model training, model deployment , model inference, model monitoring, model updating

6> Dataset(s) for training, validation, testing, and inference

7> Identify common notation and terminology for AI/ML related functions, procedures and interfaces

8> Consider the work done for FS_NR_ENDC_data_collect when appropriate

For the use cases under consideration:

- Evaluate performance benefits of AI/ML based algorithms for the agreed use cases in the final representative set:

a) Methodology based on statistical models, for link and system level simulations.

i. Extensions of 3GPP evaluation methodology for better suitability to AI/ML based techniques should be considered as needed.

ii. Whether field data are optionally needed to further assess the performance and robustness in real-world environments should be discussed as part of the study.

iii. Need for common assumptions in dataset construction for training, validation and test for the selected use cases.

iv. Consider adequate model training strategy, collaboration levels and associated implications

v. Consider agreed-upon base AI model(s) for calibration

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

b) KPIs: Determine the common KPIs and corresponding requirements for the AI/ML operations. Determine the use-case specific KPIs and benchmarks of the selected use-cases.

i. Performance, inference latency and computational complexity of AI/ML based algorithms should be compared to that of a state-of-the-art baseline

ii. Overhead, power consumption (including computational), memory storage, and hardware requirements (including for given processing delays) associated with enabling respective AI/ML scheme, as well as generalization capability should be considered.

- Assess potential specification impact, specifically for the agreed use cases in the final representative set and for a common framework:

c) PHY layer aspects,

i. Consider aspects related to, e.g., the potential specification of the AI Model lifecycle management, and dataset construction for training, validation and test for the selected use cases

ii. Use case and collaboration level specific specification impact, such as new signalling, means for training and validation data assistance, assistance information, measurement, and feedback

d) Protocol aspects, e.g., (RAN2) - RAN2 only starts the work after there is sufficient progress on the use case study in RAN1

i. Consider aspects related to, e.g., capability indication, configuration and control procedures (training/inference), and management of data and AI/ML model, per RAN1 input

ii. Collaboration level specific specification impact per use case

e) Interoperability and testability aspects, e.g., (RAN4) - RAN4 only starts the work after there is sufficient progress on use case study in RAN1 and RAN2

i. Requirements and testing frameworks to validate AI/ML based performance enhancements and ensuring that UE and gNB with AI/ML meet or exceed the existing minimum requirements if applicable

ii. Consider the need and implications for AI/ML processing capabilities definition

- specific AI/ML models are not expected to be specified and are left to implementation. User data privacy needs to be preserved.

- The study on AI/ML for air interface is based on the current RAN architecture and new interfaces shall not be introduced.

The application of AI/ML techniques to NR air interface has been studied in FS_NR_AIML_Air.

In this work item, we provide the normative support for the general framework for AI/ML for air interface, as well as, enable the recommended use cases in the preceding study. In addition, a number of study objectives in this project will tackle some outstanding issues identified during the study in an attempt to deepen the understanding in view of future normative work.

Objective of SI or Core part WI or Testing part WI

Provide specification support for the following aspects:

1> AI/ML general framework for one-sided AI/ML models within the realm of what has been studied in the FS_NR_AIML_Air project [RAN2]:

2> Signalling and protocol aspects of Life Cycle Management (LCM) enabling functionality and model (if justified) selection, activation, deactivation, switching, fallback

3> Identification related signalling is part of the above objective

2> Necessary signalling/mechanism(s) for LCM to facilitate model training, inference, performance monitoring, data collection (except for the purpose of CN/OAM/OTT collection of UE-sided model training data) for both UE-sided and NW-sided models

2> Signalling mechanism of applicable functionalities/models

1> Beam management - DL Tx beam prediction for both UE-sided model and NW-sided model, encompassing [RAN1/RAN2]:

2> Spatial-domain DL Tx beam prediction for Set A of beams based on measurement results of Set B of beams ("BM-Case1")

2> Temporal DL Tx beam prediction for Set A of beams based on the historic measurement results of Set B of beams ("BM-Case2")

2> Specify necessary signalling/mechanism(s) to facilitate LCM operations specific to the Beam Management use cases, if any

2> Enabling method(s) to ensure consistency between training and inference regarding NW-side additional conditions (if identified) for inference at UE

- Strive for common framework design to support both BM-Case1 and BM-Case2

1> Positioning accuracy enhancements, encompassing [RAN1/RAN2/RAN3]:

2> Direct AI/ML positioning:

3> (1^st priority) Case 1: UE-based positioning with UE-side model, direct AI/ML positioning

3> (2^nd priority) Case 2b: UE-assisted/LMF-based positioning with LMF-side model, direct AI/ML positioning

3> (1^st priority) Case 3b: NG-RAN node assisted positioning with LMF-side model, direct AI/ML positioning

2> AI/ML assisted positioning

3> (2^nd priority) Case 2a: UE-assisted/LMF-based positioning with UE-side model, AI/ML assisted positioning

3> (1^st priority) Case 3a: NG-RAN node assisted positioning with gNB-side model, AI/ML assisted positioning

2> Specify necessary measurements, signalling/mechanism(s) to facilitate LCM operations specific to the Positioning accuracy enhancements use cases, if any

2> Investigate and specify the necessary signalling of necessary measurement enhancements (if any)

2> Enabling method(s) to ensure consistency between training and inference regarding NW-side additional conditions (if identified) for inference at UE for relevant positioning sub use cases

1> Core requirements for the above two use cases for AI/ML LCM procedures and UE features [RAN4]:

2> Specify necessary RAN4 core requirements for the above two use cases.

2> Specify necessary RAN4 core requirements for LCM procedures including performance monitoring.

Study objectives with corresponding checkpoints in RAN#105 (Sept '24):

1> CSI feedback enhancement [RAN1]:

2> For CSI compression (two-sided model), further study ways to:

3> Improve trade-off between performance and complexity/overhead

4> e.g., considering extending the spatial/frequency compression to spatial/temporal/frequency compression, cell/site specific models, CSI compression plus prediction (compared to Rel-18 non-AI/ML based approach), etc.

3> Alleviate/resolve issues related to inter-vendor training collaboration.

while addressing other aspects requiring further study/conclusion as captured in the conclusions section of the TR 38.843.

2> For CSI prediction (one-sided model), further study performance gain over Rel-18 non-AI/ML based approach and associated complexity, while addressing other aspects requiring further study/conclusion as captured in the conclusions section of the TR 38.843 (e.g., cell/site specific model could be considered to improve performance gain).

1> Necessity and details of model Identification concept and procedure in the context of LCM [RAN2/RAN1]

1> CN/OAM/OTT collection of UE-sided model training data [RAN2/RAN1]:

2> For the FS_NR_AIML_Air study use cases, identify the corresponding contents of UE data collection

2> Analyse the UE data collection mechanisms identified during the FS_NR_AIML_Air study along with the implications and limitations of each of the methods

1> Model transfer/delivery [RAN2/RAN1]:

2> Determine whether there is a need to consider standardised solutions for transferring/delivering AI/ML model(s) considering at least the solutions identified during the FS_NR_AIML_Air study

1> Testability and interoperability [RAN4]:

2> Finalize the testing framework and procedure for one-sided models and further analyse the various testing options for two-sided models, in collaboration with RAN1, and including at least:

3> Relation to legacy requirements

3> Performance monitoring and LCM aspects considering use-case specifics

3> Generalization aspects

3> Static/non-static scenarios/conditions and propagation conditions for testing (e.g., CDL, field data, etc.)

3> UE processing capability and limitations

3> Post-deployment validation due to model change/drift

2> RAN5 aspects related to testability and interoperability to be addressed on a request basis

- offline training is assumed for the purpose of this project.

- the outcome of the study objectives should be captured in TR 38.843 for future reference.

- Coordination with SA/SA WGs of the ongoing study/work as it may relate to their required work.

With existing L3 handover mechanism, handover is triggered and executed based on reported historical measurement result and/or measurement event(s) i.e., it is kind of reactive scheme by its nature. It may work well among macro cells when UE's mobility is low for existing services. But it could be problematic when either UE's mobility is high or among micro cells of high density or both for existing services or future services e.g. XR, where such reactive scheme may result in more unintended event e.g., handover failure, radio link failure, Ping-Pong phenomenon, throughput loss or too early/late handover etc. To improve handover robustness conditional handover is introduced in Rel-16. And to reduce interruption time of frequent handover among small cells LTM HO is introduced in Rel-18. However, these two mechanisms are not sufficient because they are still reactive scheme by design. On the other hand, mechanism based on AI/ML algorithm has the potential to enable proactive scheme.

In Rel-18 SID called FS_NR_AIML_air was studied extensively on physical layer centric use cases including spatial and temporal beam prediction. Temporal prediction within serving cell is mainly to predict the best or top-K beam(s) or beam pair(s) in time domain in order to improve UE throughput. While predict the best or top-K beam(s) or beam pair(s) among a set of beams by measuring a smaller set of beams could help reduce RS signalling overhead, measurement efforts and UE power consumption etc. By extended L1 beam measurement from serving cell to neighbouring cell, majority of the RAN1 work can be reused for e.g. LTM HO study. Since L3 measurement is based on filtering of L1 measurement, the study of AI/ML for air can be leveraged for mobility purpose e.g., temporal prediction can also be used to predict beam(s)/cell(s) becoming worse so that unintended event like radio link failure or short-stay handover can be avoided.

Mobility enhancement was also studied in RAN3 in Rel-17 in SID called FS_NR_ENDC_data_collect and is now specified in Rel-18 WID NR_AIML_NGRAN-Core. In these RAN3 items the study and normative work on mobility enhancement is based on information available in network side e.g. handover and stay of time in history among cells to predict UE's trajectory in single hop and hence potential candidates. In Rel-19 RAN3 will further work on UE's trajectory for multiple hops. The predicted UE's trajectory could be helpful for study on AI/ML mobility over air interface to some extent.

Based on progress made in RAN1 and RAN3 so far and assumption on UE's trajectory it is feasible to predict RRM measurement and/or event and hence candidate target cell in UE side. In network side new assistant information, if necessary, and statistics information based on measurement report from UE and/or neighbouring nodes can be also used for smart prediction. If some prediction information could be known by network, handover and/or RRM performance can be improved by proactive measures to either make a better decision or avoid unintended event.

Objective of SI or Core part WI or Testing part WI

The study will focus on mobility enhancement in RRC_CONNECTED mode over air interface by following existing mobility framework, i.e., handover decision is always made in network side. Mobility use cases focus on standalone NR PCell change. UE-side and network-side AI/ML model can be both considered, respectively.

Study and evaluate potential benefits and gains of AI/ML aided mobility for network triggered L3-based handover, considering the following aspects:

1> AI/ML based RRM measurement and event prediction,

2> Cell-level measurement prediction including intra and inter-frequency (UE sided and NW sided model) [RAN2]

3> Inter-cell Beam-level measurement prediction for L3 Mobility (UE sided and NW sided model) [RAN2]

2> HO failure/RLF prediction (UE sided model) [RAN2]

2> Measurement events prediction (UE sided model) [RAN2]

1> Study the need/benefits of any other UE assistance information for the network side model [RAN2]

1> The evaluation of the AI/ML aided mobility benefits should consider HO performance KPIs (e.g., Ping-pong HO, HOF/RLF, Time of stay, Handover interruption, prediction accuracy, and measurement reduction) etc.) and complexity tradeoffs [RAN2]

1> Potential AI mobility specific enhancement should be based on the Rel19 AI/ML-air interface WID general framework (e.g. LCM, performance monitoring etc) [RAN2]

- This would only be treated after sufficient progress is made in the Rel-19 AI/ML air interface WID

1> Potential specification impacts of AI/ML aided mobility [RAN2]

1> Evaluate testability, interoperability, and impacts on RRM requirements and performance

- RAN1/3 work can be triggered via LS

- RAN4 scope/work can be defined and confirmed by RAN#105 after some RAN2 discussions (within the RAN4 pre-allocated TUs)

- To avoid duplicate study with "AI/ML for NG-RAN" led by RAN3

- Two-sided model is not included

FIG. 10 shows an example of a Functional Framework for RAN Intelligence.

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

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

>> Training Data: Data needed as input for the AI/ML Model Training function.

>> Inference Data: Data needed as input for the AI/ML Model Inference function.

> Model Training is a function that performs the AI/ML model training, validation, and testing which may generate model performance metrics as part of the model testing procedure. The Model Training function is also responsible for data preparation (e.g., data pre-processing and cleaning, formatting, and transformation) based on Training Data delivered by a Data Collection function, if required.

>> Model Deployment/Update: Used to initially deploy a trained, validated, and tested AI/ML model to the Model Inference function or to deliver an updated model to the Model Inference function.

> Model Inference is a function that provides AI/ML model inference output (e.g., predictions or decisions). Model Inference function may provide Model Performance Feedback to Model Training function when applicable. The Model Inference function is also responsible for data preparation (e.g., data pre-processing and cleaning, formatting, and transformation) based on Inference Data delivered by a Data Collection function, if required.

>> Output: The inference output of the AI/ML model produced by a Model Inference function.

>>> Details of inference output are use case specific.

>> Model Performance Feedback: It may be used for monitoring the performance of the AI/ML model, when available.

> Actor is a function that receives the output from the Model Inference function and triggers or performs corresponding actions. The Actor may trigger actions directed to other entities or to itself.

>> Feedback: Information that may be needed to derive training data, inference data or to monitor the performance of the AI/ML Model and its impact to the network through updating of KPIs and performance counters.

Hereinafter, technical features related to Mobility Optimization are described.

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

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

- Reduction of the probability of unintended events

- UE Location/Mobility/Performance prediction

- Traffic Steering

Reduction of the probability of unintended events associated with mobility.

Examples of such unintended events are:

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

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

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

- Successful Handover: During a successful handover, there is underlying issue.

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

UE Location/Mobility/Performance Prediction

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

Traffic Steering

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

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

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

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

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

Locations for AI/ML Model Training and AI/ML Model Inference

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

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

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

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

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

gNB is also allowed to continue model training based on AI/ML model trained in the OAM.

Step 0. NG-RAN node 2 is assumed to optionally have an AI/ML model, which can generate required input such as resource status and utilization prediction/estimation etc.

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

Step 2. The UE collects the indicated measurement, e.g., UE measurements related to RSRP, RSRQ, SINR of serving cell and neighbouring cells.

Step 3. The UE sends measurement report message to NG-RAN node 1 including the required measurement.

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

Step 5. The NG-RAN node 2 sends the input data for training to 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 can include the corresponding inference result from the NG-RAN node 2.

Step 6. Model Training. Required measurements are leveraged to training AI/ML model for UE mobility optimization.

Step 7. OAM sends AI/ML Model Deployment Message to deploy the trained/updated AI/ML model into the NG-RAN node(s). The NG-RAN node can also continue model training based on the received AI/ML model from OAM.

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

Step 9. The NG-RAN node 1 obtains the input data for inference from the NG-RAN node 2 for UE mobility optimization, where the input data for inference 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 inference can include the corresponding inference result from the NG-RAN node 2.

Step 10. Model Inference. Required measurements are leveraged into Model Inference to output the prediction, e.g., UE trajectory prediction, target cell prediction, target NG-RAN node prediction, etc.

Step 11. The NG-RAN 1 sends the model performance feedback to OAM if applicable.

This step is out of RAN3 scope.

Step 12: According to the prediction, recommended actions or configuration, the NG-RAN node 1, the target NG-RAN node (represented by NG-RAN node 2 of this step in the flowchart), and UE perform the Mobility Optimization / handover procedure to hand over UE from NG-RAN node 1 to the target NG-RAN node.

Step 13. The NG-RAN node 1 sends the feedback information to OAM.

Step 14. The NG-RAN node 2 sends the feedback information to OAM.

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

Step 2. UE collects the indicated measurement, e.g., UE measurements related to RSRP, RSRQ, SINR of serving cell and neighbouring cells.

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

Step 4. The NG-RAN node 1 obtains the input data for training from the NG-RAN node2, 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 can include the corresponding inference result from the NG-RAN node 2.

Step 5. Model training. Required measurements are leveraged to training AI/ML model for mobility optimization.

Step 6. NG-RAN node1 obtains the measurement report as inference data for real-time UE mobility optimization.

Step 7. The NG-RAN node 1 obtains the input data for inference from the NG-RAN node 2 for UE mobility optimization, where the input data for inference 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 inference can include the corresponding inference result from the NG-RAN node 2.

Step 8. Model Inference. Required measurements are leveraged into Model Inference to output the prediction, including e.g., UE trajectory prediction, target cell prediction, target NG-RAN node prediction, etc.

Step 9: According to the prediction, recommended actions or configuration, the NG-RAN node 1, the target NG-RAN node (represented by NG-RAN node 2 of this step in the flowchart), and UE perform the Mobility Optimization / handover procedure to hand over UE from NG-RAN node 1 to the target NG-RAN node.

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

For example, UE mobility information for training purposes is only sent to gNBs that requested such information or when triggered.

Input of AI/ML-based Mobility Optimization

The following data is required as input data for mobility optimization.

From the UE:

- UE location information (e.g., coordinates, serving cell ID, moving velocity) interpreted by gNB implementation when available.

- Radio measurements related to serving cell and neighbouring cells associated with UE location information, e.g., RSRP, RSRQ, SINR.

- UE Mobility History Information.

From the neighbouring RAN nodes:

- UE's history information from neighbour

- Position, QoS parameters and the performance information of historical HO-ed UE (e.g., loss rate, delay, etc.)

- Current/predicted resource status

- UE handovers in the past that were successful and unsuccessful, including too-early, too-late, or handover to wrong (sub-optimal) cell, based on existing SON/RLF report mechanism.

From the local node:

- UE trajectory prediction

- Current/predicted resource status

- Current/predicted UE traffic

Output of AI/ML-based Mobility Optimization

AI/ML-based mobility optimization can generate following information as output:

- UE trajectory prediction (Latitude, longitude, altitude, cell ID of UE over a future period of time)

Whether the UE trajectory prediction is an external output to the node hosting the Model Inference function should be discussed during the normative work phase.

- Estimated arrival probability in CHO and relevant confidence interval

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

- Priority, handover execution timing, predicted resource reservation time window for CHO.

- UE traffic prediction (will be used by the RAN node internally and the details are left to normative work phase)

- Model output validity time will be discussed during R18 normative work per inference output.

Feedback of AI/ML-based Mobility Optimization

The following data is required as feedback data for mobility optimization.

- QoS parameters such as throughput, packet delay of the handed-over UE, etc.

- Resource status information updates from target NG-RAN.

- Performance information from target NG-RAN. The details of performance information are to be discussed during normative work phase.

Standard impact

To improve the mobility decisions at a gNB (gNB-CU), a gNB can request mobility feedback from a neighbouring node. Details of the procedure will be determined during the normative phase.

If existing UE measurements are needed by a gNB for AI/ML-based mobility optimization, RAN3 shall reuse the existing framework (including MDT and RRM measurements). Whether new UE measurements are needed is left to normative phase based on the use case description.

MDT procedure enhancements should be discussed during the normative phase.

Potential Xn interface impact:

- Predicted resource status info and performance info from candidate target NG-RAN node to source NG-RAN node

- New signaling procedure or existing procedure to retrieve input information via Xn interface.

- New signaling procedure or existing procedure to retrieve feedback information via Xn interface.

Hereinafter, technical features related to AI and ML are described.

Artificial Intelligence (AI)/Machine Learning (ML) is being used in a range of application domains across industry sectors, realizing significant productivity gains. In particular, in mobile communications 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 behaviour prediction) with AI/ML models to enable applications like enhanced photography, intelligent personal assistants, VR/AR, video gaming, video analytics, personalized shopping recommendation, autonomous driving/navigation, smart home appliances, mobile robotics, mobile medicals, as well as mobile finance.

Artificial Intelligence (AI) is the science and engineering to build intelligent machines capable of carrying out tasks as humans do.

Deep neural network

Within the ML field, there is an area that is often referred to as brain-inspired computation, which is a program aiming to emulate some aspects of how we understand the brain to operate. Since it is believed that the main computational elements a human brain are 86 billion neurons, the two subareas of brain-inspired computation are both inspired by the architecture of a neuron, as shown in FIG. 13.

Compared to spiking computing approaches, the more popular ML approaches are using "neural network" as the model. Neural networks (NN) take their inspiration from the notion that a neuron's computation involves a weighted sum of the input values. But instead of simply outputting the weighted sum, a NN applies a nonlinear function to generate an output only if the inputs cross some threshold, as shown in FIG. 13.

FIG. 14 shows a diagrammatic picture of a computational neural network. The neurons in the input layer receive some values and propagate them to the neurons in the middle layer of the network, which is also called a "hidden layer". The weighted sums from one or more hidden layers are ultimately propagated to the output layer, which presents the final outputs of the network.

Neural networks having more than three layers, i.e., more than one hidden layer are called deep neural networks (DNN). In contrast to the conventional shallow-structured NN architectures, DNNs, also referred to as deep learning, made amazing breakthroughs since 2010s in many essential application areas because they can achieve human-level accuracy or even exceed human accuracy. Deep learning techniques use supervised and/or unsupervised strategies to automatically learn hierarchical representations in deep architectures for classification. With a large number of hidden layers, the superior performance of DNNs comes from its ability to extract high-level features from raw sensory data after using statistical learning over a large amount of data to obtain an effective representation of an input space. In recent years, thanks to the big data obtained from the real world, the rapidly increased computation capacity and continuously-evolved algorithms, DNNs have become the most popular ML models for many AI applications.

Training and inference

Training is a process in which a AI/ML model learns to perform its given tasks, more specifically, by optimizing the value of the weights in the DNN. A DNN is trained by inputting a training set, which are often correctly-labelled training samples. Taking image classification for instance, the training set includes correctly-classified images. When training a network, the weights are usually updated using a hill-climbing optimization process called gradient descent. The gradient indicates how the weights should change in order to reduce the loss (the gap between the correct outputs and the outputs computed by the DNN based on its current weights). The training process is repeated iteratively to continuously reduce the overall loss. Until the loss is below a predefined threshold, the DNN with high precision is obtained.

There are multiple ways to train the network for different targets. The introduced above is supervised learning which uses the labelled training samples to find the correct outputs for a task. Unsupervised learning uses the unlabelled training samples to find the structure or clusters in the data. Reinforcement learning can be used to output what action the agent should take next to maximize expected rewards. Transfer learning is to adjust the previously-trained weights (e.g. weights in a global model) using a new training set, which is used for a faster or more accurate training for a personalized model.

FIG. 15 shows an example of an AI/ML inference.

After a DNN is trained, it can perform its task by computing the output of the network using the weights determined during the training process, which is referred to as inference. In the model inference process, the inputs from the real world are passed through the DNN. Then the prediction for the task is output, as shown in FIG. 15. For instance, the inputs can be pixels of an image, sampled amplitudes of an audio wave or the numerical representation of the state of some system or game. Correspondingly, the outputs of the network can be a probability that an image contains a particular object, the probability that an audio sequence contains a particular word or a bounding box in an image around an object or the proposed action that should be taken.

The performance of DNNs is gained at the cost of high computational complexity. Hence more efficient compute engines are often used, e.g. graphics processing units (GPU) and network processing units (NPU). Compared to the inference which only involves the feedforward process, the training often requires more computation and storage resources because it involves also the backpropagation process.

Widely-used DNN models and algorithms

FIG. 16 shows an example of an MLP DNN model.

Many DNN models have been developed over the past two decades. Each of these models has a different "network architecture" in terms of number of layers, layer types, layer shapes (i.e., filter size, number of channels and filters), and connections between layers. FIG. 16 presents three popular structures of DNNs: multilayer perceptrons (MLPs), convolution neural networks (CNNs), and recurrent neural networks (RNNs). Multilayer perceptrons (MLP) model is the most basic DNN, which is composed of a series of fully connected layers. In a fully connected layer, all outputs are connected to all inputs, as shown in FIG. 16. Hence MLP requires a significant amount of storage and computation.

FIG. 17 shows an example of a CNN model.

An approach to limiting the number of weights that contribute to an output is to calculate the output only using a function of a fixed-size window of inputs. An extremely popular window-based DNN model uses a convolution operation to structure the computation, hence is named as convolution neural network (CNN). A CNN is composed of multiple convolutional layers, as shown in FIG. 17. Applying various convolutional filters, CNN models can capture the high-level representation of the input data, making it popular for image classification and speech recognition tasks.

FIG. 18 shows an example of an RNN model.

Recurrent neural network (RNN) models are another type of DNNs, which use sequential data feeding. The input of RNN consists of the current input and the previous samples. Each neuron in an RNN owns an internal memory that keeps the information of the computation from the previous samples. As shown in FIG. 18, the basic unit of RNN is called cell, and further, each cell consists of layers and a series of cells enables the sequential processing of RNN models. RNN models have been widely used in the natural language processing task on mobile devices, e.g., language modelling, machine translation, question answering, word embedding, and document classification.

FIG. 19 shows an example of Reinforcement learning.

Deep reinforcement learning (DRL) is not another DNN model. It is composed of DNNs and reinforcement learning. As illustrated in FIG. 19, the goal of DRL is to create an intelligent agent that can perform efficient policies to maximize the rewards of long-term tasks with controllable actions. The typical application of DRL is to solve various scheduling problems, such as decision problems in games, rate selection of video transmission, and so on.

The following metrics/methods for AI/ML model monitoring in lifecycle management per use case are considered:

- Monitoring based on inference accuracy, including metrics related to intermediate KPIs

- Monitoring based on system performance, including metrics related to system performance KPIs

- Other monitoring solutions, at least the following 2 options.

- Monitoring based on data distribution

- Input-based: e.g., Monitoring the validity of the AI/ML input, e.g., out-of-distribution detection, drift detection of input data, or SNR, delay spread, etc.

- Output-based: e.g., drift detection of output data

- Monitoring based on applicable condition

- Model monitoring metric calculation may be done at NW or UE

Methods to assess/monitor the applicability and expected performance of an inactive model/functionality, including the following examples for the purpose of activation/selection/switching of UE-side models/UE-part of two-sided models /functionalities (if applicable):

- Assessment/Monitoring based on the additional conditions associated with the model/functionality

- Assessment/Monitoring based on input/output data distribution

- Assessment/Monitoring using the inactive model/functionality for monitoring purpose and measuring the inference accuracy

- Assessment/Monitoring based on past knowledge of the performance of the same model/functionality (e.g., based on other UEs)

Signalling procedures for model and functionality life cycle management

In this clause the signalling procedures for different scenarios for model-ID-based management and/or functionality-based management are exemplified. The procedures can at least be considered for UE-side models. These can include scenarios for which the management decision is taken by the network or by the UE. For network-side decision, this can be either network-initiated, or UE-initiated and requested to the network. While for UE-side decision, this can be either event-triggered as configured by the network and where the UE's decision is reported to the network, or UE-autonomous, with or without UE's decision being reported to the network.

- The mapping of these scenarios to specific use cases can be left to RAN1.

- The scenarios discussed below shall not imply support for all potential functionality and/or model Management Instructions (e.g., (de)activation, selection, switching, fallback, etc.) for every use case.

- Management Request/Management Instruction/Management Decision Report may include details about the model/functionality selection, (de)activation, switching or fallback.

- Decision by the network

- Network-initiated

The case where the LCM decision is taken and initiated by the network is depicted in FIG. 20.

- The Management Instruction may be a result of model/functionality performance monitoring at the network.

- The Management Instruction may include information about the model or functionality.

- UE-initiated and requested to the network

FIG. 21 shows an example of network decision, UE-initiated AI/ML management.

The case where the LCM decision is taken by the network but where the request is initiated by the UE is depicted in FIG. 21.

- The Management Request may be a result of model/functionality monitoring at the UE.

- In response to the Management Request, the network may send a Management Instruction to the UE.

- The Management Request may include information about the model or functionality.

- The network may accept or reject the Management Request from the UE.

- The Management Request may include information related to model/functionality performance metrics.

- Decision by the UE

- Event-triggered as configured by the network, UE's decision is reported to the network

The case where the LCM decision is taken by the UE according to prior network configuration is depicted in FIG. 22.

- Use case-specific events/conditions may be configured by the network for event-triggered AI/ML management at the UE.

- UE may send a Management Decision Report to the network following event-triggered AI/ML management at the UE.

- The Management Decision Report may include information about the model or functionality.

- UE-autonomous, UE's decision is reported to the network

FIG. 23 shows an example of UE autonomous, decision reported to the network.

The case where the LCM decision can autonomously be taken by the UE is depicted in FIG. 23.

- The UE may be configured to send a Management Decision Report to the network upon performing a model/functionality Management Decision.

- UE-autonomous, UE's decision is not reported to the network

For the case where the LCM decision can autonomously be taken by the UE and where the decision is not reported to the network, the AI/ML management is transparent from a network perspective.

Hereinafter, technical features related to measurements are described. Sections of 3GPP TS 38.331 v17.5.0 may be referred.

The network may configure an RRC_CONNECTED UE to perform measurements. The network may configure the UE to report them in accordance with the measurement configuration or perform conditional reconfiguration evaluation in accordance with the conditional reconfiguration. The measurement configuration is provided by means of dedicated signalling i.e. using the RRCReconfiguration or RRCResume .

The network may configure the UE to perform the following types of measurements:

- NR measurements;

- Inter-RAT measurements of E-UTRA frequencies;

- Inter-RAT measurements of UTRA-FDD frequencies;

- NR sidelink measurements of L2 U2N Relay UEs.

The network may configure the UE to report the following measurement information based on SS/PBCH block(s):

- Measurement results per SS/PBCH block;

- Measurement results per cell based on SS/PBCH block(s);

- SS/PBCH block(s) indexes.

The network may configure the UE to report the following measurement information based on CSI-RS resources:

- Measurement results per CSI-RS resource;

- Measurement results per cell based on CSI-RS resource(s);

- CSI-RS resource measurement identifiers.

The network may configure the UE to perform the following types of measurements for NR sidelink and V2X sidelink:

- CBR measurements.

The network may configure the UE to report the following CLI measurement information based on SRS resources:

- Measurement results per SRS resource;

- SRS resource(s) indexes.

The network may configure the UE to report the following CLI measurement information based on CLI-RSSI resources:

- Measurement results per CLI-RSSI resource;

- CLI-RSSI resource(s) indexes.

The network may configure the UE to report the following Rx-Tx time difference measurement information based on CSI-RS for tracking or PRS:

- UE Rx-Tx time difference measurement result.

The measurement configuration includes the following parameters:

1. Measurement objects: A list of objects on which the UE shall perform the measurements.

- For intra-frequency and inter-frequency measurements a measurement object indicates the frequency/time location and subcarrier spacing of reference signals to be measured. Associated with this measurement object, the network may configure a list of cell specific offsets, a list of 'exclude-listed' cells and a list of 'allow-listed' cells. Exclude-listed cells are not applicable in event evaluation or measurement reporting. Allow-listed cells are the only ones applicable in event evaluation or measurement reporting.

- The measObjectId of the MO which corresponds to each serving cell is indicated by servingCellMO within the serving cell configuration.

- For inter-RAT E-UTRA measurements a measurement object is a single E-UTRA carrier frequency. Associated with this E-UTRA carrier frequency, the network can configure a list of cell specific offsets and a list of 'exclude-listed' cells. Exclude-listed cells are not applicable in event evaluation or measurement reporting.

- For inter-RAT UTRA-FDD measurements a measurement object is a set of cells on a single UTRA-FDD carrier frequency.

- For NR sidelink measurements of L2 U2N Relay UEs, a measurement object is a single NR sidelink frequency to be measured.

- For CBR measurement of NR sidelink communication, a measurement object is a set of transmission resource pool(s) on a single carrier frequency for NR sidelink communication.

- For CBR measurement of NR sidelink discovery, a measurement object is a set of discovery dedicated resource pool(s) or transmission resource pool(s) also used for NR sidelink discovery on a single carrier frequency for NR sidelink discovery.

- For CLI measurements a measurement object indicates the frequency/time location of SRS resources and/or CLI-RSSI resources, and subcarrier spacing of SRS resources to be measured.

2. Reporting configurations: A list of reporting configurations where there can be one or multiple reporting configurations per measurement object. Each measurement reporting configuration consists of the following:

- Reporting criterion: The criterion that triggers the UE to send a measurement report. This can either be periodical or a single event description.

- RS type: The RS that the UE uses for beam and cell measurement results (SS/PBCH block or CSI-RS).

- Reporting format: The quantities per cell and per beam that the UE includes in the measurement report (e.g. RSRP) and other associated information such as the maximum number of cells and the maximum number beams per cell to report.

In case of conditional reconfiguration, each configuration consists of the following:

- Execution criteria: The criteria the UE uses for conditional reconfiguration execution.

- RS type: The RS that the UE uses for obtaining beam and cell measurement results (SS/PBCH block-based or CSI-RS-based), used for evaluating conditional reconfiguration execution condition.

3. Measurement identities: For measurement reporting, a list of measurement identities where each measurement identity links one measurement object with one reporting configuration. By configuring multiple measurement identities, it is possible to link more than one measurement object to the same reporting configuration, as well as to link more than one reporting configuration to the same measurement object. The measurement identity is also included in the measurement report that triggered the reporting, serving as a reference to the network. For conditional reconfiguration triggering, one measurement identity links to exactly one conditional reconfiguration trigger configuration. And up to 2 measurement identities can be linked to one conditional reconfiguration execution condition.

4. Quantity configurations: The quantity configuration defines the measurement filtering configuration used for all event evaluation and related reporting, and for periodical reporting of that measurement. For NR measurements, the network may configure up to 2 quantity configurations with a reference in the NR measurement object to the configuration that is to be used. In each configuration, different filter coefficients can be configured for different measurement quantities, for different RS types, and for measurements per cell and per beam.

5. Measurement gaps: Periods that the UE may use to perform measurements.

A UE in RRC_CONNECTED maintains a measurement object list, a reporting configuration list, and a measurement identities list according to signalling and procedures in this specification. The measurement object list possibly includes NR measurement object(s), CLI measurement object(s), inter-RAT objects, and L2 U2N Relay objects. Similarly, the reporting configuration list includes NR, inter-RAT, and L2 U2N Relay reporting configurations. Any measurement object can be linked to any reporting configuration of the same RAT type. Some reporting configurations may not be linked to a measurement object. Likewise, some measurement objects may not be linked to a reporting configuration.

The measurement procedures distinguish the following types of cells:

1. The NR serving cell(s) - these are the SpCell and one or more SCells.

2. Listed cells - these are cells listed within the measurement object(s).

3. Detected cells - these are cells that are not listed within the measurement object(s) but are detected by the UE on the SSB frequency(ies) and subcarrier spacing(s) indicated by the measurement object(s).

For NR measurement object(s), the UE measures and reports on the serving cell(s)/serving Relay UE (for L2 U2N Remote UE), listed cells and/or detected cells. For inter-RAT measurements object(s) of E-UTRA, the UE measures and reports on listed cells and detected cells and, for RSSI and channel occupancy measurements, the UE measures and reports on the configured resources on the indicated frequency. For inter-RAT measurements object(s) of UTRA-FDD, the UE measures and reports on listed cells. For CLI measurement object(s), the UE measures and reports on configured measurement resources (i.e. SRS resources and/or CLI-RSSI resources). For L2 U2N Relay object(s), the UE measures and reports on the serving NR cell(s), as well as the discovered L2 U2N Relay UEs.

Whenever the procedural specification, other than contained in clause 5.5.2, refers to a field it concerns a field included in the VarMeasConfig unless explicitly stated otherwise i.e. only the measurement configuration procedure covers the direct UE action related to the received measConfig.

In NR-DC, the UE may receive two independent measConfig:

- a measConfig, associated with MCG, that is included in the RRCReconfiguration message received via SRB1; and

- a measConfig, associated with SCG, that is included in the RRCReconfiguration message received via SRB3, or, alternatively, included within a RRCReconfiguration message embedded in a RRCReconfiguration message received via SRB1.

In this case, the UE maintains two independent VarMeasConfig and VarMeasReportList, one associated with each measConfig, and independently performs all the procedures in clause 5.5 for each measConfig and the associated VarMeasConfig and VarMeasReportList, unless explicitly stated otherwise.

The configurations related to CBR measurements are only included in the measConfig associated with MCG.

The configurations related to Rx-Tx time difference measurement are only included in the measConfig associated with MCG.

Examples of events for measurement reporting are described.

Event A1 (Serving becomes better than threshold)

Event A2 (Serving becomes worse than threshold)

Event A3 (Neighbour becomes offset better than SpCell)

Event A4 (Neighbour becomes better than threshold)

Event A5 (SpCell becomes worse than threshold1 and neighbour becomes better than threshold2)

Event A6 (Neighbour becomes offset better than SCell)

Event B1 (Inter RAT neighbour becomes better than threshold)

Event B2 (PCell becomes worse than threshold1 and inter RAT neighbour becomes better than threshold2)

Event I1 (Interference becomes higher than threshold)

Event C1 (The NR sidelink channel busy ratio is above a threshold)

Event C2 (The NR sidelink channel busy ratio is below a threshold)

Event D1 (Distance between UE and referenceLocation1 is above threshold1 and distance between UE and referenceLocation2 is below threshold2)

CondEvent T1 (Time measured at UE is within a duration from threshold)

Event X1 (Serving L2 U2N Relay UE becomes worse than threshold1 and NR Cell becomes better than threshold2)

Event X2 (Serving L2 U2N Relay UE becomes worse than threshold)

Event Y1 (PCell becomes worse than threshold1 and candidate L2 U2N Relay UE becomes better than threshold2)

Event Y2 (Candidate L2 U2N Relay UE becomes better than threshold)

FIG. 24 shows an example of measurement reporting.

The purpose of this procedure is to transfer measurement results from the UE to the network. The UE shall initiate this procedure only after successful AS security activation.

FIG. 25 shows an example of location measurement indication.

The purpose of this procedure is to indicate to the network that the UE is going to start/stop location related measurements towards E-UTRA or NR (eutra-RSTD, nr-RSTD, nr-UE-RxTxTimeDiff, nr-PRS-RSRP) which require measurement gaps or start/stop detection of subframe and slot timing towards E-UTRA (eutra-FineTimingDetection) which requires measurement gaps. UE shall initiate this procedure only after successful AS security activation.

Hereinafter technical features related to an AI/ML model are described.

For example, for the existing (under discussion) AI/ML use cases, proprietary models may be supported and/or open format may be supported.

For example, from Management or Control point of view mainly some meta info about a model may need to be known.

For example, a model is identified by a model ID.

For example, a model ID can be used to identify which AI/ML model is being used in Life Cycle Management (LCM) including model delivery.

For example, a model ID can be used to identify a model (or models) during model selection/activation/deactivation/switching.

For example, the wording "model transfer/delivery" may be used.

For example, aim to at least analyze the feasibility and benefits of model/transfer solutions based on the following:

- gNB can transfer/deliver AI/ML model(s) to UE via RRC signalling.

- CN (except LMF) can transfer/deliver AI/ML model(s) to UE via NAS signalling.

- LMF can transfer/deliver AI/ML model(s) to UE via LPP signalling.

- gNB can transfer/deliver AI/ML model(s) to UE via UP data.

- CN (except LMF) can transfer/deliver AI/ML model(s) to UE via UP data.

- LMF can transfer/deliver AI/ML model(s) to UE via UP data.

- Server (e.g. OAM, OTT) can transfer/delivery AI/ML model(s) to UE (e.g. transparent to 3GPP).

For example, Model ID is unique "globally", e.g. in order to manage test certification each retrained version need to be identified.

For the CSI compression and beam management use cases, model/function selection/(de)activation/switching/fallback can be UE-initiated or gNB-initiated. For the positioning use case, model/function selection/(de)activation/switching/fallback can be UE-initiated or LMF-/ gNB-initiated.

For example, Information such as FFS:vendor info, applicable conditions, model performance indicators, etc. may be required for model management and control, and should, as a starting point, be part of meta information.

The general AI/ML framework consist of, (i) Data Collection, (ii) Model Training, (iii) Model Management, (iv) Model Inference, and (v) Model Storage.

Model ID can be used to identify model or models for the following LCM purposes:

model selection/activation/deactivation/switching (or identification, if that will be supported as a separate step).

(for example, for so called "model ID based LCM")

If model transfer/delivery is supported, model ID can be used for model transfer/delivery LCM purpose.

Extend the previously endorsed table with 3 columns: Inference, Monitoring and Training, and explain in free text the applicability of the data collection method to the LCM purpose and the use case(s).

For example, intention is to cover functional arch in general, e.g. covering both be model based and/or functionality based LCM

"Model Storage" in the figure is only intended as a reference point (if any) for protocol terminations etc for model transfer/delivery etc. It is not intended to limit where models are actually stored. Add a note for this.

Remove "Model" in Model Managemt and Model Inference and for the actions/the arrow form Management to Inference (to reduce the risk for misunderstanding).

Management may be model based management, or functionality based management. Add a mote for this.

For example, that for the data collection in some scenarios (e.g., internal data up to implementation or the existing data are enough), possibly no RAN2 specification effort is needed in some scenarios, e.g. (not exhaustive):

- For model inference of UE-sided model, input data for model inference is available inside the UE.

- For UE-side (real time) monitoring of UE-sided model, performance metrics are available inside the UE. UE can independently monitor a model's performance without any data input from NW.

For CSI enhancement and beam management use cases:

- For model training, training data can be generated by UE/gNB and terminated at gNB/OAM/OTT server.

- For NW-sided model inference, input data can be generated by UE and terminated at gNB.

- For UE-side model inference, input data/assistance information can be generated by gNB and terminated at UE.

- For model monitoring at NW side, performance metrics can be generated by UE and terminated at gNB.

For positioning enhancement use case:

- For model training, training data can be generated by UE/gNB and terminated at LMF/OTT server.

- For NW-sided model inference, input data can be generated by UE/gNB and terminated at LMF and/or gNB.

- For UE-side model inference, input data/assistance information can be generated by LMF/gNB and terminated at the UE.

- For model monitoring at NW side, performance metrics can be generated by UE/gNB and terminated at LMF.

Architecture and General (Functional Architecture)

The Table 5 can be used as starting point for discussion on mapping of AI/ML functions to physical entities for CSI compression with two-sided model.

Table 5 shows the mapping of functions to physical entities for CSI compression with two-sided model.

	AL/ML functions (if applicable)	Mapped entities
a)	Model training(offline training)	gNB, OAM, OTT server, UE,
b)	Model transfer/delivery	For training Type 1: gNB->UE, or OAM->gNB&UE, or OTT server->gNB&UE, or UE->gNB, For training Type 3: For UE part of two-sided model: OTT server->UE,; For NW part of two-sided model: OAM->gNB;
c)	Inference	NW part of two-sided model: gNBUE part of two-sided model: UE
d)	Model/functionality monitoring	NW-side: NW monitors the performanceUE-side: UE monitors the performance and may report to NW
e)	Model/functionality control (selection, (de)activation, switching, updating, fallback)	gNB,

For a), only data collection part may be further discussed, how to perform the model training is up to implementation.

For b), no model transfer/delivery is expected if the entity for model training and model inference is the same one.

Whether/how OAM is to be involved may need to consult RAN3, SA5.

Whether/how CN is to be involved may need to consult RAN3, SA2.

For beam management:

The Table 6 can be used as starting point for discussion on mapping of AI/ML functions to physical entities for beam management with UE -side model.

Table 6 shows the mapping of AI/ML functions to physical entities for beam management with UE-side model.

	AL/ML functions (if applicable)	Mapped entities
a)	Model training(offline training)	UE-side OTT server, UE
b)	Model transfer/delivery	UE-side OTT server->UE
c)	Inference	UE
d)	Model/functionality monitoring	UE (UE monitors the performance, and may report to gNB), gNB (gNB monitors the performance)
e)	Model/functionality control (selection, (de)activation, switching, fallback)	gNB if monitoring resides at UE or gNB, UE if monitoring resides at UE

Whether/how OAM is to be involved may need to consult RAN3, SA5.

Whether/how CN is to be involved may need to consult RAN3, SA2.

The Table 7 can be used as starting point for discussion on mapping of AI/ML functions to physical entities for beam management with NW-side model.

Table 7 shows the mapping of functions to physical entities for beam management with NW-side model.

	AL/ML functions (if applicable)	Mapped entities
a)	Model training (offline training)	gNB, OAM
b)	Model transfer/delivery	OAM->gNB
c)	Inference	gNB
d)	Model/functionality monitoring	gNB
e)	Model/functionality control (selection, (de)activation, switching, fallback)	gNB

Whether/how OAM is to be involved may need to consult RAN3, SA5.

Whether/how CN is to be involved may need to consult RAN3, SA2.

For Positioning accuracy enhancement:

The Table 8 can be used as starting point for discussion on mapping of AI/ML functions to physical entities for positioning with UE -side model (case 1 and 2a).

Table 8 shows the mapping of functions to physical entities for positioning with UE-side model (case 1 and 2a).

Use case	AL/ML functions (if applicable)	Mapped entities
a)	Model training (offline training)	UE-side OTT server, UE
b)	Model transfer/delivery	UE-side OTT server->UE
c)	Inference	UE
d)	Model/functionality monitoring	UE, LMF
e)	Model/functionality control (selection, (de)activation, switching, fallback)	UE if monitoring resides at UE, LMF if monitoring resides at UE or LMF

Whether/how OAM is to be involved may need to consult RAN3, SA5.

Whether/how CN/LMF is to be involved may need to consult RAN3, SA2.

R2 confirms that for UE-side AIML functionality, current UE capability reporting by RRC and LPP and handling is expected to be used, at least for static capabilities.

AIML algorithm for a certain use case may be tailored towards and applicable to certain scenarios/location/configuration/deployment etc. AIML algorithm may be updated, e.g. by model change (these are observations):

- RAN2 assumes that for UE-side AIML, the UE informs the RAN about applicability conditions of AIML algorithm(s) available to the UE, to support RAN control (e.g. activation/deactivation/switching).

- RAN2 further observes that current UE capability reporting and handling is designed for Capabilities that do not dynamically change. The procedure for UE reporting of AIML applicability conditions is FFS.

Model Transfer/Delivery

Model transfer/delivery can be initiated in following two ways:

- Reactive model transfer/delivery: an AI/ML model is downloaded when it is needed due to changes in scenarios, configurations, or sites.

- FFS: Proactive model transfer/delivery: AI/ML models are pre-download to UE, and a model switch is performed when changes in scenarios, configurations, or sites occur.

Architecture and General (Functional Architecture)

1. The legacy UE capability framework serves as the baseline to report UE's supported AI/ML-enabled Feature/FG:

- For CSI and beam management use cases, it is indicated in UE AS capability in RRC (i.e., UECapabilityEnquiry/UECapabilityInformation).

- For positioning use case, it is indicated in positioning capability in LPP.

2. RAN2 confirm that stage 3 details of AI/ML-enabled Feature/FG (e.g. granularity of Feature/FG) in legacy UE capability are postponed to discuss in the normative phase.

3. For additional condition reporting, the existing capability reporting framework cannot be used. To report these conditions (if needed), UAI can be used as an example. This can be defined and discussed in normative phase. FSS signaling of additional conditions from network to UE

4. Capture in the TR the reactive and proactive approaches, i.e., the UE reacts to NW's configuration, or the UE proactively informs the NW of updates/changes to its supported models/functionalities. Review the definition by email during TP review phase.

Data Collection

On NW-side data collection

For CSI and beam management

1. For training of NW-side models, both gNB- and OAM-centric data collection are considered in the study.

2. For training of NW-side models, the gNB-centric data collection implies that the gNB configures the UE to initiate/terminate the data collection procedure. To further study the details of the data collection configuration

3. For training of NW-side models, an OAM-centric data collection implies that the OAM provides the configuration (via the gNB) needed for the UE to initiate/terminate the data collection procedure. MDT framework can be considered.

4. Related to gNB-centric data collection for NW-side model training, RAN2 studies the potential impact on L3 signalling for the reporting of collected data, taking into account RAN1 further inputs/progress.

5. Related to OAM-centric data collection for NW-side model training, RAN2 studies the potential impact at on the MDT for connected mode, taking into account RAN1 further inputs/progress

Positioning

For LMF sided inference (case 2b, case 3b), RAN2 assumes LPP protocol should be applied to the data collected by UE and terminated at LMF, while the NRPPa protocol should be applied to the data collected by gNB and terminated at LMF.

For LMF sided performance monitoring, RAN2 assumes LPP protocol should be applied to the data collected by UE and terminated at LMF, while the NRPPa protocol should be applied to the data collected by gNB and terminated at LMF.

General

Principles in proposal 4 and 9 will be captured as one combined set of principles for NW-side data collection:

logging is supported

periodic, event based reporting, on demand report

The UE memory, processing power, energy consumption, signalling overhead should be taken into account.

The above principles, can be revised depending on RAN1 progress/requirements

Architecture and General

1. For CN and OAM FFSs, we will remove it and add a NOTE indicating that it was identified but RAN2 didn't study as it is out of scope of RAN2

2. For the following FFS: LMF and gNB, and CSI compression for UE control, we will remove it and add a NOTE indicating that it was identified but RAN2 didn't study or conclude as it depends on RAN1 progress

3. Update TP to indicated that CSI prediction use case for the functional mapping is the same as beam management for UE side model

4. Capture the following text:

The following proposals were discussed in RAN2

UE collects and directly transfers training data to the OTT server

1a) OTT (3GPP transparent)

1b) OTT (non-3GPP transparent)

UE collects training data and transfers it to CN. CN transfers the training data to the OTT server.

UE collects training data and transfers it to OAM. OAM transfers the needed data to the OTT server.

RAN2 did not study or analyze the proposals and did not agree to requirements or recommendations.

Meanwhile, UE can evaluate and monitor the applicability and expected performance of an inactive model/functionality for purpose of activation/selection/switching of UE-side models/UE-part of two-sided models/functionalities. UE can evaluate applicability and performance based on additional conditions, input/output data distribution, monitoring conditions, and the inference accuracy.

Thus, studies for monitoring inference function are required.

Hereinafter, a method for monitoring inference function, according to some embodiments of the present disclosure, will be described with reference to the following drawings.

The following drawings are created to explain specific embodiments of the present disclosure. The names of the specific devices or the names of the specific signals/messages/fields shown in the drawings are provided by way of example, and thus the technical features of the present disclosure are not limited to the specific names used in the following drawings. Herein, a wireless device may be referred to as a user equipment (UE).

FIG. 26 shows an example of a method for monitoring inference function.

In particular, FIG. 26 shows an example of a method performed by a wireless device in a wireless communication system.

In step S2601, a wireless device may receive, from a network, a configuration related to at least one inference function.

For example, the configuration may include information related to at least one validity condition.

For example, the configuration related to at least one inference function may include information related to at least one reporting condition.

For example, the wireless device may initiate transmission of the monitoring report based on the at least one reporting condition being satisfied.

For example, the information related to at least one reporting condition may include periodic-based reporting condition and/or event-based reporting condition.

For example, the information related to at least one validity condition may include information related to valid time and/or information related to valid area.

In step S2602, a wireless device may activate a first inference function.

For example, the wireless device may activate the first inference function and deactivate a second inference function.

In step S2603, a wireless device may determine that monitoring results related to the first inference function does not satisfy the at least one validity condition.

For example, the wireless device may determine that the monitoring results related to the first inference function does not satisfy the at least one validity condition based on the monitoring results related to the first inference function.

For example, the wireless device may determine that the monitoring results related to the first inference function does not satisfy the at least one validity condition based on the information related to valid time and/or the information related to valid area.

In step S2604, a wireless device may transmit, to the network, a monitoring report related to a second inference function which is not activated.

For example, the monitoring report may include information related to performance of the second inference function. For example, information related to performance of the second inference function may include information related to intermediate results of the second inference function, eventual results of the second inference function, and/or data distribution related to the second inference function.

For example, the monitoring report may include information related to applicability of the second inference function.

For example, the wireless device may evaluate the second inference function which is not activated based on that the monitoring results related to the first inference function does not satisfy the at least one validity condition. For example, the wireless device may not evaluate the inactive second inference function while the monitoring results related to the first inference function satisfies the at least one validity condition. When the monitoring results related to the first inference function becomes not satisfying the at least one validity condition, the wireless device may initiate evaluating the second inference function. When the evaluation of the second inference function is performed, the wireless device may transmit the monitoring report.

For example, the monitoring report related to the second inference function may be transmitted to the network based on that the monitoring results related to the first inference function does not satisfy the at least one validity condition. In this example, the wireless may perform evaluation of the second inference function when the monitoring results related to the monitoring results related to the first inference function satisfied the at least one validity condition. The wireless device may transmit the monitoring report only when the monitoring results related to the first inference function does not satisfy the at least one validity condition.

For example, based on that the monitoring results related to the first inference function satisfies the at least one validity condition, the wireless device may skip reporting the monitoring report related to the second inference function.

For example, the wireless device may evaluate the first inference function which is activated. The wireless device may transmit a monitoring report related to the first inference function based on at least one reporting condition being satisfied. For example, the monitoring report may include both monitoring results for the first inference function and monitoring results for the second inference function.

For example, the wireless device may receive, from the network, at least one management instruction related to the first inference function and/or the second inference function.

According to some embodiments of the present disclosure, the wireless device may be in communication with at least one of a user equipment, a network, or an autonomous vehicle other than the wireless device.

Hereinafter, some embodiments of a method for inactive model monitoring based on active model validity.

For example, a method of inactive model monitoring based on active model validity is provided. Active model validity can be determined by active model accuracy (based on performance KPI, inference accuracy, input/output data distribution, and applicable condition, etc), and/or active model valid time/location. UE may utilize the active model validity (i) when evaluating the inactive model and/or (ii) when reporting the inactive model. Based on the UE's report, NW may determine suitable model and management instruction, e.g., model (de)activation, switch, update, fallback, etc, and/or suitable model condition/configuration.

In particular, FIG. 27 shows an example of a method performed by a user equipment (UE) and a network (NW) in a wireless communication system.

In step S2701, NW may configure model validity related information and at least one model and/or functionality information, wherein the model validity related information may include applicability related condition, monitoring related condition, valid information.

For example, UE may receive model validity configuration from network.

In step S2702, UE may activate at least one model or functionality.

For example, at least one model or functionality may be activated.

In step S2703 and S2704, UE may evaluate inactive model and report the inactive model evaluation results based on model validity related information.

Case 1 Evaluation based on model validity related information

In step S2703a, UE may evaluate active model based on model validity related information. Depending on the evaluation results of the active model/functionality.

There may be two types of evaluation of inactive model(s)/functionality(ies):

- Option 1. Reactive way: When active model is regarded as invalid, UE may start evaluating the inactive model(s)

- Option 2. Proactive way: When it appears that the active model will be regarded as invalid within a certain time, UE may start evaluating the inactive model(s)

In step S2704a, UE may report the inactive model evaluation results.

Case 2 Reporting based on model validity related information

In step S2703b, UE may evaluate the inactive model/functionality.

In step S2704b, UE may report the inactive model evaluation results of inactive model/functionality based on model validity related information.

There may also be two types of reporting of inactive model(s):

- Option 1. Reactive way: When active model is regarded as invalid, UE may report the inactive model's evaluation results

- Option 2. Proactive way: When it appears that the active model will be regarded as invalid within a certain time, UE may report the inactive model's evaluation results

In step S2705, NW may determine functionality/model selection, and/or functionality/model management instruction, such as model/functionality (de)activation, switching, update, fallback, etc.

In step S2706, based on determination, NW may configure model/ functionality management instruction.

For example, UE may receive, from NW, configuration including management instruction for inference function (model/functionality).

Regarding step S2701,

1> NW may configure model validity related information and at least one model and/or functionality information, wherein the model validity related information may include applicability related condition, monitoring related condition, valid information

Regarding applicability related condition

2> Additional conditions may include conditions under which the UE can perform functionality/model related operations

3> Additional condition may include following information (KPI):

4> e.g., Specific location related information (e.g., polygon type, latitude/longitude, altitude, angle, indoor/outdoor, etc)

4> e.g., Specific time related information (e.g., date, time window, start time, stop time, etc)

4> e.g., Specific speed related information (e.g., 10km/h, 30km/h, 60km/h, 120km/h, etc)

4> e.g., Specific radio quality condition (e.g., RSRP, RSRQ, SINR, etc)

4> e.g., Specific deployment scenario (e.g., UMa, UMi, InH, etc)

4> e.g., Specific cell/frequency related information, (e.g., bandwidth, size of subband, carrier frequency, numerologies, etc)

4> e.g., Specific antenna related information (e.g., antenna port layouts, antenna port numbers, rank numbers/layers, antenna spacing, antenna virtualization, etc)

3> Additional condition may consist of one condition or a combination of several conditions

3> Additional condition may have a specific condition ID

3> Additional condition may be linked to a specific functionality/functionality group or a specific model/model group

3> Additional condition may include periodic-based report configuration, e.g., periodic report related to KPI

3> Additional condition may include event-based report configuration, e.g, report when (not) satisfying KPI related threshold/condition

Regarding monitoring related condition

2> Monitoring conditions may include conditions to evaluate model/functionality accuracy

3> Performance KPI monitoring

4> Prediction Accuracy, e.g., Beam prediction accuracy related KPIs such as Top-K/1 beam prediction accuracy

4> Eventual KPIs, e.g., Throughput, BLER, NACK/ACK, L1/L3-RSRP, L1/L3-SINR

4> Other solutions, e.g., Input/output data-based monitoring (such as data drift between training dataset and observed dataset and out-of-distribution detection), L1-RSRP difference evaluated by comparing measured RSRP and predicted RSRP

4> Legacy operation-based monitoring: e.g., schemes using additional legacy CSI reporting

4> Benchmark/reference for the performance comparison, including:

5> e.g., The best beam(s) obtained by measuring beams of a set indicated by gNB (e.g., beams from Set A)

5> e.g., Measurement of the predicted best beam(s) corresponding to model output (e.g., comparison between actual L1-RSRP and predicted RSRP of predicted Top-1/K Beams)

3> Intermediate KPI based model monitoring

4> Intermediate KPIs, e.g., SGCS (Squared Generalized Cosine Similarity)

4> e.g., NW-side monitoring based on the target CSI with realistic channel estimation associated to the CSI report, reported by the UE or obtained from the UE-side.

4> e.g., UE-side monitoring based on the output of the CSI reconstruction model, subject to the aligned format, associated to the CSI report, indicated by the NW or obtained from the network side.

5> Network may configure a threshold criterion to facilitate UE to perform model monitoring.

4> e.g., UE-side monitoring based on the output of the CSI reconstruction model at the UE-side

5> Note: CSI reconstruction model at the UE-side can be the same or different comparing to the actual CSI reconstruction model used at the NW-side. Network may configure a threshold criterion to facilitate UE to perform model monitoring.

3> Monitoring related condition may consist of one condition or a combination of several conditions

3> Monitoring related condition may have a specific condition ID

3> Monitoring related condition may be linked to a specific functionality/functionality group or a specific model/model group

3> Monitoring related condition may include periodic-based report configuration, e.g., periodic report related to KPI

3> Monitoring related condition may include event-based report configuration, e.g, report when (not) satisfying KPI related threshold/condition

Regarding valid information

2> The information may include time related information, e.g.,

3> The expiration date and/or time

3> The valid time window,

4> e.g., start date/time and stop date/time; or

4> e.g., valid duration

2> The information may include location related information,

3> Reference point information, e.g., polygon type, latitude/longitude, altitude, angle, indoor/outdoor, etc)

3> Reference cell/frequency/reference signal/SSB/BWP information

2> Distance threshold information, e.g., a distance threshold

2> Time threshold information, e.g., a time threshold

2> Valid information may be delivered during model identification, transfer, configuration, etc.

2> Valid information may consist of one information or a combination of time and location information

2> Valid information may have a specific ID

2> Valid information may be linked to a specific functionality/functionality group or a specific model/model group

1> NW may configure a monitoring time for inactive model(s)/functionality(ies)

Regarding step 2702,

1> At least one model or functionality may be activated

2> Model or Functionality may be activated/switched by UE decision or NW decision

2> In case of UE decision, UE may inform NW of UE decision, i.e., activation/switching

3> UE may perform UE decision by its own. After performing UE decision, UE may inform NW of its decision

3> UE may perform UE decision when NW confirm it.

2> In case of NW decision, NW may request UE to do NW decision, i.e., activation/switching

Case 1 Inactive Model Evaluation based on Active Model Validity related information

Regarding step S2703a ,

1> UE may evaluate active model based on model validity related information. Depending on the evaluation results of the active model/functionality, there may be two types of evaluation of inactive model(s)/functionality(ies):

2> Option 1. Reactive way: When active model/functionality is regarded as invalid, UE may start evaluating the inactive model(s)/functionality(ies)

Active model's validity based on current model performance

3> E.g., Model performance based: UE may evaluate whether the active model/functionality's accuracy is below a threshold or does not satisfy a condition based on monitoring related condition and/or applicability related condition

In FIG. 28, UE starts evaluating the inactive model based on active model performance.

Active model's validity based on valid information

3> E.g., Model validity information based: UE may evaluate whether the active model/functionality is invalid based on the valid information

4> If the current time is not within the time specified by time related information, UE may consider the active model/functionality is invalid

5> e.g., when expiration date/time is expired

5> e.g., when current time, t1, is out of valid time window

5> e.g., when the time threshold amount of time has passed from expiration date/time or valid time window

4> If the current location is not available at the location specified by location related information, UE may consider the active model/functionality is invalid

5> e.g., when current location, l1, is out of valid reference point/cell/frequency/reference signal/SSB/BWP

5> e.g., when the distance threshold amount of location has passed from reference point/cell/frequency/reference signal/SSB/BWP

5> e.g., when the time threshold amount of time has passed since the UE passed from reference point/cell/frequency/reference signal/SSB/BWP

2> Option 2. Proactive way: When it appears that the active model/functionality will be regarded as invalid within a specific time, i.e., for the future time, UE may start evaluating the inactive model(s)/functionality(ies)

Active model's validity based on predictive model performance

3> E.g., Model performance based: Active model's validity in the future may be determined by the accuracy prediction for the model

4> UE may derive the active model/functionality's accuracy for T at t1, wherein t1 precedes T

4> If the active model/functionality's accuracy will be below a threshold or will not satisfy a condition based on monitoring related condition and/or applicability related condition for the future time (T), UE may consider the active model/functionality is invalid

5> T may be configured by NW

4> When deriving active model/functionality accuracy prediction, UE may utilize AIML operation. UE may be configured with a more prediction model/functionality configuration. Based on the AIML model operation, UE may derive accuracy prediction.

4> The AIML model configuration may include prediction model structure information,

5> Network may configure a machine learning model to be used by UE.

6> Network may include a machine learning type, such as reinforcement learning, supervised learning, or unsupervised learning.

6> Network may include a machine learning model, such as DNN, CNN, RNN, and DRL.

6> The configured ML model may be a pre-trained ML model that has been already trained by network

7> The configured ML model is described by a model description information including model structure and parameters.

7> For example, neural-network based model may comprise input layer, output layer, and hidden layer(s), where each layer comprises one or more neurons (equivalently nodes).

8> Different layers are connected based on the connections between neurons of different layers

8> Each connection of two different neurons in two different layers may be directive (e.g. neuron A to neuron B, meaning that the output of neuron A is fed into the neuron B)

8> Each neuron may provide input to one or several connected neurons (1 to N connection).

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

8> Each neuron may take input from one or several connected neurons (N to 1 connection), and combines the input from the connected neurons, and produces an output based on activation function.

6> The configured ML model may be a ML model to be trained.

7> The configured ML model is described by a model description information including model structure and initial parameters that are to be trained.

7> When network configures the ML model to be trained, it may also configure training parameters such as optimization objective(s) and optimization-related configuration parameters.

6> Network may include machine learning input parameters for the machine learning model, such as UE location information, radio measurements related to serving cell and neighbouring cells, UE mobility history.

6> Network may include machine learning output, such as UE trajectory prediction, predicted target cell, prediction time for handover, and UE traffic prediction.

5> UE may perform a machine learning model training, validation, and testing which may generate model performance metrics based on the prediction model configuration.

6> UE may perform a model training with the machine learning input parameters.

5> UE may use the configured ML model to perform ML task such as predictions of measurements.

6> UE may derive machine learning output(s).

6> UE may infer from the outputs and use the outputs as feedback for the machine learning model.

5> UE may send feedback to network about the results related to machine learning outputs and the accuracy of the machine learning model.

6> Network may update the machine learning model and parameters related to the machine learning model.

Active model's validity based on valid information

3> E.g., Model validity information based: Active model's validity in the future may be determined by active model validity based on valid information.

4> If the remaining time is equal or below a time threshold, UE may consider the model/functionality is invalid

5> e.g., remaining time = expiration date/time - current time

5> e.g., remaining time = stop date/time - current time

5> e.g., remaining time = start date/time + duration - current time

5> e.g., remaining time = time difference from reference point/cell/frequency/reference signal/SSB/BWP coverage to current location, when UE is within the reference coverage

4> If the remaining distance is equal or below a location threshold, UE may consider the model/functionality is invalid

5> e.g., remaining distance = reference point/cell/frequency/reference signal/SSB/BWP coverage - current location, when UE is within the reference point coverage

In FIG. 29, UE starts evaluating the inactive model before expiring valid time of active model.

2> UE may start evaluating the inactive model(s)/functionality(ies) when active model/functionality is considered as invalid

2> UE may evaluate the inactive model(s)/functionality(ies) based on additional related condition, monitoring related condition, and/or valid information associated with corresponding inactive model(s)/functionality(ies)

2> If the monitoring time for inactive model(s)/functionality(ies) is configured in step S2701, UE may evaluate the inactive model(s)/functionality(ies) during the monitoring time

2> UE may select the inactive model(s)/functionality(ies) for evaluation considering the valid time information for each inactive model(s)/functionality(ies)

3> e.g., UE may order/select inactive model/functionality which has the longest remaining valid time

2> UE may select the inactive model(s)/functionality(ies) for evaluation considering the priority of each inactive model(s)/functionality(ies)

3> Priority may be configured by NW

2> e.g., UE may order/select inactive model/functionality which has the highest priority

Regarding step S2704a ,

1> UE may report the current model/functionality and/or the inactive model(s)/functionality(ies) evaluation results periodically or event-based depending on NW configuration

2> Report may be an applicability related report

2> Report may be a monitoring related report

Case 2 Reporting based on Active Model Validity related information

Regarding Step S2703b ,

1> UE may evaluate the inactive model(s)/functionality(ies) based on additional related condition, monitoring related condition, and/or valid information associated with corresponding inactive model(s)/functionality(ies)

Regarding Step S2704b ,

1> UE may report the current model/functionality and/or the inactive model evaluation results of inactive model/functionality based on model validity related information. There may also be two types of reporting of inactive model(s):

2> Option 1. Reactive way: When active model is regarded as invalid, UE may report the inactive model's evaluation results

Active model's validity based on current model performance

Active model's validity based on valid information

5> e.g., when expiration date/time is expired

5> e.g., when current time, t1, is out of valid time window

2> Option 2. Proactive way: When it appears that the active model will be regarded as invalid within a certain time, UE may report the inactive model's evaluation results

Regarding active model's validity based on predictive model performance

5> T may be configured by NW

5> Network may configure a machine learning model to be used by UE.

6> The configured ML model may be a ML model to be trained.

6> UE may perform a model training with the machine learning input parameters.

6> UE may derive machine learning output(s).

Regarding active model's validity based valid information

5> e.g., remaining time = expiration date/time - current time

5> e.g., remaining time = stop date/time - current time

e.g., remaining time = start date/time + duration - current time

2> For option 1 and option 2, UE may report the inactive model(s)/functionality(ies) evaluation results when the active model/functionality is considered as invalid

3> Report may be an applicability related report

3> Report may be a monitoring related report

2> UE may report the inactive model(s)/functionality(ies) evaluation results periodically or event-based depending on NW configuration

2> If the monitoring time for inactive model(s)/functionality(ies) is configured in step S2701, UE may report the inactive model(s)/functionality(ies) during the monitoring time

2> UE may select the inactive model(s)/functionality(ies) for report considering the valid time information for each inactive model(s)/functionality(ies)

3> e.g., UE may report evaluation results for inactive model/functionality which has the longest remaining valid time

2> UE may select the inactive model(s)/functionality(ies) for report considering the priority of each inactive model(s)/functionality(ies)

3> Priority may be configured by NW

3> e.g., UE may report evaluation results for inactive model/functionality which has the highest priority

Regarding step S2705,

1> NW may determine functionality/model management related results, such as model (de)activation, switching, update, fallback, etc

1> NW may determine whether the functionality/model configuration needs to be updated

Regarding step S2706,

1> Based on determination in step S2704, NW may configure functionality/model management related results

1> Based on determination in step S2704, NW may reconfigure functionality/model.

Alternatively, after the step S2703, UE may decide on model/functionality management by its own. (i) UE may perform model/functionality management related operation, such as (de)activation, fallback, switch, etc. After performing management related operation, UE may inform NW of the UE decision-based operation. Or (ii) UE may inform NW of UE's decision. NW may confirm (accept or reject) UE's decision, and UE may follow the NW's confirmation.

Alternatively, the UE may transmit model/function applicability information based on additional conditions related to the model training environment. Applicability information may include whether a particular model/function that is not activated is applicable in the current environment or situation.

In particular, FIG. 30 shows an example of a method performed by a wireless device in a wireless communication system.

In step S3001, the wireless device may receive information on a configuration related to an Artificial Intelligence (AI) and/or Machine Learning (ML) model, wherein the configuration includes monitoring report condition and model validity condition.

In step S3002, the wireless device may activate a first model.

In step S3003, the wireless device may evaluate the monitoring results of the first model.

In step S3004, the wireless device may report the monitoring results based on the monitoring report condition.

In step S3005, the wireless device may evaluate the monitoring results of a second model if the monitoring results of the first model does not satisfy model validity condition.

In step S3006, the wireless device may report the monitoring results of the second model based on monitoring report condition.

For example, monitoring results may include performance related results, such as intermediate results (e.g., SGCS (Squared Generalized Cosine Similarity)), eventual results (e.g., throughput, BLER, NACK/ACK rate, measurement results), and data distribution related results.

For example, the report may include the monitoring results of the first model.

Some of the detailed steps shown in the examples of FIGS. 26-30 may not be essential steps and may be omitted. In addition to the steps shown in FIGS. 26-30, other steps may be added, and the order of the steps may vary. Some of the above steps may have their own technical meaning.

Hereinafter, an apparatus for monitoring inference function, according to some embodiments of the present disclosure, will be described. Herein, the apparatus may be a wireless device (100 or 200) in FIGS. 2, 3, and 5.

For example, a wireless device may perform the methods described above. The detailed description overlapping with the above-described contents could be simplified or omitted.

Referring to FIG. 5, a wireless device 100 may include a processor 102, a memory 104, and a transceiver 106.

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

The processor 102 may be adapted to perform operations. The operations comprise: receiving, from a network, a configuration related to at least one inference function, wherein the configuration includes information related to at least one validity condition; activating a first inference function; determining that monitoring results related to the first inference function does not satisfy the at least one validity condition; and transmitting, to the network, a monitoring report related to a second inference function which is not activated.

For example, the configuration related to at least one inference function includes information related to at least one reporting condition.

For example, the operations further comprises: initiating transmission of the monitoring report based on the at least one reporting condition being satisfied.

For example, the information related to at least one reporting condition includes periodic-based reporting condition and/or event-based reporting condition.

For example, the operations further comprises: evaluating the second inference function which is not activated based on that the monitoring results related to the first inference function does not satisfy the at least one validity condition.

For example, the monitoring report related to the second inference function is transmitted to the network based on that the monitoring results related to the first inference function does not satisfy the at least one validity condition.

For example, the monitoring report includes information related to performance of the second inference function.

For example, the information related to performance of the second inference function includes information related to intermediate results of the second inference function, eventual results of the second inference function, and/or data distribution related to the second inference function.

For example, the monitoring report includes information related to applicability of the second inference function.

For example, the operations further comprises: based on that the monitoring results related to the first inference function satisfies the at least one validity condition, skipping reporting the monitoring report related to the second inference function.

For example, the operations further comprises: evaluating the first inference function which is activated; and transmitting a monitoring report related to the first inference function based on at least one reporting condition being satisfied.

For example, the information related to at least one validity condition includes information related to valid time and/or information related to valid area.

For example, the operations further comprises: receiving, from the network, at least one management instruction related to the first inference function and/or the second inference function.

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

Hereinafter, a processor for a wireless device for monitoring inference function, according to some embodiments of the present disclosure, will be described.

The processor may be configured to control the wireless device to perform operations. The operations comprise: receiving, from a network, a configuration related to at least one inference function, wherein the configuration includes information related to at least one validity condition; activating a first inference function; determining that monitoring results related to the first inference function does not satisfy the at least one validity condition; and transmitting, to the network, a monitoring report related to a second inference function which is not activated.

For example, the processor may be configured to control the wireless device to be in communication with at least one of a user equipment, a network, or an autonomous vehicle other than the wireless device.

Hereinafter, a non-transitory computer-readable medium has stored thereon a plurality of instructions for monitoring inference function, according to some embodiments of the present disclosure, will be described.

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

Some example of storage medium is coupled to the processor such that the processor can read information from the storage medium. In the alternative, the storage medium may be integral to the processor. The processor and the storage medium may reside in an ASIC. For other example, the processor and the storage medium may reside as discrete components.

The computer-readable medium may include a tangible and non-transitory computer-readable storage medium.

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 above.

In addition, the method described herein may be realized 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.

According to some embodiment of the present disclosure, a non-transitory computer-readable medium has stored thereon a plurality of instructions. The stored plurality of instructions may be executed by a processor of a wireless device.

The stored plurality of instructions may cause the wireless device to perform operations.

The operations comprise: receiving, from a network, a configuration related to at least one inference function, wherein the configuration includes information related to at least one validity condition; activating a first inference function; determining that monitoring results related to the first inference function does not satisfy the at least one validity condition; and transmitting, to the network, a monitoring report related to a second inference function which is not activated.

According to some embodiments of the present disclosure, the stored plurality of instructions may cause the wireless device to be in communication with at least one of a user equipment, a network, or an autonomous vehicle other than the wireless device.

Hereinafter, a method performed by a base station (BS) for monitoring inference function, according to some embodiments of the present disclosure, will be described.

The BS may transmit, to the wireless device, a configuration related to at least one inference function, wherein the configuration includes information related to at least one validity condition. The wireless device activates a first inference function. The wireless device determines that monitoring results related to the first inference function does not satisfy the at least one validity condition. The BS may receive, from the wireless device, a monitoring report related to a second inference function which is not activated.

Hereinafter, a base station (BS) for monitoring inference function, according to some embodiments of the present disclosure, will be described.

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

The processor may be configured to control the transceiver to transmit, to the wireless device, a configuration related to at least one inference function, wherein the configuration includes information related to at least one validity condition. The wireless device activates a first inference function. The wireless device determines that monitoring results related to the first inference function does not satisfy the at least one validity condition. The processor may be configured to control the transceiver to receive, from the wireless device, a monitoring report related to a second inference function which is not activated.

The present disclosure can have various advantageous effects.

Claims in the present disclosure can be combined in a various way. For instance, technical features in method claims of the present disclosure can be combined to be implemented or performed in an apparatus, and technical features in apparatus claims can be combined to be implemented or performed in a method. Further, technical features in method claim(s) and apparatus claim(s) can be combined to be implemented or performed in an apparatus. Further, technical features in method claim(s) and apparatus claim(s) can be combined to be implemented or performed in a method. Other implementations are within the scope of the following claims.

Claims

A method, comprising:

receiving, by a wireless device from a network, a configuration related to at least one inference function, wherein the configuration includes information related to at least one validity condition;

activating, by the wireless device, a first inference function;

determining, by the wireless device, that monitoring results related to the first inference function does not satisfy the at least one validity condition; and

transmitting, by the wireless device to the network, a monitoring report related to a second inference function which is not activated.
The method of claim 1,

wherein the configuration related to at least one inference function includes information related to at least one reporting condition.
The method of claim 2, wherein the method further comprising:

initiating, by the wireless device, transmission of the monitoring report based on the at least one reporting condition being satisfied.
The method of claim 2,

wherein the information related to at least one reporting condition includes periodic-based reporting condition and/or event-based reporting condition.
The method of claim 1, wherein the method further comprising:

evaluating, by the wireless device, the second inference function which is not activated based on that the monitoring results related to the first inference function does not satisfy the at least one validity condition.
The method of claim 1,

wherein the monitoring report related to the second inference function is transmitted to the network based on that the monitoring results related to the first inference function does not satisfy the at least one validity condition.
The method of claim 1,

wherein the monitoring report includes information related to performance of the second inference function.
The method of claim 7,

wherein the information related to performance of the second inference function includes information related to intermediate results of the second inference function, eventual results of the second inference function, and/or data distribution related to the second inference function.
The method of claim 1,

wherein the monitoring report includes information related to applicability of the second inference function.
The method of claim 1, wherein the method further comprising:

based on that the monitoring results related to the first inference function satisfies the at least one validity condition, skipping, by the wireless device, reporting the monitoring report related to the second inference function.
The method of claim 1, wherein the method further comprising:

evaluating, by the wireless device, the first inference function which is activated; and

transmitting, by the wireless device, a monitoring report related to the first inference function based on at least one reporting condition being satisfied.
The method of claim 1,

wherein the information related to at least one validity condition includes information related to valid time and/or information related to valid area.
The method of claim 1, wherein the method further comprising:

receiving, by the wireless device from the network, at least one management instruction related to the first inference function and/or the second inference function.
The method of claim 1,

wherein the wireless device is in communication with at least one of a user equipment, a network, or an autonomous vehicle other than the wireless device.
A wireless device, comprising:

a transceiver;

a memory; and

at least one processor operatively coupled to the transceiver and the memory, and adapted to perform operations, the operations comprising:

receiving, from a network, a configuration related to at least one inference function, wherein the configuration includes information related to at least one validity condition;

activating a first inference function;

determining that monitoring results related to the first inference function does not satisfy the at least one validity condition; and

transmitting, to the network, a monitoring report related to a second inference function which is not activated.
The wireless device of claim 15,

wherein the configuration related to at least one inference function includes information related to at least one reporting condition.
The wireless device of claim 16, wherein the operations further comprising:

initiating transmission of the monitoring report based on the at least one reporting condition being satisfied.
The wireless device of claim 16,

wherein the information related to at least one reporting condition includes periodic-based reporting condition and/or event-based reporting condition.
The wireless device of claim 15, wherein the operations further comprising:

evaluating the second inference function which is not activated based on that the monitoring results related to the first inference function does not satisfy the at least one validity condition.
The wireless device of claim 15,

wherein the monitoring report related to the second inference function is transmitted to the network based on that the monitoring results related to the first inference function does not satisfy the at least one validity condition.
The wireless device of claim 15,

wherein the monitoring report includes information related to performance of the second inference function.
The wireless device of claim 21,

wherein the information related to performance of the second inference function includes information related to intermediate results of the second inference function, eventual results of the second inference function, and/or data distribution related to the second inference function.
The wireless device of claim 15,

wherein the monitoring report includes information related to applicability of the second inference function.
The wireless device of claim 15, wherein the operations further comprising:

based on that the monitoring results related to the first inference function satisfies the at least one validity condition, skipping reporting the monitoring report related to the second inference function.
The wireless device of claim 15, wherein the operations further comprising:

evaluating the first inference function which is activated; and

transmitting a monitoring report related to the first inference function based on at least one reporting condition being satisfied.
The wireless device of claim 15,

wherein the information related to at least one validity condition includes information related to valid time and/or information related to valid area.
The wireless device of claim 15, wherein the operations further comprising:

receiving, from the network, at least one management instruction related to the first inference function and/or the second inference function.
The wireless device of claim 15,

wherein the wireless device is in communication with at least one of a user equipment, a network, or an autonomous vehicle other than the wireless device.
A processor for a wireless device in a wireless communication system, wherein the processor is configured to control the wireless device to perform operations comprising:

receiving, from a network, a configuration related to at least one inference function, wherein the configuration includes information related to at least one validity condition;

activating a first inference function;

determining that monitoring results related to the first inference function does not satisfy the at least one validity condition; and

transmitting, to the network, a monitoring report related to a second inference function which is not activated.
A non-transitory computer-readable medium having stored thereon a plurality of instructions, which, when executed by a processor of a wireless device, cause the wireless device to perform operations, the operations comprising:

receiving, from a network, a configuration related to at least one inference function, wherein the configuration includes information related to at least one validity condition;

activating a first inference function;

determining that monitoring results related to the first inference function does not satisfy the at least one validity condition; and

transmitting, to the network, a monitoring report related to a second inference function which is not activated.
A method, comprising,

transmitting, by a base station to a wireless device, a configuration related to at least one inference function, wherein the configuration includes information related to at least one validity condition,

wherein the wireless device activates a first inference function,

wherein the wireless device determines that monitoring results related to the first inference function does not satisfy the at least one validity condition; and

receiving, by the base station from the wireless device, a monitoring report related to a second inference function which is not activated.
A base station in a wireless communication system comprising:

a transceiver;

a memory; and

a processor operatively coupled to the transceiver and the memory, and adapted to:

transmit, to a wireless device, a configuration related to at least one inference function, wherein the configuration includes information related to at least one validity condition,

wherein the wireless device activates a first inference function,

wherein the wireless device determines that monitoring results related to the first inference function does not satisfy the at least one validity condition; and

receive, from the wireless device, a monitoring report related to a second inference function which is not activated.