Method and device for ai / ML time synchronization in wireless communication system

Abstract

The present disclosure relates to a 5G or 6G communication system for supporting higher data transmission rates beyond a 4G communication system such as LTE. A UE according to an embodiment of the present disclosure: receives, from a base station, a radio resource control (RRC) message including configuration information pertaining to a dynamic AI / ML capability report; transmits a dynamic AI / ML capability report message to the base station; receives an AI / ML synchronization command message from the base station in response to the dynamic AI / ML capability report message; and performs AI / ML time synchronization with the base station on the basis of the AI / ML synchronization command message.

Description

Method and device for AI / ML time synchronization in a wireless communication system

[0001] The present disclosure relates to a wireless communication system, and specifically, to a method and apparatus for performing AI / ML time synchronization between a base station and a UE to improve communication performance.

[0002] Looking back at the evolution of wireless communication through successive generations, technologies have been developed primarily for human-oriented services, such as voice, multimedia, and data. Following the commercialization of 5G (5th-generation) communication systems, connected devices, which have been increasing explosively, are expected to be connected to communication networks. Examples of networked objects include vehicles, robots, drones, home appliances, displays, smart sensors installed in various infrastructures, construction machinery, and factory equipment. Mobile devices are expected to evolve into various form factors, such as augmented reality glasses, virtual reality headsets, and holographic devices. In the 6G (6th-generation) era, efforts are underway to develop improved 6G communication systems to connect hundreds of billions of devices and objects to provide diverse services. For this reason, 6G communication systems are being referred to as "beyond 5G" systems.

[0003] In the 6G communication system predicted to be realized around 2030, the maximum transmission speed is tera (i.e., 1,000 gigabit) bps, and the wireless latency is 100 microseconds (μsec). In other words, compared to the 5G communication system, the transmission speed in the 6G communication system is 50 times faster, and the wireless latency is reduced to one-tenth.

[0004] To achieve such high data transmission speeds and ultra-low latency, 6G communication systems are being considered for implementation in the terahertz band (e.g., the 95 GHz to 3 terahertz (3 THz) band). In the terahertz band, due to more severe path loss and atmospheric absorption compared to the millimeter wave (mmWave) band introduced in 5G, the importance of technology capable of guaranteeing signal reach, or coverage, is expected to increase. As key technologies to ensure coverage, radio frequency (RF) devices, antennas, new waveforms that offer better coverage than orthogonal frequency division multiplexing (OFDM), beamforming, and multi-antenna transmission technologies such as massive multiple-input and multiple-output (massive MIMO), full-dimensional MIMO (FD-MIMO), array antennas, and large-scale antennas must be developed. In addition, new technologies such as metamaterial-based lenses and antennas, high-dimensional spatial multiplexing technology using orbital angular momentum (OAM), and reconfigurable intelligent surface (RIS) are being discussed to improve coverage of terahertz band signals.

[0005] In addition, to improve frequency efficiency and system network, development is underway in 6G communication systems for full duplex technology, in which uplink and downlink simultaneously utilize the same frequency resources at the same time; network technology that integrates satellites and HAPS (high-altitude platform stations); network structure innovation technology that supports mobile base stations and enables network operation optimization and automation; dynamic spectrum sharing technology through collision avoidance based on spectrum usage prediction; AI-based communication technology that utilizes AI (artificial intelligence) from the design stage and internalizes end-to-end AI support functions to realize system optimization; and next-generation distributed computing technology that realizes services of complexity exceeding the limits of terminal computing capabilities by utilizing ultra-high performance communication and computing resources (mobile edge computing (MEC), cloud, etc.). In addition, attempts are continuing to further strengthen connectivity between devices, further optimize networks, promote the softwareization of network entities, and increase the openness of wireless communication through the design of new protocols to be used in 6G communication systems, the implementation of hardware-based security environments, the development of mechanisms for the safe utilization of data, and the development of technologies regarding privacy maintenance methods.

[0006] Due to the research and development of such 6G communication systems, it is expected that a new dimension of hyper-connected experience will become possible through the hyper-connectivity of 6G communication systems, which encompasses not only connections between objects but also connections between people and objects. Specifically, it is projected that 6G communication systems will enable the provision of services such as truly immersive extended reality (truly immersive XR), high-fidelity mobile holograms, and digital replicas. Furthermore, services such as remote surgery, industrial automation, and emergency response, which are provided through 6G communication systems with enhanced security and reliability, will be applied in various fields including industry, healthcare, automotive, and home appliances.

[0007] A method of a UE in a wireless communication system according to one embodiment is provided. The method of the UE may include: receiving a Radio Resource Control (RRC) message from a base station containing configuration information regarding a dynamic AI / ML capability report; transmitting a dynamic AI / ML capability report message to the base station; receiving an AI / ML synchronization command message from the base station in response to the dynamic AI / ML capability report message; and performing AI / ML time synchronization with the base station based on the AI / ML synchronization command message.

[0008] A method of a base station in a wireless communication system according to one embodiment is provided. The method of the base station may include the steps of: transmitting a Radio Resource Control (RRC) message containing configuration information regarding a dynamic AI / ML capability report to a User Equipment (UE); receiving a dynamic AI / ML capability report message from the UE; transmitting an AI / ML synchronization command message to the UE in response to the dynamic AI / ML capability report message; and performing AI / ML time synchronization with the UE.

[0009] A UE is provided in a wireless communication system according to one embodiment. The UE may include a memory for storing one or more instructions and at least one processor. The at least one processor may execute one or more instructions stored in the memory to receive a Radio Resource Control (RRC) message from a base station containing configuration information regarding dynamic AI / ML capability reporting; transmit a dynamic AI / ML capability reporting message to the base station; receive an AI / ML synchronization command message from the base station in response to the dynamic AI / ML capability reporting message; and perform AI / ML time synchronization with the base station based on the AI / ML synchronization command message.

[0010] A base station is provided in a wireless communication system according to one embodiment. The base station may include a memory for storing one or more instructions and at least one processor. The at least one processor may, by executing one or more instructions stored in the memory, transmit a Radio Resource Control (RRC) message containing configuration information regarding dynamic AI / ML capability reporting to user equipment (UE); receive a dynamic AI / ML capability reporting message from the UE; transmit an AI / ML synchronization command message to the UE in response to the dynamic AI / ML capability reporting message; and perform AI / ML time synchronization with the UE.

[0011] A computer-readable recording medium disclosed as a technical means for achieving the above-described technical task may store a program for executing at least one of the embodiments of the disclosed method on a computer.

[0012] Other technical features can be easily made clear to a person skilled in the art from the following drawings, descriptions, and claims.

[0013] FIG. 1a is a drawing for explaining the technical field and purpose of the present disclosure.

[0014] FIG. 1b is a drawing for explaining the technical field and purpose of the present disclosure.

[0015] FIG. 2 is a diagram illustrating a method for a UE and a base station to perform AI / ML time synchronization in a wireless communication system according to one embodiment of the present disclosure.

[0016] FIG. 3 is a diagram illustrating information included in a dynamic AI / ML capability report message according to one embodiment of the present disclosure.

[0017] FIG. 4a is a diagram illustrating information included in an AI / ML synchronization command message according to one embodiment of the present disclosure and a specific example of a UE performing AI / ML time synchronization.

[0018] FIG. 4b is a diagram illustrating information included in an AI / ML synchronization command message according to one embodiment of the present disclosure and a specific example of a UE performing AI / ML time synchronization.

[0019] FIG. 4c is a diagram illustrating information included in an AI / ML synchronization command message according to one embodiment of the present disclosure and a specific example of a UE performing AI / ML time synchronization.

[0020] FIG. 4d is a diagram illustrating information included in an AI / ML synchronization command message according to one embodiment of the present disclosure and a specific example of a UE performing AI / ML time synchronization.

[0021] FIG. 5 is a diagram illustrating a method for a UE and a base station to perform AI / ML time synchronization based on an event-triggered method in a wireless communication system according to one embodiment of the present disclosure.

[0022] FIG. 6 is a diagram illustrating a method for a UE and a base station to perform AI / ML time synchronization based on a Periodic method in a wireless communication system according to one embodiment of the present disclosure.

[0023] FIG. 7 is a diagram illustrating a method for a UE and a base station to perform AI / ML time synchronization based on a NW-initiated method in a wireless communication system according to one embodiment of the present disclosure.

[0024] FIG. 8 is a diagram illustrating a method for a UE and a base station to perform AI / ML model switching or AI / ML model deactivation operations in a wireless communication system according to one embodiment of the present disclosure.

[0025] FIG. 9a is a diagram illustrating a method for determining whether a base station in a wireless communication system according to one embodiment of the present disclosure performs an AI / ML model switching or AI / ML model deactivation operation based on a Delay A parameter and a Delay C parameter.

[0026] FIG. 9b is a diagram illustrating a method for determining whether a base station in a wireless communication system according to one embodiment of the present disclosure performs an AI / ML model switching or AI / ML model deactivation operation based on a Delay A parameter and a Delay C parameter.

[0027] FIG. 10a is a diagram illustrating a method for determining whether a base station in a wireless communication system according to one embodiment of the present disclosure performs an AI / ML model switching or AI / ML model deactivation operation based on a Delay B parameter.

[0028] FIG. 10b is a diagram illustrating a method for determining whether a base station in a wireless communication system according to one embodiment of the present disclosure performs an AI / ML model switching or AI / ML model deactivation operation based on a Delay B parameter.

[0029] FIG. 11 is a diagram illustrating a method for a UE to perform AI / ML time synchronization in a wireless communication system according to one embodiment of the present disclosure.

[0030] FIG. 12 is a diagram illustrating a method for a base station to perform AI / ML time synchronization in a wireless communication system according to one embodiment of the present disclosure.

[0031] FIG. 13 is a block diagram of a UE according to one embodiment of the present disclosure.

[0032] FIG. 14 is a block diagram of a base station according to one embodiment of the present disclosure.

[0033] The present disclosure is subject to various modifications and may have various embodiments; specific embodiments are illustrated in the drawings and described in detail in the detailed description. However, FIGS. 1 through 12 discussed below and the various embodiments used to explain the principles of the present disclosure in this specification are merely illustrative and should not be construed as limiting the scope of the present disclosure in any way. Those skilled in the art will understand that the principles of the present disclosure may be implemented in any suitably arranged system or device. Furthermore, those skilled in the art will understand that the principles of the present disclosure may be implemented in any suitably configured wireless communication system.

[0034] For the same reason, some components in the attached drawings have been exaggerated, omitted, or schematically depicted. Additionally, the dimensions of each component do not entirely reflect their actual dimensions. Identical or corresponding components in each drawing have been assigned the same reference numbers.

[0035] In addition, numbers used in the description process of the specification (e.g., 1st, 2nd, etc.) are merely identifiers to distinguish one component from another.

[0036] The advantages and features of the present disclosure, and the methods for achieving them, will become clear by referring to the embodiments described below in detail together with the accompanying drawings. However, the present disclosure is not limited to the embodiments disclosed below but may be implemented in various different forms. These embodiments are provided merely to ensure that the disclosure is complete and to fully inform those skilled in the art of the scope of the disclosure, and the present disclosure is defined only by the scope of the claims. Throughout the specification, the same reference numerals refer to the same components. Furthermore, in describing the present disclosure, if it is determined that a detailed description of a related function or configuration might unnecessarily obscure the essence of the present disclosure, such detailed description is omitted. Additionally, the terms described below are defined considering their functions in the present disclosure, and these may vary depending on the intentions or conventions of the user or operator. Therefore, their definitions should be based on the content throughout the specification.

[0037] Hereinafter, a base station (BS) is an entity that performs resource allocation for terminals and may be at least one of an NG-RAN, gNode B, eNode B, Node B, or xNode B (where x is an alphabet including g and e), a radio access unit, a base station controller, a satellite, an airborn, or a node on a network. A distributed base station may be separated into a centralized unit (CU) and a distributed unit (DU). The CU provides support for higher protocol layers such as SDAP (service data adaptation protocol), RRC (radio resource control), and PDCP (packet data convergence protocol), and the DU may provide support for lower protocol layers such as RLC (radio link control), MAC (medium access control), and PHY (physical layer). A single CU may exist for each gNodeB, and multiple DUs may be connected to each CU. The DU includes both baseband processing and RF functions and can support various mobility scenarios.

[0038] Hereinafter, the terminal (user equipment, UE) may include a Mobile Station (MS), a Vehicle, a Satellite, an Airborne, a Cellular Phone, a Smartphone, a Computer, or a Multimedia System capable of performing communication functions.

[0039] In addition, while LTE, LTE-A, or 5G systems may be described below as examples, embodiments of the present disclosure may also be applied to other communication systems having similar technical backgrounds or channel types. For example, 5G-Advance or NR-Advance or 6th generation mobile communication technology (6G) developed after 5G mobile communication technology (or new radio, NR) may be included, and the 5G below may be a concept that includes existing LTE, LTE-A, and other similar services. Furthermore, the present disclosure may be applied to other communication systems with some modifications made at the discretion of a person with skilled technical knowledge, without significantly departing from the scope of the present disclosure.

[0040] At this point, it will be understood that each block of the process flow diagrams and combinations of the flow diagrams can be executed by computer program instructions. Since these computer program instructions can be loaded into the processor of a general-purpose computer, a special-purpose computer, or other programmable data processing equipment, the instructions executed through the processor of the computer or other programmable data processing equipment create means to perform the functions described in the flow diagram block(s). Since these computer program instructions can also be stored in computer-available or computer-readable memory that can be directed toward the computer or other programmable data processing equipment to implement the function in a specific way, the instructions stored in computer-available or computer-readable memory can also produce a manufactured item containing instruction means to perform the function described in the flow diagram block(s). Since computer program instructions can be loaded onto a computer or other programmable data processing equipment, instructions that perform a series of operation steps on the computer or other programmable data processing equipment to create a process executed by the computer can also provide steps for executing the functions described in the flowchart block(s).

[0041] Additionally, each block may represent a module, segment, or part of code containing one or more executable instructions for executing a specified logical function(s). It should also be noted that in some alternative practices, the functions mentioned in the blocks may occur out of order. For example, two blocks described in succession may actually be executed substantially simultaneously, or the blocks may be executed in reverse order according to their corresponding functions.

[0042] In this embodiment, the term "part" refers to a software or hardware component such as an FPGA (Field Programmable Gate Array) or an ASIC (Application Specific Integrated Circuit), and the "part" performs certain roles. However, the meaning of "part" is not limited to software or hardware. The "part" may be configured to reside in an addressable storage medium or configured to run one or more processors. Thus, as an example, the "part" includes components such as software components, object-oriented software components, class components, and task components, as well as processes, functions, attributes, procedures, subroutines, segments of program code, drivers, firmware, microcode, circuits, data, databases, data structures, tables, arrays, and variables. The functions provided within the components and "parts" may be combined into a smaller number of components and "parts" or further separated into additional components and "parts." In addition, the components and 'parts' may be implemented to utilize one or more CPUs within the device or secure multimedia card. Also, in the embodiments, 'parts' may include one or more processors.

[0043] Terms used in the following description to refer to broadcast information, control information, communication coverage, state changes (e.g., events), network entities, messages, and device components are examples provided for the convenience of explanation. Accordingly, the present disclosure is not limited to the terms described below, and other terms having equivalent technical meanings may be used.

[0044] For the convenience of the following explanation, the present invention uses terms and names defined in the LTE and NR specifications, which are the most recent standards defined by the 3GPP (The 3rd Generation Partnership Project) among currently existing communication standards. However, the present invention is not limited by the above terms and names and can be applied in the same way to systems conforming to other standards.

[0045] FIGS. 1a and 1b are drawings for explaining the technical field and purpose of the present disclosure.

[0046] Referring to FIG. 1a, a UE (10) and a base station (20) according to one embodiment of the present disclosure can perform time synchronization of artificial intelligence (AI) / machine learning (ML) mounted on both sides.

[0047] In a wireless communication system, the UE (10) and the base station (20) may share and use the same AI / ML model in a specific use case. In a narrow sense, the same AI / ML model refers to two models having the same model structure and the same model parameter values, and in a broad sense, it refers to two models that produce the same (or highly similar) output when the same input is input into the model.

[0048] For example, referring to FIG. 1b, the UE (10) and the base station (20) can share an AI / ML model for Tx-Rx beam pair prediction. This is a use case for predicting a Tx-Rx beam pair (Tx-Rx beam pair) by using an AI / ML model, which combines the base station (20)'s Tx beam and the UE (10)'s Rx beam into a single pair. When the UE (10) and the base station (20) share an AI / ML model, if the same input is input to both AI / ML models, the same Tx-Rx beam pair is obtained, and based on this, the base station (20) can select the Tx beam and the UE (10) can select the Rx beam. Therefore, when the UE (10) and the base station (20) share an AI / ML model, they can predict the other party's operation (e.g., Tx beam selection, Rx beam selection) in real time without separate signal exchange.

[0049] Additionally, although not illustrated in FIG. 1b, for example, the UE (10) and the base station (20) can share an AI / ML model for reducing UE power consumption. This is a use case in which the base station (20) uses the AI / ML model to predict whether the base station (20) will schedule DCI (downlink control information) and whether the UE (10) needs to monitor PDCCH (physical downlink control channel). When the UE (10) and the base station (20) share an AI / ML model, if the same input is input to both AI / ML models, the same True / False result is obtained. Based on this, the base station (20) can determine whether to schedule DCI in a specific slot, and the UE (10) can determine whether to monitor PDCCH in that slot. Therefore, the UE (10) can selectively monitor PDCCH only in slots where DCI is scheduled and reduce unnecessary PDCCH monitoring in slots where DCI is not scheduled, thereby reducing power consumption. Therefore, when the UE (10) and the base station (20) share an AI / ML model, they can predict and operate the other side's operation (e.g., DCI scheduling, PDCCH monitoring) in real time without separate signal exchange.

[0050] Additionally, although not illustrated in FIG. 1b, for example, the UE (10) and the base station (20) can share an AI / ML model for link adaptation. This is a use case in which the signal-to-interference-plus-noise ratio (SINR) probability distribution is predicted using an AI / ML model in the UE (10) and the base station (20). When the UE (10) and the base station (20) share an AI / ML model, if the same input is input to both AI / ML models, the same SINR probability distribution is obtained. Based on this, the UE (10) can generate an optimal channel quality indicator (CQI), and the base station (20) can select a modulation and coding scheme (MCS) suitable for it. Therefore, when the UE (10) and the base station (20) share an AI / ML model, they can predict and operate the other side's actions (e.g., CQI generation, MCS selection) in real time without separate signal exchange.

[0051] However, since the time required for AI / ML processing on the UE (10) side varies depending on the real-time computing environment of the UE (10) (e.g., battery, memory, throughput, etc.), if the AI / ML capability of the UE (10) related to the AI / ML processing time is not dynamically reported to the base station (20), the following problems may occur.

[0052] - Problem 1: In a use case where the UE (10) and the base station (20) share the same AI / ML model, the UE (10) and the base station (20) may apply different inference outputs at the same time. This is because the same inference output obtained with the same input is applied at different times due to the different processing times of the AI / ML models on the UE (10) and the base station (20) sides. This leads to a degradation in the communication performance of the UE or base station using the AI / ML model.

[0053] - Problem 2: In use cases where the UE uses an AI / ML model, if the AI / ML processing time of the UE (10) becomes long, the UE (10) applies an outdated inference output. This leads to a decrease in the communication performance of the UE using the AI / ML model.

[0054] In addition, if the AI / ML capability of the UE (10) related to the update (or measurement) of the input used in the AI / ML model is not dynamically reported, the following problems may occur:

[0055] - Problem 1: In a use case where the UE (10) and the base station (20) share the same AI / ML model, the UE (10) and the base station (20) may apply different inference outputs at the same time. This is because the UE (10) and the base station (20) obtain different inference results because they use different inputs. This leads to a degradation in the communication performance of the UE or base station using the AI / ML model.

[0056] - Problem 2: In use cases where the UE uses an AI / ML model, if the AI / ML input update cycle is long, the UE (10) obtains an outdated output because it uses an outdated input. This leads to a decrease in communication performance for the UE using the AI / ML model.

[0057] To solve the above problems, the UE (10) and the base station (20) must use synchronized AI / ML input / output values. That is, the UE (10) and the base station (20) must perform inference using the same input through AI / ML time synchronization and apply the inference output at the same time. To this end, the UE (10) needs to report to the base station (20) the AI / ML capability related to the dynamically changing AI / ML processing time and the update of the AI / ML input. Additionally, the base station (20) needs to instruct the UE (10) to an offset that takes into account the jitter of the UE (10)'s computing delay.

[0058] The present disclosure aims to improve the performance of an AI / ML-based communication system by synchronizing the AI / ML processing times of a UE and a base station in a wireless communication system.

[0059] Specifically, in one embodiment of the present disclosure, a method is proposed in which a base station transmits configuration information regarding a dynamic AI / ML capability report to a UE, a method in which a UE reports information regarding a dynamically changing AI / ML capability (i.e., dynamic AI / ML capability) to a base station, a method in which a base station transmits an offset for AI / ML time synchronization to a UE in consideration of the UE's dynamic AI / ML capability, and a method in which a base station determines AI / ML model switching or AI / ML model deactivation to a UE in consideration of the UE's dynamic AI / ML capability.

[0060] The embodiments proposed in the present disclosure will be described in detail below with reference to FIGS. 2 to 14.

[0061] FIG. 2 is a diagram illustrating a method for a UE (10) and a base station (20) to perform AI / ML time synchronization in a wireless communication system according to one embodiment of the present disclosure.

[0062] Referring to FIG. 2, in step 210, a base station (20) according to an embodiment of the present disclosure may transmit an RRC message to a UE (10) that includes configuration information regarding a dynamic AI / ML capability report. The UE (10) may receive an RRC message (e.g., RRC reconfiguration message) from the base station (20) that includes configuration information regarding a dynamic AI / ML capability report.

[0063] In one embodiment, configuration information regarding a dynamic AI / ML capability may include at least one of: identifier (ID) information; triggering conditions for a dynamic AI / ML capability report; a dynamic AI / ML capability report period; or a type of a dynamic AI / ML capability report.

[0064] ■ The identifier (ID) information may refer to the identifier (ID) of an AI / ML model or AI / ML functionality. In one embodiment, the UE (10) may measure and report dynamic AI / ML capability for each AI / ML model or AI / ML functionality.

[0065] ■ The triggering conditions for dynamic AI / ML capability reporting may refer to the conditions required for dynamic AI / ML capability reporting to be triggered in the UE. This will be explained in detail with reference to Fig. 5.

[0066] ■ The dynamic AI / ML capability reporting period may refer to the period during which the UE transmits dynamic AI / ML capability messages to the base station. This will be explained in detail with reference to Fig. 6.

[0067] ■ The type of dynamic AI / ML capability report is that the AI / ML input is multidimensional ( When ), it may be indicated whether to report the Delay B parameter (defined in step 220 below) and the Delay D parameter (defined in step 220 below) in the form of a scaler or in the form of a sequence, respectively.

[0068] In step 220, a UE (10) according to an embodiment of the present disclosure may transmit a dynamic AI / ML capability report message to a base station (20). The UE (10) may transmit a dynamic AI / ML capability report message to a base station (20) based on configuration information regarding the dynamic AI / ML capability report. The base station (20) may receive the dynamic AI / ML capability report message from the UE (10). In one embodiment, the dynamic AI / ML capability report message may be transmitted via MAC CE (medium access control control element) or DCI (downlink control information) signaling.

[0069] Conventionally, the UE (10) could report static information related to hardware constraints, such as computing power, storage, and battery, to the base station (20) as an AI / ML capability. The dynamic AI / ML capability reporting of the present disclosure is distinguished from the conventional AI / ML capability.

[0070] In one embodiment, the UE (10) can measure and report dynamic AI / ML capability for each AI / ML model or AI / ML functionality.

[0071] In one embodiment, the dynamic AI / ML capability report message may include at least one of the average value of the UE's AI / ML processing time (i.e., Delay A parameter); the average delay time between the latest update time of the UE's AI / ML input and the start time of AI / ML processing (i.e., Delay B parameter); the average value of the UE's AI / ML inference time (i.e., Delay C parameter); or the average update period of the UE's AI / ML input (i.e., Delay D parameter). This will be explained in detail with reference to FIG. 3.

[0072] At step 230, a base station (20) according to an embodiment of the present disclosure may transmit an AI / ML synchronization command message (AI / ML synchronization command message) to a UE (10). The base station (20) may transmit an AI / ML synchronization command message to the UE (10) based on / in response to a dynamic AI / ML capability report message. The UE (10) may receive an AI / ML synchronization command message from the base station (20). In one embodiment, the AI / ML synchronization command message may be transmitted via MAC CE (medium access control control element) or DCI (downlink control information) signaling.

[0073] In one embodiment, the AI / ML synchronization command message may include offset information for AI / ML time synchronization between the base station and the UE.

[0074] In one embodiment, the offset information for AI / ML time synchronization between the base station and the UE may include at least one of an AI / ML input offset value or an AI / ML process offset value. This will be explained in detail with reference to FIGS. 4a to 4d.

[0075] In step 240, the UE (10) and the base station (20) according to an embodiment of the present disclosure can perform AI / ML time synchronization with each other.

[0076] In one embodiment, the UE (10) may use an AI / ML input corresponding to an AI / ML input offset value as an input to the UE-side AI / ML model based on an AI / ML input offset value received from the base station (20). In one embodiment, the UE (10) may determine the timing for applying the processing result of the UE-side AI / ML model based on an AI / ML process offset value received from the base station (20). In one embodiment, the base station (20) may use an AI / ML input identical to the AI / ML input to be used by the UE as an input to the base station-side AI / ML model based on the determined AI / ML input offset value. In one embodiment, the base station may determine the timing for applying the processing result of the base station-side AI / ML model to the same time as the UE based on the determined AI / ML process offset value. This will be explained in detail with reference to FIGS. 4a to 4d.

[0077] According to one embodiment of the present disclosure, the UE and the base station can improve the performance of an AI / ML-based communication system by synchronizing the AI / ML processing time between the UE and the base station through dynamic AI / ML capability report messages and AI / ML synchronization command messages by applying parameters related to the UE's dynamic AI / ML capability and parameters required for AI / ML time synchronization.

[0078] However, the effects obtainable from the present disclosure are not limited to those mentioned above, and other unmentioned effects will be clearly understood by those skilled in the art to which the present disclosure belongs from the description below.

[0079] FIG. 3 is a diagram illustrating information included in a dynamic AI / ML capability report message according to one embodiment of the present disclosure.

[0080] Referring to FIG. 3, a dynamic AI / ML capability reporting message according to one embodiment of the present disclosure may include at least one of: an average value of the UE's AI / ML processing time (i.e., Delay A parameter); an average delay time between the latest update time of the UE's AI / ML input and the start time of AI / ML processing (i.e., Delay B parameter); an average value of the UE's AI / ML inference time (i.e., Delay C parameter); or an average update period of the UE's AI / ML input (i.e., Delay D parameter). Additionally, Delay A to Delay D parameters may be measured / reported for each AI / ML model or AI / ML functionality.

[0081] ■ The Delay A parameter is defined as the average value of the AI / ML processing time. This can be calculated as the sum of the average value of Pre-delay, the average value of the AI / ML inference time (i.e., the Delay C parameter), and the average value of Post-delay. Here, Pre-delay may include the time taken to preprocess the input into the form of the AI / ML model's input, data transfer time between memory and the processor, and waiting time. Additionally, Post-delay may include data post-processing for the use of the AI / ML inference output, data transfer time between memory and the processor, and waiting time.

[0082] ■ The Delay B parameter is defined as the average delay between the time of the latest update of input (x) and the start of AI / ML processing. Here, the time of the latest update of input (x) refers to the time when input (x) was last updated or measured. Additionally, input (x) refers to data prior to preprocessing to match the input format of the AI / ML model, and can be updated through UE measurements or reception from a network. If input (x) is multidimensional ( In the case of ), the Delay B parameter can be defined in the form of a sequence with the average calculated for each dimension, or, can be defined in the form of a single scalar as the average value of the average for each dimension.

[0083] The Delay C parameter is defined as the average value of AI / ML inference time. AI / ML inference time refers to the time it takes to feed a preprocessed input (x) into an AI / ML model (e.g., a neural network) and produce an output through inference. Generally, heavier models (e.g., neural network models with more parameters and layers) have longer AI / ML inference times.

[0084] ■ The Delay D parameter is defined as the average update / measurement cycle of the input (x). Here, input (x) refers to data prior to preprocessing to match the input format of the AI / ML model, and can be updated through UE measurements or reception from a network. If the input (x) is multidimensional ( In the case of ), the Delay D parameter can be defined in the form of a sequence with the average calculated for each dimension, or it can be defined in the form of a single scalar as the average value of the averages for each dimension.

[0085] FIGS. 4a to 4d are drawings for explaining information included in an AI / ML synchronization command message according to one embodiment of the present disclosure and specific examples of a UE performing AI / ML time synchronization.

[0086] A base station according to one embodiment of the present disclosure may transmit an AI / ML synchronization command message to a UE. In one embodiment, the AI / ML synchronization command message may include offset information for AI / ML time synchronization between the base station and the UE.

[0087] In one embodiment, the offset information for AI / ML time synchronization between the base station and the UE may include at least one of an AI / ML input offset value or an AI / ML process offset value.

[0088] ■ AI / ML process offset value( ) can mean the time interval between the start of AI / ML processing and the application of the AI / ML processing results.

[0089] Based on the AI / ML process offset value, the UE-side AI / ML model and the base station-side AI / ML model can apply the AI / ML inference output at the same time. In the case of delays related to AI / ML processing time, AI / ML time synchronization between the two sides can be achieved by having the side with the faster AI / ML processing time match the side with the slower AI / ML processing time.

[0090] In one embodiment, the base station, based on the Delay A parameter included in the dynamic AI / ML capability report message received from the UE, determines the AI / ML process offset value ( ) can be determined.

[0091] For example, the base station has the AI / ML process offset value ( ) can be determined as in [Equation 1]. However, this is the AI / ML process offset value( This is merely one example of a method for setting ) and the present disclosure is not limited thereto.

[0092] [Mathematical Formula 1]

[0093]

[0094] Here, the Delay A parameter is the average value of the UE's AI / ML processing time, and the Delay A' parameter is the average value of the base station's AI / ML processing time. The value is a parameter representing the additional delay required for time synchronization between the UE and the base station. As previously mentioned, since the time required for AI / ML processing varies depending on the UE's real-time computing environment (e.g., battery, memory, throughput, etc.), the Delay A parameter has a time-varying characteristic. Therefore, to compensate for this and enhance the stability of time synchronization between the UE and the base station, the base station The value can be used.

[0095] The base station The value can be set arbitrarily. For example, the base station The value can be set to 1 ms, 2 ms, etc. In addition, the base station The value can be set based on the Delay A parameter. For example, if the Delay A parameter reported periodically changes significantly (i.e., if the standard deviation is large), You can set the value high.

[0096] According to [Equation 1], the base station, AI / ML process offset value ( ) to the larger value between the Delay A parameter and the Delay A' parameter It can be determined by adding the values.

[0097] In one embodiment, the base station can determine the time to apply the processing result of the base station-side AI / ML model to the same time as the UE, based on a determined AI / ML process offset value.

[0098] In one embodiment, the base station has a determined AI / ML process offset value ( ) can be included in the AI / ML synchronization command message and sent to the UE.

[0099] In one embodiment, the UE can determine the timing for applying the processing results of the UE-side AI / ML model based on an AI / ML process offset value received from a base station.

[0100] For example, a UE or base station can compare the Delay A value with the AI / ML process offset value and determine when to apply the processing result of the AI / ML model based on the comparison result. If the AI / ML process offset value is greater than the Delay A value (or, in the case of a base station, the Delay A' value), the UE or base station may apply the processing result of the AI / ML model after finishing the AI / ML process and waiting for the difference between the two values. If the AI / ML process offset value is smaller than the Delay A value (or, in the case of a base station, the Delay A' value), the UE or base station may apply the processing result of the AI / ML model as soon as the AI / ML process is finished.

[0101] ■ AI / ML input offset value( ) can refer to the M value of input (x) that is closest to the time of the Mth update from the time of the last input (x) update. For example, in the case of the most recently updated input (x), the AI / ML input offset value ( ) = 0, and for the second most recently updated input (x), the AI / ML input offset value( ) = 1.

[0102] When the input (x) is multidimensional ( ), AI / ML input offset value( ) can be set for each dimension and defined in the form of a sequence, or, can be defined in the form of a single scalar. If defined in the form of a single scalar, the same AI / ML input offset value can be applied to all Input (x) dimensions.

[0103] In the case of delays related to the update / measurement of AI / ML input, AI / ML time synchronization between the two parties can be performed by the side with the newer AI / ML input aligning with the side with the older AI / ML input.

[0104] In one embodiment, the base station, based on the Delay B parameter and Delay D parameter included in the dynamic AI / ML capability report message received from the UE, determines the AI / ML input offset value ( ) can be determined.

[0105] For example, the base station has the AI / ML input offset value ( ) can be determined as in [Equation 2]. However, this is the AI / ML input offset value( This is merely one example of a method for setting ) and the present disclosure is not limited thereto.

[0106] [Mathematical Formula 2]

[0107]

[0108] Here, the Delay B parameter is the average delay between the latest update time of the UE's AI / ML input and the start time of AI / ML processing, and the Delay B' parameter is the average delay between the latest update time of the base station's AI / ML input and the start time of AI / ML processing. The Delay D parameter is the average update / measurement cycle of the UE's input (x).

[0109] According to [Mathematical Equation 2], the base station, In this case, the AI / ML input offset value ( Determine ) to 0, and In this case, the delay time of the input to be used by the UE ( The value of M that minimizes the difference between ) and the base station delay time (Delay B') is the AI / ML input offset value ( It can be decided as ).

[0110] In one embodiment, the AI / ML input offset value may be defined based on the order (number, index) of the updated input as in the example above, as well as based on a unit representing time (e.g., g, 1ms, 2ms,,,,).

[0111] In one embodiment, the base station has a determined AI / ML input offset value ( Based on ), the same AI / ML input (x) that the UE will use can be used as the input to the base station-side AI / ML model.

[0112] In one embodiment, the base station has a determined AI / ML input offset value ( ) can be included in the AI / ML synchronization command message and sent to the UE.

[0113] In one embodiment, the UE can use an input (x) corresponding to / corresponding to an AI / ML input offset value as an input to an AI / ML model, based on an AI / ML input offset value received from a base station.

[0114] Hereinafter, with reference to FIGS. 4a to 4d, examples of setting AI / ML process offset values ​​and AI / ML input offset values ​​according to the relationship between Delay A parameter, Delay B parameter, Delay A' parameter, and Delay B' parameter will be explained.

[0115] Figure 4a illustrates an example of setting the AI / ML process offset value and the AI / ML input offset value when Delay A > Delay A' and Delay B < Delay B'.

[0116] Referring to FIG. 4a, the base station (20) can determine the AI / ML process offset value by adding the Offset value to the Delay A parameter according to [Equation 1] and transmit this to the UE (10) as an AI / ML synchronization command message.

[0117] The UE (10) and the base station (20) can, based on the AI / ML process offset value, start a timer with a length equal to the AI / ML process offset value when starting the AI / ML process, and apply the processing result of the UE-side AI / ML model after the timer expires. The base station (20) can determine the time to apply the processing result of the base station-side AI / ML model to the same time as the UE based on the determined AI / ML process offset value. Accordingly, the UE (10) and the base station (20) can apply the processing result of the AI / ML model at the same time.

[0118] The base station (20), according to [Equation 2], sets the AI / ML input offset value to be used by the UE (10) as the delay time of the input ( The difference between the delay time (Delay B') of the base station (20) and the base station is determined to be 1, which is the value of M that minimizes the difference, and this can be transmitted to the UE (10) as an AI / ML synchronization command message.

[0119] Based on the AI / ML input offset value (= 1), the UE (10) can use the input (x) that is closest to the time of the last update of the input (x) as the input to the AI / ML model. Based on the AI / ML input offset value (= 1), the base station (20) can predict the input (x) that the UE (10) will use and use it as the input to the AI / ML model. Accordingly, the UE (10) and the base station (20) can perform inference using the same or most similar AI / ML input (= most similar update / measurement time).

[0120] Referring to Fig. 4b, examples of setting AI / ML process offset values ​​and AI / ML input offset values ​​are illustrated when Delay A > Delay A' and Delay B > Delay B'.

[0121] Referring to FIG. 4b, the base station (20) can determine the AI / ML process offset value by adding the Offset value to the Delay A parameter according to [Equation 1] and transmit this to the UE (10) as an AI / ML synchronization command message.

[0122] The UE (10) can, based on the AI / ML process offset value, start a timer with a length equal to the AI / ML process offset value when starting the AI / ML process, and apply the processing result of the UE-side AI / ML model after the timer expires. The base station (20) can determine the time to apply the processing result of the base station-side AI / ML model at the same time as the UE, based on the determined AI / ML process offset value. Accordingly, the UE (10) and the base station (20) can apply the processing result of the AI / ML model at the same time.

[0123] The base station (20) can determine the AI / ML input offset value to 0 according to [Equation 2] and transmit it to the UE (10) as an AI / ML synchronization command message.

[0124] UE (10) can use the last (i.e., most recently) updated / measured input (x) as an input to the AI / ML model based on the AI / ML input offset value (= 0). Base station (20) can predict the input (x) that UE (10) will use based on the AI / ML input offset value (= 0) and use it as an input to the AI / ML model. Accordingly, UE (10) and base station (20) can perform inference using the same or most similar (= most similar update / measurement time) AI / ML input.

[0125] Referring to Fig. 4c, examples of setting AI / ML process offset values ​​and AI / ML input offset values ​​are shown when Delay A < Delay A' and Delay B > Delay B'.

[0126] Referring to FIG. 4c, the base station (20) can determine the AI / ML process offset value by adding the Offset value to the Delay A' parameter according to [Equation 1] and transmit this to the UE (10) as an AI / ML synchronization command message.

[0127] The UE (10) can start an AI / ML process based on an AI / ML process offset value, and when starting the AI / ML process, it can operate a timer with a length equal to the AI / ML process offset value, and after the timer expires, apply the processing result of the UE-side AI / ML model. The base station (20) can determine the time to apply the processing result of the base station-side AI / ML model at the same time as the UE based on the determined AI / ML process offset value. Accordingly, the UE (10) and the base station (20) can apply the processing result of the AI / ML model at the same time.

[0128] The base station (20) can determine the AI / ML input offset value to 0 according to [Equation 2] and transmit it to the UE (10) as an AI / ML synchronization command message.

[0129] UE (10) can use the last (i.e., most recently) updated / measured input (x) as an input to the AI / ML model based on the AI / ML input offset value (= 0). Base station (20) can predict the input (x) that UE (10) will use based on the AI / ML input offset value (= 0) and use it as an input to the AI / ML model. Accordingly, UE (10) and base station (20) can perform inference using the same or most similar (= most similar update / measurement time) AI / ML input.

[0130] Referring to Fig. 4d, examples of setting AI / ML process offset values ​​and AI / ML input offset values ​​are shown when Delay A < Delay A' and Delay B < Delay B'.

[0131] Referring to FIG. 4d, the base station (20) can determine the AI / ML process offset value by adding the Offset value to the Delay A' parameter according to [Equation 1] and transmit this to the UE (10) as an AI / ML synchronization command message.

[0132] The UE (10) can start an AI / ML process based on an AI / ML process offset value, and when starting the AI / ML process, it can operate a timer with a length equal to the AI / ML process offset value, and after the timer expires, apply the processing result of the UE-side AI / ML model. The base station (20) can determine the time to apply the processing result of the base station-side AI / ML model at the same time as the UE based on the determined AI / ML process offset value. Accordingly, the UE (10) and the base station (20) can apply the processing result of the AI / ML model at the same time.

[0133] The base station (20), according to [Equation 2], sets the AI / ML input offset value to be used by the UE (10) as the delay time of the input ( The difference between the delay time (Delay B') of the base station (20) and the base station is determined to be 1, which is the value of M that minimizes the difference, and this can be transmitted to the UE (10) as an AI / ML synchronization command message.

[0134] UE (10) can use an input (x) that is closest to the time of the first update from the time of the last update of input (x) as an input to the AI / ML model based on the AI / ML input offset value (= 1). Base station (20) can predict the input (x) that UE (10) will use based on the AI / ML input offset value (= 1) and use it as an input to the AI / ML model. Accordingly, UE (10) and base station (20) can perform inference using the same or most similar (= most similar update / measurement time) AI / ML input.

[0135] According to one embodiment of the present disclosure, the UE and the base station can improve the performance of an AI / ML-based communication system by synchronizing the AI / ML processing time between the UE and the base station through dynamic AI / ML capability report messages and AI / ML synchronization command messages by applying parameters related to the UE's dynamic AI / ML capability and parameters required for AI / ML time synchronization.

[0136] However, the effects obtainable from the present disclosure are not limited to those mentioned above, and other unmentioned effects will be clearly understood by those skilled in the art to which the present disclosure belongs from the description below.

[0137] In the present disclosure, the dynamic AI / ML capability reporting of a UE may be triggered in an event-triggered, periodic, or network-triggered manner, but is not limited thereto. Hereinafter, with reference to FIGS. 5 to 7, the triggering method of the dynamic AI / ML capability reporting of a UE will be described in detail.

[0138] FIG. 5 is a diagram illustrating a method for a UE and a base station to perform AI / ML time synchronization based on an event-triggered method in a wireless communication system according to one embodiment of the present disclosure.

[0139] Referring to FIG. 5, in step 510, a base station (20) according to one embodiment of the present disclosure may transmit an RRC message (e.g., RRC reconfiguration message) containing triggering conditions for dynamic AI / ML capability reporting to a UE (10). The UE (10) may receive an RRC message (e.g., RRC reconfiguration message) containing triggering conditions for dynamic AI / ML capability reporting from the base station (20). Step 510 may operate similarly to step 210 of FIG. 2, and redundant details are omitted here with reference to FIG. 2.

[0140] ■ In one embodiment, the triggering condition for a dynamic AI / ML capability report may include an event condition.

[0141] For example, the event condition can be defined as [Equation 3].

[0142] [Mathematical Formula 3]

[0143]

[0144] Here, the Delay A parameter is the average value of the UE's AI / ML processing time, and is the AI / ML Process offset value. is a threshold related to event conditions, which can be set by the base station. For example, It can be set from the base station via RRC signaling along with event conditions.

[0145] In one embodiment, the UE may trigger a dynamic AI / ML capability report when the ratio of the AI / ML Process offset value to the Delay A value exceeds a specific threshold. In one embodiment, the UE may trigger a dynamic AI / ML capability report when an event satisfying an event condition is satisfied N consecutive times.

[0146] ■ In one embodiment, the triggering condition for a dynamic AI / ML capability report may include an entering condition and a leaving condition.

[0147] For example, the entry condition can be defined as [Equation 4].

[0148] [Mathematical Formula 4]

[0149]

[0150] Here, the Delay A parameter is the average value of the UE's AI / ML processing time, and is the AI / ML Process offset value. is a threshold related to the entry condition, which can be set by the base station. For example, It can be set from the base station via RRC signaling along with entry conditions.

[0151] For example, the departure condition can be defined as [Equation 5].

[0152] [Mathematical Formula 5]

[0153]

[0154] Here, the Delay A parameter is the average value of the UE's AI / ML processing time, and is the AI / ML Process offset value. is a threshold related to the exit condition, which can be set by the base station. For example, It can be set from the base station via RRC signaling along with entry conditions.

[0155] In one embodiment, the UE may trigger a dynamic AI / ML capability report if, after satisfying the entry condition, the exit condition is not satisfied during N AI / ML processes. In one embodiment, the UE may reset the count if it determines that the exit condition is satisfied.

[0156] ■ The triggering conditions for dynamic AI / ML capability reporting are not limited to the examples mentioned above and may be defined in other ways.

[0157] In step 520, the UE (10) according to one embodiment of the present disclosure can determine whether the triggering condition for a dynamic AI / ML capability report is satisfied.

[0158] In step 530, if the UE (10) according to one embodiment of the present disclosure determines that the triggering condition is satisfied, it may transmit a dynamic AI / ML capability report message to the base station (20). The base station (20) may receive the dynamic AI / ML capability report message from the UE (10) when the UE (10) satisfies the triggering condition. Step 530 may operate similarly to step 220 of FIG. 2, and redundant content is omitted here with reference to FIG. 2.

[0159] In step 540, the base station (20) according to an embodiment of the present disclosure may transmit an AI / ML synchronization command message to the UE (10). The base station (20) may transmit an AI / ML synchronization command message to the UE (10) based on / in response to a dynamic AI / ML capability report message. The UE (10) may receive an AI / ML synchronization command message from the base station (20). Step 540 may operate similarly to step 230 of FIG. 2, and redundant content is omitted here with reference to FIG. 2.

[0160] In step 550, the UE (10) and the base station (20) according to an embodiment of the present disclosure can perform AI / ML time synchronization with each other. Step 540 can operate similarly to step 240 of FIG. 2, and redundant details are omitted here with reference to FIG. 2.

[0161] FIG. 6 is a diagram illustrating a method for a UE and a base station to perform AI / ML time synchronization based on a Periodic method in a wireless communication system according to one embodiment of the present disclosure.

[0162] Referring to FIG. 6, in step 610, a base station (20) according to one embodiment of the present disclosure may transmit an RRC message (e.g., RRC reconfiguration message) including a dynamic AI / ML capability reporting period to a UE (10). The UE (10) may receive an RRC message (e.g., RRC reconfiguration message) including a dynamic AI / ML capability reporting period from the base station (20). Step 610 may operate similarly to step 210 of FIG. 2, and redundant details are omitted here with reference to FIG. 2.

[0163] In one embodiment, a UE (10) according to one embodiment of the present disclosure may periodically transmit a dynamic AI / ML capability report message to a base station (20) based on a dynamic AI / ML capability reporting period. For example, in FIG. 6, the UE (10) transmits a dynamic AI / ML capability report message to the base station (20) in step 620, and after a specific period has passed, transmits a dynamic AI / ML capability report message to the base station (20) in step 640, and continues to repeat this.

[0164] In steps 620 and 640, a UE (10) according to one embodiment of the present disclosure may transmit a dynamic AI / ML capability report message to a base station (20). The base station (20) may receive a dynamic AI / ML capability report message from the UE (10). Steps 620 and 640 may operate similarly to step 220 of FIG. 2, and overlapping content is omitted here with reference to FIG. 2.

[0165] In steps 630 and 650, the base station (20) according to an embodiment of the present disclosure may transmit an AI / ML synchronization command message to the UE (10). The base station (20) may transmit an AI / ML synchronization command message to the UE (10) based on / in response to a dynamic AI / ML capability report message. The UE (10) may receive an AI / ML synchronization command message from the base station (20). Steps 630 and 650 may operate similarly to step 230 of FIG. 2, and redundant details are omitted here with reference to FIG. 2.

[0166] In step 660, the UE (10) and the base station (20) according to an embodiment of the present disclosure can perform AI / ML time synchronization with each other. Step 660 can operate similarly to step 240 of FIG. 2, and redundant details are omitted here with reference to FIG. 2.

[0167] FIG. 7 is a diagram illustrating a method for a UE and a base station to perform AI / ML time synchronization based on a NW-initiated method in a wireless communication system according to one embodiment of the present disclosure.

[0168] Referring to FIG. 7, in step 710, a base station (20) according to one embodiment of the present disclosure can detect a performance degradation of the base station (20) side AI / ML model (e.g., when the base station side AI / ML capability degrades or when the KPI (key performance indicator) of the AI / ML-based communication system degrades). For example, the KPI (key performance indicator) of the AI / ML-based communication system may include network availability, data throughput, latency, packet loss rate, etc.

[0169] In step 720, a base station (20) according to one embodiment of the present disclosure may transmit an AI / ML synchronization request message to a UE (10). The UE (10) may receive an AI / ML synchronization request message from the base station (20). In one embodiment, the AI / ML synchronization request message may be transmitted via MAC CE (medium access control control element) or DCI (downlink control information) signaling.

[0170] In step 730, a UE (10) according to one embodiment of the present disclosure may transmit a dynamic AI / ML capability report message to a base station (20) in response to an AI / ML synchronization request message. The base station (20) may receive the dynamic AI / ML capability report message from the UE (10) in response to the AI / ML synchronization request message. Step 730 may operate similarly to step 220 of FIG. 2, and redundant details are omitted here with reference to FIG. 2.

[0171] In step 740, the base station (20) according to an embodiment of the present disclosure may transmit an AI / ML synchronization command message to the UE (10). The base station (20) may transmit an AI / ML synchronization command message to the UE (10) based on / in response to a dynamic AI / ML capability report message. The UE (10) may receive an AI / ML synchronization command message from the base station (20). Step 740 may operate similarly to step 230 of FIG. 2, and redundant content is omitted here with reference to FIG. 2.

[0172] In step 750, the UE (10) and the base station (20) according to an embodiment of the present disclosure can perform AI / ML time synchronization with each other. Step 750 can operate similarly to step 240 of FIG. 2, and redundant details are omitted here with reference to FIG. 2.

[0173] According to one embodiment of the present disclosure, the UE and the base station can improve the efficiency of an AI / ML-based communication system in synchronizing AI / ML processing times between the UE and the base station by triggering the UE's dynamic AI / ML capability reporting in various ways.

[0174] However, the effects obtainable from the present disclosure are not limited to those mentioned above, and other unmentioned effects will be clearly understood by those skilled in the art to which the present disclosure belongs from the description below.

[0175] Meanwhile, in the present disclosure, when a base station receives a dynamic AI / ML capability report message from a UE, it may perform an operation of transmitting an AI / ML model switching command message or an AI / ML model deactivation command message in addition to transmitting an AI / ML synchronization command message. Hereinafter, the above operation will be described in detail with reference to FIGS. 8 to 10b.

[0176] FIG. 8 is a diagram illustrating a method for a UE and a base station to perform AI / ML model switching or AI / ML model deactivation operations in a wireless communication system according to one embodiment of the present disclosure.

[0177] Referring to FIG. 8, in step 810, a base station (20) according to an embodiment of the present disclosure may transmit an RRC message containing configuration information regarding dynamic AI / ML capability reporting to a UE (10). The UE (10) may receive an RRC message (e.g., RRC reconfiguration message) containing configuration information regarding dynamic AI / ML capability reporting from the base station (20). Step 810 may operate similarly to step 210 of FIG. 2, and redundant content is omitted here with reference to FIG. 2.

[0178] In step 820, the UE (10) according to an embodiment of the present disclosure may transmit a dynamic AI / ML capability report message to the base station (20). The UE (10) may transmit a dynamic AI / ML capability report message to the base station (20) based on configuration information regarding the dynamic AI / ML capability report. The base station (20) may receive the dynamic AI / ML capability report message from the UE (10). Step 820 may operate similarly to step 220 of FIG. 2, and redundant details are omitted here with reference to FIG. 2. Step 820 may be triggered by the trigger methods of the UE's dynamic AI / ML capability report described in FIG. 5 to 7 of the present disclosure (e.g., event-triggered method, periodic method, or NW-triggered method), and redundant details are omitted here with reference to FIG. 5 to 7.

[0179] In step 830, the base station (20) according to an embodiment of the present disclosure may determine whether to perform AI / ML model switching or AI / ML model deactivation.

[0180] In one embodiment, the base station (20) may decide whether to replace the UE's AI / ML model with a lighter model (e.g., a neural network model having fewer parameters and layers, which has a shorter AI / ML processing time) if there is a lighter model that can be transmitted to the UE (10). This will be explained in more detail with reference to FIGS. 9a and 9b.

[0181] In one embodiment, the base station (20) may decide whether to stop using the AI / ML model in the UE if there is no lighter model available to transmit to the UE (10) or if other factors exist. This will be explained in more detail with reference to FIGS. 10a and 10b.

[0182] At step 840, a base station (20) according to an embodiment of the present disclosure may transmit an AI / ML model switching / switching command message (AI / ML model switching command message) or an AI / ML model deactivation command message (AI / ML model deactivation command message) to a UE (10) in response to a dynamic AI / ML capability report message. The UE (10) may receive the AI / ML model switching command message or the AI / ML model deactivation command message from the base station (20). In one embodiment, the AI / ML model switching command message may be transmitted via radio resource control (RRC) signaling or user plane (UP) data, and the AI / ML model deactivation command message may be transmitted via RRC signaling.

[0183] In one embodiment, the AI / ML model switching command message may instruct the UE to replace its AI / ML model with a lighter model (e.g., a neural network model having fewer parameters and layers, resulting in a shorter AI / ML processing time). In one embodiment, the AI / ML model switching command message may include the lighter model itself or an identifier representing the lighter model.

[0184] In one embodiment, the AI / ML model deactivation command message may instruct the UE to stop using the AI / ML model. In one embodiment, the AI / ML model deactivation command message may instruct the UE to stop using the AI / ML model by using an identifier of the AI / ML model or AI / ML functionality and a corresponding 1 bit (e.g., if set to '1', instruct to stop).

[0185] In step 850, the UE (10) according to an embodiment of the present disclosure may perform AI / ML model switching or AI / ML model deactivation.

[0186] In one embodiment, when the UE (10) receives an AI / ML model switching command message from the base station (20), it may operate using a lighter model as indicated by the AI / ML model switching command message.

[0187] In one embodiment, when the UE (10) receives an AI / ML model deactivation command message from the base station (20), it may stop using the AI / ML model and fall back to a non-AI operation.

[0188] FIGS. 9a and 9b are drawings for explaining a method in which a base station in a wireless communication system according to one embodiment of the present disclosure determines whether to perform AI / ML model switching or AI / ML model deactivation operations based on Delay A parameter and Delay C parameter.

[0189] The Delay A parameter is the average value of the UE's AI / ML processing time, and the Delay C parameter is the average value of the UE's AI / ML inference time. Refer to Figure 3 for the definitions of the Delay A and C parameters, and redundant explanations are omitted.

[0190] FIG. 9a illustrates an example in which, in a scenario where a UE and a base station share an AI / ML model according to one embodiment of the present disclosure, the base station determines whether to perform an AI / ML model switching or AI / ML model deactivation operation based on the Delay A parameter and the Delay C parameter.

[0191] Referring to FIG. 9a, in step 910a, a base station according to one embodiment of the present disclosure may receive a dynamic AI / ML capability report message from a UE. Step 910 may operate similarly to step 220 of FIG. 2, and redundant details are omitted here with reference to FIG. 2.

[0192] In step 920a, a base station according to one embodiment of the present disclosure has a Delay A parameter in a dynamic AI / ML capability report message received from a UE, which is a specific threshold value ( It can determine whether it exceeds a specific threshold ( ) is a threshold value related to the Delay A parameter, which the base station can set arbitrarily. If the Delay A parameter has a specific threshold value ( Exceeding ) may mean that the UE's AI / ML performance is degraded.

[0193] If, at step 920a, the base station, the Delay A parameter is at a specific threshold ( If it is determined that it does not exceed ) (meaning there is no or not significant degradation of the UE's AI / ML performance), the base station may perform the step 960a operation (=time synchronization operation with the UE and AI / ML).

[0194] If, at step 920a, the base station, the Delay A parameter is at a specific threshold ( If it is determined that ) exceeds (=meaning that the UE's AI / ML performance is significantly degraded), the base station may perform step 930a operation.

[0195] In step 930a, a base station according to one embodiment of the present disclosure can determine whether there is a model (i.e., AI / ML model k) lighter than the current AI / ML model of the UE.

[0196] For example, the base station can determine whether there exists a model (i.e., AI / ML model k) that is lighter than the UE's current AI / ML model according to [Equation 6].

[0197] [Mathematical Formula 6]

[0198]

[0199] The parameter is the average value of the UE's AI / ML processing time estimated assuming the UE uses AI / ML model k. The parameter is the average value of the UE's estimated AI / ML inference time, assuming the UE uses AI / ML model k. A specific threshold ( ) is a threshold value related to the Delay A parameter, which can be arbitrarily set by the base station.

[0200] If the base station at step 930a, Parameters have a specific threshold ( If it is determined that it is less than or not equal to (= exceeds) (= meaning that there is no model lighter than the current AI / ML model of the UE (i.e., AI / ML model k)), the base station may perform the operation of step 940a.

[0201] If the base station at step 930a, Parameters have a specific threshold ( If it is determined that it is less than or equal to ) (meaning that there exists a model lighter than the current AI / ML model of the UE (i.e., AI / ML model k), the base station can perform the operation of step 950a.

[0202] In step 940a, a base station according to one embodiment of the present disclosure is, Parameters have a specific threshold ( If it is determined that ) is less than or not equal to (= exceeds), it is determined that there is no model (i.e., AI / ML model k) lighter than the UE's current AI / ML model, and accordingly, an AI / ML model deactivation command message can be sent to the UE. For a description of the AI / ML model deactivation command message, refer to FIG. 8, and redundant descriptions are omitted here. Based on the AI / ML model deactivation command message, the UE can stop using the AI / ML model and fallback to non-AI operation.

[0203] In step 950a, a base station according to one embodiment of the present disclosure is, Parameters have a specific threshold ( If it is determined that ) is smaller than or equal to, it is determined that a lighter model (i.e., AI / ML model k) exists than the UE's current AI / ML model, and accordingly, an AI / ML model switching command message can be sent to the UE. For a description of the AI / ML model switching command message, refer to FIG. 8, and redundant descriptions are omitted here. The UE can operate using the lighter model (i.e., AI / ML model k) indicated by the AI / ML model switching command message.

[0204] In step 960a, a base station according to one embodiment of the present disclosure may transmit an AI / ML synchronization command message to the UE. If, in step 920a, the base station, the Delay A parameter is a specific threshold ( If it is determined that it does not exceed ) (meaning that there is no or not significant degradation in the UE's AI / ML performance), the base station performs the operation of step 960a to synchronize AI / ML time with the UE. Step 960a may operate similarly to step 230 of FIG. 2, and any overlapping details are omitted here and refer to FIG. 2.

[0205] FIG. 9b illustrates an example in which, in a scenario according to one embodiment of the present disclosure where the UE and the base station do not share an AI / ML model and only the UE side uses the AI / ML model, the base station determines whether to perform an AI / ML model switching or AI / ML model deactivation operation based on the Delay A parameter and the Delay C parameter.

[0206] Referring to Fig. 9b, since the scenario in Fig. 9b is one where the UE and the base station do not share an AI / ML model and only the UE side uses the AI / ML model, in step 920a, if the base station, the Delay A parameter is at a specific threshold ( Even if it is determined that it does not exceed ) (meaning there is no or not significant degradation of the UE's AI / ML performance), the base station may not perform the operation for AI / ML time synchronization with the UE and may perform step 910b or step 920b again.

[0207] The operations of steps 910b to 950b of FIG. 9b, excluding this, may operate similarly to steps 910a to 950a of FIG. 9a, and refer to the description of FIG. 9a, and redundant descriptions are omitted here.

[0208] FIGS. 10a and 10b are drawings for explaining a method in which a base station in a wireless communication system according to one embodiment of the present disclosure determines whether to perform AI / ML model switching or AI / ML model deactivation operation based on the Delay B parameter.

[0209] The Delay B parameter is the average delay time between the latest update time of the UE's input (x) and the start time of AI / ML processing. Refer to Figure 3 for the definition of the Delay B parameter, and redundant explanations are omitted.

[0210] FIG. 10a illustrates an example in which, in a scenario where a UE and a base station share an AI / ML model according to one embodiment of the present disclosure, the base station determines whether to perform an AI / ML model switching or AI / ML model deactivation operation based on the Delay B parameter.

[0211] Referring to FIG. 10a, in step 1010a, a base station according to one embodiment of the present disclosure may receive a dynamic AI / ML capability report message from a UE. Step 1010 may operate similarly to step 220 of FIG. 2, and redundant details are omitted here with reference to FIG. 2.

[0212] In step 1020a, a base station according to one embodiment of the present disclosure has a Delay B parameter in a dynamic AI / ML capability report message received from a UE that is at a specific threshold ( It can determine whether it exceeds a specific threshold ( ) is a threshold value related to the Delay B parameter, which the base station can set arbitrarily. If the Delay B parameter has a specific threshold value ( Exceeding ) can mean that the UE's AI / ML performance is significantly degraded.

[0213] If, in step 1020a, the base station, the Delay B parameter is at a specific threshold ( If it is determined that it does not exceed ) (=meaning there is no or not significant degradation of the UE's AI / ML performance), the base station can perform the step 1060a operation (=time synchronization operation with the UE and AI / ML).

[0214] If, in step 1020a, the base station, the Delay B parameter is at a specific threshold ( If it is determined that ) exceeds (=meaning that the UE's AI / ML performance is significantly degraded), the base station may perform step 1030a operation.

[0215] In step 1030a, a base station according to one embodiment of the present disclosure may determine whether a higher model (i.e., AI / ML model Z) exists than the current AI / ML model of the UE and whether it is possible to change to said higher model. The higher model (i.e., AI / ML model Z) refers to a model having higher accuracy (e.g., having more parameters or a complex layer structure). The ranking among the AI / ML models used by the UE or the base station may be predefined.

[0216] If, in step 1030a, the base station determines that there is no higher-level model (i.e., AI / ML model Z) than the UE's current AI / ML model, the base station may perform the operation in step 1040a.

[0217] If, in step 1030a, the base station determines that there is a higher model (i.e., AI / ML model Z) than the UE's current AI / ML model, the base station may perform the operation in step 1050a.

[0218] In step 1040a, if the base station according to one embodiment of the present disclosure determines that there is no changeable upper model (i.e., AI / ML model Z), it may transmit an AI / ML model deactivation command message to the UE. For a description of the AI / ML model deactivation command message, refer to FIG. 8, and redundant descriptions are omitted here. Based on the AI / ML model deactivation command message, the UE may stop using the AI / ML model and fallback to a non-AI operation.

[0219] In step 1050a, if a base station according to one embodiment of the present disclosure determines that a changeable upper model (i.e., AI / ML model Z) exists, it may transmit an AI / ML model switching command message to the UE. For a description of the AI / ML model switching command message, refer to FIG. 8, and redundant descriptions are omitted here. The UE may operate using the upper model (i.e., AI / ML model Z) indicated by the AI / ML model switching command message.

[0220] In step 1060a, a base station according to one embodiment of the present disclosure may transmit an AI / ML synchronization command message to the UE. If, in step 1020a, the base station, the Delay B parameter is a specific threshold ( If it is determined that it does not exceed ) (meaning that there is no or not significant degradation in the UE's AI / ML performance), the base station performs the operation of step 1060a to synchronize AI / ML time with the UE. Step 1060a may operate similarly to step 230 of FIG. 2, and any overlapping details are omitted here and refer to FIG. 2.

[0221] FIG. 10b illustrates an example in which, in a scenario according to one embodiment of the present disclosure where the UE and the base station do not share an AI / ML model and only the UE side uses the AI / ML model, the base station determines whether to perform an AI / ML model switching or AI / ML model deactivation operation based on the Delay B parameter.

[0222] Referring to FIG. 10b, since the scenario in FIG. 10b is one where the UE and the base station do not share an AI / ML model and only the UE side uses the AI / ML model, in step 1020a, if the base station, the Delay B parameter is at a specific threshold ( Even if it is determined that it does not exceed ) (meaning there is no or not significant degradation of the UE's AI / ML performance), the base station may not perform the operation for AI / ML time synchronization with the UE and may perform step 1010b or step 1020b again.

[0223] Except for this, the operation of steps 1010b to 1050b of FIG. 10b may operate similarly to steps 1010a to 1050a of FIG. 10a, and refer to the description of FIG. 10a, and redundant descriptions are omitted here.

[0224] According to one embodiment of the present disclosure, the UE and the base station can improve the performance of an AI / ML-based communication system by replacing or discontinuing the use of the UE's AI / ML model, taking into account the UE's dynamic AI / ML capability.

[0225] However, the effects obtainable from the present disclosure are not limited to those mentioned above, and other unmentioned effects will be clearly understood by those skilled in the art to which the present disclosure belongs from the description below.

[0226] FIG. 11 is a diagram illustrating a method for a UE to perform AI / ML time synchronization in a wireless communication system according to one embodiment of the present disclosure.

[0227] Referring to FIG. 11, in step 1110, a UE according to one embodiment of the present disclosure may receive a Radio Resource Control (RRC) message from a base station containing configuration information regarding a dynamic AI / ML capability report. Step 1110 may operate similarly to step 210 of FIG. 2, and refer to the description of FIG. 2, and redundant descriptions are omitted here.

[0228] In step 1120, a UE according to one embodiment of the present disclosure may transmit a dynamic AI / ML capability report message to a base station. Step 1120 may operate similarly to step 220 of FIG. 2, and refer to the description of FIG. 2, and redundant descriptions are omitted herein.

[0229] In step 1130, a UE according to one embodiment of the present disclosure may receive an AI / ML synchronization command message from a base station in response to a dynamic AI / ML capability report message. Step 1130 may operate similarly to step 230 of FIG. 2, and refer to the description of FIG. 2, and redundant descriptions are omitted herein.

[0230] In step 1140, a UE according to one embodiment of the present disclosure may perform AI / ML time synchronization with a base station based on an AI / ML synchronization command message. Step 1140 may operate similarly to step 240 of FIG. 2, and refer to the description of FIG. 2, and redundant descriptions are omitted here.

[0231] A UE according to one embodiment of the present disclosure can improve the performance of an AI / ML-based communication system by synchronizing the AI / ML processing time with a base station through dynamic AI / ML capability report messages and AI / ML synchronization command messages by applying parameters related to the UE's dynamic AI / ML capability and parameters required for AI / ML time synchronization.

[0232] However, the effects obtainable from the present disclosure are not limited to those mentioned above, and other unmentioned effects will be clearly understood by those skilled in the art to which the present disclosure belongs from the description below.

[0233] FIG. 12 is a diagram illustrating a method for a base station to perform AI / ML time synchronization in a wireless communication system according to one embodiment of the present disclosure.

[0234] Referring to FIG. 12, in step 1210, a base station according to one embodiment of the present disclosure may transmit a Radio Resource Control (RRC) message to a UE that includes configuration information regarding a dynamic AI / ML capability report. Step 1210 may operate similarly to step 210 of FIG. 2, and reference is made to the description of FIG. 2, and redundant descriptions are omitted here.

[0235] In step 1220, a base station according to one embodiment of the present disclosure may receive a dynamic AI / ML capability report message from a UE. Step 1220 may operate similarly to step 220 of FIG. 2, and refer to the description of FIG. 2, and redundant descriptions are omitted herein.

[0236] In step 1230, a base station according to one embodiment of the present disclosure may transmit an AI / ML synchronization command message to a UE in response to a dynamic AI / ML capability report message. Step 1230 may operate similarly to step 230 of FIG. 2, and refer to the description of FIG. 2, and redundant descriptions are omitted here.

[0237] In step 1240, a base station according to one embodiment of the present disclosure may perform AI / ML time synchronization with a UE. Step 1240 may operate similarly to step 240 of FIG. 2, and refer to the description of FIG. 2, and redundant descriptions are omitted here.

[0238] A base station according to one embodiment of the present disclosure can improve the performance of an AI / ML-based communication system by synchronizing the AI / ML processing time with the UE through dynamic AI / ML capability report messages and AI / ML synchronization command messages, by applying parameters related to the UE's dynamic AI / ML capability and parameters required for AI / ML time synchronization.

[0239] However, the effects obtainable from the present disclosure are not limited to those mentioned above, and other unmentioned effects will be clearly understood by those skilled in the art to which the present disclosure belongs from the description below.

[0240] FIG. 13 is a block diagram of a UE (1300) according to one embodiment of the present disclosure. UE 1300 may correspond to UE 10 of the present disclosure.

[0241] Referring to FIG. 13, the UE (1300) may be composed of a transceiver (1310), a processor (1320), and a memory (1330). Depending on the communication method of the UE (1300) described above, the transceiver (1310), the processor (1320), and the memory (1330) of the UE (1300) may operate. However, the components of the UE (1300) are not limited to the examples described above. For example, the UE (1300) may include more components or fewer components than the components described above. In one embodiment, the transceiver (1310), the processor (1320), and the memory (1330) may be implemented in the form of a single chip. Additionally, the processor (1320) may include one or more processors.

[0242] The transceiver (1310) is a collective term for the receiver and the transmitter of the UE (1300), and can transmit and receive signals with a network entity including a base station (1400). The signals transmitted and received with the network entity including the base station (1400) may include control information and data. To this end, the transceiver (1310) may be composed of an RF transmitter that up-converts and amplifies the frequency of a transmitted signal, and an RF receiver that low-noise amplifies a received signal and down-converts the frequency. However, this is one embodiment of the transceiver (1310), and the components of the transceiver (1310) are not limited to an RF transmitter and an RF receiver.

[0243] Additionally, the transceiver (1310) can perform functions for transmitting and receiving signals through a wireless channel. For example, the transceiver (1310) can receive a signal through a wireless channel and output it to a processor (1320), and transmit the signal output from the processor (1320) through a wireless channel.

[0244] Memory (1330) can store programs and data necessary for the operation of the UE (1300). Additionally, memory (1330) can store control information or data included in signals obtained from the UE (1300). Memory (1330) may be composed of a storage medium or a combination of storage media such as ROM, RAM, hard disk, CD-ROM, and DVD. Additionally, memory (1330) may not exist separately but may be configured to be included in the processor (1320). Memory (1330) may be composed of volatile memory, non-volatile memory, or a combination of volatile and non-volatile memory. Furthermore, memory (1330) can provide stored data upon the request of the processor (1320). Computer programs, code, or instructions that can be executed by the processor (1320) may be stored in memory (1330). According to one embodiment, a computer program, code, or instruction that can be executed by a processor (1320) may be stored in a single memory device or may be separated and distributed across two or more memory devices. The processor (1320) may perform various functions according to an embodiment of the present disclosure by executing instructions stored in memory (1330). According to one embodiment of the present disclosure, the operation of the UE (1300) may be caused to be performed based on at least one processor (or processing circuit), a processing circuitry not configured to execute instructions, and / or a component of a processing circuitry not configured to execute instructions, configured to perform the features of the present disclosure individually, collectively, or in any combination based on the execution of instructions (or computer program or code) stored in memory (1330).

[0245] The processor (1320) can control a series of processes to enable the UE (1300) to operate according to the embodiments of the present disclosure described above. For example, the processor (1320) can receive control signals and data signals through the transceiver (1310) and process the received control signals and data signals. The processor (1320) can transmit the processed control signals and data signals through the transceiver (1310). Additionally, the processor (1320) can write or read data to or from memory (1330). The processor (1320) can perform the functions of a protocol stack required by a communication standard. To this end, the processor (1320) may include at least one processor or microprocessor. In one embodiment, a part of the transceiver (1310) or the processor (1320) may be referred to as a communication processor (CP).

[0246] The processor (1320) may be composed of one or more processors. In this case, the one or more processors may be general-purpose processors such as CPUs, APs, and DSPs (Digital Signal Processors), graphics-dedicated processors such as GPUs and VPUs (Vision Processing Units), or artificial intelligence-dedicated processors such as NPUs. For example, if one or more processors are artificial intelligence-dedicated processors, the artificial intelligence-dedicated processors may be designed with a hardware structure specialized for processing a specific artificial intelligence model. The processor (1320) may include at least one processor (or processing circuitry), and at least one processor may perform the following operations individually, collectively, or in any combination. For example, the processor (1320) may include a communication processor (CP) that controls communication operations and an application processor (AP) that controls the execution of an upper layer (e.g., an application layer). In a specific embodiment, at least one part of the processor (1320) may be included in one chip, and another part of the processor (1320) may be included in a separate chip. Alternatively, at least one processor may be included in other components, such as a transceiver (1310) or memory (1330). To this end, the processor (1320) may control other components of the UE (1300) to perform various operations by executing computer programs, code, or instructions stored in memory (1330).

[0247] In one embodiment, by the processor (1320) executing instructions stored in memory (1330) alone or in combination, the UE (1300) can receive a Radio Resource Control (RRC) message from a base station containing configuration information regarding a dynamic AI / ML capability report. By the processor (1320) executing instructions stored in memory (1330) alone or in combination, the UE (1300) can transmit a dynamic AI / ML capability report message to the base station. By the processor (1320) executing instructions stored in memory (1330) alone or in combination, the UE (1300) can receive an AI / ML synchronization command message from the base station in response to the dynamic AI / ML capability report message. By the processor (1320) executing instructions stored in memory (1330) alone or in combination, the UE (1300) can perform AI / ML time synchronization with the base station based on the AI / ML synchronization command message.

[0248] In one embodiment, the dynamic AI / ML capability report message may include at least one of: the average value of the UE's AI / ML processing time; the average delay time between the latest update time of the UE's AI / ML input and the start time of AI / ML processing; the average value of the UE's AI / ML inference time; or the average update period of the UE's AI / ML input.

[0249] In one embodiment, the AI / ML synchronization command message may include at least one of an AI / ML input offset value or an AI / ML process offset value.

[0250] In one embodiment, by the processor (1320) executing instructions stored in memory (1330) alone or in combination, the UE (1300) can use an input corresponding to an AI / ML input offset value as an input to an AI / ML model based on an AI / ML input offset value. By the processor (1320) executing instructions stored in memory (1330) alone or in combination, the UE (1300) can determine the timing for applying the processing result of the AI / ML model based on an AI / ML process offset value.

[0251] In one embodiment, configuration information regarding a dynamic AI / ML capability may include at least one of: identifier (ID) information; triggering conditions for a dynamic AI / ML capability report; a dynamic AI / ML capability report period; or a type of a dynamic AI / ML capability report.

[0252] In one embodiment, by the processor (1320) executing instructions stored in memory (1330) alone or in combination, the UE (1300) can determine whether the triggering conditions for a dynamic AI / ML capability report are satisfied. By the processor (1320) executing instructions stored in memory (1330) alone or in combination, the UE (1300) can transmit a dynamic AI / ML capability report message to a base station if the triggering conditions for a dynamic AI / ML capability report are satisfied.

[0253] In one embodiment, by the processor (1320) executing instructions stored in memory (1330) alone or in combination, the UE (1300) can periodically transmit dynamic AI / ML capability report messages to the base station based on the dynamic AI / ML capability report cycle.

[0254] In one embodiment, by the processor (1320) executing instructions stored in memory (1330) alone or in combination, the UE (1300) can receive an AI / ML synchronization request message from a base station. By the processor (1320) executing instructions stored in memory (1330) alone or in combination, the UE (1300) can send a dynamic AI / ML capability report message to the base station in response to the AI / ML synchronization request message.

[0255] In one embodiment, by the processor (1320) executing a command stored in memory (1330) alone or in combination, the UE (1300) may receive an AI / ML model switching command message or an AI / ML model deactivation command message from the base station in response to a dynamic AI / ML capability report message.

[0256] FIG. 14 is a block diagram of a base station (1400) according to one embodiment of the present disclosure. Base station 1400 may correspond to base station 20 of the present disclosure.

[0257] Referring to FIG. 14, a base station (1400) may be composed of a transceiver (1410), a processor (1420), and a memory (1430). Depending on the communication method of the base station (1400) described above, the transceiver (1410), the processor (1420), and the memory (1430) of the base station (1400) may operate. However, the components of the base station (1400) are not limited to the examples described above. For example, the base station (1400) may include more components or fewer components than the components described above. In one embodiment, the transceiver (1410), the processor (1420), and the memory (1430) may be implemented in the form of a single chip. Additionally, the processor (1420) may include one or more processors.

[0258] The transceiver (1410) is a collective term for the receiver and the transmitter of the base station (1400), and can transmit and receive signals with a network entity including the UE (1300). The signals transmitted and received with the network entity including the UE (1300) may include control information and data. To this end, the transceiver (1410) may be composed of an RF transmitter that up-converts and amplifies the frequency of the transmitted signal, and an RF receiver that low-noise amplifies the received signal and down-converts the frequency. However, this is one embodiment of the transceiver (1410), and the components of the transceiver (1410) are not limited to an RF transmitter and an RF receiver.

[0259] Additionally, the transceiver (1410) can perform functions for transmitting and receiving signals through a wireless channel. For example, the transceiver (1410) can receive a signal through a wireless channel and output it to a processor (1420), and transmit the signal output from the processor (1420) through a wireless channel.

[0260] The memory (1430) can store programs and data necessary for the operation of the base station (1400). Additionally, the memory (1430) can store control information or data included in signals obtained from the base station (1400). The memory (1430) may be composed of a storage medium or a combination of storage media such as ROM, RAM, hard disk, CD-ROM, and DVD. Additionally, the memory (1430) may not exist separately but may be configured to be included in the processor (1420). The memory (1430) may be composed of volatile memory, non-volatile memory, or a combination of volatile and non-volatile memory. Furthermore, the memory (1430) can provide stored data upon the request of the processor (1420). Computer programs, code, or instructions that can be executed by the processor (1420) may be stored in the memory (1430). According to one embodiment, a computer program, code, or instruction that can be executed by a processor (1420) may be stored in a single memory device or may be separated and distributed among two or more memory devices. The processor (1420) may perform various functions according to an embodiment of the present disclosure by executing instructions stored in memory (1430). According to one embodiment of the present disclosure, the operation of a base station (1400) may be caused to be performed based on at least one processor (or processing circuit), a processing circuitry not configured to execute instructions, and / or a component of a processing circuitry not configured to execute instructions, configured to perform the features of the present disclosure individually, collectively, or in any combination based on the execution of instructions (or computer program or code) stored in memory (1430).

[0261] The processor (1420) can control a series of processes to enable the base station (1400) to operate according to the embodiments of the present disclosure described above. For example, the processor (1420) can receive control signals and data signals through the transceiver (1410) and process the received control signals and data signals. The processor (1420) can transmit the processed control signals and data signals through the transceiver (1410). Additionally, the processor (1420) can write or read data to or from the memory (1430). The processor (1420) can perform the functions of the protocol stack required by the communication standard. To this end, the processor (1420) may include at least one processor or microprocessor. In one embodiment, a part of the transceiver (1410) or the processor (1420) may be referred to as a communication processor (CP).

[0262] The processor (1420) may be composed of one or more processors. In this case, the one or more processors may be general-purpose processors such as CPUs, APs, and DSPs (Digital Signal Processors), graphics-dedicated processors such as GPUs and VPUs (Vision Processing Units), or artificial intelligence-dedicated processors such as NPUs. For example, if one or more processors are artificial intelligence-dedicated processors, the artificial intelligence-dedicated processors may be designed with a hardware structure specialized for processing a specific artificial intelligence model. The processor (1420) may include at least one processor (or processor circuitry), and at least one processor may perform the following operations individually, collectively, or in any combination. In a specific embodiment, at least one part of the processor (1420) may be included in one chip, and another part of the processor (1420) may be included in a separate chip. Alternatively, at least one processor may be included in other components, such as a transceiver (1410) or memory (1430). The processor (1420) may perform, cause, or control operations of a base station for performing at least one or a combination thereof of the methods according to embodiments of the present disclosure. To this end, the processor (1420) may control other components of the base station (1400) to perform various operations by executing computer programs, code, and instructions stored in memory (1430).

[0263] In one embodiment, by the processor (1420) executing commands stored in memory (1430) alone or in combination, the base station (1400) can transmit a Radio Resource Control (RRC) message containing configuration information regarding dynamic AI / ML capability reporting to the UE (user equipment). By the processor (1420) executing commands stored in memory (1430) alone or in combination, the base station (1400) can receive a dynamic AI / ML capability reporting message from the UE. By the processor (1420) executing commands stored in memory (1430) alone or in combination, the base station (1400) can transmit an AI / ML synchronization command message to the UE in response to the dynamic AI / ML capability reporting message. By the processor (1420) executing commands stored in memory (1430) alone or in combination, the base station (1400) can perform AI / ML time synchronization with the UE.

[0264] In one embodiment, the dynamic AI / ML capability report message may include at least one of: the average value of the UE's AI / ML processing time; the average delay time between the latest update time of the UE's AI / ML input and the start time of AI / ML processing; the average value of the UE's AI / ML inference time; or the average update period of the UE's AI / ML input.

[0265] In one embodiment, by having the processor (1420) execute a command stored in memory (1430) alone or in combination, the base station (1400) can determine an AI / ML input offset value and an AI / ML process offset value in response to a dynamic AI / ML capability report message. The AI / ML synchronization command message may include at least one of an AI / ML input offset value or an AI / ML process offset value.

[0266] In one embodiment, by the processor (1420) executing instructions stored in memory (1430) alone or in combination, the base station (1400) can predict the input to be used by the UE based on the AI / ML input offset value and use it as an input to the AI / ML model. By the processor (1420) executing instructions stored in memory (1430) alone or in combination, the base station (1400) can determine the timing for applying the processing result of the AI / ML model based on the AI / ML process offset value.

[0267] In one embodiment, configuration information regarding a dynamic AI / ML capability may include at least one of: identifier (ID) information; triggering conditions for a dynamic AI / ML capability report; a dynamic AI / ML capability report period; or a type of a dynamic AI / ML capability report.

[0268] In one embodiment, by the processor (1420) executing instructions stored in memory (1430) alone or in combination, the base station (1400) can receive a dynamic AI / ML capability report message from the UE when the triggering condition for a dynamic AI / ML capability report is satisfied.

[0269] In one embodiment, by the processor (1420) executing instructions stored in memory (1430) alone or in combination, the base station (1400) can periodically receive the dynamic AI / ML capability report message from the UE based on the dynamic AI / ML capability report cycle.

[0270] In one embodiment, by the processor (1420) executing instructions stored in memory (1430) alone or in combination, the base station (1400) can transmit an AI / ML synchronization request message to the UE. By the processor (1420) executing instructions stored in memory (1430) alone or in combination, the base station (1400) can receive a dynamic AI / ML capability report message from the UE in response to the AI / ML synchronization request message.

[0271] In one embodiment, by the processor (1420) executing a command stored in memory (1430) alone or in combination, the base station (1400) can transmit an AI / ML model switching command message or an AI / ML model deactivation command message to the UE in response to a dynamic AI / ML capability report message.

[0272] The specific example for explaining the embodiment according to the present disclosure is merely one combination of each standard, method, detailed method, and operation, and through a combination of at least two of the various techniques described, the base station and the UE can improve communication performance by synchronizing AI / ML time in an AI / ML-based wireless communication system. In addition, this may be performed according to a method determined through one or a combination of at least two of the aforementioned techniques. For example, it may be possible to perform a part of the operation of one embodiment in combination with a part of the operation of another embodiment.

[0273] A device-readable storage medium may be provided in the form of a non-transitory storage medium. Here, 'non-transitory storage medium' simply means that it is a tangible device and does not contain a signal (e.g., electromagnetic waves), and the term does not distinguish between cases where data is stored semi-permanently and cases where it is stored temporarily. For example, a 'non-transitory storage medium' may include a buffer in which data is stored temporarily.

[0274] According to one embodiment, the method according to the various embodiments disclosed herein may be provided by being included in a computer program product. The computer program product may be traded between a seller and a buyer as a product. The computer program product may be distributed in the form of a device-readable storage medium (e.g., compact disc read-only memory (CD-ROM)), or distributed online (e.g., download or upload) through an application store or directly between two user devices (e.g., smartphones). In the case of online distribution, at least a portion of the computer program product (e.g., downloadable app) may be temporarily stored or temporarily created on a device-readable storage medium, such as the memory of a manufacturer's server, an application store's server, or a relay server.

[0275] Meanwhile, although specific embodiments have been described in the detailed description of the present disclosure, it is understood that various modifications are possible within the scope of the present disclosure. Therefore, the scope of the present disclosure should not be limited to the described embodiments, but should be defined by the claims set forth below as well as equivalents thereof.

Claims

1. In a method of UE (User Equipment) in a wireless communication system, A step of receiving a Radio Resource Control (RRC) message from a base station containing configuration information regarding dynamic AI / ML capability reporting; A step of transmitting a dynamic AI / ML capability report message to the above base station; Receiving an AI / ML synchronization command message from the base station in response to the dynamic AI / ML capability report message; and A method comprising the step of performing AI / ML time synchronization with the base station based on the above AI / ML synchronization command message.

2. In claim 1, the dynamic AI / ML capability report message is, Average value of the AI / ML processing time of the above UE; Average latency between the latest update time of the AI / ML input of the above UE and the start time of AI / ML processing; The average value of the AI / ML inference time of the above UE; or A method comprising at least one of the average update cycle of the AI / ML input of the above UE.

3. In claim 1, the AI / ML synchronization command message is, AI / ML input offset value; or A method comprising at least one of an AI / ML process offset value.

4. In Paragraph 3, Based on the above AI / ML input offset value, a step of using an input corresponding to the above AI / ML input offset value as an input to an AI / ML model; and A method further comprising the step of determining the timing for applying the processing result of the AI / ML model based on the AI / ML process offset value.

5. In claim 1, the configuration information regarding the dynamic AI / ML capability is, Identifier (ID) information; Triggering conditions for dynamic AI / ML capability reporting; dynamic AI / ML capability reporting cycle; or A method comprising at least one of the types of dynamic AI / ML capability reporting.

6. In claim 5, the step of transmitting the dynamic AI / ML capability report message is, A step of determining whether the triggering condition of the above dynamic AI / ML capability report is satisfied; and A method comprising the step of transmitting a dynamic AI / ML capability report message to the base station when the triggering condition of the dynamic AI / ML capability report is satisfied.

7. In claim 5, the step of transmitting the dynamic AI / ML capability report message is, A method comprising the step of periodically transmitting the dynamic AI / ML capability report message to the base station based on the dynamic AI / ML capability report cycle.

8. In claim 1, the step of transmitting the dynamic AI / ML capability report message is, A step of receiving an AI / ML synchronization request message from the base station; and A method comprising the step of transmitting the dynamic AI / ML capability report message to the base station in response to the above AI / ML synchronization request message.

9. In Paragraph 1, A method further comprising the step of receiving an AI / ML model switching command message or an AI / ML model deactivation command message from the base station in response to the dynamic AI / ML capability report message.

10. In a method of a base station in a wireless communication system, A step of transmitting a Radio Resource Control (RRC) message containing configuration information regarding dynamic AI / ML capability reporting to a User Equipment (UE); A step of receiving a dynamic AI / ML capability report message from the above UE; A step of transmitting an AI / ML synchronization command message to the UE in response to the dynamic AI / ML capability report message; and A method comprising the step of performing AI / ML time synchronization with the above UE.

11. In claim 10, the dynamic AI / ML capability reporting message is, Average value of the AI / ML processing time of the above UE; Average latency between the latest update time of the AI / ML input of the above UE and the start time of AI / ML processing; The average value of the AI / ML inference time of the above UE; or A method comprising at least one of the average update cycle of the AI / ML input of the above UE.

12. In Paragraph 10, In response to the above dynamic AI / ML capability report message, the method further includes the step of determining an AI / ML input offset value and an AI / ML process offset value. A method wherein the above AI / ML synchronization command message comprises at least one of the above AI / ML input offset value; or the above AI / ML process offset value.

13. In Paragraph 12, Based on the above AI / ML input offset value, a step of predicting the input to be used by the UE and using it as the input to the AI / ML model; and A method further comprising the step of determining the timing for applying the processing result of the AI / ML model based on the AI / ML process offset value.

14. In a wireless communication system, regarding the UE (User Equipment), One or more transceivers; One or more processors coupled to communicate with the above one or more transceivers; and One or more memories coupled to communicate with the above one or more processors; including, The above one or more memories store instructions that the above one or more processors can execute alone or in combination, and the instructions are configured so that the UE performs the following: Receive a Radio Resource Control (RRC) message from a base station containing configuration information regarding dynamic AI / ML capability reporting; Transmit a dynamic AI / ML capability report message to the above base station; In response to the above dynamic AI / ML capability report message, receive an AI / ML synchronization command message from the base station; and A UE that performs AI / ML time synchronization with the base station based on the above AI / ML synchronization command message.

15. In a base station of a wireless communication system, One or more transceivers; One or more processors coupled to communicate with the above one or more transceivers; and One or more memories coupled to communicate with the above one or more processors; including, The above one or more memories store instructions that the above one or more processors can execute alone or in combination, and the instructions are configured so that the base station performs the following: Transmit an RRC (Radio Resource Control) message containing configuration information regarding dynamic AI / ML capability reporting to the UE (user equipment); Receive a dynamic AI / ML capability report message from the above UE; In response to the above dynamic AI / ML capability report message, transmit an AI / ML synchronization command message to the UE; and A base station that performs AI / ML time synchronization with the above-mentioned UE.

Images