Method and device for managing artificial intelligence-related memory and storage in wireless communication system

Abstract

The present disclosure relates to a 5G or 6G communication system for supporting a data transmission rate higher than that of a 4G communication system such as LTE. A method of a base station in a wireless communication system, according to one embodiment of the present disclosure, comprises the steps of: receiving, from a terminal, a first message including artificial intelligence (AI)-related memory information and / or AI-related storage information; determining AI model information on the basis of the AI-related memory information and / or the AI-related storage information; and transmitting the AI model information to the terminal.

Description

Method and device for managing memory and storage related to artificial intelligence in a wireless communication system

[0001] The present disclosure relates to a wireless communication system, and more specifically, to a method and apparatus for managing memory and storage related to artificial intelligence.

[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 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 gigabits) 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] The present disclosure relates to a method and apparatus for managing AI-related memory and storage information for the smooth operation of artificial intelligence (AI) technology in a wireless communication system.

[0008] According to one embodiment of the present disclosure, a method of a base station in a wireless communication system comprises: receiving a first message from a terminal comprising at least one of artificial intelligence (AI) related memory information or AI related storage information; determining AI model information; and transmitting the AI ​​model information to the terminal.

[0009] According to one embodiment of the present disclosure, a method of a terminal in a wireless communication system comprises: transmitting a first message to a base station comprising at least one of artificial intelligence (AI) related memory information or AI related storage information; and receiving AI model information from the base station.

[0010] According to one embodiment of the present disclosure, a base station in a wireless communication system comprises: a transceiver; and at least one processor; wherein the at least one processor is configured to receive a first message from a terminal comprising at least one of artificial intelligence (AI) related memory information or AI related storage information, determine AI model information, and transmit the AI ​​model information to the terminal.

[0011] According to one embodiment of the present disclosure, a terminal in a wireless communication system comprises: a transceiver; and at least one processor; wherein the at least one processor is configured to transmit a first message to a base station comprising at least one of artificial intelligence (AI) related memory information or AI related storage information, and to receive AI model information from the base station.

[0012] The method and apparatus according to the embodiment of the present disclosure have the effect of defining signaling between a network and a terminal to streamline the transmission of an AI model between the terminal and the network, and enabling scheduling in the AI ​​use case of the terminal, thereby improving communication performance.

[0013] 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.

[0014] Figure 1 is a diagram illustrating an example of using artificial intelligence (AI) in a wireless communication system based on terminals and base stations.

[0015] Figure 2 is a diagram illustrating an example of a model using artificial intelligence (AI) in a wireless communication system.

[0016] FIG. 3 is a block diagram showing a terminal including an artificial intelligence (AI) related memory and storage unit according to one embodiment of the present disclosure.

[0017] FIG. 4 is a flowchart illustrating an example of an operation in which a terminal manages an artificial intelligence (AI) model using a timer according to one embodiment of the present disclosure.

[0018] FIG. 5 is a flowchart illustrating an example of an operation in which a terminal manages an artificial intelligence (AI) model by means of a trigger of a base station according to one embodiment of the present disclosure.

[0019] FIG. 6 is a flowchart illustrating an example of an operation in which a terminal manages an artificial intelligence (AI) model by means of a trigger of the terminal according to one embodiment of the present disclosure.

[0020] FIG. 7 is a flowchart illustrating an example of an operation in which a terminal manages an artificial intelligence (AI) model by means of an event-based trigger of the terminal according to one embodiment of the present disclosure.

[0021] FIG. 8 is a flowchart illustrating an example of an operation in which a source base station handles the delivery of an artificial intelligence (AI) model during a handover operation according to one embodiment of the present disclosure.

[0022] FIG. 9 is a flowchart illustrating an example of an operation in which a target base station handles the delivery of an artificial intelligence (AI) model during a handover operation according to one embodiment of the present disclosure.

[0023] FIG. 10 is a flowchart illustrating an example of an operation in which a base station manages an artificial intelligence (AI) model of a terminal according to one embodiment of the present disclosure.

[0024] FIG. 11 is a flowchart illustrating an example of an operation in which a base station schedules a use case of an artificial intelligence (AI) model of a terminal according to one embodiment of the present disclosure.

[0025] FIG. 12 is a flowchart illustrating an example of scheduling based on hard-coded criteria in a terminal for an artificial intelligence (AI) use case operating in a terminal according to one embodiment of the present disclosure.

[0026] FIG. 13 is a flowchart illustrating the operation of a base station according to one embodiment of the present disclosure.

[0027] FIG. 14 is a flowchart illustrating the operation of a terminal according to one embodiment of the present disclosure.

[0028] FIG. 15 is a structural diagram illustrating an example of the structure of a base station according to one embodiment of the present disclosure.

[0029] FIG. 16 is a structural diagram illustrating an example of the structure of a terminal according to one embodiment of the present disclosure.

[0030] Hereinafter, embodiments of the present invention will be described in detail together with the accompanying drawings.

[0031] In describing the embodiments, technical details that are well known in the technical field to which the present invention belongs and are not directly related to the present invention are omitted. This is intended to convey the essence of the present invention more clearly without obscuring it by omitting unnecessary explanations.

[0032] For the same reason, some components in the attached drawings have been emphasized, omitted, or depicted schematically. Additionally, the dimensions of each component do not fully reflect their actual dimensions. Identical or corresponding components in each drawing have been assigned the same reference numbers.

[0033] The advantages and features of the present invention 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 invention is not limited to the embodiments disclosed below but can be implemented in various different forms. These embodiments are provided merely to ensure that the disclosure of the present invention is complete and to fully inform those skilled in the art of the scope of the invention, and the present invention is defined only by the scope of the claims. Throughout the specification, the same reference numerals refer to the same components.

[0034] At this time, 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 computer also creates means for the instructions executed through the processor of other programmable data processing equipment 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 such computer-available or computer-readable memory can also produce a manufactured item containing means of instruction that 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).

[0035] 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 execution examples, the functions mentioned in the blocks may occur out of order. For instance, 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.

[0036] In this embodiment, the term "part" refers to a software or hardware component, such as an FPGA or ASIC, 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 operate one or more processors. Accordingly, 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." Furthermore, the components and "parts" may be implemented to operate one or more CPUs within a device or secure multimedia card.

[0037] Hereinafter, a base station is an entity that performs resource allocation for terminals and may be at least one of a gNode B (gNB), eNode B (eNB), Node B, BS (Base Station), wireless access unit, base station controller, or a node on a network. A terminal may include a UE (User Equipment), MS (Mobile Station), cellular phone, smartphone, computer, or a multimedia system capable of performing communication functions. In this disclosure, a downlink (DL) refers to a wireless transmission path of a signal transmitted by a base station to a terminal, and an uplink (UL) refers to a wireless transmission path of a signal transmitted by a terminal to a base station. Furthermore, while LTE, LTE-A, or 5G systems may be described as examples below, embodiments of this disclosure may also be applied to other communication systems having similar technical backgrounds or channel types. For example, 5th generation mobile communication technology (5G, new radio, NR) developed after LTE-A may be included therein, and the 5G below may be a concept that includes existing LTE, LTE-A, and other similar services. In addition, 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.

[0038] Hereinafter, various embodiments are described in detail with reference to the attached drawings. It should be noted that identical components in the attached drawings are indicated by the same reference numerals whenever possible. Furthermore, it should be noted that the drawings of the present invention attached below are provided to aid in understanding the present invention, and that the present invention is not limited to the forms or arrangements exemplified in the drawings. Additionally, detailed descriptions of known functions and configurations that may obscure the essence of the present invention will be omitted. It should be noted that in the following description, only the parts necessary for understanding the operation according to various embodiments of the present invention will be explained, and descriptions of other parts will be omitted so as not to distract from the essence of the present invention.

[0039] Figure 1 is a diagram illustrating an example of using artificial intelligence (AI) in a wireless communication system based on terminals and base stations.

[0040] With three use cases—channel state indicator (CSI), beam management (BM), and positioning—being discussed in the field of AI / machine learning (ML) in relation to 3GPP (3rd generation partnership project) standards, discussions are continuing regarding the use of artificial intelligence (AI) technology in wireless communication systems (e.g., 5G or 6G systems).

[0041] Referring to FIG. 1, artificial intelligence (AI) technology can be applied to, for example, CSI prediction (120), CSI compression (130), and positioning (140) based on the terminal (100). Additionally, artificial intelligence (AI) technology can be applied to, for example, CSI prediction (125), CSI compression (135), and beam prediction (145) based on the base station (110).

[0042] Furthermore, new use cases, including CSI and mobility, are being discussed in relation to 3GPP standards.

[0043] Figure 2 is a diagram illustrating an example of a model using artificial intelligence (AI) in a wireless communication system.

[0044] Cases where AI inference is performed at either the network or the terminal can be classified as a one-sided type, and cases where AI inference is performed at both the terminal and the network, with the output of one location being used as the input of the other, can be classified as a two-sided type.

[0045] Figure 2 (a) shows a two-sided type, and Figure 2 (b) shows a one-sided type.

[0046] For example, among the use cases being discussed at 3GPP, beam management (BM), positioning, and CSI prediction are representative one-sided cases illustrated in Figure 2 (b), while CSI compression may correspond to a two-sided type in which an encoder and a decoder exist at the terminal and the base station, respectively.

[0047] Scenarios for AI / ML operations in wireless communication systems are being actively discussed, and due to the characteristics of wireless communication systems consisting of multiple nodes, methods and definitions for transmitting AI models from one node to another are being discussed.

[0048] When an AI model to be used by a terminal is transmitted from the network (or when an AI model to be used by the network is transmitted from the terminal), it is necessary to define the process of transmitting the AI ​​model between the terminal and the network. This process is defined as AI model transfer / delivery and is being discussed in relation to 3GPP standards. For example, the definition of AI model transfer / delivery discussed in 3GPP standards is as follows.

[0049] AI / ML model transfer: Refers to the transmission of an AI / ML model over a wireless interface, which is either parameters of a model structure known at the receiving end or a new model containing such parameters. The transmission may include a full model or a partial model.

[0050] AI / ML model delivery: A generic term referring to the delivery of an AI / ML model from one entity to another entity in any manner. Among these, the form of transmitting an AI model between a terminal and a base station via an air interface is defined as Option 1, and can be defined as Option 1 (a) when using the control plane (CP) and Option 1 (b) when using the user plane (UP).

[0051] Meanwhile, the emergence of AI technology and the definition of use cases may require devices capable of operating AI on-device. Although the demand for GPUs (graphic processing units), which are chipsets for parallel processing, is skyrocketing for the training of AI technology that requires increased computational power compared to existing ones, the use of GPUs in AI technology, which requires high computational power compared to existing algorithms even during inference, is not suitable for on-device devices (e.g., smartphones, smartwatches, etc.) that require low-power performance. To address this, a new type of chipset called an NPU (neural processing unit) has emerged.

[0052] An NPU is a device designed to maximize low-power performance and accelerate computations for Deep Neural Networks (DNNs). In wireless communication systems, support for utilizing AI technology is being discussed, including methods such as incorporating an NPU within the wireless communication modem chip, using CPU accelerators that offer enhanced acceleration performance for existing CPUs, and utilizing external NPU resources for wireless communication AI use cases.

[0053] The situations in which the above-mentioned AI model transfer / delivery may occur are diverse, including model update situations within a cell, such as when an update is required due to performance degradation of the AI ​​model; model transfers resulting from the use of different AI models across multiple vendors; model transfers caused by the nature of cell-specific AI models; handover process-related situations, such as model transfers resulting from differences in AI support use cases by base station; and AI model transfer situations upon initial entry into a cell.

[0054] In addition, the model size may vary for the same use case, and even within the same use case, the model size may vary depending on the model backbone, complexity, and model quantization level.

[0055] However, in the case of a terminal, the size of the storage for storing the AI ​​model and the memory required for executing and computing the AI ​​model are finite and limited depending on the performance of the terminal.

[0056] FIG. 3 is a block diagram showing a terminal including an artificial intelligence (AI) related memory and storage unit according to one embodiment of the present disclosure.

[0057] Referring to FIG. 3, the terminal may include at least one of an AI (NPU) memory (300) and an AI model storage (310).

[0058] The above AI (NPU) memory (300) may refer to a space capable of running the AI ​​model to perform operations such as inferencing or learning the AI ​​model. That is, the AI ​​model must be run in the memory (300) space to perform inferencing and learning operations. Additionally, depending on the type of model backbone, model parameters on actual storage may run while occupying the same space on memory, but the size may vary as some parameters run in memory in a repeating or spread-out form. Therefore, the size of the space occupied in storage and memory for the same model may differ. Since the memory space for the terminal's AI can be utilized for internal terminal implementation operations other than use cases defined in standards that the base station can recognize and control, the size of the space may fluctuate over time independently of the base station's operations or signaling.

[0059] The above AI model storage (310) is a space capable of storing an AI model, and may refer to a space where a terminal or a specific node stores a model for the purpose of using the model later.

[0060] The chipset of the terminal may include storage (300) of a finite size allocated to store an AI model. Additionally, it may include memory (300) allocated to an NPU located within the terminal's communication chipset, an external NPU, an AI accelerator existing within the chipset, or an AI accelerator located in the CPU.

[0061] For a model that exceeds the allocated AI memory (300) or the currently available AI memory (305), operation on the terminal is not possible even if the size of the storage (310) is sufficient. Additionally, for a model that exceeds the available storage (315), it cannot be transmitted from the base station even if it does not exceed the range of the total storage (310).

[0062] The base station cannot check the status of the AI-related storage and memory of the terminal described earlier.

[0063] The AI ​​chipset of a terminal (e.g., NPU) may include an accelerated model backbone for each chipset. The availability of acceleration support and acceleration performance of backbones such as Convolutional Neural Networks (CNN) and Multi-Layer Perceptrons (MLP) Transformers may vary depending on the terminal vendor and the chipset of the terminal.

[0064] Accordingly, the present disclosure proposes a method for selecting models requiring transmission and preventing unnecessary model transmission by defining signaling information, such as the terminal's AI-related storage, memory, and backbone preferences, during AI model transfer / delivery. Through the definition and exchange of information proposed in the present disclosure, a base station can adaptively schedule the terminal's standard-related AI operations.

[0065] The embodiments of the invention proposed in this disclosure can be classified into the following three categories.

[0066] Type 1: Operation in which a terminal manages the terminal's AI model (Figs. 4 to 9)

[0067] Type 2: Operation of a base station managing a terminal's AI model (Figs. 10 to 11)

[0068] Type 3: Operation when an AI use case running on the terminal is hard-coded in the terminal (Fig. 12)

[0069] FIG. 4 is a flowchart illustrating an example of an operation in which a terminal manages an artificial intelligence (AI) model using a timer according to one embodiment of the present disclosure.

[0070] FIG. 4 illustrates an example of an operation in which a terminal directly maintains and deletes the terminal's AI model and updates AI-related storage and memory information based on a timer.

[0071] Referring to FIG. 4, in step 420, the base station (410) can transmit a message containing timer information for AI-related memory and storage updates to the terminal (400).

[0072] In one embodiment, the base station (410) may transmit a parameter setting message for updating at least one of the available AI memory size and AI storage size of the terminal (400) based on a timer. In one embodiment, the parameter setting message may be included in a radio resource control (RRC) reconfiguration message.

[0073] In one embodiment, the timer information may include timer information for available AI memory. In one embodiment, the timer information may include bitmap or index information indicating time information for available AI memory, for example, 'without timer', '20ms', or '40ms'.

[0074] In one embodiment, the timer information may include timer information for available AI storage. In one embodiment, the timer information may include bitmap or index information indicating time information for available AI storage, for example, 'without timer', '20ms', or '40ms'.

[0075] In one embodiment, the timer information may be included in the update timer field of the setting message.

[0076] In one embodiment, the information indicating timer information for available AI memory or timer information for available AI storage included in the timer information may include an index value included in the update timer field of the setting message, and an example of the index value is as shown in Table 1.

[0077] IndexUpdate timer0Timer for Available AI memory1Timer for Available AI storage

[0078] In one embodiment, at least one of the timer information for the available AI memory or the timer information for the AI ​​storage may include an index value of a timer setting for the update timer field of the setting message, and an example of the index value is shown in Table 2.

[0079] Indextimer0Without timer120ms240ms

[0080] In step 425, the terminal (400) can transmit AI-related memory and storage information to the base station (410). In one embodiment, in step 425, the terminal (400) can transmit at least one of AI-related memory information or AI-related storage information to the base station (410).

[0081] In one embodiment, at least one of the AI-related memory information or AI-related storage information of the terminal (400) may include at least one of the total AI memory size information, available AI memory size information, total AI storage size information, and available AI storage size information.

[0082] In one embodiment, at least one of the AI-related memory information or AI-related storage information may be included in a message transmitted from the terminal (400) to the base station (410). For example, a message containing at least one of the AI-related memory information or AI-related storage information may include an RRC message, a MAC-CE (medium access control-control element) message, or a UCI (uplink control information) message, etc.

[0083] In one embodiment, the AI-related memory information may include AI memory size information. In one embodiment, the AI ​​memory size information may include NPU memory size information when inferring wireless communication-related use cases in the NPU within the terminal's modem. In one embodiment, the AI ​​memory size information may include memory size information allocated to the AI ​​when allocating a portion of the CPU within the terminal's modem to the AI. In one embodiment, when the terminal uses the NPU of an external AP, the AI ​​memory size information may include memory size information allocated to the wireless communication AI among the NPUs of the external AP.

[0084] In one embodiment, the AI ​​memory size information may include at least one of total AI memory size information or available AI memory size information. In one embodiment, the total AI memory size information may be transmitted once when the terminal connects to the Cell. In one embodiment, the total AI memory size information may include the closest value smaller than the actual total AI memory size among discontinuous size values ​​such as 20 MB, 40 MB, 60 MB, etc. In one embodiment, the total AI memory size information may be transmitted using a format such as actual values ​​of data size units (FP32, FP16, INT8), such as 'MB', 'GB', etc.

[0085] In one embodiment, the available AI memory size information may include a value that changes over time and needs to be updated periodically or non-periodically by the base station after the terminal connects to the Cell. In one embodiment, the available AI memory size information may include the closest value smaller than the actual available AI memory size among discontinuous size values ​​such as 20 MB, 40 MB, 60 MB, etc. In one embodiment, the total AI memory size information may be transmitted using a format such as actual values ​​in data size units (FP32, FP16, INT8) such as 'MB', 'GB', etc. In one embodiment, the available AI memory size information may be transmitted using a 'Current and total ratio' field as a ratio value of the currently available memory relative to the total memory, an index, or a bitmap format.

[0086] In one embodiment, the AI-related storage information may include AI storage size information. In one embodiment, the AI ​​storage size information may include NPU storage size information when inferring wireless communication-related use cases in the NPU within the terminal's modem. In one embodiment, the AI ​​storage size information may include storage size information allocated to the AI ​​when allocating a portion of the CPU within the terminal's modem to the AI. In one embodiment, when the terminal uses the NPU of an external AP, the storage size information allocated to the wireless communication AI among the NPUs of the external AP may be included.

[0087] In one embodiment, the AI ​​storage size information may include at least one of the total AI storage size information or available AI storage size information. In one embodiment, the total AI storage size information may be transmitted once when the terminal connects to the Cell. In one embodiment, the total AI storage size information may include the closest value smaller than the actual total AI storage size among discontinuous size values ​​such as, for example, 100 MB, 200 MB, 300 MB, etc. In one embodiment, the total AI storage size information may be transmitted using a format such as an actual value of a data size unit (FP32, FP16, INT8), for example, 'MB', 'GB'.

[0088] In one embodiment, the available AI storage size information may include a value that changes over time and needs to be updated periodically or non-periodically to the base station after the terminal connects to the Cell. In one embodiment, the available AI storage size information may include the closest value smaller than the actual available AI memory size among discontinuous size values ​​such as, for example, 100 MB, 200 MB, 300 MB. In one embodiment, the total AI storage size information may be transmitted using a format such as actual values ​​of data size units (FP32, FP16, INT8), for example, 'MB', 'GB'. In one embodiment, the available AI storage size information may be transmitted in the form of a ratio value of the currently available storage relative to the total storage, an index, or a bitmap using a 'Current and total ratio' field.

[0089] In one embodiment, an index value indicating total AI memory size information or available AI memory size information included in the AI-related memory and storage information may be included in the AI-related memory and storage information field, and an example of the index value is shown in Table 3.

[0090] IndexAI memory and storage info0Total AI memory size1Available AI memory size2Total AI storage size3Available AI storage size

[0091] In one embodiment, the AI-related memory and storage information may include an index value of an AI memory size setting for the AI-related memory and storage information field, and an example of the index value is shown in Table 4.

[0092] IndexAI memory size [MB]020140260

[0093] In one embodiment, the AI-related memory and storage information may include an index value of an AI storage size setting for the AI-related memory and storage information field, and an example of the index value is as shown in Table 5.

[0094] IndexAI storage size [MB]010012002300

[0095] In one embodiment, the AI-related memory and storage information may include an index value of the ratio of currently available memory or storage to the total for the AI-related memory and storage information field, and an example of the index value is shown in Table 6.

[0096] IndexCurrent and total ratio0110.720.330.1

[0097] In step 430, the terminal (400) can transmit backbone type preference information to the base station (410). The backbone type preference information may be related to the AI ​​backbone computing function of the terminal.

[0098] In one embodiment, the backbone type preference information may include information on the backbone type of an AI model (e.g., CNN, Transformer, MLP, LSTM, etc.) that the chipset of the terminal (400) supports and prefers. In one embodiment, the message containing the backbone type preference information may include a message for the terminal (400) to transmit to the base station (410), e.g., an RRC message, a MAC-CE message, or a UCI message.

[0099] In one embodiment, a chipset such as an NPU of a terminal may support accelerated computation of some of the backbones. In this case, even if the terminal supports accelerated computation of various backbones, the accelerated computation performance for each backbone may differ, and the terminal may define and transmit preferred backbone information or priority information within the preferred backbone.

[0100] In one embodiment, the backbone type preference information may be included in a bitmap message that transmits at least one of information indicating whether accelerated computation is possible for each backbone (e.g., a value of '1' (possible) or '0' (impossible), or information indicating the priority of backbones capable of accelerated computation (e.g., a 2-bit value ('00', '01', '10', and '11') if there are 4 types of backbones), and an example of the bitmap message is shown in Table 7.

[0101] BackboneAcceleratorPriorityCNN100Transformer010MLP101LSTM011..., etc.

[0102] In one embodiment, the information indicating the priority of the accelerated computation-capable backbone may include only information indicating the best backbone. In one embodiment, the information indicating the priority of the accelerated computation-capable backbone may include information indicating the best backbone and information indicating the next-highest backbone. For example, if there are four types of backbones, for each backbone, it may include 4 bits indicating whether accelerated computation is possible and 2 bits indicating the best backbone. For example, if there are four types of backbones, for each backbone, it may include 4 bits indicating whether accelerated computation is possible, 2 bits indicating the best backbone, and 2 bits indicating the next-highest backbone.

[0103] In one embodiment, the backbone type preference information may include only information indicating whether accelerated computation is possible for each backbone, and in this case, the base station may determine one of the backbone models capable of accelerated computation and transmit it to the terminal.

[0104] In one embodiment, the possibility of acceleration operation may be indicated not by using values ​​of '1' or '0' to indicate 'yes' or 'no', but by using an acceleration operation step feedback method. In one embodiment, the information indicating the acceleration operation step feedback method may include information indicating high, mid, and low, and may use, for example, a 2-bit value. For example, when indicating the acceleration operation step feedback method as high, mid, and low, it may be indicated using the values ​​of '10', '01', or '00', respectively, and an example thereof is shown in Table 8 below.

[0105] BackboneAcceleratorCNN01Transformer00MLP11LSTM00..., etc.

[0106] In step 435, the base station (410) may determine the delivery of an AI model. In one embodiment, the base station (410) may determine an AI use case and model to be transmitted to the terminal (400) by referring to the AI ​​memory and storage information received in step 425.

[0107] In step 440, the base station (410) can deliver the AI ​​model to the terminal (400).

[0108] In step 445, the terminal (400) can send an acknowledgment (ACK) message to the base station (410).

[0109] In one embodiment, the terminal (400) may transmit update timer information requested by the terminal (400). In one embodiment, the update timer information requested by the terminal may be included in a field included in an uplink message. The uplink message may include at least one of an RRC message, a MAC-CE message, or a UCI message. In one embodiment, when the update timer information requested by the terminal is transmitted, the timer information transmitted from the base station in step 420 may be replaced.

[0110] In one embodiment, when a base station utilizes an update timer transmission sent by the terminal, the base station may transmit a downlink message to the terminal containing information indicating that the update timer has changed. In one embodiment, the downlink message may include an RRC message.

[0111] In step 450, the terminal (400) can transmit AI memory and storage information to the base station (410) when the timer (460) corresponding to the timer information received in step 420 has expired. In one embodiment, the AI ​​memory and storage information may include at least one of an available AI memory field or an available AI storage field.

[0112] In one embodiment, the available AI storage size may change over time depending on whether the model of the existing cell (source cell) is deleted due to timer expiration after the handover, or whether the model of the target cell is received during the handover process, or whether the existing model is deleted due to timer expiration. Accordingly, the available AI memory size may change depending on the dynamic allocation of AI memory according to the state of the terminal and changes in the operation status of AI by use case, and an update may be required. Accordingly, the terminal (400) may transmit update information regarding the available AI memory field or at least one of the available AI memory fields among the AI ​​memory and storage information message fields to the base station (410) according to the timer information received in step 420.

[0113] FIG. 5 is a flowchart illustrating an example of an operation in which a terminal manages an artificial intelligence (AI) model by means of a trigger of a base station according to one embodiment of the present disclosure.

[0114] FIG. 5 is an example of an operation in which a terminal updates terminal AI-related memory and storage information to a base station triggered by the base station, among cases where the terminal manages the terminal's AI model and directly participates in maintenance / deletion. The embodiment illustrated in FIG. 5 defines a message to request the terminal to update AI-related memory and storage information in order for the base station to trigger the update of AI-related memory and storage information. In FIG. 5, configurations that are redundant with those described in FIG. 4 have been omitted, and the configurations described in FIG. 4 can also be applied to FIG. 5.

[0115] Referring to FIG. 5, in step 520, the base station (510) may transmit a message containing timer information for updating AI-related memory and storage to the terminal (500). In one embodiment, when the base station (510) triggers an update of AI-related memory and storage information to the terminal (500), the update timer field containing the timer information may include an index value or bitmap indicating 'without timer'. In one embodiment, when the base station (510) triggers an update of AI-related memory and storage information to the terminal (500), step 520 may be omitted. In one embodiment, even when the base station (510) triggers an update of AI-related memory and storage information to the terminal (500), a timer may be used as illustrated in FIG. 4.

[0116] In step 525, the terminal (500) can transmit AI-related memory and storage information to the base station (510).

[0117] In step 530, the terminal (500) can transmit backbone type preference information to the base station (510). The backbone type preference information may be related to the AI ​​backbone computing function of the terminal.

[0118] In step 535, the base station (510) may determine the delivery of an AI model. In one embodiment, the base station (510) may determine an AI use case and model to be transmitted to the terminal (500) based on the AI ​​memory and storage information received in step 525.

[0119] In step 540, the base station (510) can deliver the AI ​​model to the terminal (500).

[0120] In step 545, the terminal (500) can send an acknowledgment (ACK) message to the base station (510).

[0121] In step 550, the base station (510) can send a message to the terminal (500) requesting an update of AI memory and storage information.

[0122] In one embodiment, the base station (510) may transmit a request message to receive an update of at least one of the available AI memory size information and available AI storage information of the terminal (500). The request message may include, for example, at least one of an RRC message, a MAC-CE message, or a UCI message.

[0123] In one embodiment, when a base station (510) requires additional model transfer to a terminal (500), it may be necessary to update at least one of the available AI memory size information and available AI storage information of the terminal (500) prior to model transfer. In one embodiment, a message requesting the update of the AI ​​memory and storage information may include a 1-bit message requesting the transmission of at least one of the available AI memory size information and available AI storage information.

[0124] In one embodiment, the base station (510) can transmit information about at least one of the model size or model use case of the model that the base station intends to transmit to the terminal (500).

[0125] In step 555, the terminal (500) can transmit AI memory and storage information to the base station (510) based on the request message received in step 550.

[0126] In one embodiment, when a terminal (500) receives information from a base station (510) regarding at least one of the model size or model use case of a model that the base station intends to transmit, the terminal (500) may provide feedback on the remaining storage after deleting the existing model corresponding to the current use case when providing feedback on available AI storage size information.

[0127] In step 560, the terminal (500) and the base station (510) can perform the AI ​​model delivery procedure according to steps 535 to 545.

[0128] FIG. 6 is a flowchart illustrating an example of an operation in which a terminal manages an artificial intelligence (AI) model by means of a trigger of the terminal according to one embodiment of the present disclosure.

[0129] FIG. 6 is an example of an operation in which the terminal triggers an update of terminal AI-related memory and storage information to the base station, among cases where the terminal manages the terminal's AI model and directly participates in maintenance / deletion. The embodiment illustrated in FIG. 6 defines a UL scheduling request message for the terminal to request the transmission of AI-related memory and storage information to the base station. In FIG. 6, configurations that are redundant with those described in FIG. 4 have been omitted, and the configurations described in FIG. 4 are also applicable to FIG. 6.

[0130] Referring to FIG. 6, in step 620, the base station (610) may transmit a message containing timer information for updating AI-related memory and storage to the terminal (600). In one embodiment, when the base station (610) triggers an update of AI-related memory and storage information to the terminal (600), the update timer field containing the timer information may include an index value or bitmap indicating 'without timer'. In one embodiment, when the base station (610) triggers an update of AI-related memory and storage information to the terminal (600), step 620 may be omitted. In one embodiment, even when the base station (610) triggers an update of AI-related memory and storage information to the terminal (600), a timer may be used as illustrated in FIG. 4.

[0131] In step 625, the terminal (600) can transmit AI-related memory and storage information to the base station (610).

[0132] In step 630, the terminal (600) can transmit backbone type preference information to the base station (610). The backbone type preference information may be related to the AI ​​backbone computing function of the terminal.

[0133] In step 635, the base station (610) may determine the delivery of an AI model. In one embodiment, the base station (610) may determine an AI use case and model to be transmitted to the terminal (600) based on the AI ​​memory and storage information received in step 625.

[0134] In step 640, the base station (610) can deliver the AI ​​model to the terminal (600).

[0135] In step 645, the terminal (600) can send an acknowledgment (ACK) message to the base station (610).

[0136] In step 650, the terminal (600) can send an uplink scheduling request to the base station (610) for updating AI memory and storage information.

[0137] In one embodiment, the base station (610) may transmit an uplink scheduling request message to update at least one of the available AI memory size information and available AI storage information of the terminal (600).

[0138] In step 655, the base station (610) can send a message for uplink allocation to the terminal.

[0139] In step 660, the terminal (600) can transmit AI memory and storage information to the base station (610) based on the allocation message received in step 655.

[0140] Afterwards, the terminal (600) and the base station (610) can perform the AI ​​model delivery procedure according to steps 635 to 645 above.

[0141] FIG. 7 is a flowchart illustrating an example of an operation in which a terminal manages an artificial intelligence (AI) model by means of an event-based trigger of the terminal according to one embodiment of the present disclosure.

[0142] FIG. 7 is an operation in which memory and storage information related to the terminal AI is updated by being triggered by the terminal based on an event when the terminal manages the terminal's AI model and directly participates in maintenance / deletion. FIG. 7 defines an event trigger operation and a message that defines and transmits conditions for the event to be triggered. In FIG. 7, configurations that are redundant with those described in FIG. 4 have been omitted, and the configurations described in FIG. 4 can also be applied to FIG. 7.

[0143] Referring to FIG. 7, in step 720, the base station (710) can transmit a message containing UE triggering condition information to the terminal (700).

[0144] In one embodiment, the base station (710) can transmit an AI memory and storage update timer to the terminal (700) to operate an operation to update AI-related memory and storage information based on the timer and the UE triggering condition.

[0145] In one embodiment, the UE triggering condition information may be included in a request message for transmitting available AI memory and storage information. In one embodiment, the UE triggering condition information may be included in a condition transmission message for updating available AI memory and storage information based on a UE trigger when the size of available AI memory or storage satisfies a specific condition and transmitted. In one embodiment, the message of step 720 may include at least one of an RRC message, a MAC-CE, or a DCI message. The UE triggering condition information may be transmitted using a field within a reporting configuration message.

[0146] In one embodiment, the UE triggering condition information includes time-to-trigger (TTT) information and a threshold of available storage. storage ), available memory threshold (Threshold memory It may include at least one of the information regarding UE triggering conditions.

[0147] In one embodiment, the time-to-trigger (TTT) information may be used to request an update of available AI memory and storage information when the condition is maintained for a corresponding time length after satisfying a specific condition. In one embodiment, the time-to-trigger (TTT) information may be represented, for example, as a time value or a number of slots. For example, the time-to-trigger (TTT) information may include condition information for 40ms or more or 10 slots or more.

[0148] In one embodiment, the threshold of the available storage (Threshold storage ) may be indicated as a ratio value of available storage to total storage (e.g., 0.7). In one embodiment, the threshold of the available storage (Threshold storage ) can be indicated by the size value of available storage (e.g., 100MB).

[0149] In one embodiment, the threshold of the available memory (Threshold memory ) may be indicated as a ratio value of available memory to total memory (e.g., 0.7). In one embodiment, the threshold of the available memory (Threshold memory ) can be indicated by the size value of available memory (e.g., 50MB).

[0150] In one embodiment, the terminal (700) may determine whether to perform the operation of step 755 by determining whether the information regarding the UE triggering condition is satisfied for the TTT or longer. In one embodiment, the information regarding the UE triggering condition may include at least one of information regarding an AI memory-related triggering condition or information regarding an AI storage-related triggering condition. In one embodiment, the information regarding the UE triggering condition may include information requesting the transmission of AI-related memory information or AI-related storage information corresponding to the satisfied condition when only one of the information regarding the AI ​​memory-related triggering condition or the information regarding the AI ​​storage-related triggering condition is satisfied. In one embodiment, the information regarding the UE triggering condition may include information requesting the transmission of AI-related memory information and AI-related storage information when only one of the information regarding the AI ​​memory-related triggering condition or the information regarding the AI ​​storage-related triggering condition is satisfied. In one embodiment, the information regarding the UE triggering condition may include information requesting the transmission of AI-related memory information and AI-related storage information when both the information regarding the AI ​​memory-related triggering condition and the information regarding the AI ​​storage-related triggering condition are satisfied.

[0151] In one embodiment, the information regarding AI storage-related triggering conditions included in the information regarding the UE triggering conditions is Avail storage <Threshold storage It may include.

[0152] In one embodiment, the information regarding AI memory-related triggering conditions included in the information regarding the UE triggering conditions is Avail memory <Threshold memoryIt may include. The inequality sign ("<") used in the above triggering condition is merely an example and can be replaced with various inequality signs such as "<=", ">", ">=", etc. depending on the implementation method.

[0153] In step 725, the terminal (700) can transmit AI-related memory and storage information to the base station (710).

[0154] In step 730, the terminal (700) can transmit backbone type preference information to the base station (710). The backbone type preference information may be related to the AI ​​backbone computing function of the terminal.

[0155] In step 735, the base station (710) may determine the delivery of an AI model. In one embodiment, the base station (710) may determine an AI use case and model to be transmitted to the terminal (700) based on the AI ​​memory and storage information received in step 725.

[0156] In step 740, the base station (710) can deliver the AI ​​model to the terminal (700).

[0157] In step 745, the terminal (700) can send an acknowledgment (ACK) message to the base station (710).

[0158] In step 750, the terminal (700) can determine whether an event to report information related to the AI ​​model has been triggered based on the UE triggering condition information received in step 720.

[0159] In step 750, the terminal (700) can send an uplink scheduling request to the base station (710) for updating AI memory and storage information.

[0160] In step 760, the base station (710) can send a message for uplink allocation to the terminal.

[0161] In step 765, the terminal (700) can transmit AI memory and storage information to the base station (710) based on the allocation message received in step 760.

[0162] Afterwards, the terminal (700) and the base station (710) can perform the AI ​​model delivery procedure according to steps 735 to 745 above.

[0163] In FIGS. 8 and 9 below, an operation occurs in which information related to an AI model is exchanged during a handover process when the terminal manages the terminal's AI model and is directly involved in its maintenance / deletion. FIGS. 8 and 9 define a message that transmits AI-related memory and storage information of the terminal between base stations.

[0164] FIG. 8 is a flowchart illustrating an example of an operation in which a source base station handles the delivery of an artificial intelligence (AI) model during a handover operation according to one embodiment of the present disclosure.

[0165] In Fig. 8, configurations that are redundant with those described in Fig. 4 have been omitted, and configurations described in Fig. 4 can also be applied to Fig. 8.

[0166] In step 820, the terminal (800) can determine whether a ready event has occurred.

[0167] In step 825, the terminal (800) can transmit a measurement report to the source base station (810).

[0168] In step 830, the terminal (800) can transmit AI-related memory and storage information to the source base station (810).

[0169] In step 835, the source base station (810) can make a decision regarding the handover (e.g., conditional handover (CHO)).

[0170] In step 840, the source base station (810) can send and receive a handover request message and a confirmation message for the handover request with the target base station and / or candidate base stations (815).

[0171] In step 845, the source base station (810) can transmit a handover command to the terminal (800).

[0172] In step 850, the source base station (810) can send and receive AI model request and AI model delivery messages with the target base station and / or candidate base stations, or the UPF (817) of the target base station and / or candidate base stations. That is, the source base station (810) can deliver the AI ​​model in response to requests from the target base station and / or candidate base stations.

[0173] In step 855, the source base station (810) may transmit timer information for AI-related memory and storage updates to the terminal (800). The timer information may be transmitted when a change in the period relative to the timer defined when the terminal enters the cell is required. The timer information may be included in the update timer field for handover included in the AI ​​memory and storage update timer information described in step 420 of FIG. 4.

[0174] In step 860, the source base station (810) can deliver the AI ​​model to the terminal (800).

[0175] In step 865, the terminal (800) can send an acknowledgment (ACK) message to the source base station (810).

[0176] In step 870, the terminal (800) can transmit AI memory and storage information to the source base station (810) when the timer (880) corresponding to the timer information received in step 855 has expired. In one embodiment, the AI ​​memory and storage information may include at least one of an available AI memory field or an available AI storage field.

[0177] Afterward, the terminal (800) and the target base station (820) can proceed with a random access procedure (RACH (random access channel) procedure) after the handover procedure is completed.

[0178] FIG. 9 is a flowchart illustrating an example of an operation in which a target base station handles the delivery of an artificial intelligence (AI) model during a handover operation according to one embodiment of the present disclosure.

[0179] FIG. 9 illustrates a possible embodiment for a handover to a single base station, such as a base-handover. In this case, it is necessary to define a message that transmits AI memory and storage information received by the source base station from the terminal to the target base station. In FIG. 9, configurations that are redundant with those described in FIG. 4 have been omitted, and the configurations described in FIG. 4 are also applicable to FIG. 9.

[0180] In step 920, the terminal (900) can determine whether a ready event has occurred.

[0181] In step 925, the terminal (900) can transmit a measurement report to the source base station (910).

[0182] In step 930, the terminal (900) can transmit AI-related memory and storage information to the source base station (910).

[0183] In step 935, the source base station (910) can make a decision regarding the conditional handover (CHO).

[0184] In step 945, the source base station (910) can transmit AI memory and storage information to the target base station (915). The message for transmitting the AI ​​memory and storage information can be a message in the form of an Xn interface between base stations.

[0185] In step 950, the source base station (910) can send and receive a handover request message and a confirmation message for the handover request with the target base station (915).

[0186] In step 955, the terminal (900), source base station (910), and target base station (915) can then perform the remaining handover procedure.

[0187] FIG. 10 is a flowchart illustrating an example of an operation in which a base station manages an artificial intelligence (AI) model of a terminal according to one embodiment of the present disclosure.

[0188] FIG. 10 illustrates an example of an operation in which a base station manages an AI model of a terminal and is involved in maintaining or deleting the AI ​​model. The embodiment illustrated in FIG. 10 defines a message for the base station to delete an AI model stored in the terminal. In FIG. 10, configurations that are redundant with those described in FIG. 4 have been omitted, and the configurations described in FIG. 4 are also applicable to FIG. 10.

[0189] Referring to FIG. 10, in step 1020, the base station (1010) may transmit a message containing timer information for updating AI-related memory and storage to the terminal (1000). In one embodiment, if the base station (1010) is directly involved in deleting AI, updating the available AI storage information may be unnecessary. In this case, the timer information of the available AI storage field included in the timer information may be transmitted in the form of an index or bitmap indicating 'without timer'.

[0190] In step 1025, the terminal (1000) can transmit AI-related memory and storage information to the base station (1010).

[0191] In step 1030, the terminal (1000) can transmit backbone type preference information to the base station (1010). The backbone type preference information may be related to the AI ​​backbone computing function of the terminal.

[0192] In step 1035, the base station (1010) may determine the delivery of an AI model. In one embodiment, the base station (1010) may determine an AI use case and model to be transmitted to the terminal (1000) based on the AI ​​memory and storage information received in step 1025.

[0193] In step 1040, the base station (1010) can deliver the AI ​​model to the terminal (1000).

[0194] In step 1045, the terminal (1000) can send an acknowledgment (ACK) message to the base station (1010).

[0195] In step 1050, the base station (1010) can send an AI model deletion command to the terminal (1000).

[0196] In one embodiment, the base station (1010) may transmit a message to delete a specific model of a use case designated by the base station. The message may include at least one of an RRC message, a MAC-CE message, or a DCI message. In one embodiment, the message transmitted in step 1050 may include at least one of index or bitmap information representing the use case and index or bitmap information representing the AI ​​model to be deleted. In one embodiment, the message transmitted in step 1050 may include the model identifier (ID) of the AI ​​model to be deleted.

[0197] In step 1055, the terminal (1000) can transmit AI memory and storage information to the base station (1010) when the timer (1060) corresponding to the timer information received in step 1020 has expired. In one embodiment, the terminal (1000) can transmit AI memory and storage information to the base station (1010) after performing a deletion operation based on the AI ​​model deletion command received in step 1050.

[0198] In one embodiment, the terminal (1000) may delete the corresponding AI model according to a predefined condition by the base station or a specific condition hard-coded in the terminal. For example, the condition may include a case where the terminal is outside the PLMN area of ​​the network operator. In this case, the terminal may flush the model corresponding to the existing operator in order to receive the base model of the new operator. In one embodiment, when the terminal deletes the model, it may send an uplink message to the base station notifying that the model has been deleted. The uplink message may include at least one of an RRC message, a MAC-CE message, or a UCI message. In one embodiment, the uplink message may include the model ID deleted from the terminal.

[0199] The embodiment illustrated in FIG. 10 relates to an operation of deleting a model stored in a terminal and may be combined with at least one of the operations illustrated in FIG. 4 to 7. FIG. 11 is a flowchart illustrating an example of an operation in which a base station schedules a use case of an artificial intelligence (AI) model of a terminal according to an embodiment of the present disclosure.

[0200] FIG. 11 defines an operation and message in which a base station schedules an AI use case of a terminal based on information received from the terminal and transmits the result to the terminal. For example, scheduling the AI ​​use case may include at least one of the operations of turning on / off the operation of an AI model for an AI use case (e.g., CSI prediction, beam management, positioning, etc.) or selecting an AI model for an AI use case. In FIG. 11, configurations that are redundant with those described in FIG. 4 have been omitted, and the configurations described in FIG. 4 can also be applied to FIG. 11.

[0201] Referring to FIG. 11, in step 1120, the terminal (1100) can transmit AI-related memory and storage information to the base station (1110).

[0202] In step 1125, the terminal (1100) can transmit backbone type preference information to the base station (1110). The backbone type preference information may be related to the AI ​​backbone computing function of the terminal.

[0203] In step 1130, the base station (1110) may determine the delivery of an AI model. In one embodiment, the base station (1110) may determine an AI use case and model to be transmitted to the terminal (1100) based on the AI ​​memory and storage information received in step 1125.

[0204] In step 1140, the base station (1110) can deliver the AI ​​model to the terminal (1100).

[0205] In step 1145, the terminal (1100) can send an acknowledgment (ACK) message to the base station (1110).

[0206] In step 1150, the base station (1110) may determine scheduling information for an AI use case based on at least one of the AI ​​memory information and cell state information of the terminal. In one embodiment, the base station (1110) may schedule an AI use case for the operation of the terminal based on at least one of the base station state (e.g., base station capacity, base station cell throughput, base station legacy block performance, etc.) or the available memory information of the terminal. In one embodiment, the scheduling may include, for example, turning on / off the operation of an AI model for an AI use case or selecting an AI model.

[0207] In step 1155, the base station (1110) can transmit scheduling information for an AI use case to the terminal (1100).

[0208] In one embodiment, the base station (1110) may transmit a message to control whether an AI model is driven for each terminal AI use case, and the operation model, etc. In one embodiment, the message may include at least one of an RRC message, a MAC-CE message, or a DCI.

[0209] In one embodiment, the message may transmit at least one model ID (or a bundle of model IDs) of a use case required for operation. In one embodiment, the message may transmit at least one model ID (or a bundle of model IDs) of a use case required for operation, and, if there are multiple model IDs, priority information among the model IDs. The priority information may be used when the terminal cannot use all of the models provided by the base station due to its current state.

[0210] The embodiment illustrated in FIG. 11 is an operation in which a base station schedules an AI use case that operates at a terminal during initial connection, and can be combined with at least one of the operations illustrated in FIGs. 4 to 7.

[0211] FIG. 12 is a flowchart illustrating an example of scheduling based on hard-coded criteria in a terminal for an artificial intelligence (AI) use case operating in a terminal according to one embodiment of the present disclosure.

[0212] FIG. 12 defines an example of a hard-coded standard in a terminal, a scheduling operation of the terminal, an operation of transmitting scheduling information to a base station, and a message. For example, scheduling for the AI ​​use case may include at least one of the operations of turning on / off the operation of an AI model for an AI use case (e.g., CSI prediction, beam management, positioning, etc.) or selecting an AI model for an AI use case. In FIG. 12, configurations that are redundant with those described in FIG. 4 have been omitted, and the configurations described in FIG. 4 can also be applied to FIG. 12.

[0213] Since multiple terminals are connected to a base station, micro-control, such as scheduling resources for the terminals, can act as overhead. Accordingly, information regarding the operation of an AI use case (e.g., an AI use case defined in the standard) in which an AI model is executed with the base station involved at the terminal can be distributed and operated in a form that is hard-coded into the terminal's chip.

[0214] In one embodiment, information hardcoded in the chip of the terminal may include priority information for an AI use case. In one embodiment, information hardcoded in the chip of the terminal may include model priority information for the use case. In one embodiment, information hardcoded in the chip of the terminal may include condition information regarding a comparison between a legacy operation and an operation to which an AI model is applied. In one embodiment, the condition information between the legacy operation and the operation to which the model is applied may include condition information for a use case to which an AI model is applied by comparing legacy performance and AI performance. In one embodiment, the condition information between the legacy operation and the operation to which the model is applied may include condition information for a use case to which an AI model is applied by comparing the AI ​​performance with the result of adding an offset value to the legacy performance.

[0215] Referring to FIG. 12, in step 1220, the terminal (1200) can transmit AI-related memory and storage information to the base station (1210).

[0216] In step 1225, the terminal (1200) can transmit backbone type preference information to the base station (1210). The backbone type preference information may be related to the AI ​​backbone computing function of the terminal.

[0217] In step 1230, the base station (1210) may determine the delivery of an AI model. In one embodiment, the base station (1210) may determine an AI use case and model to be transmitted to the terminal (1200) based on the AI ​​memory and storage information received in step 1225.

[0218] In step 1235, the base station (1210) can deliver the AI ​​model to the terminal (1200).

[0219] In step 1240, the terminal (1200) can send an acknowledgment (ACK) message to the base station (1210).

[0220] In step 1245, the terminal (1200) can perform scheduling of AI use cases based on information hardcoded in the terminal (1200). In one embodiment, the terminal (1200) can perform scheduling of AI use cases (e.g., AI use cases defined in the specification) using condition information hardcoded in the terminal described above.

[0221] In step 1250, the terminal (1200) can transmit AI use case scheduling information to the base station (1210).

[0222] In one embodiment, at step 1250, the terminal (1200) may transmit a message to the base station (1210) containing an AI use case or model identifier (ID) that the terminal is using. In one embodiment, the message may include, for example, an RRC message, a MAC-CE message, or a UCI message. Information regarding the AI ​​use case that the terminal is using (or operating) may include, for example, CSI prediction or beam management.

[0223] The embodiment described in FIG. 12 is an operation in which a base station schedules an AI use case that operates at a terminal during initial connection, and can be operated in combination with at least one of the operations shown in FIG. 4 to 7.

[0224] In one embodiment, when the terminal possesses a model without model transmission from the base station, the terminal can schedule AI use cases and models without exchanging information regarding AI-related memory, storage, or models. In one embodiment, since configurations such as the prediction time length may vary even within an AI use case, the base station transmits the configuration to the terminal upon initial connection, and the terminal can schedule AI use cases and models based on the information.

[0225] FIG. 13 is a flowchart illustrating the operation of a base station according to one embodiment of the present disclosure.

[0226] Referring to FIG. 13, in step 1300, the base station may receive a first message from a terminal containing at least one of artificial intelligence (AI) related memory information or AI related storage information.

[0227] In step 1310, the base station can determine AI model information based on at least one of the AI-related memory information or AI-related storage information.

[0228] In step 1320, the base station can transmit the AI ​​model information to the terminal.

[0229] In one embodiment, the base station may receive a second message containing backbone type preference information from the terminal.

[0230] In one embodiment, the AI-related memory information may include memory size information in which the terminal can run an AI model.

[0231] In one embodiment, the AI-related storage information may include size information of the storage in which the terminal can store an AI model.

[0232] In one embodiment, the base station may transmit timer information related to updating at least one of AI-related memory information or AI-related storage information to the terminal.

[0233] In one embodiment, when a timer based on the timer information expires, the base station may receive at least one of AI-related updated memory information or AI-related updated storage information from the terminal.

[0234] In one embodiment, the base station may transmit a request message to the terminal for updating at least one of AI-related memory information or AI-related storage information.

[0235] In one embodiment, the base station may receive at least one of AI-related updated memory information or AI-related updated storage information from the terminal in response to the request message.

[0236] In one embodiment, the base station may receive an uplink scheduling request message from the terminal for updating at least one of AI-related memory information or AI-related storage information. In one embodiment, the base station may transmit an uplink scheduling message to the terminal. In one embodiment, the base station may receive at least one of AI-related updated memory information or AI-related updated storage information from the terminal in response to the uplink scheduling message.

[0237] In one embodiment, the base station may transmit to the terminal a message containing condition information that triggers updating at least one of AI-related memory information or AI-related storage information. In one embodiment, the base station may base the uplink scheduling request message on the condition information.

[0238] In one embodiment, the base station may transmit a handover request message to a second base station. In one embodiment, the base station may transmit the AI ​​model information to the second base station.

[0239] In one embodiment, the base station may transmit a command message to the terminal to delete the stored AI model.

[0240] In one embodiment, the base station according to claim 1 may determine whether to run an AI model for an AI usage case of the terminal or at least one of the type of an AI model for an AI usage case based on at least one of the AI-related memory information or the cell state information of the base station.

[0241] In one embodiment, the base station may transmit to the terminal at least one of information regarding whether the determined AI model is running or information regarding the type of the AI ​​model.

[0242] FIG. 14 is a flowchart illustrating the operation of a terminal according to one embodiment of the present disclosure.

[0243] Referring to FIG. 14, in step 1400, the terminal can transmit a first message to a base station that includes at least one of artificial intelligence (AI) related memory information or AI related storage information.

[0244] In step 1410, the terminal can receive AI model information based on at least one of the AI-related memory information or AI-related storage information from the base station.

[0245] In one embodiment, the terminal can transmit a second message containing backbone type preference information to the base station.

[0246] In one embodiment, the AI-related memory information may include memory size information in which the terminal can run an AI model. In one embodiment, the AI-related storage information may include storage size information in which the terminal can store an AI model.

[0247] In one embodiment, the terminal may receive timer information from the base station related to updating at least one of AI-related memory information or AI-related storage information. In one embodiment, when the timer based on the timer information expires, the terminal may transmit at least one of the AI-related updated memory information or AI-related updated storage information to the base station.

[0248] In one embodiment, the terminal may receive a request message from the base station for updating at least one of AI-related memory information or AI-related storage information.

[0249] In one embodiment, the terminal may transmit at least one of AI-related updated memory information or AI-related updated storage information to the base station in response to the request message.

[0250] In one embodiment, the terminal may transmit an uplink scheduling request message to the base station for updating at least one of AI-related memory information or AI-related storage information.

[0251] In one embodiment, the terminal can receive an uplink scheduling message from the base station.

[0252] In one embodiment, the terminal may transmit at least one of AI-related updated memory information or AI-related updated storage information to the base station in response to the uplink scheduling message.

[0253] In one embodiment, the terminal may transmit a message from the base station that includes condition information for triggering the update of at least one of AI-related memory information or AI-related storage information.

[0254] In one embodiment, the uplink scheduling request message may be based on the condition information.

[0255] In one embodiment, the terminal may receive a command message from the base station to delete a stored AI model.

[0256] In one embodiment, the terminal can delete a stored AI model based on the command message.

[0257] In one embodiment, the terminal may receive from the base station at least one of information regarding whether an AI model is running or information regarding the type of the AI ​​model, which is determined based on at least one of the AI-related memory information or the cell status information of the base station.

[0258] FIG. 15 is a drawing illustrating the structure of a base station in a wireless communication system according to one embodiment of the present disclosure.

[0259] Referring to FIG. 15, a base station may include a transceiver (1502) referring to a base station receiver and a base station transmitter, a memory (1503), and a base station processor (1501, or a base station control unit or processing unit). Depending on the communication method of the base station described above, the transceiver (1502), memory (1503), and base station processor (1501) of the base station may operate. However, the components of the base station are not limited to the examples described above. For example, the base station may include more components or fewer components than the components described above. In addition, the transceiver, memory, and processor may be implemented in the form of a single chip.

[0260] The transceiver can transmit and receive signals with a terminal. Here, the signal may include control information and data. To this end, the transceiver 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 its frequency. However, this is merely one embodiment of the transceiver, and the components of the transceiver are not limited to an RF transmitter and an RF receiver.

[0261] In addition, the transceiver receives a signal through a wireless channel and outputs it to a processor, and can transmit the signal output from the processor through a wireless channel.

[0262] Memory can store programs and data necessary for the operation of the base station. Additionally, memory can store control information or data included in signals transmitted and received by the base station. Memory can be composed of storage media or combinations of storage media such as ROM, RAM, hard disk, CD-ROM, and DVD. Additionally, there may be multiple memories.

[0263] A processor can control a series of processes to enable a base station to operate according to the embodiments of the present disclosure described above. For example, the processor can control each component of the base station to configure two layers of DCIs containing allocation information for a plurality of PDSCHs and to transmit them. There may be multiple processors, and the processors can perform control operations on the components of the base station by executing a program stored in memory.

[0264] FIG. 16 is a drawing illustrating the structure of a terminal in a wireless communication system according to one embodiment of the present disclosure.

[0265] Referring to FIG. 16, the terminal may include a transceiver (1602) referring to a terminal receiver and a terminal transmitter, a memory (1603), and a terminal processor (1601, or a terminal control unit or processing unit). Depending on the communication method of the terminal described above, the transceiver (1602), memory (1603), and terminal memory (1601) of the terminal may operate. However, the components of the terminal are not limited to the examples described above. For example, the terminal may include more components or fewer components than the components described above. Furthermore, the transceiver, memory, and processor may be implemented in the form of a single chip.

[0266] The transceiver can transmit and receive signals with a base station. Here, the signal may include control information and data. To this end, the transceiver 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 its frequency. However, this is merely one embodiment of the transceiver, and the components of the transceiver are not limited to an RF transmitter and an RF receiver.

[0267] In addition, the transceiver can receive a signal through a wireless channel and output it to a processor, and transmit the signal output from the processor through a wireless channel.

[0268] Memory can store programs and data necessary for the operation of the terminal. Additionally, memory can store control information or data included in signals transmitted and received by the terminal. Memory may be composed of storage media or combinations of storage media such as ROM, RAM, hard disk, CD-ROM, and DVD. Additionally, there may be multiple memories.

[0269] In addition, the processor can control a series of processes to enable the terminal to operate according to the aforementioned embodiment. For example, the processor can receive a DCI composed of two layers and control the components of the terminal to receive multiple PDSCHs simultaneously. There may be multiple processors, and the processors can perform the operation of controlling the components of the terminal by executing a program stored in memory.

[0270] Methods according to the embodiments described in the claims or specification of the present disclosure may be implemented in the form of hardware, software, or a combination of hardware and software. When implemented in software, a computer-readable storage medium may be provided for storing one or more programs (software modules). One or more programs stored in the computer-readable storage medium are configured for execution by one or more processors within an electronic device. One or more programs include instructions that cause the electronic device to execute the methods according to the embodiments described in the claims or specification of the present disclosure.

[0271] Such programs (software modules, software) may be stored in random access memory, non-volatile memory including flash memory, ROM (Read Only Memory), EEPROM (Electrically Erasable Programmable Read Only Memory), magnetic disc storage devices, CD-ROM (Compact Disc-ROM), Digital Versatile Discs (DVDs), or other forms of optical storage devices, magnetic cassettes. Alternatively, they may be stored in memory composed of some or all of these. Additionally, each constituent memory may include multiple units.

[0272] Additionally, the above program may be stored on an attachable storage device that can be accessed via a communication network such as the Internet, Intranet, Local Area Network (LAN), Wide LAN (WLAN), or Storage Area Network (SAN), or a combination thereof. Such a storage device may be connected to a device performing an embodiment of the present disclosure through an external port. Additionally, a separate storage device on a communication network may be connected to a device performing an embodiment of the present disclosure.

[0273] In the specific embodiments of the present disclosure described above, the components included in the invention are expressed in a singular or plural form according to the specific embodiments presented. However, the singular or plural expression is selected to suit the situation presented for convenience of explanation, and the present disclosure is not limited to singular or plural components; even if a component is expressed in the plural form, it may be composed of a singular form, or even if a component is expressed in the singular form, it may be composed of a plural form.

[0274] 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 a base station in a wireless communication system, A step of receiving a first message from a terminal comprising at least one of artificial intelligence (AI) related memory information or AI-related storage information; A step of determining AI model information based on at least one of the above AI-related memory information or AI-related storage information; and A method characterized by including the step of transmitting the AI ​​model information to the terminal.

2. In Paragraph 1, A method further comprising the step of receiving a second message containing backbone type preference information from the terminal.

3. In Paragraph 1, The above AI-related memory information includes memory size information in which the terminal can run an AI model, and A method characterized in that the above AI-related storage information includes storage size information in which the terminal can store an AI model.

4. In Paragraph 1, A step of transmitting timer information related to updating at least one of AI-related memory information or AI-related storage information to the above terminal; and A method characterized by including the step of receiving at least one of AI-related updated memory information or AI-related updated storage information from the terminal when the timer based on the above timer information has expired.

5. In Paragraph 1, A step of transmitting a request message to the above terminal for updating at least one of AI-related memory information or AI-related storage information; and A method characterized by including the step of receiving at least one of AI-related updated memory information or AI-related updated storage information in response to the request message from the terminal.

6. In Paragraph 1, A step of receiving an uplink scheduling request message from the terminal for updating at least one of AI-related memory information or AI-related storage information; and A step of transmitting an uplink scheduling message to the above terminal; A method characterized by including the step of receiving at least one of AI-related updated memory information or AI-related updated storage information from the terminal in response to the uplink scheduling message.

7. In Paragraph 6, The method further includes the step of transmitting a message to the terminal containing condition information that triggers updating at least one of AI-related memory information or AI-related storage information. A method characterized in that the above-mentioned uplink scheduling request message is based on the above-mentioned condition information.

8. In Paragraph 4, A step of transmitting a handover request message to a second base station; and A method characterized by further including the step of transmitting the AI ​​model information to the second base station.

9. In Paragraph 1, A method characterized by further including the step of transmitting a command message to delete a stored AI model to the above terminal.

10. In claim 1, a step of determining whether to run an AI model for an AI usage case of the terminal or at least one of the type of an AI model for an AI usage case based on at least one of the AI-related memory information or the cell status information of the base station; and A method further comprising the step of transmitting to the terminal, at least one of information regarding whether the determined AI model is running or information regarding the type of the AI ​​model.

11. In a method of a terminal in a wireless communication system, A step of transmitting a first message to a base station comprising at least one of artificial intelligence (AI) related memory information or AI-related storage information; and A method characterized by including the step of receiving AI model information based on at least one of the AI-related memory information or AI-related storage information from the base station.

12. In Paragraph 11, A method further comprising the step of transmitting a second message containing backbone type preference information to the base station.

13. In Paragraph 11, The above AI-related memory information includes memory size information in which the terminal can run an AI model, and A method characterized in that the above AI-related storage information includes storage size information in which the terminal can store an AI model.

14. In a base station of a wireless communication system, Transmitter / receiver; and It includes at least one processor; and the at least one processor, Receiving a first message from a terminal comprising at least one of artificial intelligence (AI) related memory information or AI-related storage information, and Determine AI model information, and A base station characterized by being configured to transmit the AI ​​model information to the above terminal.

15. In a terminal of a wireless communication system, Transmitter / receiver; and It includes at least one processor; and the at least one processor, Transmitting a first message to a base station comprising at least one of artificial intelligence (AI) related memory information or AI-related storage information, and A terminal characterized by being configured to receive AI model information from the above-mentioned base station.

Images