Terminal, base station, communication method and integrated circuit

Abstract

A terminal according to the present invention is provided with: a control circuit that determines control information relating to at least one of a calculation amount and power consumption caused by the use of an artificial intelligence model on a terminal side; and a transmission circuit that transmits the control information.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to a terminal, a base station, and a communication method. BACKGROUND

[0002] In recent years, against the background of expansion and diversification of wireless services, the leap development of Internet of Things (IoT) is expected, and the use of mobile communication is expanding to all fields such as information terminals such as smartphones, vehicles, houses, home appliances, or industrial equipment. In order to support the diversification of services, in addition to increasing system capacity, various necessary conditions such as an increase in the number of connected devices or low latency are required, and a substantial improvement in the performance and functions of the mobile communication system is required. The fifth generation mobile communication system (5G: 5th Generation mobile communication systems) has features such as large capacity and ultra-high speed (eMBB: enhanced Mobile Broadband), multi-device connection (mMTC: massive Machine Type Communication), and ultra-high reliability and low latency communication (URLLC: Ultra Reliable and Low Latency Communication), and provides wireless communication flexibly according to a wide variety of needs.

[0003] The 3rd Generation Partnership Project (3GPP), which is an international standard organization, is working on standardization of New Radio (NR) as one of the 5G wireless interfaces.

[0004] Prior Art Documents

[0005] Non-Patent Literature

[0006] Non-Patent Literature 1: RP-213599, “New SI: Study on Artificial Intelligence (AI) / Machine Learning (ML) for NR Air Interface,” Qualcomm (Moderator), December 2021. SUMMARY

[0007] However, there is room for research on methods for improving the efficiency of wireless communication.

[0008] The non-limiting embodiments of the present disclosure contribute to provide a base station, a terminal, and a communication method that can improve efficiency of wireless communication.

[0009] The terminal of one embodiment of the present disclosure includes a control circuit that decides control information related to at least one of a calculation processing amount and power consumption caused by use of an artificial intelligence model on the terminal side, and a transmission circuit that transmits the control information.

[0010] Note that these general and specific integrated circuit, computer program, and recording medium can be implemented by any combination of systems or apparatuses, methods, integrated circuits, computer programs, and recording media.

[0011] According to one embodiment of the present disclosure, it is possible to improve efficiency of wireless communication.

[0012] Further advantages and effects of one embodiment of the present disclosure will be clarified by the description and drawings. These advantages and / or effects are provided by the features described in the specification and drawings, but it is not necessary to provide all of them in order to obtain one or more of the same features. BRIEF DESCRIPTION OF DRAWINGS

[0013] Figure 1 FIG. 1 is a diagram illustrating an example of a combination of processing sets related to AI / ML processing.

[0014] Figure 2 FIG. 2 is a block diagram illustrating a structure example of a base station.

[0015] Figure 3 FIG. 3 is a block diagram illustrating a structure example of a terminal.

[0016] Figure 4 FIG. 4 is a diagram illustrating an example of a combination of processing sets related to AI / ML processing.

[0017] Figure 5 FIG. 5 is a diagram illustrating an example of a combination of processing sets related to AI / ML processing.

[0018] Figure 6 FIG. 6 is a diagram illustrating an example of a combination of processing sets related to AI / ML processing.

[0019] Figure 7 FIG. 7 is a diagram illustrating an example of a combination of processing sets related to AI / ML processing.

[0020] Figure 8This is a diagram illustrating examples of actions taken by the terminal and the base station.

[0021] Figure 9 This is a diagram representing an example of an architecture.

[0022] Figure 10 This is a block diagram representing a structural example of a base station.

[0023] Figure 11 This is a block diagram representing a structural example of a terminal.

[0024] Figure 12 This is a diagram illustrating the architecture of a 3GPP NR (3rd generation partnership project new radio) system.

[0025] Figure 13 This is a schematic diagram illustrating the functional separation between NG-RAN (Next Generation-Radio Access Network) and 5GC (5th Generation Core).

[0026] Figure 14 This is a timing diagram of the setting / resetting process for an RRC (Radio Resource Control) connection.

[0027] Figure 15 This is a schematic diagram illustrating the application scenarios of high-capacity high-speed communication (eMBB: enhanced mobile broadband), massive machine-type communications (mMTC: massive machine-type communications), and ultra-reliable and low-latency communications (URLLC: ultra-reliable and low-latency communications).

[0028] Figure 16 This is a block diagram representing an exemplary 5G system architecture for non-roaming scenarios. Detailed Implementation

[0029] The embodiments of this disclosure will now be described in detail with reference to the accompanying drawings.

[0030] In NR, for example, an access method based on Orthogonal Frequency Division Multiplexing (OFDM) is used in downlink transmission. Furthermore, Multiple-Input Multiple-Output (MIMO) is employed to increase data rates. To effectively utilize MIMO performance, closed-loop control is introduced, feeding back Channel State Information (CSI) from the terminal (e.g., also known as user equipment (UE)) to the base station (e.g., also known as gNB). The aim is to achieve high performance with less control information in the CSI feedback.

[0031] In 3GPP Release 18 (e.g., also known as Rel. 18), artificial intelligence (AI) technologies such as machine learning (ML) (hereinafter also referred to as "AI / ML") are discussed for applying to the radio interface of NR (e.g., see Non-Patent Document 1). Examples of use cases utilizing AI / ML technologies are being investigated, such as CSI feedback, beam control, and location estimation.

[0032] For example, regarding CSI feedback, methods are being researched to compress the information content of CSI in the spatial and frequency domains using AI / ML techniques (hereinafter also referred to as CSI compression). For instance, in machine learning, by training AI / ML models (e.g., artificial intelligence models) using large amounts of training data, their features can be automatically extracted. In CSI compression, algorithms primarily used for dimensionality compression and reconstruction of image data, such as those called autoencoders, can be utilized. Figure 1 As shown, the CSI matrix, represented by a two-dimensional region in the spatial and frequency domains, is regarded as an image. The encoding part (encoder) of the autoencoder is applied to the CSI compression process on the terminal side, and the decoding part (decoder) for reconstruction is applied to the CSI reconstruction process on the base station or network (e.g., also referred to as base station / network).

[0033] Generally, high-performance AI / ML models are large in size, which can lead to significant computational complexity and power consumption during processing on the terminal. Therefore, in the practical application of AI / ML models, it is necessary to appropriately balance the performance improvements (e.g., communication performance) brought by using AI / ML with the computational load and power consumption of the terminal.

[0034] In use cases where AI / ML technology is applied to wireless interfaces and AI / ML models (or parts thereof) are configured on the terminal side, the use of AI / ML models via Model-ID-based Life Cycle Management (LCM) is being considered. In Model-ID-based LCM, the decision regarding the use of the AI / ML model can be made by the base station / network side, regardless of whether the model is trained on the terminal side or the network side.

[0035] For example, it is being considered to include a function to monitor the performance of AI / ML models (referred to as performance monitoring) as part of LCM. Examples of AI / ML model performance monitoring include, for instance, performance monitoring to confirm the behavior of unverified new models or parameters (e.g., also known as Category 1 monitoring), and performance monitoring to confirm whether the currently used AI / ML model or parameters are suitable for the current environment (e.g., also known as Category 2 monitoring).

[0036] For example, in performance monitoring of the behavior of new, unverified models or parameters, the actions taken after monitoring the performance of AI / ML models (groups) may include not changing (or updating) the AI / ML models (groups), updating the parameters of the AI / ML models (groups), updating the AI / ML models (groups) themselves, or deactivating or stopping the preconfigured settings of the AI / ML models (groups).

[0037] Furthermore, for example, in performance monitoring to confirm whether the currently used AI / ML model or parameters are suitable for the current environment, as actions taken after monitoring the performance of the AI / ML model, the currently running (activated) AI / ML model may not be changed (or updated), the parameters of the currently running (activated) AI / ML model may be updated, the currently running (activated) AI / ML model may be changed (switched) to another AI / ML model in the pre-configured AI / ML model group, or the operation of the currently running (activated) AI / ML model may be stopped (deactivated) and reverted to the CSI report method that does not use AI / ML.

[0038] In model ID-based LCM where the use of AI / ML models is determined by the base station / network side, in order to determine whether the communication performance brought by the use of AI / ML models has been properly balanced with the computing power and power consumption of the terminal when using AI / ML models, it is expected that the base station / network can determine information related to the computing power and power consumption of the terminal.

[0039] For example, one approach could be to implement the AI / ML model on the terminal side within the base station / network, and determine the usage of the AI / ML model on the terminal based on its computational load or power consumption. However, the actual implementation of the AI / ML model may vary depending on the terminal's hardware or software. Therefore, it is difficult to determine the actual computational load or power consumption of the terminal-side AI / ML model in the base station / network without information from the terminal.

[0040] For example, there is still room for research into methods for terminals to report information related to at least one of computing power and power consumption (hereinafter also referred to as "computing power consumption / power consumption") to base stations / networks.

[0041] In one non-limiting embodiment of this disclosure, a method for a terminal to report computational processing load / power consumption to a base station / network is described. For example, a method is described where, in a use case where an AI / ML model (or a portion thereof) is configured on the terminal side, when the decision regarding the use of the AI / ML model is made by the base station / network, the base station / network considers the balance between the communication performance resulting from the use of the AI / ML model and the computational processing load / power consumption of the terminal, and appropriately determines the use of the AI / ML model on the terminal side. For example, the terminal assists the base station / network in determining the use of the AI / ML model by reporting information related to the computational processing load / power consumption resulting from the use of the AI / ML model on the terminal side.

[0042] The following describes non-limiting embodiments of this disclosure.

[0043] [Overview of Communication Systems]

[0044] One aspect of the communication system disclosed herein includes, for example, at least one base station and at least one terminal.

[0045] Figure 2 This is a block diagram illustrating a structural example of a base station 100 according to an embodiment of the present disclosure. Figure 3 This is a block diagram illustrating a structural example of a terminal 200 according to an embodiment of the present disclosure.

[0046] exist Figure 2 In the base station 100 shown, a receiving unit (e.g., corresponding to a receiving circuit) receives control information relating to at least one of the computational processing volume and power consumption resulting from the use of the artificial intelligence model (AI / ML model) on the terminal side. A control unit (e.g., corresponding to a control circuit) controls the use of the artificial intelligence model on the terminal side based on the control information.

[0047] exist Figure 3 In the terminal 200 shown, the control unit (e.g., corresponding to the control circuit) determines control information related to at least one of the computational processing volume and power consumption resulting from the use of the artificial intelligence model (AI / ML model) on the terminal side. The transmitting unit (e.g., corresponding to the transmitting circuit) transmits the control information.

[0048] (Implementation Method 1)

[0049] In this embodiment, terminal 200 reports information related to the amount of computing power / power consumption in terminal 200 to base station 100 (e.g., base station / network). Based on the information reported by terminal 200, base station / network makes a determination on the use of AI / ML models in terminal 200.

[0050] In this way, by utilizing information related to the computational processing volume / power consumption resulting from the use of the terminal-side AI / ML model in terminal 200, the base station / network can be assisted in making judgments about the use of the AI / ML model.

[0051] Furthermore, the information reported from terminal 200 may include one of the information related to the computing power of terminal 200 and the information related to the power consumption of terminal 200, or it may include both the information related to the computing power of terminal 200 and the information related to the power consumption of terminal 200.

[0052] For example, information related to the computational workload resulting from the use of the terminal-side AI / ML model in terminal 200 may include information related to at least one of FLOPs (Floating-point Operations) and memory size for inference processing using the AI / ML model.

[0053] In addition, for example, information related to power consumption resulting from the use of the terminal-side AI / ML model in terminal 200 may include information related to the absolute value of power consumption of terminal 200 when using the AI / ML model.

[0054] According to this embodiment, the base station / network can determine (or understand) the computational processing volume / power consumption of the terminal 200 caused by the use of the AI / ML model on the terminal side based on the information related to computational processing volume / power consumption reported from the terminal 200, thereby enabling appropriate judgment on the use of the AI / ML model on the terminal side.

[0055] In addition, reports of information related to computing power / power consumption from terminal 200 can be sent via user plane (U-plane), higher-layer signaling (e.g., Radio Resource Control (RRC)), Medium Access Control-Control Element (MAC-CE), or Uplink Control Information (UCI).

[0056] (Implementation Method 2)

[0057] For example, the power consumption of a device (e.g., terminal 200) may depend on various factors such as operating conditions like temperature, or other actions performed in parallel with the action of using the AI / ML model in terminal 200. Therefore, for example, even if the absolute value of power consumption when using the AI / ML model in terminal 200 is reported, the base station / network may not be able to properly determine the use of the AI / ML model, and may not be able to properly determine the next action after the use of the AI / ML model.

[0058] Furthermore, for example, processing with high computational load but no frequent memory access may consume the same power as processing with low computational load but frequent memory access in some terminal implementations, but different power in other terminal implementations. Thus, the power consumption of a terminal may vary depending on its implementation. Therefore, for example, even when reporting the computational load or memory size of terminal 200, the relationship between power consumption and computational load or memory size when using an AI / ML model remains unclear. The base station / network may be unable to properly determine the use of the AI / ML model and thus the appropriate next action after such a determination.

[0059] Therefore, in this embodiment, the terminal 200 reports information, for example, a comparison of computational processing volume / power consumption when using an AI / ML model with that when not using an AI / ML model to the base station / network.

[0060] Thus, in this embodiment, information comparing the computational processing volume / power consumption of the terminal 200 when using the AI / ML model with that when not using the AI / ML model is used to assist the base station / network in determining the use of the AI / ML model.

[0061] Furthermore, the control information (or comparison object) reported from the terminal 200 to the base station / network can be information related to either computational processing volume or power consumption, or information related to both computational processing volume and power consumption.

[0062] For example, terminal 200 can report information to the base station / network about whether the computational processing volume / power consumption when using an AI / ML model is greater (or less) than that when not using an AI / ML model.

[0063] Furthermore, in order for the base station / network to correctly understand the information related to the comparison of computational processing volume / power consumption, it is desirable to share knowledge related to the calculation method (e.g., the method for determining control information) used by the terminal 200 for comparison of computational processing volume / power consumption between the terminal 200 and the base station / network.

[0064] Sharing of knowledge related to the calculation method for computational processing power / power consumption used for comparison can be achieved, for example, by pre-defining it through standards. Furthermore, for example, terminal 200 can report the calculation method for computational processing power / power consumption used for comparison to the base station / network when reporting information related to its capabilities or the application capabilities of AI / ML models. Additionally, the base station / network can configure the calculation method for computational processing power / power consumption used for comparison to terminal 200 through semi-static configuration such as RRC. Furthermore, terminal 200 can report the calculation method for computational processing power / power consumption used for comparison along with information related to the comparison of computational processing power / power consumption when reporting it to the base station / network.

[0065] For example, the calculation method for computational processing volume / power consumption can be based on at least one of the following: the average computational processing volume / power consumption within a specified interval, the peak (maximum) computational processing volume / power consumption occurring within the specified interval, and the total computational processing volume / power consumption within the specified interval. For example, the specified interval can be the CSI computation time defined in the standard, or it can be set by the terminal 200 or the base station / network.

[0066] Furthermore, for example, comparisons of computational processing power can be made based on at least one of the FLOPs of inference processing using AI / ML models and memory size.

[0067] According to this embodiment, the base station / network can determine (or grasp) the relationship between the computational workload / power consumption of the terminal 200 under the same conditions, between actions based on the use of the terminal-side AI / ML model and existing actions (e.g., actions without using the terminal-side AI / ML model). In other words, by comparing the computational workload / power consumption with that without using the terminal-side AI / ML model, the base station / network can accurately determine the computational workload / power consumption when using the terminal-side AI / ML model, independent of the implementation details of the terminal 200.

[0068] In this way, the base station / network can appropriately determine the use of AI / ML models on the terminal side. Furthermore, the base station / network can appropriately select the next action after determining the use of AI / ML models on the terminal side (e.g., actions such as continuing to use the AI / ML model, stopping the currently running AI / ML model and reverting to a method that does not use AI / ML, etc.).

[0069] In addition, reports of information related to the comparison of computing power / power consumption from terminal 200 can be sent via one of the user plane, RRC, MAC-CE, and UCI.

[0070] (Implementation Method 3)

[0071] In this embodiment, terminal 200, for example, reports information about the difference between the computational processing volume / power consumption when using an AI / ML model and when not using an AI / ML model to the base station / network.

[0072] Thus, in this embodiment, information about the difference between the computational processing volume / power consumption of the terminal 200 when using the AI / ML model and when not using the AI / ML model is used to assist the base station / network in determining the use of the AI / ML model.

[0073] Furthermore, the control information (or the object for calculating differential values) reported from the terminal 200 to the base station / network can be information related to either the amount of computation or the power consumption, or information related to both the amount of computation and the power consumption.

[0074] For example, as information about the difference between computational processing volume / power consumption when using an AI / ML model and when not using an AI / ML model, the terminal 200 may report to the base station / network a value representing the absolute value of the difference, or a value representing one of a range of values ​​obtained by dividing (or quantizing) the difference by a certain level.

[0075] Furthermore, in order for the base station / network to correctly understand the information related to the difference in computational processing volume / power consumption, it is desirable to share knowledge related to the calculation method (e.g., the method for determining control information) of the difference in computational processing volume / power consumption in the terminal 200 between the terminal 200 and the base station / network.

[0076] Sharing of knowledge related to the calculation method of the difference between computational processing power and power consumption can be achieved, for example, by pre-defining it through standards. Furthermore, for example, terminal 200 can report the calculation method of the difference between computational processing power and power consumption to the base station / network when reporting information related to its own capabilities or the application capabilities of AI / ML models. Additionally, the base station / network can configure the calculation method of the difference between computational processing power and power consumption to terminal 200 through semi-static configuration such as RRC. Furthermore, terminal 200 can report the calculation method of the difference between computational processing power and power consumption along with information related to this difference when reporting it to the base station / network.

[0077] For example, the method for calculating the difference between computational processing volume and power consumption can be based on at least one of the following: the average computational processing volume / power consumption within a specified interval, the peak (maximum) computational processing volume / power consumption occurring within the specified interval, and the total computational processing volume / power consumption within the specified interval. For example, the specified interval can be the CSI computation time defined in the standard, or it can be set by the terminal 200 or the base station / network.

[0078] Furthermore, for example, the difference in computational processing volume can be calculated based on at least one of the FLOPs of inference processing using AI / ML models and the memory size.

[0079] According to this embodiment, the base station / network can determine (or grasp) the relative value of the computational processing volume / power consumption of the terminal 200 between actions based on the use of the terminal-side AI / ML model and existing actions (e.g., actions without using the terminal-side AI / ML model). In other words, by comparing the computational processing volume / power consumption with that without using the terminal-side AI / ML model, the base station / network can accurately determine the computational processing volume / power consumption of the terminal-side AI / ML processing without relying on the implementation of the terminal 200.

[0080] In this way, the base station / network can appropriately determine the use of AI / ML models on the terminal side. Furthermore, the base station / network can appropriately select the next action after determining the use of AI / ML models on the terminal side (e.g., actions such as continuing to use the AI / ML model, stopping the currently running AI / ML model and reverting to a method that does not use AI / ML, etc.).

[0081] In addition, reports of information related to the difference in computing power / power consumption from terminal 200 can be sent via one of the user plane, RRC, MAC-CE, and UCI.

[0082] (Implementation Method 4)

[0083] In this embodiment, for example, similar to embodiments 2 and 3, the terminal 200 may report information about the comparison or difference between the computational processing volume / power consumption when using the AI / ML model and when not using the AI / ML model to the base station / network.

[0084] Furthermore, in this embodiment, the terminal 200 may, for example, report information to the base station / network regarding a comparison or difference between the computational processing volume / power consumption when using a certain AI / ML model (e.g., a first AI / ML model) and when using another AI / ML model (e.g., a second AI / ML model). Additionally, in this embodiment, for example, the terminal 200 may report information to the base station / network regarding a comparison or difference between the computational processing volume / power consumption when using a certain parameter (e.g., a first parameter) of an AI / ML model and when using another parameter (e.g., a second parameter) of the same AI / ML model.

[0085] Thus, in this embodiment, information about the computational processing volume / power consumption when using a certain AI / ML model in the terminal 200 and the computational processing volume / power consumption when using another AI / ML model or different parameters of the same AI / ML model is used to assist the base station / network in determining the use of AI / ML models.

[0086] Furthermore, the control information (or the object of comparison or the object of calculation of differential values) reported from the terminal 200 to the base station / network can be information related to either the amount of computation or the power consumption, or information related to both the amount of computation and the power consumption.

[0087] For example, terminal 200 may report information to the base station / network about whether the computational processing volume / power consumption when using the AI / ML model currently pre-configured in terminal 200 or the AI / ML model currently used by terminal 200 is greater (or smaller) than the computational processing volume / power consumption when not using the AI / ML model, or information related to the difference in computational processing volume / power consumption between the two cases.

[0088] In addition, the terminal 200 may report to the base station / network information regarding whether the computational processing volume / power consumption when using the AI / ML model currently pre-configured in the terminal 200 or the AI / ML model currently in use by the terminal 200 is greater (or smaller) than when using an AI / ML model that may be pre-configured to the terminal 200 in the future, an AI / ML model that the terminal 200 is not currently using, or different parameters (unused parameters) of the AI / ML model currently pre-configured in the terminal 200 or the AI / ML model currently in use by the terminal 200.

[0089] Furthermore, in order for the base station / network to correctly understand the information related to the comparison or difference value of computing processing volume / power consumption, it is desirable to share knowledge related to the calculation method of computing processing volume / power consumption used for comparison or the calculation method of the difference value of computing processing volume / power consumption in the terminal 200 (e.g., the method for determining control information) between the terminal 200 and the base station / network.

[0090] Sharing of knowledge related to the calculation method for computational processing volume / power consumption used for comparison or the calculation method for the difference in computational processing volume / power consumption in terminal 200 (hereinafter referred to as "comparison or difference-related calculation method") can be achieved, for example, by being predefined by a standard. Furthermore, for example, terminal 200 may report the comparison or difference-related calculation method to the base station / network when reporting information related to the capabilities of terminal 200 or the application capabilities of AI / ML models. Furthermore, the base station / network can set the comparison or difference-related calculation method to terminal 200 through semi-static configuration such as RRC. Additionally, terminal 200 may report the comparison or difference-related calculation method along with information related to the comparison or difference in computational processing volume / power consumption when reporting such information to the base station / network.

[0091] For example, the calculation method related to the comparison or difference value can be based on at least one of the average computational processing volume / power consumption within a specified interval, the peak (maximum) computational processing volume / power consumption occurring within the specified interval, and the total computational processing volume / power consumption within the specified interval. For example, the specified interval can be the CSI computation time defined in the standard, or it can be set by the terminal 200 or the base station / network.

[0092] Furthermore, for example, the difference in computational processing volume can be calculated based on at least one of the FLOPs of inference processing using AI / ML models and the memory size.

[0093] According to this embodiment, the base station / network can determine (or grasp) the magnitude or relative value of the computational processing volume / power consumption of the terminal 200 between actions based on the use of the terminal-side AI / ML model, existing actions (e.g., actions without using the terminal-side AI / ML model), or actions when the AI / ML model or parameters are changed. That is, similar to embodiments 2 and 3, the base station / network can accurately determine the computational processing volume / power consumption of the terminal-side AI / ML processing without relying on the implementation of the terminal 200, and can determine the magnitude or relative value of the computational processing volume / power consumption that may be set between AI / ML models or between the parameters of the AI / ML models of the terminal 200.

[0094] In this way, the base station / network can appropriately determine the use of AI / ML models on the terminal side. Furthermore, the base station / network can appropriately select the next action after determining the use of AI / ML models on the terminal side (e.g., actions such as continuing to use the AI / ML model, switching the pre-configured or used AI / ML model or parameters, stopping the use of the AI / ML model or the operation of the currently running AI / ML model and reverting to a method that does not use AI / ML, etc.).

[0095] In addition, reports of information related to comparisons or differential values ​​of computing power / power consumption from terminal 200 can be sent via one of the user plane, RRC, MAC-CE, and UCI.

[0096] Furthermore, in this embodiment, the comparison of computational processing volume / power consumption in the control information (e.g., information related to comparison results or differential values) reported from the terminal 200 to the base station / network is not limited to the comparison of computational processing volume / power consumption when using an AI / ML model versus when not using an AI / ML model, or the comparison of computational processing volume / power consumption when using a certain AI / ML model versus when using another AI / ML model or different parameters of the same AI / ML model.

[0097] For example, it can also report the comparison or difference between the computational processing volume / power consumption when not using a certain AI / ML model and when using another AI / ML model or different parameters of that AI / ML model. Alternatively, it can also report the comparison or difference between the computational processing volume / power consumption of multiple AI / ML models or parameters that are not currently being used by terminal 200.

[0098] Alternatively, a reference model can be defined for calculating the computational processing volume / power consumption comparison or differential value in the control information (e.g., information related to comparison results or differential values) reported from the terminal 200 to the base station / network in this embodiment. The following comparisons are performed and the results or differential values ​​are reported: a comparison of computational processing volume / power consumption using a certain AI / ML model with that of a reference model; a comparison of computational processing volume / power consumption using another AI / ML model or different parameters of the same AI / ML model with that of the reference model; and a comparison of computational processing volume / power consumption without using an AI / ML model with that of the reference model. The structure or parameters of the reference model used for the computational processing volume / power consumption comparison can be notified or set by the base station / network to the terminal 200, or the terminal 200 can determine and notify the base station / network. Furthermore, the structure or parameters of the reference model can also be defined by a standard.

[0099] (Implementation Method 5)

[0100] In this embodiment, the terminal 200, for example, reports information related to the utilization or occupancy rate of computing processing volume / power consumption to the base station / network. Therefore, in this embodiment, the information related to the utilization or occupancy rate of computing processing volume / power consumption in the terminal 200 assists the base station / network in determining the use of AI / ML models.

[0101] In addition, the information reported from the terminal 200 may be related to the utilization or occupancy of either computing power or power consumption, or it may be related to the utilization or occupancy of both computing power and power consumption.

[0102] For example, when performing AI / ML processing using a general-purpose processor, information related to the utilization or occupancy of computing power may be information related to the CPU (Central Processing Unit) utilization rate. For instance, information related to the utilization or occupancy of computing power may be at least one of the following: the current overall CPU utilization rate of terminal 200, the CPU utilization rate of computing power when using an AI / ML model, the CPU utilization rate of computing power when not using an AI / ML model, or the CPU utilization rate of computing power when using another AI / ML model or different parameters of the same AI / ML model.

[0103] Furthermore, for example, when using a dedicated AI / ML processing accelerator, information related to the utilization or occupancy of computational processing could be the utilization or resource occupancy of computational processing within that accelerator when using an AI / ML model.

[0104] Alternatively, the utilization or occupancy rate of computing power and the utilization or occupancy rate of power consumption can be calculated and reported separately. Or, a combined utilization or occupancy rate can be defined based on both the utilization or occupancy rate of computing power and the utilization or occupancy rate of power consumption, and information related to the combined utilization or occupancy rate can be reported.

[0105] Furthermore, in order for the base station / network to correctly understand the information related to the utilization or occupancy of computing power / power consumption, it is desirable to share knowledge related to the calculation method (e.g., the method for determining control information) of the computing power / power consumption utilization or occupancy in the terminal 200 between the terminal 200 and the base station / network.

[0106] Sharing of knowledge related to the calculation method of computing power utilization or occupancy can be achieved, for example, by pre-defining it through standards. Furthermore, for example, terminal 200 can report the calculation method of computing power utilization or occupancy to the base station / network when reporting information related to its own capabilities or the application capabilities of AI / ML models. Additionally, the base station / network can configure the calculation method of computing power utilization or occupancy to terminal 200 through semi-static configuration such as RRC. Furthermore, terminal 200 can report the calculation method of computing power utilization or occupancy along with information related to computing power utilization or occupancy when reporting such information to the base station / network.

[0107] For example, the calculation method for the utilization or occupancy rate of computational processing volume / power consumption can be based on at least one of the following: the average computational processing volume / power consumption within a specified interval, the peak (maximum) computational processing volume / power consumption occurring within the specified interval, and the total computational processing volume / power consumption within the specified interval. For example, the specified interval can be the CSI computation time defined in the standard, or it can be set by the terminal 200 or the base station / network.

[0108] According to this embodiment, the base station / network can appropriately determine the use of the AI / ML model by considering the utilization rate or occupancy rate of the terminal 200's computing power / power consumption. Furthermore, the base station / network can appropriately select the next action after the terminal-side AI / ML model usage determination (e.g., actions such as continuing to use the AI / ML model, switching the pre-configured or used AI / ML model or parameters, stopping the use of the AI / ML model or the operation of the currently running AI / ML model and reverting to a method that does not use AI / ML, etc.).

[0109] For example, the base station / network could determine the use of AI / ML models in a way that ensures the utilization or occupancy rate of the terminal 200's computing power / power consumption does not exceed 100%.

[0110] Furthermore, the base station / network may be configured to utilize or occupy more than 100% of the computing power / power consumption on the terminal side. In this case, processing may be achieved in a best-effort manner.

[0111] In addition, reports from terminal 200 related to the utilization or occupancy of computing power / power consumption can be sent via one of the user plane, RRC, MAC-CE, and UCI.

[0112] The above describes the various embodiments of this disclosure.

[0113] (Variation Example 1)

[0114] In the case of application performance monitoring processing of terminal 200 (e.g., in the case of performance monitoring of AI / ML model of terminal 200), the above-mentioned information related to the computing power / power consumption of terminal 200 and the information related to performance monitoring processing (e.g., information related to KPI of AI / ML model, information related to failure of AI / ML model) can be reported from terminal 200 to base station / network.

[0115] In addition, the aforementioned information related to the computing power / power consumption of the terminal 200, along with signals used to notify the terminal 200 of its AI / ML-related capabilities, can be reported from the terminal 200 to the base station / network.

[0116] In addition, for example, if a notification is triggered from the base station / network, the aforementioned information related to the computing power / power consumption of the terminal 200 can be reported from the terminal 200 to the base station / network.

[0117] Furthermore, for example, when a set timer expires, or when information related to computational processing volume / power consumption exceeds a threshold, the aforementioned information related to the computational processing volume / power consumption of the terminal 200 can be reported from the terminal 200 to the base station / network. Moreover, the timer setting and the threshold for the information related to computational processing volume / power consumption can be predetermined by a standard, or set by the terminal 200 and reported to the base station / network, or set to the terminal 200 through semi-static configuration such as RRC performed by the base station / network.

[0118] (Variation Example 2)

[0119] Furthermore, the above embodiments describe a scenario where the terminal 200 reports information related to the comparison results or difference values ​​of computational processing volume / power consumption between two processes that differ in at least one of the following: the use of an AI / ML model, the AI / ML model, and the parameters. However, this is not a limitation. For example, the terminal 200 may also report the comparison results (or difference values) of computational processing volume / power consumption between three or more processes that differ in at least one of the following: the use of an AI / ML model, the AI / ML model, and the parameters. For example, the terminal 200 may also report the largest (or smallest) value of computational processing volume / power consumption among the three or more processes.

[0120] (Variation Example 3)

[0121] The computational workload / power consumption reported by terminal 200 when using AI / ML models can be the computational workload / power consumption required for inference processing using AI / ML models, the computational workload / power consumption required for performance monitoring, or the computational workload / power consumption for other purposes. Furthermore, it can report the total of multiple processes (e.g., the computational workload / power consumption required for inference processing and the computational workload / power consumption required for performance monitoring) or report them separately. Moreover, the timing of reporting the computational workload / power consumption for each process can be the same or different.

[0122] [Performance monitoring and examples of actions following performance monitoring]

[0123] Performance monitoring of AI / ML models includes, for example, performance monitoring that validates the behavior of unverified new models or parameters (e.g., referred to as Category 1 monitoring) and performance monitoring that validates whether the currently used AI / ML models or parameters are suitable for the current environment (e.g., referred to as Category 2 monitoring).

[0124] For example, Category 1 monitoring can be applied to the performance monitoring of AI / ML model groups pre-configured in terminal 200. Furthermore, Category 2 monitoring can be applied to the performance monitoring of currently running (activated) AI / ML models. Additionally, different implementation methods and variations corresponding to the categories of performance monitoring can also be applied.

[0125] In Category 1 monitoring, actions taken after monitoring the performance of AI / ML models (groups) can prevent the AI / ML models (groups) from being changed (or updated), update the parameters of the AI / ML models (groups), update the AI / ML models (groups) themselves, or cancel or stop the pre-configured settings of the AI / ML models (groups).

[0126] In Category 2 monitoring, actions taken after monitoring the performance of an AI / ML model can include: preventing the currently running (activated) AI / ML model from being changed (or updated); updating the parameters of the currently running (activated) AI / ML model; changing the currently running (activated) AI / ML model to another AI / ML model in a pre-configured AI / ML model group; or stopping (deactivating) the currently running (activated) AI / ML model and reverting to the CSI reporting method that does not use AI / ML.

[0127] Furthermore, in functionality-based LCM (Life Cycle Management), the terminal 200 can determine which action to perform after performance monitoring based on the implementation. On the other hand, in model ID-based LCM, the base station / network can determine which action to perform after performance monitoring.

[0128] [Training methods for AI / ML models]

[0129] In CSI compression using a two-sided model, i.e., using an encoder for compression on the terminal side and a decoder for reconstruction on the base station / network side, one of the following methods can be applied to train the encoder (CSI compression unit) on the terminal side and the decoder (CSI reconstruction unit) on the base station / network side.

[0130] <Training Method 1>

[0131] Training method 1 involves joint training of a two-sided model on one side (either the terminal side or the network side). In training method 1, for example, on the network side, the CSI compression unit and the CSI reconstruction unit are trained jointly. The relevant information of the trained CSI compression unit is then distributed / transmitted to the terminal. Alternatively, the CSI compression unit and the CSI reconstruction unit can be trained jointly on the terminal side.

[0132] <Training Method 2>

[0133] Training method 2 involves joint training of the two-sided models on both the network and terminal sides. In training method 2, for example, by exchanging forward propagation (FP) or backward propagation (BP) information between the terminal and the network, the terminal and network sides jointly train the CSI compression and CSI reconstruction parts, respectively. Furthermore, training method 2 also includes a method where, after joint training of the two-sided models on one side (terminal or network side) using training method 1, the model on one side is fixed, and FP or BP information is exchanged between the terminal and the network to train the model on the other side.

[0134] <Training Method 3>

[0135] Training method 3 involves training the CSI compression unit on the terminal side and the CSI reconstruction unit on the network side separately. In training method 3, for example, the CSI compression unit and the CSI reconstruction unit are jointly trained on the network side. After the network side training is completed, the network side shares the training dataset (e.g., the input data and output data used for training) with the terminal 200. The terminal 200 uses the dataset shared from the network side to train the CSI compression unit on the terminal side.

[0136] For example, network-side performance monitoring based on the target CSI reported from terminal 200 is useful for cases where model training is performed entirely on the network side, and for Category 2 monitoring with LCM based on model ID. On the other hand, terminal-side performance monitoring based on the CSI reconstruction output notified by the network is useful when the CSI reconstruction unit on the network side is trained by terminal 200.

[0137] [Use Case]

[0138] The use cases applying the above-described implementation methods or variations are not limited to CSI compression (CSI feedback enhancement), but can also be applied to any use case where AI / ML models are configured on the terminal side. Examples of use cases for configuring AI / ML models on the terminal side include CSI prediction based on terminal-side AI / ML models, beam prediction in the spatial or temporal domain based on terminal-side AI / ML models, positioning accuracy enhancement based on terminal-side AI / ML models, and optimization of parameters and configuration of AI / ML-based wireless interfaces.

[0139] Furthermore, when the processor or AI / ML accelerator is general for AI / ML processing, AI / ML processing for different use cases can be executed simultaneously, or multiple AI / ML processing options with different use cases can be selected.

[0140] [AI / ML Models]

[0141] In the non-limiting embodiments of this disclosure, "AI / ML model (or model)" may include a physical model, a binary model, an executable model, a converted model, a source code model, a non-executable model, a raw model, a logical model, or other model forms.

[0142] [Notification related to the AI / ML processing capabilities of Terminal 200]

[0143] Information indicating whether terminal 200 supports the functions, actions, or processes shown in the above embodiments or variations can be sent (or notified) from terminal 200 to base station 100 as terminal capability information or capability parameters.

[0144] As described above, when CSI compression is performed using a two-sided model—that is, an encoder for compression on the terminal side and a decoder for reconstruction on the base station / network side—it is desirable to monitor the performance of the AI / ML model to ensure the availability of the AI / ML model in both the terminal-side CSI compression unit and the base station / network-side CSI reconstruction unit. For example, the following scenario can be considered where the availability of the AI / ML model deteriorates: although an AI / ML model trained in a certain environment (e.g., a specific cell, hereinafter referred to as cell #A) is used in the same environment (e.g., cell #A), changes in the environment due to factors such as the mobility of the terminal 200 (e.g., cell #B) cause the model to no longer match the training environment (e.g., cell #A), leading to performance degradation of the AI / ML model. In this case, it is expected that the terminal 200 will perform appropriate processing such as switching to use an AI / ML model suitable for the changed environment (e.g., an AI / ML model trained in the environment of cell #B) or stopping the CSI compression processing that utilizes AI / ML.

[0145] As mentioned above, in order to run AI / ML models properly, it is desirable to monitor the performance of the currently running AI / ML model, or the performance of an AI / ML model that is not currently running but may be switched to.

[0146] Generally, high-performance AI / ML models are large in size, and high computing power is desired when implementing them in terminal 200. Therefore, terminal 200 may have capabilities or limitations regarding the AI / ML actions and processing it can support (or execute simultaneously). Furthermore, ideally, the network side should have control over the AI / ML processing capabilities of terminal 200.

[0147] On the other hand, the more AI / ML models there are, or the more processing procedures associated with each AI / ML model (e.g., inference-based CSI reports (e.g., compressed CSI reports) and performance monitoring), the more likely the amount of information notifying the terminal 200 of its AI / ML processing capabilities will become large and complex.

[0148] The following is an example illustrating the scenario where terminal 200 has the ability to implement each of the following: CSI reporting / monitoring based on "Model A", CSI reporting / monitoring based on "Model B", and existing CSI reporting / monitoring without applying AI / ML.

[0149] In this scenario, terminal 200 is expected to notify, in addition to the capability information regarding the executable models and existing CSI reports / monitoring, of the following: Whether performance monitoring of CSI reports using model A can be performed simultaneously when executing CSI reports using model A; whether performance monitoring of existing CSI reports without AI / ML can be performed simultaneously when executing CSI reports using model A; whether performance monitoring of CSI reports using model A can be performed simultaneously when executing CSI reports using model B; whether performance monitoring of existing CSI reports without AI / ML can be performed simultaneously when executing CSI reports using model B; and whether, when executing existing CSI reports without AI / ML, performance monitoring is... Can performance monitoring of CSI reports using Model A be performed simultaneously? When executing existing CSI reports without AI / ML, can performance monitoring of CSI reports using Model B be performed simultaneously? When executing CSI reports using Model A, can performance monitoring of CSI reports using Model B and existing CSI reports without AI / ML be performed simultaneously? When executing CSI reports using Model B, can performance monitoring of CSI reports using Model A and existing CSI reports without AI / ML be performed simultaneously? When executing existing CSI reports without AI / ML, can performance monitoring of CSI reports using Model A and CSI reports using Model B be performed simultaneously?

[0150] For example, to simplify notifications related to the AI / ML processing capabilities of terminal 200, notifications of AI / ML processing capabilities based on "AI / ML processing set" and "process set combination" can be applied.

[0151] For example, a set of processing related to AI / ML processing can be defined for each AI / ML model. For instance, within the AI / ML processing set, such as... Figure 4 As shown, each entry corresponds to a category of AI / ML processing, such as CSI reports (e.g., CSI reports based on inference (denoted as "Inference") or existing CSI reports (denoted as "Reporting")) and performance monitoring (e.g., denoted as "Monitoring"). Furthermore, notification of process set combinations can be achieved, for example, by notifying combinations of process sets. Figure 5 As shown, when notifying the combination (processing set A, processing set B), i.e., (A+B), it indicates that a CSI report based on inference processing using model A (“Model A inference”) can be executed simultaneously as the first entry for processing set A, and a performance monitoring report using model B (“Model B monitoring”) as the second entry for processing set B. Furthermore, if a combination of more than three processing sets needs to be notified, this can be achieved by adding entries for each processing set.

[0152] In addition, for example, it could also be, such as Figure 6 As shown, for each AI / ML model, a set of processes associated with AI / ML processes is defined. Each process set does not have entries; instead, it contains the executable AI / ML processes within each AI / ML model. Furthermore, for example, the combination of process sets can be defined by models mapped to individual entries. For example, it could indicate which AI / ML process (e.g., CSI reporting or performance monitoring) differs based on the position of the entry in the combination of process sets. For example, it could be as follows... Figure 7 As shown, the first entry for the combination of processing sets is defined as a CSI report, and the second entry is defined as performance monitoring. In this case, when notifying the combination (processing set A, processing set B), i.e., (A+B), it means that both the CSI report based on inference processing using model A and the performance monitoring based on CSI reports using model B can be executed simultaneously. Furthermore, if more than three combinations need to be notified, this can be achieved by adding entries to the combination of processing sets.

[0153] Furthermore, the AI / ML processing notified through AI / ML processing sets and combinations thereof is not limited to inference-based CSI reports, existing CSI reports without AI / ML application, and performance monitoring; it can also include processing required for other AI / ML actions (e.g., data acquisition). Moreover, AI / ML processing sets and combinations thereof can include multiple capabilities related to performance monitoring (e.g., multiple entries). For example, capabilities related to performance monitoring using the CSI reconstruction unit on the terminal side and capabilities related to performance monitoring without using the CSI reconstruction unit on the terminal side can be represented separately.

[0154] Furthermore, the AI / ML processing capabilities of the terminal 200 are not limited to simply notifying whether a certain capability is supported; for example, it can also notify the priority information of the capability. Moreover, when AI / ML processing capabilities are assigned priorities, each AI / ML process can be executed according to the set priorities.

[0155] [Example of operations of base station 100 and terminal 200]

[0156] Figure 8 This is a flowchart illustrating the actions of base station 100 (base station / network) and terminal 200.

[0157] exist Figure 8 In the process, the terminal 200 calculates information related to the amount of computing processing / power consumption in the terminal 200 (S101).

[0158] Terminal 200, for example, determines whether the conditions for reporting the calculated information related to computational processing volume / power consumption to the base station / network are met (reporting conditions) (S102). If the reporting conditions are not met (S102: No), terminal 200 terminates the process. Figure 8 The processing shown.

[0159] If the reporting conditions are met (S102: Yes), the terminal 200 reports information related to the amount of computing processing / power consumption to the base station / network, for example, according to the method described in the above embodiments or variations (S103).

[0160] The base station / network determines the use of the AI / ML model on the terminal side based on information related to computing power / power consumption reported from the terminal 200 (S104). Then, the base station / network notifies the terminal 200 of the information related to the determination of the use of the AI / ML model on the terminal side (S105).

[0161] [About the architecture]

[0162] Figure 9This represents an example of an architecture that includes AI / ML processing capabilities. Alternatively, it could be a part of a network architecture. Figure 9 The structure is shown. Additionally, AI / ML processing capabilities can be included in the Access and Mobility Management Function (AMF) or the RAN (gNB).

[0163] [Base station structure]

[0164] Figure 10 This is a block diagram representing a structural example of base station 100. Figure 10 In this base station 100, there are a control unit 101, a signal generation unit 102, a transmission unit 103, a receiving unit 104, an extraction unit 105, a demodulation unit 106, and a decoding unit 107.

[0165] Alternatively, it could be, Figure 10 At least one of the control unit 101, signal generation unit 102, extraction unit 105, demodulation unit 106, and decoding unit 107 shown is included in Figure 2 The control unit shown. Alternatively, it could be, Figure 10 The receiving unit 104 shown is included in Figure 2 The receiving unit shown.

[0166] Control unit 101 determines, for example, control information related to AI / ML-based actions of terminal 200 and outputs the determined information to signal generation unit 102. The control information related to AI / ML-based actions of terminal 200 may include, for example, at least one of the following: information related to reports of computational processing volume / power consumption, information related to the use of the AI / ML model on the terminal side, and information related to performance monitoring of the AI / ML model. Furthermore, control unit 101 may output control information related to the AI / ML model of terminal 200 to signal generation unit 102 when it receives control information related to the AI / ML model from the AI / ML processing function (e.g., when it is input from decoding unit 107).

[0167] In addition, the control unit 101 can, for example, output information related to the computing power / power consumption of the terminal 200 or a performance monitoring report from the decoding unit 107 to the AI / ML processing function.

[0168] Additionally, the control unit 101 may, for example, determine the information used by the terminal 200 to receive downlink signals and output the determined information to the signal generation unit 102. The information used by the terminal 200 to receive downlink signals may include, for example, information related to resource allocation of the downlink data channel (e.g., PDSCH: Physical Downlink Shared Channel) or downlink control channel (e.g., PDCCH: Physical Downlink Control Channel), and information related to the coding / modulation scheme (e.g., MCS: Modulation and Coding Scheme).

[0169] Additionally, the control unit 101 determines, for example, the information used by the terminal 200 to transmit uplink signals (e.g., CSI reports), and outputs the determined information to the signal generation unit 102, the extraction unit 105, the demodulation unit 106, and the decoding unit 107. The information used by the terminal 200 to transmit uplink signals may include, for example, information related to resource allocation of the uplink data channel (e.g., PUSCH: Physical Uplink Shared Channel) or the uplink control channel (e.g., PUCCH: Physical Uplink Control Channel), information related to the coding / modulation method (e.g., MCS), and information related to CSI reports.

[0170] The signal generation unit 102 generates a data signal or control signal bit sequence using information input from the control unit 101, and encodes it as needed. Furthermore, the signal generation unit 102 modulates the encoded bit sequence to generate a modulated signal (e.g., a symbol sequence) and maps it to a radio resource indicated by the control unit 101. The signal generation unit 102 outputs the mapped signal to the transmission unit 103.

[0171] The transmitting unit 103 performs OFDM-like transmission waveform generation processing on the signal input from the signal generation unit 102. Additionally, for example, in the case of OFDM transmission using a cyclic prefix (CP), the transmitting unit 103 performs an Inverse Fast Fourier Transform (IFFT) on the signal and appends a CP to the IFFT-derived signal. Furthermore, the transmitting unit 103 performs RF (Radio Frequency) processing on the signal, such as D / A (Digital / Analog) conversion or up-conversion, and transmits the wireless signal to the terminal 200 via an antenna.

[0172] The receiving unit 104 performs RF processing, such as down-conversion or A / D (Analog / Digital) conversion, on the uplink signal received from the terminal 200 via the antenna. Alternatively, in the case of OFDM transmission, the receiving unit 104 performs Fast Fourier Transform (FFT) processing on the received signal and outputs the obtained frequency domain signal to the extraction unit 105.

[0173] The extraction unit 105 extracts the radio resource portion of the uplink signal (e.g., PUSCH or PUCCH) from the received signal input from the receiving unit 104 based on information input from the control unit 101, and outputs the extracted radio resource portion to the demodulation unit 106.

[0174] The demodulation unit 106 demodulates the uplink signal (e.g., PUSCH or PUCCH) input from the extraction unit 105 based on information input from the control unit 101. The demodulation unit 106 outputs the demodulation result to the decoding unit 107, for example.

[0175] The decoding unit 107 performs error correction decoding on the uplink signal (e.g., PUSCH or PUCCH) based on information input from the control unit 101 and demodulation results input from the demodulation unit 106, to obtain the decoded received bit sequence. Furthermore, if the decoded received bit sequence contains information related to computational processing volume / power consumption from the terminal 200, the decoding unit 107 outputs this information to the control unit 101.

[0176] [Terminal Structure]

[0177] Figure 11 This is a block diagram illustrating a structural example of a terminal 200 according to an embodiment of the present disclosure. For example, in Figure 11In the terminal 200, there are receiving units 201, extraction units 202, demodulation units 203, decoding units 204, control units 205, signal generation units 206, and transmission units 207.

[0178] Alternatively, it could be, Figure 11 At least one of the extraction unit 202, demodulation unit 203, decoding unit 204, control unit 205, and signal generation unit 206 shown is included in Figure 3 The control unit shown. Alternatively, it could be, Figure 11 The transmitting unit 207 shown is included in Figure 3 The transmitting unit is shown.

[0179] The receiving unit 201 receives downlink signals (e.g., downlink data signals or downlink control signals) from the base station 100 via an antenna, and performs RF processing such as down-conversion or A / D conversion on the received wireless signal to obtain a received signal (baseband signal). Additionally, when receiving OFDM signals, the receiving unit 201 performs FFT processing on the received signal to convert it to the frequency domain. The receiving unit 201 outputs the received signal to the extraction unit 202.

[0180] For example, based on radio resource information related to downlink control signals input from control unit 205, extraction unit 202 extracts radio resource portions that may contain downlink control signals from the received signal input from receiving unit 201, and outputs them to demodulation unit 203. Additionally, based on radio resource information related to data signals input from control unit 205, extraction unit 202 extracts radio resource portions containing downlink data signals, and outputs them to demodulation unit 203.

[0181] The demodulation unit 203 demodulates the signal (e.g., PDCCH or PDSCH) input from the extraction unit 202 based on information input from the control unit 205, and outputs the demodulation result to the decoding unit 204.

[0182] The decoding unit 204 uses, for example, information input from the control unit 205 and demodulation results input from the demodulation unit 203 to perform error correction decoding on the PDCCH or PDSCH to obtain, for example, control signals or downlink data signals. The decoding unit 204 outputs control signals to the control unit 205.

[0183] Control unit 205, for example, determines information related to downlink transmission based on information obtained from control signals input from decoding unit 204, and outputs it to extraction unit 202, demodulation unit 203, and decoding unit 204. Additionally, control unit 205, for example, determines information related to uplink transmission based on information obtained from control signals input from decoding unit 204, and outputs it to signal generation unit 206. Furthermore, control unit 205 generates information related to computational processing volume / power consumption using the above method and outputs it to signal generation unit 206.

[0184] The signal generation unit 206 generates an uplink data signal or an uplink control signal based on information related to computational processing volume / power consumption or information related to uplink transmission input from the control unit 205. It then encodes and modulates the bit sequence of the generated signal and maps it to radio resources. For example, the signal generation unit 206 outputs the mapped uplink signal to the transmission unit 207.

[0185] The transmitting unit 207 generates a transmit signal waveform, such as OFDM, from the signal input from the signal generation unit 206. Additionally, in cases of OFDM transmission using CP or DFT-s-OFDM transmission, the transmitting unit 207 performs IFFT processing on the signal and adds CP to the IFFT-derived signal. Alternatively, when generating a single-carrier waveform such as DFT-s-OFDM, the transmitting unit 207 may, for example, add a DFT section (not shown) before the signal generation unit 206. Furthermore, the transmitting unit 207 performs RF processing on the transmit signal, such as D / A conversion and up-conversion, and transmits the wireless signal to the base station 100 via an antenna.

[0186] The above describes various implementations of one non-limiting embodiment of this disclosure.

[0187] In addition, different AI / ML models may be used depending on parameters such as cell, site, Transmission and Reception Point (TRP), beam, location, terminal movement speed, multipath and other wireless channels, cell congestion status, and transmitted traffic.

[0188] Furthermore, the "mobility" used in the above embodiments can include handover in RRC connected mode, reselection in RRC idle or RRC inactive mode. Additionally, "mobility" can also include movement between TRPs, beam switching, and location movement.

[0189] Additionally, in this disclosure, the signal / message / signaling used for notification can be a control plane message (e.g., UCI or MAC-CE), an RRC signal, or a notification using DCI as a physical layer signaling.

[0190] (Replenish)

[0191] Information indicating whether terminal 200 supports the functions, actions, or processes shown in the above embodiments and supplements can also be sent (or notified) by terminal 200 to base station 100 as capability information or capability parameters of terminal 200.

[0192] The capability information may also include the following information elements (IE: Information Element), each individually indicating whether the terminal 200 supports at least one of the functions, actions, and processes shown in the above-described embodiments, modifications, and supplements. Alternatively, the capability information may also include the following information elements, indicating whether the terminal 200 supports two or more combinations of the functions, actions, and processes shown in the above-described embodiments, modifications, and supplements.

[0193] Base station 100 can, for example, determine (or decide or envision) the functions, actions, or processes supported (or not supported) by the source terminal 200, based on capability information received from terminal 200. Base station 100 can implement actions, processes, or controls corresponding to the determination results based on the capability information. For example, base station 100 can control processing related to AI / ML models based on the capability information received from terminal 200.

[0194] It should be noted that terminal 200 does not support some of the functions, operations, or processes shown in the above embodiments, modifications, and supplements. Alternatively, in terminal 200, such a portion of the functions, operations, or processes may be limited. For example, information or requests related to such limitations may also be notified to base station 100.

[0195] Information related to the capabilities or limitations of terminal 200 may be defined in a standard, or may be implicitly communicated to base station 100 in association with information known to base station 100 or information sent to base station 100.

[0196] The above describes various implementations, modifications, and additions of a non-limiting embodiment of this disclosure.

[0197] (Control signal)

[0198] In this disclosure, the downlink control signal (information) associated with this disclosure can be a signal (information) transmitted in the physical layer PDCCH, or a signal (information) transmitted in the higher layer MAC (Medium Access Control) CE (Control Element) or RRC. Alternatively, a predefined signal (information) can also be used as the downlink control signal.

[0199] The uplink control signals (information) associated with this disclosure can be signals (information) transmitted in the physical layer PUCCH, or signals (information) transmitted in the higher layer MAC CE or RRC. Alternatively, the uplink control signals can be predefined signals (information). Furthermore, the uplink control signals can be replaced with UCI (Uplink Control Information), first-stage SCI (Sidelink Control Information), or second-stage SCI.

[0200] (Base station)

[0201] In this disclosure, a base station can be a TRP (Transmission Reception Point), cluster head, access point, RRH (Remote Radio Head), eNodeB (eNB), gNodeB (gNB), BS (BaseStation), BTS (Base Transceiver Station), host, gateway, etc. Additionally, in sidelink communication, the base station can be replaced with a terminal. A base station can also be a relay device for communication between a high-level relay node and a terminal. Furthermore, a base station can also be a roadside device.

[0202] (Uplink / Downlink / Sidelink)

[0203] This disclosure can be applied to any link in the uplink, downlink, and sidelink. For example, this disclosure can be applied to the uplink PUSCH, PUCCH, PRACH (Physical Random Access Channel), the downlink PDSCH, PDCCH, PBCH (Physical Broadcast Channel), and the sidelink PSSCH (Physical Sidelink Shared Channel), PSCCH (Physical Sidelink Control Channel), and PSBCH (Physical Sidelink Broadcast Channel).

[0204] It should be noted that PDCCH, PDSCH, PUSCH, and PUCCH are examples of downlink control channels, downlink data channels, uplink data channels, and uplink control channels, respectively. PSCCH and PSSCH are examples of sidelink control channels and sidelink data channels, respectively. PBCH and PSBCH are examples of broadcast channels, and PRACH is an example of a random access channel.

[0205] (Data Channel / Control Channel)

[0206] This disclosure can be applied to any channel in the data channel and the control channel. For example, the channels of this disclosure can also be replaced with PDSCH, PUSCH, PSSCH of the data channel and PDCCH, PUCCH, PBCH, PSCCH, PSBCH of the control channel.

[0207] (Reference signal)

[0208] In this disclosure, the reference signal is a signal known to both the base station and the terminal, and is sometimes referred to as "RS (Reference Signal)" or "pilot signal". The reference signal can also be one of DMRS (Demodulation Reference Signal), CSI-RS (Channel State Information-Reference Signal), TRS (Tracking Reference Signal), PTRS (Phase Tracking Reference Signal), CRS (Cell-specific Reference Signal), or SRS (Sounding Reference Signal).

[0209] (Time interval)

[0210] In this disclosure, the unit of time resource is not limited to one or a combination of time slots and symbols. For example, it can be a frame, superframe, subframe, time slot, sub-time slot, micro-time slot, or symbol, OFDM (Orthogonal Frequency Division Multiplexing) symbol, SC-FDMA (Single Carrier - Frequency Division Multiple Access) symbol, or other time resource units. Furthermore, the number of symbols contained in one time slot is not limited to the number of symbols exemplified in the above embodiments, and can also be other numbers of symbols.

[0211] (frequency band)

[0212] This disclosure can be applied to any band domain, whether it is an authorized band domain or an unauthorized band domain.

[0213] (communication)

[0214] This disclosure can be applied to any communication in base station-terminal communication (Uu link communication), terminal-to-terminal communication (sidelink communication), and V2X (Vehicle to Everything) communication. For example, the channels in this disclosure can be replaced with PSCCH, PSSCH, PSFCH (Physical Sidelink Feedback Channel), PSBCH, PDCCH, PUCCH, PDSCH, PUSCH, and PBCH.

[0215] Furthermore, this disclosure can be applied to any network, including terrestrial networks and non-terrestrial networks (NTNs) that use satellites or High Altitude Pseudo Satellites (HAPS). Additionally, this disclosure can also be applied to terrestrial networks with transmission delays greater than the symbol length or time slot length, such as networks with large cell sizes and ultra-wideband transmission networks.

[0216] (Antenna Port)

[0217] An antenna port refers to a logical antenna (antenna array) consisting of one or more physical antennas. That is, an antenna port does not necessarily refer to a single physical antenna; sometimes it refers to an array antenna composed of multiple antennas. For example, instead of specifying the number of physical antennas constituting an antenna port, it is defined as the smallest unit that the terminal can transmit a reference signal. Additionally, an antenna port is sometimes defined as the smallest unit multiplied by a precoding vector.

[0218] <5G NR System Architecture and Protocol Stack>

[0219] To realize the next version of fifth-generation mobile phone technology (also known simply as "5G"), which includes the development of a new radio access technology (NR) operating in the frequency range up to 100 GHz, 3GPP is continuing its work. The first version of the 5G standard was completed at the end of 2017, thus enabling the transition to the trial production of terminals (e.g., smartphones) according to the 5G NR standard and commercial deployment.

[0220] For example, the overall system architecture envisions a gNB-RAN (Next Generation Radio Access Network). The gNB provides UE-side termination for the NG radio access user plane (SDAP (Service Data Adaptation Protocol) / PDCP (Packet Data Convergence Protocol) / RLC (Radio Link Control) / MAC / PHY (Physical Layer)) and control plane (RRC) protocols. gNBs are interconnected via the Xn interface. Additionally, gNBs are connected to the NGC (Next Generation Core) via the Next Generation (NG) interface, and more specifically, to the AMF (Access and Mobility Management Function) (e.g., a specific core entity implementing the AMF) via the NG-C interface, and to the UPF (User Plane Function) (e.g., a specific core entity implementing the UPF) via the NG-U interface. Figure 12 This refers to the NG-RAN architecture (e.g., refer to 3GPP TS 38.300 v15.6.0, section 4).

[0221] The user plane protocol stack for NR (e.g., see 3GPP TS 38.300, section 4.4.1) comprises the PDCP (Packet Data Convergence Protocol, see TS 38.300, section 6.4) sublayer, RLC (Radio Link Control, see TS 38.300, section 6.3) sublayer, and MAC (Media Access Control, see TS 38.300, section 6.2) sublayer, which terminates on the network side in the gNB. Additionally, a new Access Stratum (AS) sublayer (SDAP: Service Data Adaptation Protocol) has been incorporated into PDCP (e.g., see 3GPP TS 38.300, section 6.5). Furthermore, a control plane protocol stack is defined for NR (e.g., see TS 38.300, section 4.4.2). A summary of Layer 2 functionality is described in Section 6 of TS 38.300. The functions of the PDCP sublayer, RLC sublayer, and MAC sublayer are listed in Sections 6.4, 6.3, and 6.2 of TS 38.300, respectively. The functions of the RRC layer are listed in Section 7 of TS 38.300.

[0222] For example, the media access control layer handles the multiplexing of logical channels, the scheduling of processing involving various parameter sets, and the various functions associated with scheduling.

[0223] For example, the Physical Layer (PHY) is responsible for encoding, PHY HARQ (Physical Layer Hybrid Automatic Repeat Request) processing, modulation, multi-antenna processing, and mapping signals to appropriate physical time-frequency resources. Additionally, the Physical Layer handles the mapping of physical channels to transport channels. The Physical Layer provides services to the MAC Layer in the form of transport channels. A physical channel corresponds to a set of time-frequency resources used to transmit a specific transport channel; each transport channel is mapped to a corresponding physical channel. For example, in physical channels, uplink physical channels include PRACH (Physical Random Access Channel), PUSCH (Physical Uplink Shared Channel), and PUCCH (Physical Uplink Control Channel), while downlink physical channels include PDSCH (Physical Downlink Shared Channel), PDCCH (Physical Downlink Control Channel), and PBCH (Physical Broadcast Channel).

[0224] In NR use cases / extended scenarios, enhanced mobile broadband (eMBB), ultra-reliable low-latency communications (URLLC), and massive machine-type communications (mMTC) can be included, each with multiple requirements in terms of data rate, latency, and coverage. For example, eMBB is expected to support peak data rates approximately three times that of IMT-Advanced (20Gbps in downlink and 10Gbps in uplink) and effective (user-experienced) data rates. On the other hand, in the case of URLLC, ultra-low latency (0.5ms in both UL and DL for the user plane) and high reliability (within 1ms, 1-10) are expected. -5 This introduces more stringent requirements. Finally, in mMTC, a high connection density is preferably required (1,000,000 devices / km in urban environments). 2 ), wide coverage in harsh environments and extremely long battery life (15 years) for inexpensive devices.

[0225] Therefore, a set of OFDM parameters suitable for one use case (e.g., subcarrier spacing (SCS), OFDM symbol length, cyclic prefix (CP) length, number of symbols per scheduling interval) may be ineffective for other use cases. For example, in low-latency services, it is preferable to require a shorter symbol length than in mMTC services (therefore, a larger subcarrier spacing) and / or fewer symbols per scheduling interval (also known as "TTI"). Moreover, in extended scenarios with large channel delay spread, it is preferable to require a longer CP length than in scenarios with shorter delay spread. The subcarrier spacing can also be optimized depending on the situation to maintain the same CP overhead. NR supports more than one subcarrier spacing value. Correspondingly, subcarrier spacings of 15kHz, 30kHz, 60kHz... are currently considered. The symbol length Tu and the subcarrier spacing Δf are directly related according to the formula Δf = 1 / Tu. Similarly, the LTE system can use the term "resource element" to represent the smallest unit of resources consisting of a subcarrier of the length of one OFDM / SC-FDMA (Single-Carrier Frequency Division Multiple Access) symbol.

[0226] In the new 5G-NR wireless system, resource grids for subcarriers and OFDM symbols are defined in both the uplink and downlink for each parameter set and each carrier. Each element of the resource grid is called a "resource element," which is determined based on the frequency index in the frequency domain and the symbol position in the time domain (refer to 3GPP TS 38.211 v15.6.0).

[0227] <Functional Separation between NG-RAN and 5GC in 5G NR>

[0228] Figure 13 This indicates the functional separation between NG-RAN and 5GC. The logical node of NG-RAN is either gNB or ng-eNB. 5GC has logical nodes AMF, UPF, and SMF (Session Management Function).

[0229] For example, gNB and ng-eNB host the following main functions:

[0230] - Functions such as Radio Bearer Control, Radio Admission Control, Connection Mobility Control, and Radio Resource Management (RRM) that dynamically allocates (schedules) resources to the UE in both the uplink and downlink links;

[0231] - Data IP (Internet Protocol) header compression, encryption, and integrity protection;

[0232] - Selection of AMF when attaching a UE in situations where the route to the AMF cannot be determined based on the information provided by the UE;

[0233] - Routing to user plane data towards UPF;

[0234] - Routing of control plane information toward AMF;

[0235] - Setting and canceling connections;

[0236] - Scheduling and sending paging messages;

[0237] - The scheduling and transmission of system broadcast information (originating from AMF or Operation, Admission, and Maintenance functions (OAM));

[0238] - Setting up measurements and measurement reports for mobility and scheduling;

[0239] - Packet markings for transmission class in the uplink;

[0240] -Session management;

[0241] -Support for network slicing;

[0242] - QoS (Quality of Service) flow management and mapping to data radio bearers;

[0243] Support for UEs in RRC_INACTIVE (RRC inactive) state;

[0244] - NAS (Non-Access Stratum) message distribution function;

[0245] - Sharing of wireless access networks;

[0246] - Dual connectivity;

[0247] - Close collaboration between NR and E-UTRA (Evolved Universal Terrestrial Radio Access).

[0248] The Access and Mobility Management Function (AMF) administers the following main functions:

[0249] - Function to terminate Non-Access Stratum (NAS) signaling;

[0250] -Security of NAS signaling;

[0251] - Security controls at the access layer (AS);

[0252] - Core Network (CN) inter-node signaling for mobility between 3GPP access networks;

[0253] - The possibility of a UE reaching idle mode (including control and execution of paging retransmission);

[0254] - Management of the registered area;

[0255] - Support for intra-system mobility and inter-system mobility;

[0256] -Access authentication;

[0257] - Access licenses that include roaming permission checks;

[0258] - Mobility management controls (subscription and policies);

[0259] -Support for network slicing;

[0260] - Selection of Session Management Function (SMF).

[0261] In addition, the User Face Function (UPF) hosts the following main functions:

[0262] - Anchor points for intra-RAT (Radio Access Technology) mobility / inter-RAT (where applicable) mobility;

[0263] - External PDU (Protocol Data Unit) session points used for interconnection with data networks;

[0264] - Packet routing and forwarding;

[0265] - Enforcement of policy rules in group checks and user-facing aspects;

[0266] - Reports on business usage;

[0267] - Uplink classifier used to support routing of service flows toward the data network;

[0268] - Branching points used to support multi-homed PDU sessions;

[0269] - For user plane QoS processing (e.g., packet filtering, gating, UL / DL rate enforcement);

[0270] - Uplink service verification (SDF (Service Data Flow) mapping to QoS flow);

[0271] - Downlink packet buffering and downlink data notification triggering functions.

[0272] Finally, the Session Management Function (SMF) administers the following main functions:

[0273] -Session management;

[0274] - The allocation and management of UE IP addresses;

[0275] -Selection and control of UPF;

[0276] - A function for setting traffic steering in the User Plane Function (UPF) to direct traffic to the appropriate destination;

[0277] - Enforcing policies and QoS in the control section;

[0278] - Notification of downlink data.

[0279] <The process of setting up and resetting RRC connection>

[0280] Figure 14 This refers to several interactions between the UE, gNB, and AMF (5GC entity) when the UE in the NAS part transitions from RRC_IDLE (RRC idle) to RRC_CONNECTED (RRC connected) (refer to TS 38.300 v15.6.0).

[0281] RRC is a higher-level signaling (protocol) used for UE and gNB configuration. Through this transition, the AMF prepares UE context data (which may include, for example, PDU session context, security keys, UE radio capabilities, UE security capabilities, etc.) and sends it to the gNB along with an initial context setting request. Next, the gNB and UE activate AS security together. This is done by the gNB sending a Security Mode Command message to the UE, which responds with a Security Mode Complete message. Then, the gNB sends an RRC Reconfiguration message to the UE, and receives an RRC Reconfiguration Complete message from the UE for this message, thereby enabling the reconfiguration of Signaling Radio Bearer 2 (SRB2) and Data Radio Bearer (DRB). For signaling-only connections, since SRB2 and DRB are not configured, the steps related to RRC reconfiguration can be omitted. Finally, the gNB notifies the AMF that the configuration process is complete using the Initial Context Setup Reply.

[0282] Therefore, this disclosure provides an entity for a fifth-generation core network (5GC) (e.g., AMF, SMF, etc.), comprising: a control circuit that, upon operation, establishes a Next Generation (NG) connection with a gNodeB; and a transmission unit that, upon operation, sends an initial context setting message to the gNodeB via the NG connection to configure the signaling radio bearer between the gNodeB and the User Equipment (UE). Specifically, the gNodeB transmits Radio Resource Control (RRC) signaling containing an Information Element (IE) to the UE via the signaling radio bearer. The UE then performs uplink transmission or downlink reception based on the resource allocation settings.

[0283] <Application Scenarios of IMT after 2020>

[0284] Figure 15This section outlines several use cases for 5G NR. Within the 3rd Generation Partnership Project New Radio (3GPP NR), three use cases supporting a wide variety of services and applications, conceived through IMT-2020, have been studied. Planning for the first phase of specifications for enhanced mobile broadband (eMBB) has been completed. Current and future work, in addition to gradually expanding eMBB support, includes standardization for ultra-reliable and low-latency communications (URLLC) and massive machine-type communications (mMTC). Figure 15 Several examples illustrating conceptual application scenarios for IMT after 2020 (e.g., referring to ITU-R M.2083). Figure 2 ).

[0285] URLLC use cases have strict requirements related to performance aspects such as throughput, latency, and availability. URLLC is conceived as a key technology for enabling wireless control of future industrial production or manufacturing processes, remote medical surgery, automation of power transmission and distribution in smart grids, and traffic safety applications. Ultra-high reliability of URLLC is supported by defining technologies that meet the requirements set by TR38.913. In NR URLLC version 15, a crucial requirement is a target user plane latency of 0.5ms in the UL (uplink) and 0.5ms in the DL (downlink). For a single packet transmission, the overall requirement for URLLC is a block error rate (BLER) of 1E-5 for a 32-byte packet size with a user plane latency of 1ms.

[0286] Considering the physical layer, numerous methods are available to improve reliability. Current possibilities for reliability enhancement include defining additional CQI (Channel Quality Indicator) tables for URLLC, a more compact DCI format, and PDCCH iteration. However, as NR (a crucial prerequisite for NR URLLC) becomes more stable and is further developed, this scope can be expanded to achieve ultra-high reliability. Specific use cases for NR URLLC in version 15 include augmented reality / virtual reality (AR / VR), e-health, e-safety, and other critical applications.

[0287] Furthermore, the technical enhancements for NR URLLC aim to improve latency and reliability. Latency enhancements include configurable parameter sets, non-slot-based scheduling utilizing flexible mapping, unlicensed (already licensed) uplinks, slot-level repetition in the data channel, and pre-emption in the downlink. Pre-emption refers to stopping a transmission with allocated resources and using those resources for a later-requested transmission that requires lower latency / higher priority. Therefore, a permitted transmission is replaced by a subsequent transmission. Pre-emption can be applied regardless of the specific service type. For example, a transmission in service type A (URLLC) can be replaced by a transmission in service type B (eMBB, etc.). Reliability enhancements include a dedicated CQI / MCS table for a target BLER of 1E-5.

[0288] The use cases for mMTC (massive machine-type communications) are characterized by a large number of connected devices that transmit relatively small amounts of data that are not easily affected by latency. These devices require low cost and very long battery life. From NR's perspective, utilizing very narrow bandwidth segments is a solution to save UE power and extend its battery life.

[0289] As mentioned above, the potential for improved reliability in NR is further expanded. It is one of the essential conditions for all situations; for example, high or ultra-high reliability is a crucial requirement related to URLLC and mMTC. From both wireless and network perspectives, reliability can be improved through several mechanisms. Generally, there are two to three important areas that could potentially contribute to improved reliability. These areas include compact control channel information, data / control channel iteration, and diversity related to the frequency, time, and / or spatial domains. These areas can be universally used to improve reliability, independent of specific communication scenarios.

[0290] Regarding NR URLLC, further use cases with more stringent requirements are envisioned, such as factory automation, transportation, and power transmission. Strict requirements refer to high reliability (achieving 10...). -6 High reliability, high availability, packet size up to 256 bytes, and time synchronization up to several microseconds (μs) (capable of setting the value to 1 microsecond or several microseconds depending on the use case, frequency range, and short latency of about 0.5ms to 1ms (e.g., 0.5ms latency in the target user plane)).

[0291] Furthermore, from a physical layer perspective, there are several technical enhancements to NR URLLC. These enhancements include strengthening the PDCCH (Physical Downlink Control Channel) associated with compact DCI, PDCCH repetition, and increased PDCCH monitoring. Additionally, enhancements to UCI (Uplink Control Information) are related to enhanced HARQ (Hybrid Automatic Repeat Request) and CSI feedback. Furthermore, there may be enhancements to PUSCH and retransmission / repetition related to mini-slot-level frequency hopping. The term "mini-slot" refers to a transmission time interval (TTI) containing fewer symbols than a time slot (a time slot has 14 symbols).

[0292] <QoS Control>

[0293] 5G's QoS (Quality of Service) model is based on QoS flows, supporting both QoS flows that require guaranteed bit rate (GBR) and QoS flows that do not require guaranteed bit rate (non-GBR QoS flows). Therefore, at the NAS level, QoS flows represent the finest granular QoS classification within a PDU session. QoS flows are determined within a PDU session based on the QoS Flow ID (QFI) transmitted via the encapsulation header through the NG-U interface.

[0294] For each UE, the 5GC establishes one or more PDU sessions. For each UE, in conjunction with the PDU session, the NG-RAN establishes at least one Data Radio Bearer (DRB), for example, as described above with reference to FIG21. Additionally, DRBs can be subsequently configured in QoS flows added to that PDU session (when to configure depends on the NG-RAN). The NG-RAN maps packets belonging to various PDU sessions to various DRBs. NAS-level packet filters in the UE and 5GC are used to associate UL packets and DL packets with QoS flows, while AS-level mapping rules in the UE and NG-RAN associate UL QoS flows and DL QoS flows with DRBs.

[0295] Figure 16 This refers to the non-roaming reference architecture of 5G NR (refer to TS 23.501 v16.1.0, section 4.23). Application Function (AF) (e.g., hosting). Figure 15The external application server for the illustrated 5G service interacts with the 3GPP core network to provide services. For example, it may access a Network Exposure Function (NEF) to support applications that impact service routing, or it may interact with a policy framework (see Policy Control Function (PCF)) for policy control (e.g., QoS control). Based on operator deployment, operators deem trusted application functions capable of directly interacting with associated network functions. Application functions not permitted by the operator to directly access network functions interact with associated network functions via the NEF, using an open framework accessible to the outside world.

[0296] Figure 16 It also indicates further functional units of the 5G architecture, namely, the Network Slice Selection Function (NSSF), the Network Repository Function (NRF), Unified Data Management (UDM), the Authentication Server Function (AUSF), the Access and Mobility Management Function (AMF), the Session Management Function (SMF), and the Data Network (DN: Data Network, such as services provided by operators, internet access, or services provided by third parties). All or part of the core network's functions and application services can also be deployed and operate in a cloud computing environment.

[0297] Therefore, this disclosure provides an application server (e.g., an AF in a 5G architecture) comprising: a transmitting unit that, in order to establish a PDU session containing a radio bearer between a gNodeB and a UE corresponding to QoS requirements, sends, during operation, a request containing at least one of the QoS requirements for URLLC service, eMMB service, and mMTC service to at least one of the functions of 5GC (e.g., NEF, AMF, SMF, PCF, UPF, etc.); and a control circuit that, during operation, performs services using the established PDU session.

[0298] The term “...department” as used in this disclosure may be interchanged with other terms such as “...circuitry”, “...device”, “...unit” or “...module”.

[0299] This disclosure can be implemented in software, hardware, or software in cooperation with hardware. The functional blocks used in the above embodiments are implemented partially or wholly as LSIs (Large Scale Integration), and the processes described in the above embodiments can also be controlled partially or wholly by a single LSI or a combination of LSIs. An LSI can be composed of individual chips, or it can be composed of a single chip containing some or all of the functional blocks. An LSI may also include data input and output. Depending on the degree of integration, an LSI may also be referred to as an "IC (Integrated Circuit)," "System LSI," "Super LSI," or "Ultra LSI."

[0300] The method of integrating the LSI is not limited to LSI; it can also be implemented using dedicated circuits, general-purpose processors, or special-purpose processors. Alternatively, it can utilize a programmable FPGA (Field Programmable Gate Array) manufactured using the LSI, or a reconfigurable processor that allows reconfiguration of the connections or settings of the circuit blocks within the LSI. This disclosure can also be implemented for digital or analog processing.

[0301] Furthermore, if advancements in semiconductor technology or the emergence of other derivative technologies lead to integrated circuit technologies that can replace LSIs, these technologies could also be used to integrate functional blocks. There are also possibilities for applications such as biotechnology.

[0302] This disclosure can be implemented in all kinds of devices, apparatuses, and systems with communication capabilities (collectively referred to as "communication devices"). A communication device may also include a wireless transceiver and processing / control circuitry. The wireless transceiver may also include a receiving unit and a transmitting unit, or perform the functions of these units. The wireless transceiver (transmitting unit, receiving unit) may also include an RF (Radio Frequency) module and one or more antennas. The RF module may also include an amplifier, an RF modulator / demodulator, or similar devices. Non-limiting examples of communication devices include: telephones (mobile phones, smartphones, etc.), tablet computers, personal computers (PCs) (laptops, desktops, laptops, etc.), cameras (digital cameras, digital camcorders, etc.), digital players (digital audio / video players, etc.), wearable devices (wearable cameras, smartwatches, tracking devices, etc.), game consoles, e-book readers, remote health / telemedicine (remote healthcare / medical prescription) devices, vehicles or transportation vehicles with communication capabilities (cars, airplanes, ships, etc.), and combinations of the various devices described above.

[0303] Communication devices are not limited to portable or movable devices, but also include all kinds of devices, equipment, and systems that cannot be carried or fixed. Examples include: smart home devices (home appliances, lighting equipment, smart meters or meters, control panels, etc.), vending machines, and all other "things" that can exist on the IoT (Internet of Things) network.

[0304] In addition to data communication via cellular systems, wireless LAN (Local Area Network) systems, and communication satellite systems, communication also includes data communication via a combination of these systems.

[0305] In addition, the communication device also includes devices such as controllers or sensors that are connected or linked to a communication device performing the communication functions described in this disclosure. For example, it includes a controller or sensor that generates control signals or data signals used by the communication device to perform the communication functions of the communication device.

[0306] In addition, the communication device includes infrastructure equipment that communicates with or controls the various devices described above (not limited to these), such as base stations, access points, and all other devices, equipment, and systems.

[0307] One embodiment of the present disclosure includes a terminal comprising: a control circuit that determines control information relating to at least one of computational processing volume and power consumption resulting from the use of an artificial intelligence model on the terminal side; and a transmission circuit that transmits the control information.

[0308] In one embodiment of this disclosure, the control information includes information related to a comparison between at least one of the computational processing power and power consumption when using the artificial intelligence model and at least one of the computational processing power and power consumption when not using the artificial intelligence model.

[0309] In one embodiment of this disclosure, the control information includes information related to a difference value that is the difference between at least one of the computational processing volume and power consumption when using the artificial intelligence model and at least one of the computational processing volume and power consumption when not using the artificial intelligence model.

[0310] In one embodiment of this disclosure, the control information includes information related to a comparison or difference value that is at least one of the computational processing power and power consumption when using a first artificial intelligence model and at least one of the computational processing power and power consumption when using a second artificial intelligence model.

[0311] In one embodiment of this disclosure, the control information includes information related to a comparison or difference value that is at least one of the computational processing volume and power consumption when using a first parameter of the artificial intelligence model and at least one of the computational processing volume and power consumption when using a second parameter of the artificial intelligence model.

[0312] In one embodiment of this disclosure, the control information includes information related to the utilization or occupancy of at least one of the computing power and power consumption in the terminal.

[0313] In one embodiment of this disclosure, there is knowledge related to the method of sharing and determining the control information between the network and the terminal.

[0314] In one embodiment of this disclosure, information related to the method is predefined by a standard, or is reported from the terminal to the network when reporting the terminal's capabilities or capabilities related to the application of the artificial intelligence model, or is set from the network to the terminal, or is reported from the terminal to the network when reporting the control information.

[0315] In one embodiment of this disclosure, when the terminal performs performance monitoring processing of the artificial intelligence model, the control information and information related to the performance monitoring processing are reported from the terminal to the network.

[0316] In one embodiment of this disclosure, the control information and the capabilities of the terminal associated with the artificial intelligence model are reported from the terminal to the network.

[0317] In one embodiment of this disclosure, the control information is reported from the terminal to the network when a notification is triggered from the network, a timer expires, or information related to at least one of the computing power and power consumption exceeds a threshold.

[0318] A base station according to an embodiment of this disclosure includes: a receiving circuit that receives control information relating to at least one of computational processing volume and power consumption resulting from the use of an artificial intelligence model on the terminal side; and a control circuit that controls the use of the artificial intelligence model on the terminal side based on the control information.

[0319] In a communication method according to an embodiment of this disclosure, the terminal performs the following steps: determining control information related to at least one of the computational processing volume and power consumption resulting from the use of the artificial intelligence model on the terminal side; and sending the control information.

[0320] In a communication method according to an embodiment of this disclosure, the base station performs the following steps: receiving control information related to at least one of computational processing volume and power consumption resulting from the use of an artificial intelligence model on the terminal side; and controlling the use of the artificial intelligence model on the terminal side based on the control information.

[0321] The entire contents of the specification, drawings and abstract of the specification contained in Japanese Patent Application No. 2023-129364, filed on August 8, 2023, are incorporated herein by reference.

[0322] Industrial applicability

[0323] One embodiment of this disclosure is useful for wireless communication systems.

[0324] Explanation of reference numerals in the attached figures

[0325] 100 base stations

[0326] 101, 205 Control Department

[0327] Signal generation units 102 and 206

[0328] 103, 207 Sending Department

[0329] Receiving Departments 104 and 201

[0330] Extraction sections 105 and 202

[0331] Demodulation Departments 106 and 203

[0332] Decoding sections 107 and 204

[0333] 200 terminals

[0334] 251 CSI Production Department

[0335] 252 CSI Reconstruction Department

[0336] 253 Performance Monitoring Department

[0337] 254 CQI Generation Department

Claims

1. A terminal, characterized by comprising: Possessing: a control circuit that decides control information related to at least one of a calculation processing amount and power consumption brought about by use of an artificial intelligence model on the terminal side; and a transmission circuit that transmits the control information.

2. The terminal according to claim 1, wherein the control information contains information related to a comparison of at least one of the calculation processing amount and power consumption when using the artificial intelligence model and at least one of the calculation processing amount and power consumption when not using the artificial intelligence model.

3. The terminal according to claim 1, wherein the control information contains information related to a difference value of at least one of the calculation processing amount and power consumption when using the artificial intelligence model and at least one of the calculation processing amount and power consumption when not using the artificial intelligence model.

4. The terminal according to claim 1, wherein the control information contains information related to a comparison or difference value of at least one of the calculation processing amount and power consumption when using a first artificial intelligence model and at least one of the calculation processing amount and power consumption when using a second artificial intelligence model.

5. The terminal according to claim 1, wherein the control information contains information related to a comparison or difference value of at least one of the calculation processing amount and power consumption when using a first parameter of the artificial intelligence model and at least one of the calculation processing amount and power consumption when using a second parameter of the artificial intelligence model.

6. The terminal according to claim 1, wherein the control information contains information related to a usage rate or occupancy rate of at least one of the calculation processing amount and power consumption in the terminal.

7. The terminal according to claim 1, wherein cognition related to a method of deciding the control information is shared between a network and the terminal.

8. The terminal according to claim 7, wherein information related to the method is pre-specified by a standard, or is reported from the terminal to the network when reporting a capability of the terminal or capability information related to application of the artificial intelligence model, or is set from the network to the terminal, or is reported from the terminal to the network when reporting the control information.

9. The terminal according to claim 1, wherein in a case where the terminal performs a performance monitoring process of the artificial intelligence model, the control information and information related to the performance monitoring process are reported from the terminal to a network.

10. The terminal according to claim 1, wherein the control information and a capability of the terminal related to the artificial intelligence model are reported from the terminal to a network.

11. The terminal according to claim 1, wherein in a case where a trigger is notified from the network, or a timer expires, or information related to at least one of the calculation processing amount and power consumption exceeds a threshold value, the control information is reported from the terminal to a network.

12. A base station, characterized by Possessing: receiving circuitry that receives control information related to at least one of a computation processing amount and power consumption caused by use of the artificial intelligence model on the terminal side; and control circuitry that controls the use of the artificial intelligence model on the terminal side based on the control information.

13. A communication method characterized by comprising: a terminal performing steps of: determining control information related to at least one of a computation processing amount and power consumption caused by use of an artificial intelligence model on the terminal side; and transmitting the control information.

14. A communication method characterized by comprising: a base station performing steps of: receiving control information related to at least one of a computation processing amount and power consumption caused by use of an artificial intelligence model on a terminal side; and controlling the use of the artificial intelligence model on the terminal side based on the control information.

15. An integrated circuit that controls the processing of a terminal, characterized in that, the processing includes processing of: determining control information related to at least one of a computation processing amount and power consumption caused by use of an artificial intelligence model on a terminal side; and transmitting the control information.

16. An integrated circuit that controls processing of a base station, the integrated circuit comprising: the processing includes processing of: receiving control information related to at least one of a computation processing amount and power consumption caused by use of an artificial intelligence model on a terminal side; and controlling the use of the artificial intelligence model on the terminal side based on the control information.

Images