Method and system for managing artificial intelligence / machine learning (AI / ML) models

Abstract

Methods and systems for managing artificial intelligence / machine learning (AI / ML) in a communication system are provided. According to an embodiment, a method for implementing AI / ML for air interface optimization in a communication system is provided. The method may include: determining, by a first node, a coordination level for AI / ML coordination between a network and a user equipment (UE) from among a plurality of predefined coordination levels; and performing, by a second node, air interface optimization for the UE using at least one AI / ML model based on the determined coordination level. The plurality of predefined coordination levels include a first level corresponding to AI / ML coordination between the network and the UE, a second level corresponding to signaling-based AI / ML coordination between the network and the UE without model transfer, and a third level corresponding to signaling-based AI / ML coordination between the network and the UE with model transfer.

Description

[Technical field]

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS This application claims the benefit of U.S. Provisional Patent Application No. 63 / 335,916, entitled "AI / ML MODEL LIFE CYCLE MANAGEMENT," filed with the United States Patent and Trademark Office (USPTO) on April 28, 2022, and U.S. Patent Application No. 17 / 795,400, filed on July 26, 2022 as a national phase of International Application No. PCT / US2022 / 026646, filed on April 28, 2022, the disclosures of which are incorporated herein by reference in their entireties.

[0002] Methods and systems consistent with example embodiments of the present disclosure relate to managing Artificial Intelligence (AI) / Machine Learning (ML) models, and more particularly, to managing collaboration between nodes in a communication system and AI / ML model lifecycle management therefor. [Background technology]

[0003] In recent years, there has been growing interest in implementing AI / ML in communication systems. Nevertheless, frameworks and techniques for managing and / or utilizing AI / ML models in communication systems remain unclear and unclear to date. In particular, coordination between nodes (e.g., user equipment, base stations, network functions, etc.) in a communication system in managing and / or utilizing AI / ML models, as well as techniques for providing lifecycle management of AI / ML models, remain unclear and undefined.

[0004] In view of the above, there is a need to define frameworks and procedures related to the coordination or cooperation between nodes in a communication system, for example, for model information exchange, model training, providing model performance feedback, initiating or transferring models between nodes, etc.

[0005] Furthermore, to realize the cooperation or coordination between nodes, it is necessary to define procedures and signaling protocols or flows according to the AI / ML capabilities of the nodes. Similarly, it is necessary to define or classify the AI / ML capabilities, as well as classify the AI / ML model training. Summary of the Invention

[0006] According to embodiments, systems and methods are provided for facilitating collaboration between nodes in a communication system. Specifically, according to embodiments, systems and methods are provided for facilitating collaboration between nodes in a communication system for implementing AI / ML for air interface optimization and lifecycle management of AI / ML models therefor. According to embodiments, systems and methods are provided for usage in a cooperative manner of AI / ML models / inference, more specifically, systems and methods are provided for using AI / ML models / inference for optimizing air interfaces of a communication system.

[0007] According to an embodiment, a method for implementing artificial intelligence / machine learning (AI / ML) for air interface optimization in a mobile communication system may be provided. The method may include determining, by a first node, a coordination level for AI / ML coordination between a network and a user equipment (UE) from among a plurality of predefined coordination levels, and performing, by a second node, air interface optimization for the UE using at least one AI / ML model based on the determined coordination level. The plurality of predefined coordination levels may include a first level corresponding to AI / ML coordination between the network and the UE, a second level corresponding to signaling-based AI / ML coordination between the network and the UE without model transfer, and a third level corresponding to signaling-based AI / ML coordination between the network and the UE with model transfer.

[0008] According to an embodiment, according to the second level, AI / ML model training may be performed in both the network and the UE.

[0009] According to an embodiment, according to the second level, the network may provide the AI / ML model and / or inference tuning parameters to the UE.

[0010] According to an embodiment, the second level may include multiple split levels. A first split level of the multiple split levels may correspond to signaling-based AI / ML cooperation for a one-sided model without joint inference performed jointly across the UE and the network. A second split level of the multiple split levels may correspond to signaling-based AI / ML cooperation for a two-sided model with joint inference performed jointly across the UE and the network.

[0011] Alternatively (instead of the above configuration) or additionally (in addition to the above configuration), the second level may correspond to signaling-based AI / ML cooperation for a one-sided model without joint inference performed jointly across the UE and the network, and the third level may correspond to signaling-based AI / ML cooperation for a two-sided model with joint inference performed jointly across the UE and the network.

[0012] According to an embodiment, the model transfer may include at least one of transferring parameters of the AI / ML model structure or transferring a new AI / ML model having the parameters, where the new AI / ML model may be a full model or a partial model.

[0013] According to an embodiment, determining the cooperation level may include sending, by the first node, an AI / ML capability request to the UE inquiring about AI / ML capabilities of the UE, and receiving, by the first node, an AI / ML capability report from the UE in response to the AI / ML capability request. The AI / ML capability report indicates the AI / ML capabilities of the UE. Determining the cooperation level may further include determining the cooperation level from among a plurality of predefined cooperation levels based on the received AI / ML capability report.

[0014] According to an embodiment, the method may further include transmitting, by the first node, a request to the UE for information regarding at least one of the AI / ML models stored in the UE or used by the UE.

[0015] According to an embodiment, the air interface optimization may include at least one of Channel State Information (CSI) feedback enhancement, beam management, and positioning accuracy enhancement.

[0016] According to an embodiment, full-scale AI / ML model training may be performed in the UE according to the second level, and lightweight AI / ML model training may be performed in the UE according to the third level.

[0017] According to an embodiment, a system for implementing artificial intelligence / machine learning (AI / ML) for air interface optimization in a mobile communication system may be provided. The system may include a first node comprising a memory for storing instructions and at least one processor configured to execute the instructions to determine a coordination level for AI / ML coordination between a network and a user equipment (UE) from among a plurality of predefined coordination levels, and a second node comprising a memory for storing instructions and at least one processor configured to execute the instructions to perform air interface optimization for the UE using at least one AI / ML model based on the determined coordination level. The plurality of predefined coordination levels may include a first level corresponding to AI / ML coordination between the network and the UE, a second level corresponding to signaling-based AI / ML coordination between the network and the UE without model transfer, and a third level corresponding to signaling-based AI / ML coordination between the network and the UE with model transfer.

[0018] According to an embodiment, according to the second level, AI / ML model training may be performed in both the network and the UE.

[0019] According to an embodiment, according to the second level, the network can provide the AI / ML model and / or inference tuning parameters to the UE.

[0020] According to an embodiment, the second level may include a plurality of decomposition levels. A first decomposition level of the plurality of decomposition levels may correspond to a signaling-based AI / ML cooperation for a one-sided model without joint inference performed jointly across the UE and the network. A second decomposition level of the plurality of decomposition levels may correspond to a signaling-based AI / ML cooperation for a two-sided model with joint inference performed jointly across the UE and the network.

[0021] Alternatively or additionally, the second level may correspond to signaling-based AI / ML cooperation for a one-sided model without joint inference performed jointly across the UE and the network, and the third level may correspond to signaling-based AI / ML cooperation for a two-sided model with joint inference performed jointly across the UE and the network.

[0022] According to an embodiment, the model transfer may include at least one of transferring parameters of the AI / ML model structure or transferring a new AI / ML model having the parameters, where the new AI / ML model may be a full model or a partial model.

[0023] According to an embodiment, at least one processor of the first node may be configured to execute instructions for determining a coordination level by: sending, by the first node, an AI / ML capability request to the UE inquiring about the UE's AI / ML capabilities; receiving, by the first node, an AI / ML capability report from the UE in response to the AI / ML capability request, where the AI / ML capability report may indicate the UE's AI / ML capabilities; and determining a coordination level from among a plurality of predefined coordination levels based on the received AI / ML capability report.

[0024] According to an embodiment, the at least one processor of the first node may be further configured to execute instructions to send a request to the UE for information regarding at least one of an AI / ML model stored in the UE or an AI / ML model used by the UE.

[0025] According to an embodiment, the air interface optimization may include at least one of channel state information (CSI) feedback enhancement, beam management, and positioning accuracy enhancement.

[0026] According to an embodiment, according to a second level, full-scale AI / ML model training may be performed at the UE, and according to a third level, lightweight AI / ML model training may be performed at the UE.

[0027] Additional aspects will be set forth in part in the description that follows, and in part will be apparent from the description, or may be realized by practice of illustrative embodiments of the present disclosure.

[0028] The features, advantages and significance of exemplary embodiments of the present disclosure will now be described with reference to the accompanying drawings, in which like reference numerals denote like elements, and in which: [Brief description of the drawings]

[0029] [Figure 1] FIG. 1 illustrates a functional system architecture or framework for AI / ML applications in a Radio Access Network (RAN), according to one or more embodiments. [Diagram 2] FIG. 1 is a flow diagram of an example method for implementing AI / ML for air interface optimization in a mobile communication system, according to one or more embodiments. [Diagram 3] 1 is a flow diagram of an example method for determining a level of cooperation, according to one or more embodiments. [Figure 4] FIG. 1 illustrates example procedures associated with (performed in) lifecycle management of one or more AI / ML models, in accordance with one or more embodiments. [Diagram 5]FIG. 1 illustrates a signaling flow between a first node and a second node for collaborative AI / ML model usage, according to one or more embodiments. [Figure 6] FIG. 1 illustrates a breakdown of categories of AI / ML model training, according to one or more embodiments. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0030] Additional aspects will be set forth in part in the description that follows, and in part will be apparent from the description, or may be realized by practice of illustrative embodiments of the present disclosure.

[0031] The following detailed description of the exemplary embodiments refers to the accompanying drawings, in which the same reference numbers in different drawings may identify the same or similar elements.

[0032] The foregoing disclosure provides illustrations and descriptions, but is not intended to be exhaustive or to limit the implementations to the precise configurations disclosed. Modifications and variations are possible in light of the above disclosure or may result from practice of the implementations. Furthermore, one or more features or one or more components of one embodiment may be incorporated into or combined with another embodiment (or one or more features of another embodiment). In addition, in the description of operations provided below, it is understood that one or more operations may be omitted, one or more operations may be added, one or more operations may be performed (at least partially) simultaneously, and the order of one or more operations may be interchanged (changed).

[0033] It will be apparent that the systems and / or methods described herein may be implemented in various forms of hardware, firmware, or a combination of hardware and software. The actual dedicated control hardware or software code used to implement these systems and / or methods is not intended to limit the implementation. Thus, the operation and behavior of the systems and / or methods are described herein without reference to specific software code. It will be understood that software and hardware can be designed to implement the systems and / or methods based on the description herein.

[0034] Although particular combinations of features may be disclosed herein, such combinations are not intended to limit the disclosure of possible implementations, and in fact many of these features may be combined in methods, configurations, or steps not specifically disclosed herein.

[0035] As used herein, no element, act, or instruction should be construed as critical or essential unless explicitly described. Additionally, as used herein, the articles "a" and "an" are intended to include one or more items (elements, acts, instructions, nodes, modules, functions, entities, parts, components, configurations, steps) and may be used synonymously with "one or more." When only one item is intended, the term "one" or similar language is used. Additionally, as used herein, terms such as "has," "have," "having," "include," "including," and the like are intended to be open-ended terms (meaning that they may have or include more than what is stated). Additionally, the phrase "based on" is intended to mean "based at least in part on," unless specifically noted otherwise. Additionally, phrases such as "at least one of [A] and [B]" and "at least one of [A] or [B]" should be understood to include only A, only B, or both A and B.

[0036] Additionally, terms such as "AI / ML model" as used herein may refer to a data driven algorithm that applies artificial intelligence (AI) / machine learning (ML) techniques to generate one or more outputs based on one or more inputs.

[0037] General System Architecture 1 illustrates a functional system architecture or framework for AI / ML applications in a radio access network (RAN) in accordance with one or more embodiments. With reference to FIG. 1, system 100 may include a data collection module 110, a model training module 120, a model inference module 130, and an actor module 140.

[0038] According to an embodiment, the data collection module 110 may be a function (or an element / entity configured to perform a function) that collects data and provides the data to the model training module 120 and the model inference module 130.

[0039] For example, the data collection module 110 may collect data from one or more nodes and / or one or more network functions in the system, such as (but not limited to) User Equipment (UE), base stations (e.g., gNB), Location Management Function (LMF), Access and Mobility Management Function (AMF), and Operation And Management (OAM) functions or entities. The collected data may include measurement data (e.g., state measurement data results, performance measurement data results, etc.) from the one or more nodes and / or one or more network functions, feedback data from the actor module 140, output data from AI / ML models, and any other suitable data.

[0040] Additionally, data collection module 110 can provide collected data to model training module 120 so that the data can be utilized to train an AI / ML model (in this regard, the data provided to model training module 120 can be referred to herein as "training data"). According to an embodiment, training data can include offline field data (e.g., data collected from the field and used for offline training of an AI / ML model) and online field data (e.g., data collected from the field and used for online training of an AI / ML model).

[0041] Additionally, the data collection module 110 may provide the collected data to the model inference module 130 so that the data may be utilized in inference operations / calculations (in this regard, the data provided to the model inference module 130 may be referred to herein as “inference data”).

[0042] According to embodiments, the data collection module 110 may be communicatively coupled to, and configured to collect data from, one or more additional modules configured to perform one or more data preparation operations, such as (but not limited to) data pre-processing, data cleaning, data formatting, and data conversion. According to other embodiments, the data collection module 110 may be configured to perform the one or more data preparation operations.

[0043] Further, according to an embodiment, the model training module 120 may be a function (or an element / entity configured to perform a function) that performs training operations of one or more AI / ML models (which may be referred to herein as "AI / ML model training"). AI / ML model training may refer to a process of training one or more AI / ML models by learning input-output relationships in a data driven manner and obtaining trained AI / ML models therefrom. Furthermore, AI / ML model training may be performed online and / or offline, and may be performed in the UE and / or the network.

[0044] According to an embodiment, AI / ML model training may include AI / ML model validation and / or AI / ML model testing. AI / ML model validation may be the process of evaluating the quality of an AI / ML model using a dataset different from the dataset used for model training, and AI / ML model testing may be the process of evaluating the performance of a final AI / ML model using a dataset different from the dataset used for model training and validation.

[0045] For example, the model training module 120 may be configured to obtain or request data (e.g., training data) or information (e.g., from the data collection module 110), receive the requested data or information, and execute one or more AI / ML algorithms (based on the received data or information) to perform one or more of model training, model validation, and model testing. Further, the model training module 120 may be configured to provide or deploy one or more trained models (trained models), validated models (validated models), and / or tested models (tested models) to the model inference module 130.

[0046] According to an embodiment, the model training module 120 may be configured to generate model performance metrics via a model testing operation. The model training module 120 may also be configured to perform one or more data preparation operations based on the training data provided by the data collection module 110. Furthermore, the model training module 120 may be configured to receive (from other modules or entities) one or more trained models, one or more validated models, and / or one or more tested models, and to update, upgrade, and / or roll back the one or more trained models, the one or more validated models, and / or the one or more tested models.

[0047] However, according to an embodiment, model inference module 130 may be a function (or an element / entity configured to perform that function) that performs one or more inference operations associated with an AI / ML model (sometimes referred to herein as "AI / ML model inference"). AI / ML model inference may be a process of using a trained AI / ML model to generate one or more outputs based on one or more inputs. For example, AI / ML model inference may include one or more operations that utilize the trained AI / ML model and inference data to predict or determine one or more outputs of the trained AI / ML model.

[0048] As an example, model inference module 130 may be configured to obtain or request data (e.g., inference data, AI / ML models, etc.) or information (e.g., from data collection module 110, from model training module 120, etc.), receive the requested data or information, and perform one or more AI / ML model inferences based on the received data or information. According to an embodiment, model inference module 130 may be configured to perform said one or more inference operations on one or more trained models, one or more validated models, and / or one or more tested models (e.g., provided by model training module 120, etc.) based on the inference data provided by data collection module 110.

[0049] Further, model inference module 130 can be configured to provide one or more inference outputs to actor module 140. According to an embodiment, model inference module 130 can be configured to monitor the performance of the AI / ML model and provide the performance of the AI / ML model (which may be referred to herein as "model performance feedback (data, information)") to model training module 120, such that model training module 120 can utilize the model performance feedback (data, information) when performing one or more of model training, model validation, and model testing.

[0050] According to an embodiment, model inference module 130 may be configured to perform one or more data preparation operations based on the inference data provided by data collection module 110. Additionally, model inference module 130 may be configured to signal an inference output of the model to a node that requested the inference output and / or to a node that takes an action based on the inference output.

[0051] Further, according to an embodiment, actor module 140 may be a function (or an element / entity configured to perform a function) that performs one or more actions based on the inference output provided by model inference module 130. For example, actor module 140 may be configured to trigger or perform one or more actions directed at other entities (e.g., other modules, other nodes, other functions, etc.) and / or itself (actor module 140). Additionally, actor module 140 may be configured to provide feedback data or information to data collection module 110. Feedback data or information may include any suitable data that may be utilized in deriving training data, inference data, and / or in monitoring the performance of the AI / ML model and its impact on the network.

[0052] It is believed that the above-described exemplary embodiments are merely possible embodiments, and the scope of the present disclosure should not be limited thereto. Furthermore, it is believed that one or more of the modules 110-140 in the system 100 can be implemented or deployed in software form, hardware form, or a combination thereof. According to an embodiment, one or more of the modules 110-140 can be implemented or deployed in the form of computer-readable and / or computer-executable instructions that, when read or executed by at least one processor, cause the at least one processor to perform one or more operations associated with the module.

[0053] To this end, system 100 provides a framework for managing intelligence and data of one or more AI / ML models in the RAN. Accordingly, system 100 (and associated architecture) may be utilized for various purposes in a communication system. For example, one or more exemplary embodiments may apply system 100 for AI / ML optimization of an air interface in a communication system. In one or more embodiments, the air interface is a 5G New Radio (NR) air interface, and the optimization of the air interface may include channel state information (CSI) feedback enhancement, beam management, positioning accuracy enhancement, and any other suitable optimization of air interface related operations and functions / features. Furthermore, it is contemplated that system 100 may be utilized in any other possible use cases (e.g., network energy conservation, load balancing, mobility optimization, etc.).

[0054] Classification of collaboration levels using AI / ML According to example embodiments, multiple predefined levels or categories of AI / ML collaboration or cooperation are provided (e.g., as applied or standardized by the communication system) between network nodes and / or network functions in a communication system, such as user equipment (UE), base stations (e.g., gNB), location management functions (LMF), access and mobility management functions (AMF), and operation and administration (OAM) functions.

[0055] Table 1 below illustrates an example of multiple predefined levels of AI / ML cooperation or collaboration regarding AI / ML standardization and gNB / UE capabilities, according to one embodiment. [Table 1]

[0056] Although Table 1 illustrates example coordination between a UE and a gNB and the associated roles in the example coordination, it is contemplated that similar disclosure (coordination and roles) may be applicable to coordination between a UE and any suitable network (e.g., a gNB and network functions such as LMF, AMF, OAM, etc.) without departing from the scope of the present disclosure.

[0057] As shown in Table 1, the multiple pre-defined levels of AI / ML collaboration (or coordination) may include three levels or categories: Level #0, Level #1, and Level #2 (although it is understood that one or more other embodiments are not limited thereto and may include two levels or categories or more than three levels or categories).

[0058] In the first level #0, implementation-based AI / ML algorithms are used. That is, in the first level #0, the UE and / or the network (e.g., gNB, LMF, AMF, OAM, etc.) can utilize one or more specific built-in AI / ML models without exchanging information with each other. The specific built-in AI / ML models may be provided by any suitable personnel, such as the UE vendor, the network operator, etc. According to an embodiment, one or more of the AI / ML models utilized on one side are transparent to the other side. Also, one or more AI / ML models may be base models similarly built into multiple sides (e.g., the UE and the network may have similar base models, etc.).

[0059] At the second level #1, information exchange may take place. That is, at the second level #1, the UE and / or the network (e.g., gNB, LMF, AMF, OAM, etc.) utilize one or more AI / ML models for which information has been exchanged without transferring the one or more AI / ML models to each other. For example, if the cooperation is between the UE and the gNB, the UE and the gNB may exchange information for AI / ML use, but certain operations related to the AI / ML models (e.g., training, validation, testing, inference, etc.) may be performed separately at the UE and / or the gNB.

[0060] According to an embodiment, the information exchanged between the UE and the network may include AI / ML models and / or inference tuning parameters. In this regard, the UE and / or the network may be configured to train one or more AI / ML models, update one or more AI / ML models and / or associated parameters (e.g., based on the exchanged information, etc.), and provide the AI / ML models and / or inference tuning parameters to each other. For example, in an uplink use case, the UE may provide parameters or information to the gNB so that the gNB can tune its AI / ML models based on the provided parameters or information. As another example, the UE may have certain built-in parameters (e.g., provided by a UE vendor, etc.) that the UE would like to utilize, and thus the UE may provide the parameters to the gNB so that the gNB can utilize the parameters.

[0061] Furthermore, according to an embodiment, the second level #1 may include multiple split levels, such as a first split level #1A and a second split level #1B. The first split level #1A may correspond to signaling-based AI / ML cooperation of one-sided model(s) with / without collaboration, and the second split level #1B may correspond to signaling-based AI / ML cooperation of two-sided model(s) with / without collaboration.

[0062] The term "one-sided model" can refer to an AI / ML model that is utilized on one side (e.g., trained, validated, tested, inferred, etc.) and independent of the other. For example, a one-sided model may be a UE-side AI / ML model (e.g., an AI / ML model where inference is performed in the UE), a network-side model (e.g., an AI / ML model where inference is performed in the network), etc.

[0063] The term "two-sided model" may refer to an AI / ML model that is utilized on both sides (e.g., trained, validated, tested, inferred, etc.). For example, a two-sided model may be an AI / ML model that is utilized first in the UE and then in the gNB, or vice versa.

[0064] The term "joint operation" may refer to one or more operations (e.g., training, validation, testing, inference operations, etc.) that are performed jointly across nodes and / or networks. For example, a joint operation may include joint inference where one or more inference operations (e.g., prediction, decision, etc.) are performed jointly across nodes and / or networks (e.g., joint inference may be performed across a UE and a gNB, where a first portion of the inference is performed by the UE, a second portion of the inference is performed by the gNB, etc.).

[0065] In view of the above, a first partitioning level #1A may correspond to signaling-based AI / ML cooperation for a one-sided model with or without joint inference performed jointly across the UE and the network (e.g., gNB and network functions such as LMF, AMF, OAM, etc.), and a second partitioning level #1B may correspond to signaling-based AI / ML cooperation for a two-sided model with or without joint inference performed jointly across the UE and the network.

[0066] To this end, the second level #1 (and its associated decomposition levels) defines model assistance information exchange and mode / inference parameter exchange or updates between the UE and the network. Furthermore, the second level #1 (and its associated decomposition levels) defines a framework for providing feedback or information regarding model performance between the UE and the network. Furthermore, the second level #1 (and its associated decomposition levels) defines a framework for model training (e.g., based on the UE, etc.).

[0067] At the third level #2, model exchange or model transfer may take place. That is, at the third level #2, the UE and / or the network (e.g., gNB, LMF, AMF, OAM, etc.) may exchange one or more AI / ML models. For example, the network may transfer one or more AI / ML models to the UE, the UE may transfer one or more AI / ML models to the network, the UE may download one or more AI / ML models from the network, etc. The model transfer may include at least one of transferring parameters of the AI / ML model structure, transferring a new AI / ML model with parameters (e.g., a new full AI / ML model, a new partial AI / ML model, etc.), etc.

[0068] According to an embodiment, at the third level #2, the UE and / or network can obtain updates for one or more AI / ML models from one or more external repositories. For example, the UE and / or network can download updated or new models from an external repository, etc. Additionally, at the third level #2, lightweight AI / ML model training can be performed at the UE and / or network.

[0069] To this end, the third level #2 defines a framework for initiating, transferring, and validating one or more AI / ML models between the UE and the network (e.g., gNB, LMF, AMF, OAM, etc.).

[0070] According to an embodiment, the information exchange may be performed based on the ID of the associated AI / ML model. For example, the network may provide the UE with the AI / ML model and / or inference tuning parameters based on the model ID, etc. Furthermore, according to an embodiment, at the second level #1, full-scale AI / ML model training may be performed at the UE and / or the network.

[0071] Further, according to an embodiment, one or more AI / ML models may be applied to air interface optimization based on the implementation or selection of a particular level (or category) of cooperation among multiple predefined levels (or categories). For example, air interface optimization may include channel state information (CSI) feedback enhancement, beam management, positioning accuracy enhancement, and optimization of any other suitable air interface related operations and features. Furthermore, it is contemplated that one or more AI / ML models may also be utilized in any other possible use cases (e.g., network energy savings, load balancing, mobility optimization, etc.) based on the implementation or selection of a particular level (or category) of cooperation among multiple predefined levels (or categories).

[0072] AI / ML Capabilities Classification AI / ML capabilities may refer to the capabilities of a UE and / or a network (e.g., gNB, LMF, AMF, OAM, etc.) to perform one or more AI / ML related tasks.

[0073] According to an embodiment, the AI / ML capabilities of a UE may be divided into classes: Class 0, Class 1, Class 2, and Class 3.

[0074] A UE classified as class 0 may not have AI / ML capabilities. A UE classified as class 1 may only support information exchange. A UE classified as class 2 may support information exchange and model transfer. A UE classified as class 3 may support information exchange, model transfer, and model training.

[0075] In consideration of the above, class 0 UEs can perform one or more AI / ML tasks based on the above first Level #0 cooperation, class 1 UEs can perform one or more AI / ML tasks based on the above second Level #1 cooperation, and class 2 UEs and class 3 UEs can perform one or more AI / ML tasks based on the above third Level #2 cooperation.

[0076] Although the above description has been described with reference to categorizing the AI / ML capabilities of a UE, it will be appreciated that a similar categorization may be applicable to any node or network (e.g., gNB, LMF, AMF, OAM, etc.) without departing from the scope of the present disclosure.

[0077] To this end, the exemplary embodiments of the present disclosure define a classification of AI / ML capabilities for nodes.

[0078] Exemplary Use Case: AI / ML Implementation for Air Interface Optimization As mentioned above, according to one embodiment, AI / ML may be implemented for air interface optimization in a mobile communication system.

[0079] FIG. 2 illustrates a flow diagram of an example method 200 for implementing AI / ML for air interface optimization in a mobile communication system, according to one or more embodiments.

[0080] 2, in operation S210, a cooperation level is determined. Specifically, the first node may be configured to determine a cooperation level for AI / ML cooperation between the network and a user equipment (UE) from among a plurality of predefined cooperation levels.

[0081] According to an embodiment, the first node may be a UE, a network (e.g., a gNB, an LMF, an AMF, an OAM, etc.), or any other suitable node or element in a mobile communication system.

[0082] Further, the multiple predefined coordination levels may include a first level corresponding to AI / ML coordination between the network and the UE, a second level corresponding to a first signaling-based AI / ML coordination between the network and the UE, and a third level corresponding to a second signaling-based AI / ML coordination between the network and the UE.

[0083] According to the embodiment, the first level, the second level, and the third level may correspond to the above-mentioned first level #0, the above-mentioned second level #1, and the above-mentioned third level #2, respectively. For example, the second level may correspond to signaling-based AI / ML cooperation between the network and the UE without model transfer, and the third level may correspond to signaling-based AI / ML cooperation between the network and the UE with model transfer.

[0084] In this regard, the second level may include multiple decomposition levels, for example, the first decomposition level (e.g., the first decomposition level #1A described above) corresponds to signaling-based AI / ML cooperation for one-sided models with / without joint inference performed jointly across the UE and the network, and the second decomposition level (e.g., the second decomposition level #1B described above) corresponds to signaling-based AI / ML cooperation for two-sided models with / without joint inference performed jointly across the UE and the network. Furthermore, the model transfer may include at least one of transferring parameters of the AI / ML model structure or transferring a new AI / ML model with the parameters, where the new AI / ML model may be a full model or a partial model.

[0085] Alternatively, the second level and the third level may correspond to the above-mentioned first division level and the above-mentioned second division level, respectively. For example, the second level may correspond to a one-sided model signaling-based AI / ML cooperation without joint inference performed jointly across the UE and the network, and the third level may correspond to a two-sided model signaling-based AI / ML cooperation with / without joint inference performed jointly across the UE and the network.

[0086] According to an embodiment, according to the second level, AI / ML model training may be performed at both the network and the UE. Additionally or alternatively, according to the second level, the network can provide the AI / ML model and / or inference tuning parameters to the UE.

[0087] According to an embodiment, according to a second level, full-scale AI / ML training may be performed at the UE, and according to a third level, lightweight AI / ML model training may be performed at the UE.

[0088] 2, in operation S220, one or more air interface optimizations are performed. Specifically, the second node may be configured to perform one or more air interface optimizations for the UE using at least one AI / ML model based on the determined cooperation level (e.g., the cooperation level determined in operation S210). The second node may be a UE or a network (e.g., a gNB, an LMF, an AMF, an OAM, etc.), or any other suitable node or element in a mobile communication system. The one or more air interface optimizations may include channel state information (CSI) feedback enhancement, beam management, positioning accuracy enhancement, and optimization of any other suitable air interface related operations and features.

[0089] It is contemplated that method 200 may include one or more additional or alternative operations (processes) without departing from the scope of the present disclosure.

[0090] For example, according to an embodiment, method 200 may include an operation in which the first node may be configured to send a request to the UE for information regarding at least one of an AI / ML model stored in the UE or an AI / ML model used by the UE, and receive said information from the UE. Alternatively or additionally, according to an embodiment, method 200 may include an operation in which the first node may be configured to send a request to an external repository for information regarding at least one of an AI / ML model stored in the external repository or an AI / ML model used by the UE, and receive said information from the external repository.

[0091] Similarly, one or more of the operations of method 200 may include one or more additional operations without departing from the scope of this disclosure. For example, Figure 3 illustrates a flow diagram of an example method 300 for determining a level of cooperation (i.e., operation S210 in method 200) according to one or more embodiments.

[0092] 3, in operation (process) S310, one or more AI / ML capability requests are transmitted. Specifically, the first node may be configured to transmit one or more AI / ML capability requests to the UE to inquire about the UE's AI / ML capabilities. According to an embodiment, the one or more AI / ML capability requests may include a request for information defining a classification of the UE's AI / ML capabilities (e.g., class 0 to class 3 above).

[0093] In operation S320, one or more AI / ML capability reports are received. Specifically, the first node may be configured to receive one or more AI / ML capability reports from the UE in response to the AI / ML capability request, and the one or more AI / ML capability reports may indicate the AI / ML capabilities of the UE. According to an embodiment, the one or more AI / ML capability reports may include information defining a classification of the AI / ML capabilities of the UE.

[0094] In operation S330, a cooperation level is determined. Specifically, the first node may be configured to determine the cooperation level from among a plurality of predefined cooperation levels based on the received AI / ML capability report. According to an embodiment, the first node may be configured to determine the cooperation level by determining a classification of the AI / ML capability of the UE.

[0095] AI / ML model lifecycle management According to an embodiment, nodes (e.g., UEs and networks such as gNBs, LMFs, AMFs, OAMs, etc.) can cooperate with each other to perform lifecycle management of one or more AI / ML models.

[0096] FIG. 4 illustrates example procedures involved in the lifecycle management of one or more AI / ML models, according to one or more embodiments.

[0097] Referring to FIG. 4 , AI / ML model lifecycle management 400 may include (but is not limited to) the following procedures: AI / ML capability reporting procedure 410, AI / ML model synchronization procedure 420, AI / ML model update / upgrade procedure 430, AI / ML model performance monitoring procedure 440, AI / ML model training / generation procedure 450, AI / ML model deployment procedure 460, and AI / ML model rollback procedure 470.

[0098] It may be understood that steps 410-470 may be performed in any order without departing from the scope of the present disclosure. It is further contemplated that one or more of steps 410-470 may be combined with one another to form other steps (e.g., AI / ML capability reporting and AI / ML model synchronization may be combined to form a step such as AI / ML capability discovery) and / or may be included within one another (e.g., AI / ML model training / generation may be included in AI / ML model update, etc.).

[0099] Further, according to an embodiment, at least one or all of the procedures 410-470 may be applicable when one or more of “information exchange”, “model exchange”, and “model training” are supported by the UE or the network.

[0100] According to an embodiment, the AI / ML capabilities reporting procedure 410 may include one or more operations for reporting the AI / ML capabilities of the UE and / or the AI / ML capabilities of the network. For example, the network (e.g., a gNB) may inquire about the AI / ML capabilities of the UE by sending one or more AI / ML capabilities requests to the UE and obtaining one or more AI / ML capabilities reports or responses from the UE, as described above.

[0101] According to an embodiment, the AI / ML model synchronization procedure 420 may include one or more operations to discover one or more repositories that store one or more AI / ML models and associated information therewith, retrieve one or more AI / ML models or associated information from the discovered repositories, and update current AI / ML models (stored or utilized by the UE and / or network, such as a gNB) with the retrieved AI / ML models or information. For example, the network (e.g., a gNB) may request information of all stored AI / ML models or information of a particular AI / ML model based on one or more requirements (e.g., AI / ML capabilities, optimization requirements, etc.).

[0102] According to an embodiment, the AI / ML model update / upgrade procedure 430 may include one or more of the following operations (processes): model update trigger, new model update (e.g., by model download or model replacement, etc.), model training / generation, update / upgrade of existing model (e.g., by information exchange, etc.), AI / ML inference update / transfer, model / inference selection procedure, model or inference validation, and retraining or fine-tuning of one or more AI / ML models.

[0103] According to an embodiment, the AI / ML model performance monitoring procedure 440 may include one or more operations (processes) for monitoring the performance of one or more AI / ML models. For example, the procedure 440 may include operations for monitoring the inference performance of one or more AI / ML models (e.g., by receiving feedback, etc.) after the one or more AI / ML models are deployed, trained / generated, and / or updated / upgraded.

[0104] According to an embodiment, the AI / ML model training / generation procedure 450 may include one or more operations (processes) for training or generating one or more AI / ML models. The one or more operations may be any suitable operations for learning input-output relationships in a data driven manner and generating / obtaining a trained AI / ML model therefrom. For example, the one or more operations may include online training, offline training, or a combination thereof. Furthermore, the one or more operations may be performed in a UE, a network (e.g., gNB, etc.), or a combination thereof. Additionally, the one or more operations may include any suitable type of learning operation, such as, but not limited to, federated learning / training (one or more AI / ML models are trained across multiple nodes, such as UEs, gNBs, etc., with each node performing local model training using local data samples), supervised learning (one or more AI / ML models are trained based on input data and corresponding predefined parameters), unsupervised learning (one or more AI / ML models are trained without predefined parameters), semi-supervised learning (one or more AI / ML models are trained with a mixture of predefined and non-predefined data / parameters), and reinforcement learning (one or more AI / ML models are trained based on input data and feedback signals resulting from the output of the models in the environment with which the models are interacting).

[0105] According to an embodiment, the AI / ML model deployment procedure 460 may include one or more operations to deliver one or more AI / ML models (e.g., trained models, validated models, tested models, etc.) and associated information (e.g., runtime images, algorithms, etc.) to a target UE or a target network (e.g., a target gNB) where one or more inference operations are performed. The one or more operations may include a model transfer operation in which the one or more AI / ML models are delivered over an air interface.

[0106] According to an embodiment, the AI / ML model rollback procedure 470 may include one or more operations to roll back or restore one or more current AI / ML models to a previous version of the AI / ML model. For example, the one or more operations may include replacing or updating one or more current AI / ML models with the respective previous version of the AI / ML model, fine tuning or adjusting one or more current AI / ML models to the previous version, etc.

[0107] Multiple nodes (e.g., UE, gNB, etc.) may cooperate or coordinate with one another in performing one or more of the above-described procedures 410-470. For example, Figure 5 illustrates a signaling flow between a first node (e.g., gNB) and a second node (e.g., UE) for use of a collaborative AI / ML model, according to one or more embodiments.

[0108] Referring to FIG. 5, the signaling flow may include the procedures of FIG. 4, such as an AI / ML model capability reporting procedure 410, an AI / ML model synchronization procedure 420, an AI / ML model update / upgrade procedure 430, an AI / ML model performance monitoring procedure 440, and an AI / ML training / generation procedure 450.

[0109] Specifically, in operation S501, a first node may send one or more AI / ML capability requests to a second node. In response, in operation S502, the second node may provide one or more AI / ML capability reports to the first node. Operations S501 and S502 may be part of the AI / ML model capability reporting procedure 410 described above with reference to FIG. 4, and the description related to one or more AI / ML capability requests and one or more AI / ML capability reports described above with reference to FIGS. 2-4 may be similarly applicable to the description of FIG. 5.

[0110] In operation S503, the first node may send one or more requests for the AI / ML models and / or associated information or parameters to the second node. In response, in operation S504, the second node may provide the first node with one or more reports including the requested AI / ML models and / or associated information or parameters. Operations S503 and S504 may be part of the AI / ML model synchronization procedure 420 described above with reference to FIG. 4, and the process and features of the information exchange and model transfer may be similar to those described above with reference to FIGS. 2-4. It may therefore be understood that similar descriptions may be applicable to the description of FIG. 5 in similar content and manner.

[0111] In operation S505, the first node may send one or more requests to the second node to perform model inference and / or model update. In response, the second node may perform the requested model inference and / or model update and provide one or more results (e.g., completion, failure, inference value / output, etc.) to the first node (operation S506). Operations S505 and S506 may be part of the AI / ML model update / upgrade procedure 430 or may be part of the AI / ML model deployment procedure 460 described above with reference to FIG. 4. It may therefore be understood that similar descriptions may be applicable to the description of FIG. 5 in similar content and manner.

[0112] In operation S507, the first node may monitor the AI / ML model performance at the second node. For example, the first node may send one or more requests for performance data associated with the AI / ML model to the second node, after which, in operation S508, the second node may provide one or more feedbacks defining the performance of the AI / ML model. According to an embodiment in which one or more AI / ML models are also utilized at the first node (e.g., inferred, updated, etc.), the second node may also monitor the AI / ML model performance of the first node in a similar manner. Operations S507 and S508 may be part of the AI / ML model performance monitoring procedure 440 described above with reference to FIG. 4. It may therefore be understood that similar descriptions may be applicable to the description of FIG. 5 in similar content and manner.

[0113] In operation S509, the first node may determine that one or more of a model rollback, an inference update, and a model training are required at the second node (e.g., based on AI / ML performance feedback, etc.). Thus, the first node may send one or more requests to the second node to perform the model rollback, the inference update, and / or the model training. Operation S509 may include one or more of the AI / ML model update / upgrade procedure 430, the AI / ML model training / generation procedure 450, and the AI / ML model rollback procedure 470 described above with reference to FIG. 4. Thus, it may be understood that a similar description may be applicable to the description of FIG. 5 in a similar content and manner.

[0114] To this end, example embodiments of the present disclosure define procedures and signaling flows between nodes for coordination and collaboration in utilizing and managing one or more AI / ML models.

[0115] Classification of AI / ML model training According to an embodiment, categories of AI / ML model training can be defined based on use case requirements and AI / ML capabilities.

[0116] For example, according to one or more embodiments, training of one or more AI / ML models may be performed based on a particular level (or category) of cooperation among multiple predefined levels (or categories).

[0117] FIG. 6 is a diagram illustrating a breakdown of categories of AI / ML model training, according to one or more embodiments.

[0118] Referring to FIG. 6, AI / ML model training may be performed in the UE (e.g., on-UE training, etc.) and / or in the network, such as a gNB (e.g., on-network training, etc.).

[0119] Model training in the UE can be further classified into full-scale AI / ML model training and light weight AI / ML model training. Full-scale AI / ML model training can be large scale training in which all (or a majority, half) of the data or parameters are utilized for training. Conversely, light weight AI / ML model training can be small scale training in which only a portion of the data or parameters are utilized for training. Full-scale AI / ML model training and / or light weight AI / ML model training can include one or more of federated learning / training, supervised learning, unsupervised learning, semi-supervised learning, and reinforcement learning (previously described with reference to FIG. 4).

[0120] According to an embodiment, one or more full-scale AI / ML model training may be performed at the UE in accordance with a second level (e.g., second level #1 described above) of coordination between the UE and the network (e.g., gNB), while one or more lightweight AI / ML model training may be performed at the UE in accordance with a third level (e.g., third level #2 described above) of coordination between the UE and the network.

[0121] Meanwhile, model training at the network (e.g., gNB) may be further categorized into transfer AI / ML model training or learning, full-scale AI / ML model training, and lightweight AI / ML model training. Transfer AI / ML model training may include one or more training operations that utilize knowledge or output obtained from a previous task to improve generalization about the AI / ML model. For example, transfer AI / ML model training may include reusing one or more pre-trained models in a new task. Full-scale and lightweight AI / ML model training are as described above with reference to model training at the UE.

[0122] According to an embodiment, forwarded AI / ML model training and / or lightweight AI / ML model training may be performed at the network whenever the network is commanded (forced) or required to forward one or more AI / ML models to the UE. Additionally, full-scale AI / ML model training may be performed at the network whenever the network determines that lightweight AI / ML model training has been performed at the UE.

[0123] To this end, example embodiments of the present disclosure define categories of AI / ML model training according to various conditions and requirements.

[0124] Various aspects of the embodiments The foregoing disclosure provides illustration and description, but is not intended to be exhaustive or to limit the implementations to the precise forms disclosed. Modifications and variations to the above disclosure are possible in light of the above disclosure or may be acquired from practice of the implementations.

[0125] Some embodiments may relate to systems, methods, and / or computer-readable media at any possible level of technical detail of implementation. Furthermore, one or more of the above components described above may be implemented as instructions stored on a computer-readable medium and executable by at least one processor (and / or may include at least one processor). The computer-readable medium may include a computer-readable non-transitory storage medium(s) having computer-readable program instructions for causing a processor to perform operations.

[0126] A computer readable storage medium may be a tangible device capable of holding and storing instructions for use by an instruction execution device. A computer readable storage medium may be, for example, but not limited to, an electronic storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination thereof. A non-exhaustive list of more specific examples of computer readable storage media includes the following: portable computer diskettes, hard disks, random access memories (RAM), read-only memories (ROM), erasable programmable read-only memories (EPROM or flash memory), static random access memories (SRAM), portable compact disk read-only memories (CD-ROM), digital versatile disks (DVD), memory sticks, floppy disks, mechanically encoded devices such as punch cards or grooved ridge structures having instructions recorded therein, and any suitable combination thereof. Computer-readable storage media as used herein should not be interpreted as transitory signals per se, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through a waveguide or other transmission medium (e.g., light pulses passing through a fiber optic cable), or electrical signals transmitted through wires.

[0127] The computer-readable program instructions described herein can be downloaded from a computer-readable storage medium to each computing / processing device or to an external computer or external storage device via a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. The network may include copper transmission cables, optical fiber transmission, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. A network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards the computer-readable program instructions for storage in a computer-readable storage medium in the respective computing / processing device.

[0128] The computer readable program code / instructions for performing the operations may be either source code or object code written in any combination of one or more programming languages, including assembler instructions, Instruction-Set-Architecture (ISA) instructions, machine instructions, machine dependent instructions, microcode, firmware instructions, state-setting data, configuration data for an integrated circuit, or object-oriented programming languages ​​such as Smalltalk, C++, and procedural programming languages ​​such as the "C" programming language or similar programming languages. The computer readable program instructions may be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the latter scenario, the remote computer may be connected to the user's computer through any type of network, including a Local Area Network (LAN) or Wide Area Network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet Service Provider). In some embodiments, electronic circuitry including, for example, a programmable logic circuit, a field programmable gate array (FPGA), or a programmable logic array (PLA) can execute computer readable program instructions by utilizing state information of the computer readable program instructions to personalize the electronic circuitry to perform an aspect or operation.

[0129] These computer readable program instructions may be provided to a processor of a general purpose computer, a special purpose computer, or other programmable data processing apparatus to generate a machine such that the instructions, which execute on the processor of the computer or other programmable data processing apparatus, create means for performing the functions / acts specified in one or more blocks of the flowcharts and / or block diagrams. These computer readable program instructions may also be stored on a computer readable storage medium capable of directing a computer, programmable data processing apparatus, and / or other device to function in a particular manner, such that the computer readable storage medium on which the instructions are stored includes a product including instructions implementing aspects of the functions / acts specified in one or more blocks of the flowcharts and / or block diagrams.

[0130] The computer readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device to cause the computer, other programmable data processing apparatus, or other device to execute a series of operational steps to generate a computer-implemented process, such that the instructions executing on the computer, other programmable data processing apparatus, or other device perform the functions / acts specified in one or more blocks of the flowcharts and / or block diagrams.

[0131] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer-readable media according to various embodiments. In this regard, each block in the flowcharts or block diagrams may represent a module, segment, or part of instructions that includes one or more executable instructions for implementing the specified logical function. The methods, computer systems, and computer-readable media may include additional, fewer, different, or differently arranged blocks than those shown in the figures. In some alternative implementations, the functions described in the blocks may occur in a different order than that described in the figures. For example, two blocks shown in succession may actually be executed simultaneously or substantially simultaneously, or the blocks may sometimes be executed in the reverse order depending on the functionality involved. It should also be noted that each block of the block diagrams and / or flowchart diagrams, as well as combinations of blocks in the block diagrams and / or flowchart diagrams, may be implemented by a dedicated hardware-based system that performs the specified functions or operations or executes a combination of dedicated hardware and computer instructions.

[0132] It will be apparent that the systems and / or methods described herein may be implemented in various forms of hardware, firmware, or a combination of hardware and software. The actual dedicated control hardware or software code used to implement these systems and / or methods is not intended to limit the implementation. Thus, the operation and behavior of the systems and / or methods are described herein without reference to specific software code, and it will be understood that software and hardware can be designed to implement the systems and / or methods based on the description herein.

[0133] In view of the above, various further respective aspects and features of the embodiments of the present disclosure can be defined by the following clauses: Item [1]: A method for implementing artificial intelligence / machine learning (AI / ML) for air interface optimization in a mobile communication system, the method including: determining, by a first node, a coordination level for AI / ML coordination between a network and a user equipment (UE) from among a plurality of predefined coordination levels; and performing, by a second node, air interface optimization for the UE using at least one AI / ML model based on the determined coordination level, the plurality of predefined coordination levels including a first level corresponding to AI / ML coordination between the network and the UE, a second level corresponding to signaling-based AI / ML coordination between the network and the UE without model transfer, and a third level corresponding to signaling-based AI / ML coordination between the network and the UE with model transfer. Item [2]: The method according to item [1], wherein according to a second level, AI / ML model training can be performed in both the network and the UE. Item [3]: According to the second level, the network can provide the UE with an AI / ML model and / or inference adjustment parameters, according to any of items [1] to [2]. Item [4]: ​​The method according to any of items [1] to [3], wherein the second level may include multiple decomposition levels, a first decomposition level among the multiple decomposition levels may correspond to signaling-based AI / ML cooperation for a one-sided model without joint inference performed jointly across the UE and the network, and a second decomposition level among the multiple decomposition levels may correspond to signaling-based AI / ML cooperation for a two-sided model with / without joint inference performed jointly across the UE and the network. Item [5]: A method according to any of items [1] to [3], wherein the second level can correspond to signaling-based AI / ML cooperation for a one-sided model without joint inference performed jointly across the UE and the network, and the third level can correspond to signaling-based AI / ML cooperation for a two-sided model with / without joint inference performed jointly across the UE and the network. Item [6]: The method according to any of items [1] to [5], wherein the model transfer may include at least one of transferring parameters of the AI / ML model structure or transferring a new AI / ML model with the parameters, and the new AI / ML model may be a full model or a partial model. Item [7]: A method according to any of Items [1] to [6], wherein determining the cooperation level includes sending, by the first node, an AI / ML capability request to the UE inquiring about the UE's AI / ML capabilities, and receiving, by the first node, an AI / ML capability report from the UE in response to the AI / ML capability request, wherein the AI / ML capability report indicates the UE's AI / ML capabilities, and determining the cooperation level may further include determining the cooperation level from among a plurality of predetermined cooperation levels based on the received AI / ML capability report. Item [8]: A method described in any of Items [1] to [7], wherein the method may further include sending, by the first node, a request for information to the UE regarding at least one of an AI / ML model stored in the UE or an AI / ML model used by the UE. Item [9]: A method described in any of items [1] to [8], wherein the air interface optimization may include at least one of channel state information (CSI) feedback enhancement, beam management, and positioning accuracy enhancement. Item

[10] : A method according to any of items [1] to [9], wherein full-scale AI / ML model training can be performed in the UE according to the second level, and lightweight AI / ML model training can be performed in the UE according to the third level. Item

[11] : A system for implementing artificial intelligence / machine learning (AI / ML) for air interface optimization in a mobile communication system, the system including: a first node having a memory for storing instructions and at least one processor configured to execute the instructions to determine a coordination level for AI / ML coordination between a network and a user equipment (UE) from among a plurality of predetermined coordination levels; and a second node having a memory for storing instructions and at least one processor configured to execute the instructions to perform air interface optimization for the UE using at least one AI / ML model based on the determined coordination level, the plurality of predetermined coordination levels including a first level corresponding to AI / ML coordination between the network and the UE, a second level corresponding to signaling-based AI / ML coordination between the network and the UE without model transfer, and a third level corresponding to signaling-based AI / ML coordination between the network and the UE with model transfer. Item

[12] : The system described in Item

[11] , wherein according to the second level, AI / ML model training can be performed in both the network and the UE. Item

[13] : A system described in any of items

[11] to

[12] , wherein, according to the second level, the network can provide AI / ML models and / or inference tuning parameters to the UE. Item

[14] : The system described in any of Items

[11] to

[13] , wherein the second level may include multiple decomposition levels, a first decomposition level of the multiple decomposition levels may correspond to signaling-based AI / ML cooperation for a one-sided model without joint inference performed jointly across the UE and the network, and a second decomposition level of the multiple decomposition levels may correspond to signaling-based AI / ML cooperation for a two-sided model with / without joint inference performed jointly across the UE and the network. Item

[15] : A system described in any of items

[11] to

[13] , wherein the second level can correspond to signaling-based AI / ML cooperation for a one-sided model without joint inference performed jointly across the UE and the network, and the third level can correspond to signaling-based AI / ML cooperation for a two-sided model with / without joint inference performed jointly across the UE and the network. Item

[16] : The system described in any of Items

[11] to

[15] , wherein the model transfer may include at least one of transferring parameters of the AI / ML model structure or transferring a new AI / ML model with the parameters, and the new AI / ML model may be a full model or a partial model. Item

[17] : A method according to any of Items

[11] to

[16] , wherein at least one processor of the first node may be configured to execute instructions for determining a cooperation level by sending, by the first node, an AI / ML capability request to the UE inquiring about the UE's AI / ML capabilities, receiving, by the first node in response to the AI / ML capability request, an AI / ML capability report from the UE indicating the UE's AI / ML capabilities, and determining a cooperation level from among a plurality of predetermined cooperation levels based on the received AI / ML capability report. Item

[18] : A system described in any of Items

[11] to

[17] , wherein at least one processor of the first node may be further configured to execute instructions to send a request to the UE for information regarding at least one of an AI / ML model stored in the UE or an AI / ML model used by the UE. Item

[19] : A system described in any of Items

[11] to

[18] , wherein the air interface optimization may include at least one of channel state information (CSI) feedback enhancement, beam management, and positioning accuracy enhancement. Item

[20] : A system described in any of items

[11] to

[19] , wherein full-scale AI / ML model training can be performed in the UE according to the second level, and lightweight AI / ML model training can be performed in the UE according to the third level. Item

[21] : According to the second level, the network can provide the AI / ML model and / or inference adjustment parameters to the UE based on the identity (ID) of the AI / ML model, according to any of items [1] to

[10] . Item

[22] : According to the second level, the network can provide the AI / ML model and / or inference tuning parameters to the UE based on the identity (ID) of the AI / ML model, in a system described in any of items

[11] to

[20] .

[0134] It can be appreciated that many modifications and variations of the present disclosure are possible in light of the above teachings, and it will be apparent that, within the scope of the appended provisions, the present disclosure may be embodied in forms, manners, and ways other than those specifically described herein.

Claims

1. 1. A method for implementing artificial intelligence / machine learning (AI / ML) for air interface optimization in a mobile communication system, the method comprising: determining, by a first node, a coordination level for AI / ML coordination between the network and a user equipment (UE) from among a plurality of predefined coordination levels; and performing, by a second node, air interface optimization for the UE using at least one AI / ML model based on the determined cooperation level; The plurality of predetermined cooperation levels include: a first level corresponding to AI / ML cooperation between the network and the UE; a second level corresponding to signaling-based AI / ML cooperation between the network and the UE without model transfer; and a third level corresponding to signaling-based AI / ML cooperation between the network and the UE involving model transfer.

2. The method of claim 1 , wherein AI / ML model training is performed in both the network and the UE according to the second level.

3. The method of claim 1 , wherein the network provides AI / ML models and / or inference tuning parameters to the UE according to the second level.

4. the second level includes a plurality of division levels; a first decomposition level of the plurality of decomposition levels corresponds to a signaling-based AI / ML cooperation for a one-sided model without joint inference performed jointly across the UE and the network; 2. The method of claim 1 , wherein a second decomposition level of the plurality of decomposition levels corresponds to signaling-based AI / ML cooperation for a two-sided model with joint inference performed jointly across the UE and the network.

5. The second level corresponds to a signaling-based AI / ML cooperation for a one-sided model without joint inference performed jointly across the UE and the network; 2. The method of claim 1, wherein the third level corresponds to signaling-based AI / ML cooperation for a two-sided model with joint inference performed jointly across the UE and the network.

6. 2. The method of claim 1, wherein the model transfer includes at least one of transferring parameters of an AI / ML model structure or transferring a new AI / ML model having parameters, the new AI / ML model being a full model or a partial model.

7. The determining the level of cooperation may include: sending, by the first node, an AI / ML capability request to the UE inquiring about the AI / ML capabilities of the UE; receiving, by the first node, an AI / ML capabilities report from the UE in response to the AI / ML capabilities request, the AI / ML capabilities report indicating the AI / ML capabilities of the UE; and determining the cooperation level from among the plurality of predefined cooperation levels based on the received AI / ML capability report.

8. 2. The method of claim 1, further comprising: transmitting, by the first node, a request to the UE for information regarding at least one of an AI / ML model stored in the UE or an AI / ML model used by the UE.

9. 2. The method of claim 1, wherein the air interface optimization includes at least one of channel state information (CSI) feedback enhancement, beam management, and positioning accuracy enhancement.

10. According to the second level, full-scale AI / ML model training is performed in the UE; The method of claim 1 , wherein according to the third level, lightweight AI / ML model training is performed in the UE.

11. 1. A system implementing artificial intelligence / machine learning (AI / ML) for air interface optimization in a mobile communication system, the system comprising: a first node comprising a memory storing instructions and at least one processor configured to execute the instructions to determine a coordination level for AI / ML coordination between a network and a user equipment (UE) from among a plurality of predefined coordination levels; a second node comprising a memory for storing instructions and at least one processor configured to execute the instructions to perform air interface optimization for the UE using at least one AI / ML model based on the determined cooperation level; The plurality of predetermined cooperation levels include: a first level corresponding to AI / ML cooperation between the network and the UE; a second level corresponding to signaling-based AI / ML cooperation between the network and the UE without model transfer; and a third level corresponding to signaling-based AI / ML cooperation between the network and the UE with model transfer.

12. The system of claim 11 , wherein according to the second level, AI / ML model training is performed in both the network and the UE.

13. The system of claim 11 , wherein the network provides AI / ML models and / or inference tuning parameters to the UE according to the second level.

14. the second level includes a plurality of division levels; a first decomposition level of the plurality of decomposition levels corresponds to a signaling-based AI / ML cooperation for a one-sided model without joint inference performed jointly across the UE and the network; 12. The system of claim 11, wherein a second decomposition level of the plurality of decomposition levels corresponds to signaling-based AI / ML cooperation for a two-sided model with joint inference performed jointly across the UE and the network.

15. The second level corresponds to a signaling-based AI / ML cooperation for a one-sided model without joint inference performed jointly across the UE and the network; The system of claim 11 , wherein the third level corresponds to signaling-based AI / ML cooperation for a two-sided model with joint inference performed jointly across the UE and the network.

16. 12. The system of claim 11, wherein the model transfer includes at least one of transferring parameters of an AI / ML model structure or transferring a new AI / ML model having parameters, the new AI / ML model being a full model or a partial model.

17. The at least one processor of the first node executes the instructions to sending, by the first node, an AI / ML capability request to the UE inquiring about the AI / ML capabilities of the UE; receiving, by the first node, an AI / ML capabilities report from the UE in response to the AI / ML capabilities request, the AI / ML capabilities report indicating the AI / ML capabilities of the UE; determining the coordination level from among the plurality of predefined coordination levels based on the received AI / ML capability report; The system of claim 11 , configured to determine the level of cooperation by:

18. 12. The system of claim 11, wherein the at least one processor of the first node is further configured to execute the instructions to send to the UE a request for information regarding at least one of an AI / ML model stored in the UE or an AI / ML model used by the UE.

19. 12. The system of claim 11, wherein the air interface optimization includes at least one of channel state information (CSI) feedback enhancement, beam management, and positioning accuracy enhancement.

20. According to the second level, full-scale AI / ML model training is performed in the UE; and The system of claim 11 , wherein according to the third level, lightweight AI / ML model training is performed in the UE.

Images