Lifecycle management using ML model identification and ML function identification

Abstract

A method comprising: requesting the use of an ML feature based on the use of an ML model appropriate for the machine learning (ML) feature; receiving an activation message relating to at least one of the following: activation of one ML model appropriate for the ML feature, wherein the ML model is associated with at least one feature, or activation of one of the at least one features associated with one ML model appropriate for the ML feature, wherein the activation message indicates that the ML model and / or one feature is being activated; starting the use of the ML model based on the received activation message; and reporting at least one of the following: a key performance indicator (KPI) for the ML model, or a feature performance KPI for the feature associated with the ML model.

Description

Technical Field

[0001] The present disclosure relates to methods and apparatuses for life cycle management (LCM) using machine learning (ML) model identification and ML function identification.

Background Art

[0002] The following description of the background art may include insights, discoveries, understandings or disclosures, or associations with, but not necessarily known to the relevant prior art, of at least some examples of embodiments of the present disclosure. Some of such contributions of the present disclosure may be specifically pointed out below, but other such contributions of the present disclosure will become apparent from the relevant context.

[0003] In recent years, there has been a growing expansion of wireless communication networks worldwide, including wire-based communication networks such as Integrated Services Digital Network (ISDN) and Digital Subscriber Line (DSL), as well as cellular 3G networks such as cdma2000 (Code Division Multiple Access) systems and Universal Mobile Telecommunications System (UMTS), 4G or enhanced communication networks based on Long-Term Evolution (LTE) or Long-Term Evolution-Advanced (LTE-A), 5G networks, 6G networks, 2G cellular networks such as Global System for Mobile Communications (GSM), General Purpose Packet Radio System (GPRS), and Enhanced Data Rate for Global Evolution (EDGE), and other wireless communication systems such as Wireless Local Area Network (WLAN), Bluetooth, and Worldwide Interoperability for Microwave Access (WiMAX). Various organizations, including the European Telecommunications Standards Institute (ETSI), the Third Generation Partnership Project (3GPP), Telecom & Internet Converged Services & Protocols for Advanced Networks (TISPAN), the International Telecommunication Union (ITU), the Third Generation Partnership Project II (3GPP2), the Internet Engineering Task Force (IETF), the IEEE (Institute of Electrical and Electronics Engineers), the WiMAX Forum, and similar organizations, are working on standards or specifications for telecommunications networks and access environments.

[0004] In this context, within the machine learning paradigm, model lifecycle management (LCM) encompasses data processing, training, deployment, and monitoring. LCM ensures the effectiveness and robustness of models used for complex tasks. While machine learning solutions are already adapted to the 5G NR air interface, LCM interactions and signaling have not yet been discovered. As agreed in RAN1#110, different LCM components are listed below.

[0005] [Table 1]

[0006] For interoperability across different components of LCMs and models in different nodes (UEs, NWs), 3GPP proposed two different LCM-based identifications: a) model-identification-based LCM and b) function-identification-based LCM. Working hypotheses and agreements for model identification and function identification were defined in RAN1#111. Figures 1 and 2 illustrate, respectively, examples of model ID-based LCM and function (ID)-based LCM, as proposed by Nokia in 3GPP RAN1#112. Apart from LCMs for AI / ML models, 3GPP also identified three use cases as AI / ML applications. Figure 3 illustrates an example of how these two LCM mechanisms could be applied to one of the use cases. Use cases for the Rel-18 NR air interface are CSI feedback enhancement, beam management, and positioning enhancement.

[0007] [Table 2]

[0008] In addition, RAN1 also agreed on UE-NW collaboration in response to signaling via air interfaces and model transfer, as mentioned below.

[0009] [Table 3]

[0010] Generally, LCM procedures ensure the continuous integration and validity of an AI / ML model throughout its entire lifecycle. While LCM for models is clearly defined in the machine learning domain, the structural framework for the 3GPP NR air interface is still under discussion. During the discussion, two distinct LCM mechanisms have been identified: a model-based LCM mechanism and a function-based LCM mechanism. The former aims to enable continuous development of the model and ensure smooth integration with other underlying behaviors surrounding the model. The latter aims to enable continuous integration of ML functions within UE and NW communications. However, this has yet to be discovered. a) How should ML models and ML functions be organized to ensure smooth collaboration between the UE and the NW? b) How should the interaction between model-based LCM and function-based LCM be defined? c) How should different levels of collaboration (level x, level y, and level z) be ensured using these two LCM mechanisms? d) How should we provide an extensible structure to accommodate an increasing number of ML-enabled features?

[0011] The example of ML-enabled feature definition in Figure 3 illustrates just one of several different ways to organize and define the use of ML models and ML functions within all ML features.

[0012] For example, another approach is to define ML features such that one feature ID corresponds to one ML feature (use case), and different ML models (identified by the ID) are defined for different sub-use cases, or scenario configurations (spanning the columns in Figure 3).

[0013] Another approach should potentially be that both are identified by one single ID (one ML model in each column of FIG. 3) when there is a one-to-one mapping between the ML function and the ML model.

[0014] Variations by combining the above approaches are further possible.

Summary of the Invention

[0015] These options will increase the complexity of both UE specifications (UE capabilities, features, testability, requirements, etc.) for the actual implementation solution, and ultimately increase the complexity of the LCM procedure.

[0016] Therefore, there is a need for improvement. In particular, there is a need for LCM using ML model identification and ML function identification.

[0017] LCM using ML model identification and ML function identification as disclosed herein can solve such problems.

[0018] Therefore, it is an object of the present disclosure to improve the prior art.

[0019] The following meanings of the abbreviations used herein apply. 2G Second Generation 3G Third Generation 3GPP Third Generation Partnership Project 3GGP2 Third Generation Partnership Project 2 4G Fourth Generation 5G Fifth Generation 6G Sixth Generation AI Artificial Intelligence AP Access Point BS Base Station CDMA Code Division Multiple Access CSI Channel State Information DSL Digital Subscriber Line EDGE Enhanced Data Rate for Global Evolution EEPROM Electrically Erasable Programmable Read Only Memory eNB Evolved Node B ETSI European Telecommunications Standards Institute FID Function ID gNB Next Generation Node B GPRS General Packet Radio Service GSM Global System for Mobile Communications ID Identification IEEE Institute of Electrical and Electronics Engineers ISDN Integrated Services Digital Network ITU International Telecommunication Union KPI Key Performance Indicator LCM Life Cycle Management LTE Long Term Evolution LTE-A Long Term Evolution - Advanced MANET Mobile Ad Hoc Network MAC CE MAC Control Element MID Model ID ML Machine Learning NB Node B NW Network RAM Random Access Memory RAN Radio Access Network ROM Read Only Memory TISPAN Telecommunications & Internet Converged Services & Protocols for Advanced Networks UE User Equipment UMTS Universal Mobile Telecommunications System UUID Universally Unique Identifier UWB Ultra Wide Band WCDMA Wideband Code Division Multiple Access WiMAX Worldwide Interoperability for Microwave Access WLAN Wireless Local Area Network

[0020] The objective of various examples of embodiments of this disclosure is to improve upon the prior art. Accordingly, at least some examples of embodiments of this disclosure aim to address at least some of the issues and / or problems and shortcomings described above.

[0021] Various embodiments of the embodiments of this disclosure are presented in the appended claims and relate to methods, apparatus and computer program products relating to LCM using ML model identification and ML function identification.

[0022] The objective is achieved by methods, apparatus, and non-temporary storage media as specified in the attached claims. Further advantageous developments are presented in the respective dependent claims.

[0023] Any one of the embodiments referred to in accordance with the attached claims enables LCM using ML model identification and ML function identification, thereby enabling the resolution of at least some of the problems and shortcomings identified / derivable from above.

[0024] Therefore, improvements are achieved through methods, apparatus, and computer program products that enable LCM using ML model identification and ML function identification.

[0025] More specifically, this specification discloses an LCM using ML model identification and ML function identification, which is advantageous over proprietary solutions because, among other advantageous technical effects, it achieves reduced complexity for UE specifications and actual implementation solutions, as well as reduced complexity of LCM procedures.

[0026] Further advantages will become clear from the detailed explanation below.

[0027] Some embodiments of this disclosure are described below as examples with reference to the accompanying drawings. [Brief explanation of the drawing]

[0028] [Figure 1] This is a diagram of the proposed model-identification-based LCM module for ML model training and inference procedures [3GPP RAN1#112]. [Figure 2] This is a diagram of a proposed function identification-based LCM module for ML functionality with an ML inference mode procedure [3GPP RAN1#112]. [Figure 3] This diagram shows an example of an ML-enabled feature definition, along with the use of feature IDs and model IDs, and the associated information. [Figure 4] This is a diagram illustrating different illustrations of features (FID) or collections of features (FID1, FID2, ...) that will be correlated with a model (MID) or a collection of models (MID1, MID2). [Figure 5] This is a diagram (parts 1 / 2 and 2 / 2) of Framework C, which shows the call flow for feature (ID)-based and ML model ID-based LCM with feature and / or ML model switching at collaboration level z. [Figure 6] This is a diagram of Framework B (parts 1 / 2 and 2 / 2) showing the call flow for feature (ID)-based LCM monitoring and switching in the UE with Framework B at collaboration level y. [Figure 7] This is a diagram of Framework B (parts 1 / 2 and 2 / 2) showing the call flow for feature (ID)-based LCM monitoring and switching in the UE with Framework B at collaboration level z. [Figure 8] This flowchart illustrates the steps corresponding to various examples of the embodiment. [Figure 9] This flowchart illustrates the steps corresponding to various examples of the embodiment. [Figure 10] This is a block diagram illustrating the apparatus according to various examples of the embodiment. [Figure 11] This is a block diagram illustrating the apparatus according to various examples of the embodiment. [Modes for carrying out the invention]

[0029] Basically, in order to properly establish and handle communication between two or more endpoints (e.g., terminal devices, user equipment (UE), or other communication network elements such as communication stations, elements, or functions, databases, servers, hosts, etc.), one or more network elements or functions (e.g., virtualization network functions), such as access network elements such as access points (APs), radio base stations (BS), relay stations, eNBs, gNBs, etc., as well as core network elements or functions such as control nodes, support nodes, service nodes, gateways, user plane functions, access and mobility functions, may be included, and these may belong to one communication network system or different communication network systems.

[0030] In the following, different exemplary embodiments will be described using a communication network architecture based on 3GPP standards for communication networks such as 5G / NR (or 6G / NR) as an example of a communication network to which the embodiments may be applied, but the embodiments will not be limited to such an architecture. It will be apparent to those skilled in the art that the embodiments may also be applicable to other types of communication networks such as 4G and / or LTE (or 5G, 6G or even “XG”) that integrate mobile communication principles such as Wi-Fi, Worldwide Interoperability for Microwave Access (WiMAX), Bluetooth(R), Personal Communication Services (PCS), ZigBee(R), Broadband Code Division Multiple Access (WCDMA), Systems using Ultra-Wideband (UWB) technology, Mobile Ad Hoc Networks (MANET), and wired access. Furthermore, without loss of generality, although the description of some examples of embodiments relates to mobile communication networks, the principles of this disclosure can be extended and applied to any other type of communication network, such as wired communication networks or data center networking.

[0031] The following examples and embodiments should be understood as illustrative examples only. While this specification may refer to “an,” “one,” or “some” examples or embodiments in several places, this does not necessarily mean that each such reference relates to the same example or embodiment, or that the features apply only to a single example or embodiment. Furthermore, a single feature from a different embodiment may be combined to provide other embodiments. In addition, terms such as “equipment” and “include” should be understood not to limit the embodiments described to consist only of features that have been mentioned, and such examples and embodiments may also include features, structures, units, modules, etc., that have not been specifically mentioned.

[0032] A basic system architecture for a (long-range) communication network, including a mobile communication system to which some examples of embodiments may be applicable, may include one or more communication network architectures, including a wireless access network subsystem and a core network. Such architectures may include one or more communication network control elements or functions, access network elements, radio access network elements, access service network gateways, or base transceiver base stations, such as base stations (BS), access points (AP), node B (NB), eNB or gNB, distributed or centralized units (CU), which control their respective coverage areas or cells, and one or more communication stations, elements, functions or applications capable of conducting communications, such as UEs, elements or functions, or other devices with similar functions, such as modem chipsets, chips, modules, or similar devices, which may also be connected as separate elements to such elements, functions or applications, and which are capable of conducting communications, via one or more channels with one or more communication beams for transmitting several types of data in multiple access domains. Furthermore, it may include (core) network elements or network functions ((core) network control elements or network functions, (core) network management elements or network functions) such as gateway network elements / functions, mobility management entities, mobile switching centers, servers, databases, and the like.

[0033] The overall functionality and interconnections of the elements and functions described, which also depend on the actual network type, are known to those skilled in the art and described in the corresponding specifications; therefore, a detailed description thereof is omitted herein. Nevertheless, it should be noted that several additional network elements and signaling links may be employed for communication between elements, functions, or applications, such as communication endpoints, communication network control elements, and other elements of the same or other communication network, including servers, gateways, wireless network controllers, and other elements of the communication network, in addition to those described in detail herein below.

[0034] Communication network architectures, such as those considered in the examples of embodiments, may also have the ability to communicate with other networks, such as public switched telephone networks or the Internet. Communication networks may also have the ability to support the use of cloud services for virtual network elements or their functionalities, and it should be noted that the virtual network portion of a telecommunications network may also be provided by non-cloud resources, such as internal networks or similar. It should be understood that network elements of access systems, such as core networks, and / or their respective functionalities, can be implemented by using any node, host, server, access node, or entity appropriate for such use. Generally, network functions can be implemented as network elements on dedicated hardware, as software instances running on dedicated hardware, or as virtualized functions instantiated on an appropriate platform, such as cloud infrastructure.

[0035] Furthermore, core network control elements or functions, (core) network management elements or functions, and any other elements, functions or applications, such as network elements including communication elements like UEs, mobile devices, and terminal devices; control elements or functions such as access network elements like base stations (BS), eNBs / gNBs, and wireless network controllers; gateway elements; or other network elements or functions as described herein, may be implemented by software, such as computer program products for computers, and / or by hardware. To perform each of these processes, the devices, nodes, functions, or network elements used accordingly may include several means, modules, units, components, etc. (not shown) required for control, processing, and / or communication / signaling functions.Such means, modules, units, and components may include, for example, one or more processors or processor units, including one or more processing sections for executing instructions and / or programs, and / or processing data; storage or memory units or means (e.g., ROM, RAM, EEPROM, and the like) for storing instructions, programs, and / or data to be served as a work area for the processor or processing section and the like; input or interface means (e.g., floppy disks, CD-ROMs, EEPROM, and the like) for inputting data and instructions by software; user interfaces (e.g., screens, keyboards, and the like) for providing the user with monitoring and operability; other interfaces or means for establishing links and / or connections under the control of the processor unit or section (e.g., wired and wireless interface means, wireless interface means including, for example, an antenna unit or the like, means for forming a wireless communication section, etc.); and the like, where each means forming an interface such as a wireless communication section may also be located at a remote site (e.g., a wireless head or radio station, etc.). It should be noted that in this specification, the term "processing portion" should not be considered solely as representing the physical portion of one or more processors, but also as a logical division of a referred processing task performed by one or more processors.

[0036] It should be understood that the so-called “liquid” or flexible network concept can be adopted when, in some examples, the operation and functionality of network elements, network functions, or other entities in the network can be flexibly implemented in different entities or functions, such as nodes, hosts, or servers. In other words, the “division of labor” between the network elements, functions, or entities involved can vary on a case-by-case basis.

[0037] Referring here to Figure 4, Figure 4 illustrates different examples of functions (FIDs) or collections of functions (FID1, FID2, ...) that will be correlated with models (MIDs) or collections of models (MID1, MID2) according to various examples of embodiments.

[0038] In Figure 4, Framework A can be considered to represent the prior art. Furthermore, with respect to Figure 4, it is assumed that the UE provides the gNB with information about at least the features supported (in the UE), and optionally, information about the ML models supported (in the UE), and the features and ML models are associated with specific ML-enabled features, such as those depicted in Figure 3.

[0039] ML models are identified by a model ID and may optionally be supplemented by associated metadata. Functions are identified by a function ID and may optionally be supplemented by associated metadata.

[0040] According to at least some examples of the embodiments, and further with respect to Figure 4, there may be several frameworks for how FIDs and MIDs can be related. As an example, Figure 4 visualizes three different frameworks. In framework A410, a function (ID) is linked / associated with a collection of ML models (IDs). These associated ML models support the same function. In framework B420, a given ML model (ID) is linked / associated with a collection of different functions (IDs). In this case, the associated functions are supported by the same ML models. In framework C430, the ML model (ID) is independent of the function (ID). This framework supports either a combination of frameworks A410 and B420.

[0041] The frameworks disclosed herein aim to mitigate the aforementioned complexity issues and provide an extensible structure capable of accommodating an increased number of ML-enabled features (Release 19, 6G) and underlying ML models. Frameworks B420 and C430 in Figure 4 outline two identified categories of solutions proposed herein.

[0042] Two possible associations between LCM procedures for signaling ML model configurations corresponding to frameworks B420 and C430 in Figure 4 are disclosed herein. The objective is to enable effective collaboration and signaling between the UE and NW during ML-enabled features. Some of the key concepts are highlighted below.

[0043] In the case of framework B420 or C430 according to various examples of embodiments, - The first node (e.g., UE) can provide the second node (e.g., gNB) with information about the association between each of its ML models and the functions provided by these ML models. - The first node may receive an activation message from the second node to activate one of the ML models, and the activation may include at least a configuration of a desired feature output KPI, which will be monitored by the second node. - After receiving ML model activation, the first node can start the appropriate ML model management procedure (model monitoring, activation, deactivation, switching) for the associated function. - The first node can report the configured functional output KPIs to the second node. - The second node can initiate appropriate ML function management procedures (models (monitoring, activation, deactivation, switching)) based on the function output KPIs received from the first node.

[0044] In this specification, the following terms are introduced:

[0045] FID (Function ID): During the function identification procedure, a function, or a set of related functions, is uniquely identified by an ID called an FID. This FID is expected to be a string of bits that may be a collection of functions (such as activation, deactivation, and switching) for a use case. However, the format of this FID is outside the scope of this specification. Further examples of the format are shown below.

[0046] MID (Model ID): During the model identification procedure, a model is uniquely identified by an ID called the MID. This ID ensures that the model is uniquely identified throughout its lifecycle. The format of this MID is also outside the scope of this specification. Further examples of the format are shown below.

[0047] MID-LCM (MID-based LCM): Lifecycle management of ML models is identified by MID.

[0048] FID-LCM (FID-based LCM): Function lifecycle management is identified by FID.

[0049] Furthermore, according to at least some examples of the embodiments, the following should be considered:

[0050] In Framework B420, model-based and feature-based LCMs are dependent on each other, and a given ML model is associated with different features (collections of features). Interoperability between model-based and feature-based LCMs requires a specific alignment in their configuration.

[0051] In such a framework B420, as illustrated by various examples of embodiments, it is assumed that the UE has a set of ML models, and the NW requests that specific functions be used / activated by the UE. The function LCM is controlled by the NW, while the ML model LCM can be controlled by the UE, or alternatively by the NW. When the operation of the ML model LCM and the function LCM are on different nodes, their configurations still need to be aligned and compatible.

[0052] In Framework C430, model-based LCM and feature-based LCM are completely unrelated. Therefore, to ensure efficient interoperability between these two LCMs, strict alignment between models and features (configuration, monitoring, etc.) is required.

[0053] Call flows for Framework C430 by various examples of embodiments.

[0054] Referring from here to Figure 5 (which will be connected by parts 1 / 2 and 2 / 2, A-A' and B-B'), Figure 5 shows Framework C, which is a call flow for feature (ID)-based and ML model ID-based LCM with feature and / or ML model switching at collaboration level z.

[0055] A detailed example call flow is shown in Figure 5. According to various examples of the embodiment, it is assumed that the UE has a function and requests a model from the NW. Both the model and the function will be controlled by the NW. Nevertheless, although the operation of the LCM in both cases is independent of each other, the synchronization and compatibility of the operation must be carefully handled and incorporated using configuration and signaling.

[0056] The call flow steps are explained below, referring to Figure 5. Step 0 (not shown): The UE500 indicates to the NW (represented, for example, by the gNB510) that it needs to use a specific ML feature (e.g., a beam management use case). Step 1 (for Level Z collaboration only): If the UE500 does not have a suitable model for the ML features, the NW transfers a preferred model along with the MID and associated metadata. Step 2: The UE500 determines the ML function associations for the received model. Option 1 Step 3: If the UE500 accepts the ML functionality, it will send an ACK (e.g., an acknowledgment message) to the network to use the received model. Step 4: The UE500 will configure its functions for this model and exchange that configuration with the network. Once the UE500 is ready, it will request the network to enable the ML features. Step 5: The network will enable / activate the ML functionality. Step 6: UE500 will now begin using the model. Steps 7-8: Since the model-based LCM is controlled by the network, the UE500 will send periodic, aperiodic / triggered MID performance reports to the network. If the network decides to switch functions based on the reported performance, the network will signal to switch models. Step 9: If a switch is required, the UE500 can decide to continue from Step 1 or fall back. Otherwise, continue to Step 10. Steps 10-12: Since the function-based LCM is controlled by the network, the UE500 will send periodic, aperiodic, and triggered function reports to the network. If the network decides to switch functions based on the reported performance, it will signal to switch functions. Option 2 Step 13: The UE500 rejects associating the received ML model with its ML functionality. Step 14: The network selects a different model and either continues with Step 1 or falls back.

[0057] The following provides references to call flows for Framework B420 through various examples of embodiments.

[0058] Furthermore, according to at least some examples of the embodiments, the following should be considered:

[0059] In Framework B420, a model can be associated with / linked to one or more features. An example could be an intermediate KPI generated from the model. The following steps may occur: - Framework B420 can be implemented in either the UE or NW. - When Framework B420 is implemented in the network, function (ID)-based LCM and model (ID)-based LCM will be fully controlled by the network. - If Framework B420 is in UE, three possible interactions can occur depending on the level of collaboration. During Level X collaboration, the functionality and model LCM are evident to the network. During Level Y collaboration, features need to be indicated to enable requirements, configurations, and KPIs related to a specific use case. In this case, features (sets of features) are interlinked with models, and therefore the models are • To be registered with the network. • Activated by the network. This needs to be done, and then the function (set of functions) is, • Activated by the network • LCM management is performed via the network. At level z, the UE may not have the model required for a particular use case and will request the specified model from the NW. In this case, • Model LCM is maintained by NW. • Functional LCM is maintained by the network.

[0060] Figures 6 and 7 show a promising call flow for Framework B, assuming that the model LCM is maintained by the UE, while the feature-based LCM is maintained by the gNB. The gNB has more knowledge about the surrounding environment and can therefore help the UE effectively use the model through the feature-based LCM.

[0061] Referring from here to Figure 6 (which will be connected by parts 1 / 2 and 2 / 2, A-A' and B-B'), Figure 6 shows Framework B, which is the call flow for function (ID)-based LCM monitoring and switching in the UE with Framework B at collaboration level y.

[0062] A call flow for collaboration level y in Framework B420, using various examples of embodiments, is shown in Figure 6, and the steps are described below. Step 1: The UE600 selects a model for an ML-enabled feature (e.g., a beam management use case), and indicates and registers the model's metadata (identified by the MID) and the feature (or set of features) information associated with the model (identified by the FID) in the NW (represented by, for example, the gNB610). Step 2: The network determines the metadata for the feature (set of features) FID for the indicated ML-enabled feature. Steps 3-4: The network sends an ACK signal (such as an acknowledgment message) for FID. The UE600 configures the FID option for the model. Step 5: The network sends FID activation via MAC CE. Steps 6-7: The UE600 starts using the model with the activated FID. LCM operations related to the model (model monitoring, updates, etc.) are performed by the UE and are visible to the network. Step 8: The UE600 sends periodic / aperiodic / event-triggered performance KPI reports for FID to the network. Steps 9-10: The network evaluates the performance of the FID. If the performance is insufficient, the network can send a FID deactivation via MAC CE. The network can also send a new FID activation via MAC CE. Step 11: If the UE600 has a function switch, the operation in Step 6 continues or falls back.

[0063] Referring from here to Figure 7 (which will be connected by parts 1 / 2 and 2 / 2, A-A' and B-B'), Figure 7 shows Framework B, which is the call flow for function (ID)-based LCM monitoring and switching in the UE with Framework B at collaboration level z.

[0064] A call flow for collaboration level z in Framework B420, according to at least some examples of the embodiments, is shown in Figure 7, and the steps are described below. Step 0 (not shown): The UE700 indicates a network (represented by, for example, gNB710) for ML-enabled features (e.g., beam management use case). Step 1: The network transfers the model and the associated features (or sets of features). The model metadata is identified by the MID, and the feature (or set of features) information is identified by the FID associated with the model. Step 2: The UE700 determines the metadata for the feature (set of features) FID for the indicated ML-enabled feature. Steps 3-4: The UE700 sends an ACK signal (like an acknowledgment message) for both MID and FID. This signal ensures that the UE700 can use the model and that the UE700 has support for the feature configuration indicated by FID. The UE700 configures the FID option for the model. Steps 5-6: The network sends FID activation and MID activation signals to the UE via MAC CE. Steps 6-7: The UE700 starts using models with activated FIDs. LCM operations related to models (such as model monitoring and updates) and LCM operations related to functions (such as function monitoring and switching) are performed by the network. Step 8: The UE700 sends periodic / aperiodic / event-triggered performance KPI reports for both MID and FID to the network. Steps 9-10: The network evaluates the performance of the FID. If the performance is insufficient, the network can send a FID deactivation via MAC CE. The network can also send a new FID activation via MAC CE. Step 11: If the UE700 has a function switch, the operation in Step 7 continues or falls back. Step 12: When the network is also monitoring the model, the network can request a model update. In this case, the network can trigger a model switchover. Steps 13-14: The network first sends a FID deactivation signal via MAC CE, and then sends a MID deactivation signal via MAC CE. Step 15: If the UE700 receives a model switching indication, the UE700 requests to use a different model. If the model is not available in the UE700, Step 1 is repeated.

[0065] For completeness, examples of elements in FID and MID are provided, but the content is not limited to these. FID may include at least one of the following: a) Function ID: A label / tag / UUID that uniquely identifies the function and the combination of items b) to g) below. b) Additional IDs / labels / tags a. When model identification is applied, each additional ID / label corresponds to a model ID. b. Otherwise, each additional ID / label corresponds to a model that will be monitored, and the model can identify performance variations in the features supported by the model. c) Applicable scenarios / configurations / parameters / conditions made possible by the model's capabilities: including the systems and intermediate KPIs that will be used for feature-based LCM. d) Input data types / sources and preparation / preprocessing: including indications for any delay-sensitive ML-specific data processing to be performed, such as indications for the expected delay budget for such operations. e) Non-ML behavior (optional): Indication of any non-ML behavior / algorithm included in the model feature, such as indication of the expected delay budget for such behavior. f) Output data and post-processing (optional): including indications for any delay-sensitive ML-specific output post-processing to be performed, such as indications for the expected delay budget for such operations. A specific control signaling configuration that enables and (partially) controls the operations of g)b)~f) and the corresponding function-based LCM. MID may include at least one of the following: a) Model ID: A label / tag / UUID that uniquely identifies an ML model (such as an implemented version) within an ML-enabled feature (spanning several potential features) or only within a specified feature, and b) Related information about AI / ML models i. When a proprietary ML model format is used, the associated information may include the following: ● Potential additional metadata required by function-based LCM ii. When the Open ML model format is used, the associated information may include the following: ● Model input data (dimensions, features) and preparation / preprocessing, including indications for any feature extraction, feature selection, or any other delay-sensitive ML-specific data processing to be performed, such as indications for the expected delay budget for such operations. ● Model output data (dimensions, features) and post-processing (optional), including indications for any delay-sensitive ML-specific output post-processing to be performed, such as indications for the expected delay budget for such operations. ● Specific control signaling configurations that enable and (partially) control the operations of i) and ii) and the corresponding Model ID-based LCM.

[0066] Further examples of embodiments are described below with respect to the methods and / or apparatus described above.

[0067] Referring here to Figure 8, a flowchart illustrating the steps corresponding to various examples of the embodiment is shown. The method steps illustrated in Figure 8 can represent at least some of the method / process steps outlined above with reference to Figures 4 to 7. Furthermore, the method illustrated in Figure 8 can be applied to the UE500, 600, and / or 700 outlined above with reference to Figures 5 to 7.

[0068] In particular, as shown in Figure 8, in S810, the method includes requiring the use of ML features based on the use of an appropriate ML model for the ML features.

[0069] It should be noted that such a requirement may indicate the need to use specific ML features, such as steps 0 and / or 1, as outlined above with reference to Figures 5 through 7. Such ML features could, for example, be a beam management use case. Furthermore, the ML model may represent a transferred / received and / or selected (preferred) ML model, such as step 1, as outlined above with reference to Figures 5 through 7. The ML model may be represented and / or identified by a MID, as outlined above with reference to Figures 4 through 7.

[0070] Furthermore, in S820, the method is: - Activation of one ML model that is available and appropriate for the ML features, wherein one ML model is associated with at least one feature, or - Activation of at least one feature associated with one ML model that is available and suitable for ML features. This includes receiving an activation message for at least one of the following. The activation message indicates that one ML model and / or one feature is being activated.

[0071] It should be noted that such receipt of an activation message may represent at least part of enabling / activating, as outlined above with reference to Figures 5 to 7, for example, steps 5 and 6. Thus, an activation message may be provided by an access network element such as gNB510, 610, and / or 710, as outlined above with reference to Figures 5 to 7. Furthermore, one such ML model and / or one such function to be activated may represent a transferred / received and / or selected (preferred) ML model and / or (ML) function, as outlined above with reference to Figures 4 to 7, for example, steps 5 and 6. In addition, the term "available" may be understood as meaning that an ML model and / or function is available and ready for use by a device such as an endpoint terminal, which may be represented by UE500, 600, and / or 700, as outlined above with reference to Figures 5 to 7. Additionally, the expression "at least one function" may include, for example, FID1 and FID2 as outlined above with reference to Figure 4, and / or (ML) functions as outlined above with reference to Figures 5 through 7. Thus, one ML model may correspond to a MID as outlined above with reference to Figures 4 through 7.

[0072] Additionally, in the S830, the method includes starting the use of one ML model based on the received activation message.

[0073] It should be noted that this type of start may represent at least some of the starts outlined above with reference to Figures 5 through 7, such as steps 6 and 7.

[0074] Furthermore, in S840, the method includes reporting at least one of the following: a Key Performance Indicator (KPI) for the ML model performance of one ML model, or a Functional Performance KPI for one feature associated with one ML model.

[0075] Please note that such reporting may represent at least some of the reporting / sending of (performance) reports outlined above with reference to Figures 5 through 7, such as steps 7 and 10 in Figure 5, step 8 in Figure 6, and step 8 in Figure 7.

[0076] Furthermore, according to at least some examples of the embodiments, the method may further include configuring one of at least one features provided by a single ML model, and providing a feature configuration resulting from the configuration. The feature configuration indicates at least one of a feature performance KPI to be reported, or a single ML model performance KPI to be reported. The method may further include requesting the activation of a single ML model, or the activation of a single feature, and receiving an activation message in response to the request.

[0077] Please note that this configuration may represent the configuration outlined above, referring to Figure 5, Step 4. Furthermore, this request may represent the request outlined above, referring to Figure 5, Step 4.

[0078] Furthermore, according to various embodiments, the method may further include receiving an ML model appropriate to an ML feature and determining a feature association for the received ML model, wherein the feature association represents an association between the received ML model and at least one available feature. If the determined feature association is accepted, the method may further include providing an acceptance message indicating that the received ML model is available. If the determined feature association is rejected, the method may further include providing a rejection message indicating that the received ML model is not available.

[0079] Please note that such acceptance may represent the acceptance outlined above with reference to Figure 5, Step 1. Furthermore, such decision may represent the decision outlined above with reference to Figure 5, Step 2. Such acceptance may represent the acceptance outlined above with reference to Figure 5, Step 3. Such rejection may represent the rejection outlined above with reference to Figure 5, Step 13.

[0080] In addition, according to various examples of the embodiments, the method may further include configuring one of the at least one features that have been agreed to be associated with the received ML model, if the determined feature association is accepted, and providing a feature configuration resulting from the configuration. The feature configuration indicates at least one of the feature performance KPIs that will be reported, or the ML model performance KPIs that will be reported. The method may further include requesting at least one of the following: activation of the received ML model representing one ML model, or activation of one of the at least one features that have been agreed to be associated with the received ML model representing one feature associated with one ML model. The method may further include receiving an activation message in response to the request.

[0081] Please note that this configuration may represent the configuration outlined above, referring to Figure 5, Step 4. Furthermore, this request may represent the request outlined above, referring to Figure 5, Step 4.

[0082] Optionally, according to at least some examples of the embodiments, the method may further include receiving a signal to switch one ML model in response to reporting and deciding whether to use another ML model that is available and appropriate for the ML feature, wait for the receipt of a new ML model that is appropriate for the ML feature, or fall back; or receiving a signal to switch one feature and configuring another feature based on the determined feature association.

[0083] It should be noted that such receiving may represent receiving as outlined above with reference to Figure 5, steps 7 through 8, and steps 10 through 12. Furthermore, such deciding may represent deciding as outlined above with reference to Figure 5, step 9.

[0084] Furthermore, according to various examples of the embodiments, the method may further include selecting an ML model that is available and appropriate for the ML features, indicating the selected ML model, and registering information regarding at least the selected ML model and at least one feature associated with the selected ML model. The method may further include receiving an approval signal for at least one feature associated with the selected ML model, configuring one of the at least one features associated with the selected ML model, wherein configuring includes configuring at least one of the feature performance KPIs or ML model performance KPIs that will be reported, and receiving an activation message that the selected ML model represents one ML model.

[0085] It should be noted that such selection, indication, and registration may represent the selection, indication, and registration outlined above with reference to Figure 6, Step 1. Furthermore, such receiving may represent the receiving outlined above with reference to Figure 6, Steps 3 through 5. Furthermore, such configuring may represent the configuring outlined above with reference to Figure 6, Steps 3 through 4.

[0086] Furthermore, according to at least some examples of the embodiments, the method may further include controlling lifecycle management (LCM) operations with respect to a selected ML model.

[0087] Please note that this type of control can be described as the control outlined above, referring to Figure 6, steps 6 to 7.

[0088] Furthermore, according to various examples of the embodiments, the method may further include receiving a deactivation for a feature associated with a selected ML model in response to reporting, or receiving a deactivation for a feature associated with a selected ML model and an activation for another feature of at least one feature associated with the selected ML model in response to reporting.

[0089] Please note that this type of receiving can represent the receiving outlined above, with reference to Figure 6, steps 9 through 10.

[0090] In addition, according to various examples of embodiments, the method may further include receiving an appropriate ML model for an ML feature and at least one feature associated with the received ML model, and determining information about the received at least one feature relating to the ML feature. The method may further include providing an acknowledgment signal for both the received ML model and the received at least one feature, the acknowledgment signal indicating that both the received ML model and the received at least one feature are available. The method may further include configuring one of the received at least one feature, the configuration including configuring at least one of the feature performance KPIs or ML model performance KPIs to be reported, and receiving an activation message, where the received ML model represents one ML model and the received one feature represents one feature.

[0091] Please note that such receiving may represent receiving as outlined above with reference to Figure 7, Step 1. In addition, such deciding may represent deciding as outlined above with reference to Figure 7, Step 2. Furthermore, such providing may represent providing as outlined above with reference to Figure 7, Steps 3 to 4. Moreover, such structuring may represent structuring as outlined above with reference to Figure 7, Steps 3 to 4. Furthermore, such receiving may represent receiving as outlined above with reference to Figure 7, Step 5.

[0092] Optionally, according to at least some examples of the embodiments, the method may further include receiving a deactivation for a received function in response to reporting, or receiving a deactivation for a received function and an activation for another function of the at least one received function in response to reporting.

[0093] Please note that this type of receiving can represent the receiving described above, with reference to Figure 7, steps 9 through 10.

[0094] Furthermore, according to various examples of the embodiments, the method may further include, in response to reporting, receiving an indication that an update of the received ML model is required; receiving a deactivation for the received feature; using another ML model that is available and appropriate for the ML feature; or waiting for the receipt of a new ML model appropriate for the ML feature, and at least one feature associated with the new ML model.

[0095] Please note that such receiving may represent receiving as outlined above, with reference to Figure 7, steps 12 through 14. Furthermore, such implementation may represent implementing as outlined above, with reference to Figure 7, step 15.

[0096] Furthermore, according to various embodiments, the method may further include providing association information about the association between at least one available ML model and at least one function provided by the at least one ML model. Thus, if such association information is obtained at / for an endpoint terminal and provided to, for example, an access network element such as a gNB, the access network element (i.e., the network) may be made able to identify, i.e., ready for use, the ML model and the associated at least one available function at such an endpoint terminal. Furthermore, the network may therefore use such association information to determine whether to perform an action at such an endpoint terminal, for example, if it may be necessary to use a particular ML feature, such as updating at least one of the available ML models or available functions, or, for example, transferring / providing additional ML models and / or functions that become available at the endpoint terminal.

[0097] Furthermore, according to various examples of the embodiments, the method may further include controlling LCM operations (with respect to ML models and / or functions) including at least one of monitoring ML models and / or functions, activating ML models and / or functions, deactivating ML models and / or functions, or switching ML models and / or functions.

[0098] The solutions outlined above enable LCM using ML model identification and ML function identification. Therefore, the solutions outlined above are advantageous in that they enable more efficient and / or more secure and / or more robust and / or fault-tolerant and / or flexible and / or less complex LCM using ML model identification and ML function identification.

[0099] Referring here to Figure 9, Figure 9 is a flowchart illustrating steps corresponding to various examples of the embodiment. The method steps illustrated in Figure 9 may represent at least some of the method / processing steps outlined above with reference to Figures 4 to 7. Furthermore, the method illustrated in Figure 9 may be applied to access network elements such as gNB510, 610, and / or 710, as outlined above with reference to Figures 5 to 7.

[0100] It should be noted that similar terms / expressions used in Figures 8 and 9 may be understood similarly, and therefore their repetitive explanations can be omitted.

[0101] In particular, as shown in Figure 9, with respect to machine learning (ML) features that are required to be used based on the usage of an ML model appropriate for the ML feature, in S910, the method includes providing an activation message relating to at least one of the following: activation of one ML model that is available and appropriate for the ML feature, wherein the ML model is associated with at least one feature, or activation of one feature of the at least one feature associated with one ML model that is available and appropriate for the ML feature.

[0102] It should be noted that such provision may refer to providing / transferring as outlined above with reference to Figures 5 to 7, for example, steps 5 to 6. Furthermore, the term “available” may be understood as meaning that the ML model and / or functionality are available and ready for use in devices such as endpoint terminals, which may be represented by UE500, 600, and / or 700 as outlined above with reference to Figures 5 to 7.

[0103] Furthermore, in S920, the method includes activating one ML model and / or one function.

[0104] Please note that this activation process can represent enabling / activating the features outlined above, such as steps 5 to 6, with reference to Figures 5 to 7.

[0105] In addition, in S930, the method includes receiving a report that indicates at least one ML model performance key performance indicator (KPI) for one ML model, or a functional performance KPI for one feature associated with one ML model.

[0106] It should be noted that such receiving may represent at least some of the receiving outlined above with reference to Figures 5 to 7, such as steps 7 to 8 and 10 to 12.

[0107] Furthermore, in S940, the method includes taking action on one ML model and / or one function based on the received report.

[0108] It should be noted that performing such actions may represent at least a portion of the actions that can be derived from the above by referring to Figures 5 to 7, such as steps 7 to 14, which will be performed by the NW as enumerated in Figures 5 to 7.

[0109] Furthermore, according to at least some examples of the embodiments, the method may further include receiving a functional configuration that indicates at least one of the functional performance KPIs or ML model performance KPIs included in a report to be received, receiving a request for at least one of the activation of an ML model or the activation of a functional, and providing an activation message in response to the received request.

[0110] Please note that such receiving may represent receiving as outlined above, with reference to Figure 5, Step 4. Furthermore, such providing may represent providing as outlined above, with reference to Figure 5, Step 4.

[0111] Furthermore, according to various examples of the embodiments, the method may further include providing an appropriate ML model for an ML feature, receiving an approval message indicating that the provided ML model is available if an association between the provided ML model and at least one available feature is accepted, and receiving a rejection message indicating that the provided ML model is not available if the association is rejected.

[0112] Please note that such provision may represent the provision outlined above with reference to Figure 5, Step 1. Furthermore, such receipt may represent the receipt outlined above with reference to Figure 5, Step 3 and 13.

[0113] In addition, according to various examples of embodiments, the method may further include receiving a feature configuration resulting from the configuration of one of the at least one features agreed to be associated with the provided ML model, where the association is accepted, and the feature configuration indicates at least one of the feature performance KPIs included in the report to be received, or the ML model performance KPIs included in the report to be received; receiving a request for at least one of the activation of the provided ML model representing one ML model, or the activation of one of the at least one features agreed to be associated with the provided ML model representing one feature associated with one ML model; and providing an activation message in response to the received request.

[0114] Please note that this type of receiving can represent the receiving described above, referring to Figure 5, Step 4.

[0115] Optionally, according to at least some examples of the embodiments, the method is - Controlling the lifecycle management (LCM) behavior for a single ML model, or controlling the LCM behavior for a single function. It may further include at least one of the following: - Taking action means Based on controlled LCM behavior for a single ML model, it is necessary to decide whether or not to switch a single function, and if it is decided to switch a single function, to provide signaling for switching a single ML model, or Based on controlled LCM operation for a single function, determine whether or not to switch that function, and if so, provide signaling for switching that function. It may further include at least one of the following.

[0116] It should be noted that such LCM control can represent the LCM control outlined above with reference to Figure 5, steps 7 to 8, and steps 10 to 12. Furthermore, such a decision can represent the decision outlined above with reference to Figure 5, steps 7 to 8.

[0117] Furthermore, according to various examples of the embodiments, the method may further include receiving an indication of a selected ML model that is available and appropriate for the ML features; receiving a registration of information relating to at least a selected ML model and at least one feature associated with the selected ML model; determining the information received by the registration; providing an approval signal for at least one feature associated with the selected ML model based on the determination; and providing an activation message, wherein the selected ML model represents one ML model.

[0118] It should be noted that such receiving may represent receiving as outlined above with reference to Figure 6, Step 1. Furthermore, such deciding may represent making a decision as outlined above with reference to Figure 6, Step 2. Furthermore, such providing may represent providing as outlined above with reference to Figure 6, Steps 3 through 5.

[0119] Furthermore, according to at least some examples of the embodiments, performing an action may further include providing deactivation for a function associated with a selected ML model, or providing deactivation for a function associated with a selected ML model, and providing activation for another function of at least one function associated with a selected ML model.

[0120] Please note that providing in this manner may represent providing as outlined above, with reference to Figure 6, steps 9 through 10.

[0121] Furthermore, according to various examples of the embodiments, the method may further include providing an ML model appropriate to an ML feature and at least one function associated with the provided ML model, receiving an acknowledgment signal for both the provided ML model and the provided at least one function, the acknowledgment signal indicating that both the provided ML model and the provided at least one function are available, and providing an activation message, where the provided ML model represents one ML model and the provided one function represents one function.

[0122] Please note that such giving and receiving may represent the giving and receiving outlined above, with reference to Figure 7, Steps 1 through 5.

[0123] In addition, according to various examples of embodiments, the method may further include controlling LCM behavior relating to a provided ML model and at least one of the provided functions, and performing an action includes, based on the controlled LCM behavior, providing deactivation for a provided function, or providing deactivation for a provided function and providing activation for another of the at least one of the provided functions.

[0124] Please note that this type of provision may represent the provision outlined above, with reference to Figure 7 and steps 9 through 10.

[0125] Optionally, according to at least some examples of the embodiments, the method may further include controlling LCM behavior relating to at least a provided ML model and one provided function, and the action performed includes, at least one of, triggering a switch of one ML model by providing an indication that an update of the provided ML model is required based on the controlled LCM behavior, providing deactivation for the provided function, and providing deactivation for the provided ML model.

[0126] It should be noted that such triggering and providing can represent the triggering and providing outlined above, with reference to Figure 7, steps 12 through 14.

[0127] Furthermore, according to various embodiments, the method may further include receiving association information as outlined above with reference to Figure 8. Thus, if such association information is obtained at / for an endpoint terminal and received by, for example, an access network element such as a gNB, the method may further include identifying ML models and at least one associated feature that are available, i.e., ready for use, at such an endpoint terminal. Furthermore, the method may further include using such association information to determine whether actions should be taken at such an endpoint terminal, such as updating at least one of the available ML models or available features, or, for example, transferring / providing additional ML models and / or features that become available at the endpoint terminal, if it may be necessary to use a particular ML feature.

[0128] Therefore, the method may further include determining, based on the received association information, that there are no ML models and / or functions available to use a particular ML feature, and providing a suitable ML model and / or function to use the particular ML feature.

[0129] Furthermore, according to various examples of the embodiments, the method may further include the control of LCM operations (with respect to ML models and / or functions) including at least one of monitoring ML models and / or functions, activating ML models and / or functions, deactivating ML models and / or functions, or switching ML models and / or functions.

[0130] The solutions outlined above enable LCM using ML model identification and ML function identification. Therefore, the solutions outlined above are advantageous in that they enable more efficient and / or more secure and / or more robust and / or fault-tolerant and / or flexible and / or less complex LCM using ML model identification and ML function identification.

[0131] Referring here to Figure 10, Figure 10 shows a block diagram illustrating the apparatus according to various examples of the embodiment.

[0132] Specifically, Figure 10 shows a block diagram illustrating device 1000, which may represent an endpoint terminal such as a UE, as outlined above with reference to various examples of embodiments in Figures 5 to 7, which may be involved in LCM using ML model identification and ML function identification. Furthermore, even if references to endpoint terminals are made, the endpoint terminal may also be another device or function having similar tasks, such as a chipset, chip, module, or application, which may be part of a network element, or attached to a network element as a separate element, or similar. It should be understood that each block and any combination thereof may be implemented by various means or combinations thereof, such as hardware, software, firmware, one or more processors and / or circuit equipment.

[0133] The apparatus 1000 shown in Figure 10 may include processing circuit equipment, processing functions, control units, or processor 1010, such as a CPU or similar, suitable for enabling LCM using ML model identification and ML function identification. The processor 1010 may include one or more processing parts or functions dedicated to specific processing, as described below, or the processing may be performed by a single processor or processing function. The parts for performing such specific processing may also be provided as separate elements or in one or more further processors, processing functions, or processing parts, such as in one physical processor like a CPU, or in one or more physical or virtual entities. Reference numerals 1031 and 1032 represent input / output (I / O) units or functions (interfaces) connected to the processor or processing function 1010. The I / O units 1031 and 1032 may be combined units including communication equipment to several entities / elements, or they may include a distributed structure with multiple interfaces that differ among entities / elements. Reference numeral 1020 represents memory that can be used, for example, to store data and programs to be executed by the processor or processing function 1010, and / or as working storage for the processor or processing function 1010. Note that memory 1020 can be realized by using one or more memory portions of the same or different types of memory, but may also represent external memory, such as an external database provided on a cloud server.

[0134] The processor or processing function 1010 is configured to perform the processing related to the processing described above. In particular, the processor or processing circuit device or function 1010 includes one or more of the following sub-parts: Sub-part 1011 is a request part that can be used as a part for using the ML model. Part 1011 may be configured to perform the processing according to S810 in Figure 8. Furthermore, sub-part 1012 is a receiving part that can be used as a part for receiving activation messages. Part 1012 may be configured to perform the processing according to S820 in Figure 8. In addition, sub-part 1013 is a start part that can be used as a part for starting the use of the ML model. Part 1013 may be configured to perform the processing according to S830 in Figure 8. Furthermore, sub-part 1014 is a reporting part that can be used as a part for reporting KPIs. Part 1014 may be configured to perform the processing according to S840 in Figure 8.

[0135] In various examples of the embodiments, the following can be further considered with respect to the apparatus 1000 shown in Figure 10.

[0136] According to various examples of the embodiment, the apparatus 1000 includes at least one processor, and when executed by at least one processor, at least The requirement to use ML features based on the use of an appropriate ML model for machine learning (ML) features, Activation of one ML model available in device 1000 and appropriate for ML features, wherein one ML model is associated with at least one function, or Activation of one of at least one functions associated with one ML model that is available in the device and appropriate for ML features. Receiving an activation message regarding at least one of the following: An activation message is received to indicate that one ML model and / or one feature is being activated. Starting the use of one ML model based on the received activation message, and Key performance indicators (KPIs) for a single ML model, or Functional performance KPI for one feature associated with one ML model Report at least one of the following It may also include at least one memory that stores instructions for causing the device 1000 to perform the action.

[0137] According to various embodiments, the device 1000 may further configure one of at least one functions provided by one ML model, provide a function configuration resulting from the configuration, the function configuration indicating at least one of a function performance KPI to be reported or an ML model performance KPI to be reported, request at least one of the activation of one ML model or an activation of one function, and receive an activation message in response to the request.

[0138] According to various embodiments, the device 1000 may further perform the following: receive an ML model appropriate to an ML feature; determine a functional association for the received ML model, wherein the functional association represents an association between the received ML model and at least one available functional; provide an approval message indicating that the received ML model is available if the device 1000 is required to determine the functional association and the device 1000 accepts the determined functional association; and provide a rejection message indicating that the received ML model is not available if the device 1000 is required to determine the functional association and the device 1000 rejects the determined functional association.

[0139] According to various embodiments, the device 1000 may, if the device 1000 accepts a determined functional association, further perform the following: configure one of the at least one functions that have been accepted to be associated with the received ML model; provide a functional configuration resulting from the configuration, wherein the functional configuration indicates at least one of the functional performance KPIs that will be reported or the ML model performance KPIs that will be reported; request at least one of the following: activation of the received ML model representing one ML model, or activation of one of the at least one functions that have been accepted to be associated with the received ML model representing one function associated with one ML model; and receive an activation message in response to the request.

[0140] According to various embodiments, the device 1000 may, in response to a report, be prompted to receive a signal to switch one ML model and decide whether to use another ML model available in the device and suitable for the ML feature, wait for the arrival of a new ML model suitable for the ML feature, or fall back; or to receive a signal to switch one function and configure another function based on the determined function association.

[0141] According to various embodiments, the device 1000 may further configure, which includes configuring one of the at least one of the function performance KPIs or ML model performance KPIs that will be reported, which includes configuring one of the at least one of the at least one of the functions that will be reported, which includes configuring one of the function performance KPIs or ML model performance KPIs that will be reported, which includes configuring include configuring one of the function performance KPIs or ML model performance KPIs that will be reported, which include configuring one of the function performance KPIs or ML model performance KPIs

[0142] According to various embodiments, the device 1000 may be further configured to control lifecycle management (LCM) operations with respect to a selected ML model.

[0143] According to various embodiments, the device 1000 may, in response to being made to report, receive a deactivation for a function associated with a selected ML model, or receive a deactivation for a function associated with a selected ML model and further receive an activation for another function of at least one function associated with the selected ML model.

[0144] According to various embodiments, the device 1000 may further perform the following: receive an ML model appropriate to an ML feature and at least one function associated with the received ML model; determine information about the received at least one function relating to the ML feature; provide an acknowledgment signal for both the received ML model and the received at least one function, the acknowledgment signal indicating that both the received ML model and the received at least one function are available in the device; configure one of the received at least one function, which includes configuring at least one of the function performance KPIs or ML model performance KPIs that will be reported; and receive an activation message, where the received ML model represents one ML model and the received one function represents one function.

[0145] According to various embodiments, the device 1000 may, in response to being made to report, receive a deactivation for a received function, or receive a deactivation for a received function and further receive an activation for another function of the at least one received function.

[0146] According to various embodiments, the device 1000 may, in response to being prompted to report, receive an indication that an update of the received ML model is required, receive a deactivation for the received function, receive a deactivation for the received ML model, and further perform one of the following: use another ML model available in the device and appropriate for the ML feature, or wait for the receipt of a new ML model appropriate for the ML feature, and at least one function associated with the new ML model.

[0147] Referring here to Figure 11, Figure 11 shows a block diagram illustrating the apparatus according to various examples of the embodiment.

[0148] Specifically, Figure 11 shows an illustrative block diagram of an apparatus, which may represent an access network element such as a gNB, as outlined above with reference to Figures 5 to 7, for example, with reference to various examples of embodiments, which may be involved in LCM using ML model identification and ML function identification. Furthermore, even where an access network element is mentioned, the access network element may also be another device or function having a similar task, such as a chipset, chip, module, or application, which may be part of the network element, or attached to the network element as a separate element, or similar. It should be understood that each block and any combination thereof may be implemented by various means or combinations thereof, such as hardware, software, firmware, one or more processors and / or circuit equipment.

[0149] The apparatus 1100 shown in Figure 11 may include processing circuit equipment, processing functions, control units, or processor 1110, such as a CPU or similar, suitable for enabling LCM using ML model identification and ML function identification. The processor 1110 may include one or more processing parts or functions dedicated to specific processing, as described below, or the processing may be performed by a single processor or processing function. The parts for performing such specific processing may also be provided as separate elements or in one or more further processors, processing functions, or processing parts, such as in one physical processor like a CPU, or in one or more physical or virtual entities. Reference numerals 1131 and 1132 represent input / output (I / O) units or functions (interfaces) connected to the processor or processing function 1110. The I / O units 1131 and 1132 may be combined units including communication equipment to several entities / elements, or they may include a distributed structure with multiple interfaces that differ for each entity / element. Reference numeral 1120 represents memory that can be used, for example, to store data and programs to be executed by the processor or processing function 1110, and / or as working storage for the processor or processing function 1110. Note that memory 1120 can be realized by using one or more memory portions of the same or different types of memory, but may also represent external memory, such as an external database provided on a cloud server.

[0150] The processor or processing function 1110 is configured to perform the processing related to the processing described above. In particular, the processor or processing circuit device or function 1110 includes one or more of the following sub-parts: Sub-part 1111 is a providing part that can be used as a part for providing an activation message. Part 1111 may be configured to perform the processing according to S910 in Figure 9. Furthermore, sub-part 1112 is an activation part that can be used as a part for activating the ML model and / or function. Part 1112 may be configured to perform the processing according to S920 in Figure 9. Moreover, sub-part 1113 is a receiving part that can be used as a part for receiving reports. Part 1113 may be configured to perform the processing according to S930 in Figure 9. Moreover, sub-part 1114 is an action-executing part that can be used as a part for performing actions related to the ML model and / or function. Part 1114 may be configured to perform the processing according to S940 in Figure 9.

[0151] In various examples of the embodiments, the following can be further considered with respect to the apparatus 1100 shown in Figure 11.

[0152] According to various embodiments, the device 1100 may include at least one processor and at least one memory storing instructions that cause the device 1100 to perform the following actions: providing an activation message for at least one of the following with respect to machine learning (ML) features that are required to be used by another device based on usage of an ML model appropriate for the ML features, where the ML model is associated with at least one function; or activating one of the at least one functions associated with the ML model that is available in another device and appropriate for the ML features; causing the ML model and / or one function to be activated; receiving a report that indicates at least one ML model key performance indicator (KPI) for the ML model or a function performance KPI for the function associated with the ML model; and taking action with respect to the ML model and / or one function based on the received report.

[0153] According to various embodiments, the device 1100 may further receive a functional configuration that indicates at least one of the functional performance KPIs or ML model performance KPIs included in a report to be received, receive a request for at least one of the activation of an ML model or the activation of a function, and provide an activation message in response to the received request.

[0154] According to various embodiments, the device 1100 may further provide an ML model suitable for an ML feature, and receive an approval message indicating that the provided ML model is available in another device if an association between the provided ML model and at least one available feature is accepted, and receive a rejection message indicating that the provided ML model is not available in another device if the association is rejected.

[0155] According to various embodiments, the device 1100 may further receive a functional configuration resulting from the configuration of one of the at least one functions agreed to be associated with the provided ML model, provided that the association is accepted, wherein the functional configuration indicates at least one of the functional performance KPIs included in the report to be received, or the ML model performance KPIs included in the report to be received; receive a request for at least one of the activation of the provided ML model representing one ML model, or the activation of one of the at least one functions agreed to be associated with the provided ML model representing one function associated with one ML model; and provide an activation message in response to the received request.

[0156] According to various embodiments, the device 1100 may further perform at least one of controlling lifecycle management (LCM) operations for one ML model or controlling LCM operations for one function, and the device 1100 performing an action may further include the device 1100 determining whether or not to switch a function based on controlled LCM operations for one ML model, and if it decides to switch a function, providing signaling for switching one ML model, or determining whether or not to switch a function based on controlled LCM operations for one function, and if it decides to switch a function, providing signaling for switching a function.

[0157] According to various embodiments, the device 1100 may further receive an indication of a selected ML model available in another device and appropriate for the ML features; receive a registration of information relating to at least a selected ML model and at least one function associated with the selected ML model; determine the information received by the registration; provide an acknowledgment signal for at least one function associated with the selected ML model based on the determined information; and provide an activation message, wherein the selected ML model represents one ML model.

[0158] According to various embodiments, the device 1100 being made to perform an action may further include the device 1100 being made to perform at least one of the following: providing deactivation for a function associated with a selected ML model, or providing deactivation for a function associated with a selected ML model, and providing activation for another function of at least one function associated with a selected ML model.

[0159] According to various embodiments, the device 1100 may further provide an ML model appropriate to an ML feature and at least one function associated with the provided ML model, and receive an acknowledgment signal for both the provided ML model and the provided at least one function, the acknowledgment signal indicating that both the provided ML model and the provided at least one function are available in another device, and provide an activation message, where the provided ML model represents one ML model and the provided one function represents one function.

[0160] According to various embodiments, the device 1100 may further control LCM operations relating to at least one of the provided ML models and one of the provided functions, and the device 1100 performing an action may further include the device 1100 performing at least one of the following based on the controlled LCM operation: providing deactivation for a provided function, or providing deactivation for a provided function and providing activation for another of the at least one of the provided functions.

[0161] According to various embodiments, the device 1100 may further control LCM operations relating to at least one of a provided ML model and one provided function, and the device 1100 performing an action may further include the device 1100 triggering a switch of one ML model by providing an indication that an update of the provided ML model is required based on the controlled LCM operation, providing deactivation for a provided function, and providing deactivation for a provided ML model.

[0162] It should be noted that apparatuses 1000 and 1100 as outlined above with reference to Figures 10 and 11 may include further / additional sub-parts that enable apparatuses 1000 and 1100 to carry out methods / method steps outlined above with reference to Figures 5 to 7, and / or Figures 8 and 9.

[0163] Furthermore, according to various embodiments, a computer program product for a computer is provided which, when the computer program product is executed on the computer, includes a portion of software code for performing any of the steps of any of the appended claims 1 to 9 or any of the appended claims 10 to 15, and optionally the computer program product includes a computer-readable medium on which the software code portion is stored, and / or which the computer program product can be directly loaded into the internal memory of the computer, and / or which can be transmitted over a network by at least one of the upload, download, and push procedures.

[0164] Please understand the following: - The access technology used to transfer traffic between entities in a communication network may be any suitable current or future technology, such as WLAN (Wireless Local Access Network), WiMAX (Worldwide Interoperability for Microwave Access), LTE, LTE-A, 5G, Bluetooth, infrared, and similar technologies. Additionally, embodiments may also apply wired technologies, such as IP-based access technologies like cable networks or fixed lines. - Embodiments suitable for execution as software code or a part thereof, and executed using a processor or processing function, are independent of software code and can be specified using any known or future-developed programming language, such as high-level programming languages ​​like Objective-C, C, C++, C#, Java, Python, Javascript, or other scripting languages, or low-level programming languages ​​like machine code or assembler. - Implementation of the embodiments is hardware independent and may be carried out using any known or future-developed hardware technology, or a hybrid thereof, such as a microprocessor or CPU (Central Processing Unit), MOS (Metal Oxide Semiconductor), CMOS (Complementary MOS), BiMOS (Bipolar MOS), BiCMOS (Bipolar CMOS), ECL (Emitter-Coupled Logic), and / or TTL (Transistor-Transistor Logic). - Embodiments may be implemented as individual devices, apparatus, units, means, or functions, or in a distributed manner, for example, one or more processors or processing functions may be used or shared during processing, or one or more processing sections or processing parts may be used and shared during processing, or one or more physical processors or two or more physical processors may be used to implement one or more processing parts dedicated to the specific processing described. - The device may be implemented by a semiconductor chip, a chipset, or a (hardware) module including such a chip or chipset. - Embodiments may also be implemented as any combination of hardware and software, such as ASIC (Application-Specific Integrated Circuit) components, FPGA (Field-Programmable Gate Array) or CPLD (Complex-Programmable Logic Device) components, or DSP (Digital Signal Processor) components. - The embodiments may also be implemented as a computer program product including a computer-readable medium in which computer-readable program code is materialized, the computer-readable program code being adapted to perform a process as described in the embodiments, and the computer-readable medium may be a non-temporary medium.

[0165] While this disclosure has been described herein in relation to certain embodiments, it is not limited to these embodiments, and various modifications can be made to them.

Claims

1. Step (S810) requires the use of the machine learning (ML) features based on the use of an appropriate ML model for the machine learning (ML) features, Activation of an available ML model appropriate to the ML features, wherein the ML model is associated with at least one function, or Activation of one of at least one features associated with one ML model that is available and appropriate for the aforementioned ML features. The step (S820) of receiving an activation message relating to at least one of the following: Step (S820) indicates that the activation message indicates that the one ML model and / or the one function is activated. Step (S830) of starting to use the one ML model based on the activation message received, The key performance indicators (KPIs) of the aforementioned ML model, or Functional performance KPI of the one function associated with the one ML model The step of reporting at least one of the following (S840) Methods that include...

2. A step comprising constituting one of the at least one functions provided by the aforementioned ML model, A step of providing a functional configuration resulting from the above-mentioned steps, The functional configuration includes the step of indicating at least one of the functional performance KPIs or the ML model performance KPIs that will be reported, The activation of the aforementioned ML model, or Activation of the aforementioned one function A step to request at least one of the following, The steps include receiving the activation message in response to the aforementioned request step and The method according to claim 1, further comprising:

3. The steps include receiving an ML model appropriate to the aforementioned ML features, A step of determining a functional association for the received ML model, wherein the functional association represents an association between the received ML model and at least one available functional. If the determined function association is accepted, the step is to provide an approval message indicating that the received ML model is available, If the determined function association is rejected, the step is to provide a rejection message indicating that the received ML model is unavailable. The method according to claim 1, further comprising:

4. If the aforementioned determined function association is accepted, A step of constituting one of the at least one functions that has been agreed to be associated with the received ML model, A step of providing a functional configuration resulting from the above-mentioned steps, The functional configuration includes the step of indicating at least one of the functional performance KPIs or the ML model performance KPIs that will be reported, The activation of the received ML model representing the aforementioned one ML model, or Activation of one of the at least one functions that have been agreed to be associated with the received ML model, which represents the one function associated with the one ML model. A step to request at least one of the following, The steps include receiving the activation message in response to the aforementioned request, and The method according to claim 3, further comprising:

5. In response to the reporting step described above, The steps include receiving a signal to switch between the aforementioned ML models, and The step of deciding whether to use another ML model that is available and appropriate for the ML features, to wait for the arrival of a new ML model that is appropriate for the ML features, or to fall back, or The steps include receiving a signal to switch one of the aforementioned functions, and Based on the determined function association, the step of configuring another function, or A step of receiving deactivation for the function associated with the selected ML model, or The steps of receiving deactivation for the function associated with the selected ML model, and receiving activation for another function of the at least one function associated with the selected ML model. The method according to any one of claims 1 to 4, further comprising at least one of the following.

6. The steps include receiving an ML model appropriate to the ML features and at least one function associated with the received ML model, A step of determining information about the at least one function received regarding the ML feature, A step of providing an approval signal for both the received ML model and the received at least one function, The approval signal indicates that both the received ML model and at least one received function are available. A step of constituting one of the at least one functions received, The steps comprising the above-mentioned steps include a step that constitutes at least one of the functional performance KPIs or the ML model performance KPIs that will be reported, A step of receiving the activation message, wherein the received ML model represents the one ML model, and the received one function represents the one function. The method according to claim 1, further comprising:

7. In response to the reporting step, a step of receiving deactivation for the received function, or In response to the reporting step, the steps include receiving a deactivation for the received function and receiving an activation for another function among the at least one of the received functions. The method according to claim 1 or 6, further comprising:

8. In response to the reporting step, the process includes receiving an indication that an update to the received ML model is required, The steps include receiving the deactivation for the aforementioned received function, The steps include receiving the deactivation for the ML model received, The step of using another ML model that is available and suitable for the aforementioned ML features, or Steps include waiting to receive a new ML model suitable for the aforementioned ML features, and at least one function associated with the new ML model. The step of implementing one of them The method according to any one of claims 1, 6, or 7, further comprising:

9. With respect to the machine learning (ML) features, which are required to be used based on the appropriate use of ML models, Activation of an available ML model appropriate to the ML features, wherein the ML model is associated with at least one function, or Activation of one of at least one features associated with one ML model that is available and appropriate for the aforementioned ML features. The steps include providing an activation message relating to at least one of the following (S910), Step (S920) of activating the aforementioned ML model and / or the aforementioned one function, at least, The key performance indicators (KPIs) of the aforementioned ML model, or Functional performance KPI of the one function associated with the one ML model The step of receiving a report that indicates, Step (S940) of performing an action relating to the one ML model and / or the one function based on the received report. Methods that include...

10. The steps include receiving a functional configuration that indicates at least one of the functional performance KPIs included in the report to be received, or the ML model performance KPIs included in the report to be received, The activation of the aforementioned ML model, or Activation of the aforementioned one function A step to receive a request for at least one of the following, The steps of providing the activation message in response to the received request and The method according to claim 9, further comprising:

11. The steps of providing an ML model appropriate to the aforementioned ML features, and If the association between the provided ML model and at least one available function is accepted, the step is to receive an approval message indicating that the provided ML model is available. If the association is rejected, the step involves receiving a rejection message indicating that the provided ML model is unavailable. The method according to claim 9, further comprising:

12. If the aforementioned association is accepted, A step of receiving a functional configuration resulting from the configuration of one of the at least one functions that has been agreed to be associated with the provided ML model, The functional configuration includes the step of indicating at least one of the functional performance KPIs included in the report to be received, or the ML model performance KPIs included in the report to be received. Activation of the provided ML model representing the one ML model, or Activation of one of the at least one functions that have been agreed to be associated with the provided ML model, which represents the one function associated with the one ML model. A step to receive a request for at least one of the following, The steps of providing the activation message in response to the received request and It further includes, The method described above is A step of controlling the lifecycle management (LCM) operation for one ML model, or A step of controlling the LCM operation related to the aforementioned one function. It further includes at least one of the following: The step of performing the aforementioned action is A step of determining whether or not to switch the one function based on the controlled LCM operation with respect to the one ML model, and If it is decided to switch one of the aforementioned functions, the step is to provide signaling for switching one of the aforementioned ML models. or A step of deciding whether or not to switch the function based on the controlled LCM operation with respect to the function, and If it is decided to switch one of the aforementioned functions, the step of providing signaling for switching one of the aforementioned functions. The method according to claim 11, comprising at least one of the following.

13. The steps include providing an ML model suitable for the ML features and at least one function associated with the provided ML model, A step of receiving an approval signal for both the provided ML model and the provided at least one function, The approval signal indicates that both the provided ML model and at least one of the provided functions are available. A step of providing the activation message, wherein the provided ML model represents the one ML model, and the provided function represents the one function. The method according to claim 9, further comprising:

14. A step of controlling LCM operation relating to the provided ML model and one of the provided functions. It further includes, The step of performing the action is based on the controlled LCM operation, A step of providing deactivation for the aforementioned provided function, or A step of providing deactivation for the provided function, and a step of providing activation for another function among the at least one provided function. The method according to claim 9 or 13, comprising at least one of the following.

15. A step of controlling LCM operation relating to the provided ML model and one of the provided functions. It further includes, The step of performing the action is based on the controlled LCM operation, A step of triggering a switch to the aforementioned ML model by providing an indication that an update to the aforementioned ML model is required, A step of providing deactivation for the aforementioned provided function, A step of providing deactivation for the aforementioned ML model and The method according to any one of claims 9, 13, or 14, comprising at least one of the following.

16. At least one processor, The memory contains at least one memory that stores instructions, when executed by the at least one processor, to cause the device (1000) to perform the method according to at least one of claims 1 to 8, or to cause the device (1100) to perform the method according to at least one of claims 9 to 15. A device (1000; 1100) equipped with the following.

Images