Artificial intelligence / machine learning model identification for model description

Abstract

A user equipment (UE) receives, from a network, information for identifying an artificial intelligence / machine learning (AIML) model or AIML functionality. The UE identifies an AIML model or AIML func

Description

BACKGROUND Field

[0001] Certain examples of the present disclosure provide approaches for the use of Artificial Intelligence / Machine Learning (AI / ML) model identification for model description. For example, certain examples of the present disclosure provide methods, apparatus and systems for the use of Artificial Intelligence / Machine Learning (AI / ML) model identification for model description in 3rd Generation Partnership Project (3GPP) networks such as 5th Generation (5G) and / or 6th Generation (6G) networks. Description of Related Art

[0002] The content of the following documents is referred to below and / or their content provides background information and context that the following disclosure should be considered in view of: [1] 3GPP RP-213599, Study on Artificial Intelligence (Al) / Machine Learning (ML) for NR Air Interface. [2] RAN2#121bis-e meeting Chairman notes [3] R2-2304195 Report of [AT121bis-e]

[014] [AIML18] Model ID (incl meta data) progress [4] RAN 1#121 meeting Chairman notes (Note: the example versions shown for each document are non-limiting, other versions of the documents may be considered also)

[0003] Wireless or mobile (cellular) communications networks in which a mobile terminal (e.g., user equipment (UE), such as a mobile handset) communicates via a radio link with a network of base stations, or other wireless access points or nodes, have undergone rapid development through a number of generations. The 3rd Generation Partnership Project (3GPP) design, specify and standardise technologies for mobile wireless communication networks. Fourth Generation (4G) and Fifth Generation (5G) systems are now widely deployed, and development of Sixth Generation (6G) Systems is in progress.

[0004] 3GPP standards for 4G systems include an Evolved Packet Core (EPC) and an Enhanced-UTRAN (E-UTRAN: an Enhanced Universal Terrestrial Radio Access Network). The E-UTRAN uses Long Term Evolution (LTE) radio technology. LTE is commonly used to refer to the whole system including both the EPC and the E-UTRAN, and LTE is used in this sense in the remainder of this document. LTE should also be taken to include LTE enhancements such as LTE Advanced and LTE Pro, which offer enhanced data rates compared to LTE.

[0005] In 5G systems a new air interface has been developed, which may be referred to as 5G New Radio (5G NR) or simply NR. NR is designed to support the wide variety of services and use case scenarios envisaged for 5G networks, though builds upon established LTE technologies. New frameworks and architectures are also being developed as part of 5G networks in order to increase the range of functionality and use cases available through 5G networks. New frameworks and architectures are also being developed for 6th Generation (6G) networks. Artificial Intelligence (AI)ZMachine Learning (ML) in 3GPP Networks

[0006] 3GPP has also started studying the benefits of introducing Artificial Intelligence (Al) / Machine Learning (ML) solutions to communications networks, for example, enhancement of management and orchestration, performance, resource allocation, in addition to reduction of complexity and overhead in the network [1],

[0007] AI / ML for NR Air Interface is currently a study item in 3GPP working groups (WGs) RAN1 and RAN2. Both WGs have discussed the use of model ID in association with different AI / ML life-cycle-management (LCM) purposes and model transfer / delivery. The following is a list of latest agreements in RAN2 on model ID. [2] [3] Model ID can be used to identify model or models for the following LCM purposes: model selection / activation / deactivation / switching (or identification, if that will be supported as a separate step), (e.g. for so called “model ID based LCM”). If model transfer / delivery is supported, model ID can be used for model transfer / delivery LCM purpose. How to achieve globality of the Model ID is for further study (FFS). Initial discussion in RAN2: the following global unique model ID definition directions can be considered as a starting point: Direction 1: Pre-defined / hard-coded global unique model ID Directions: Assigned global unique model ID via specific ID management node. Note: Other global unique model ID definition is not precluded. Model ID structure, if any, is FFS.

[0008] The following is a list of latest agreements in RAN1 on model ID: [4] For AI / ML model identification and model-1 D-based LCM of UE-side models and / or UE-part of two-sided models: Model-1 D-based LCM operates based on identified models, where a model may be associated with specific configurations / conditions associated with UE capability of an AI / ML-enabled Feature / FG and additional conditions (e.g., scenarios, sites, and datasets) as determined / identified between UE-side and NW-side. FFS: Which aspects should be considered as additional conditions, and how to include them into model description information during model identification will be discussed in each sub-use-case agenda.

[0009] However, as shown above, the model ID structure, model description and relation to model identification remain as open points (FFS) that require further study and a suitable solution. SUMMARY

[0010] It is an aim of certain examples of the present disclosure to address, solve and / or mitigate, at least partly, at least one of the problems and / or disadvantages associated with the related art, for example at least one of the problems and / or disadvantages described herein. It is an aim of certain examples of the present disclosure to provide at least one advantage over the related art, for example at least one of the advantages described herein.

[0011] In particular, the present disclosure provides a development of AI / ML model IDs to also carry meta data about the models. This will be useful to implement in 3GPP now, before the number of AI / ML models in RAN starts to grow exponentially. The AI / ML model ID can be formulated to reflect certain properties of the model and / or certain relationships one model has with other models. By using the AI / ML model ID to carry such meta data, it will be easier for the network to identify model properties and links amongst each other models, without resorting to using a model catalogue, which will consume additional resources.

[0012] In accordance with a first aspect of the present disclosure, there is provided a method for AI / ML management in a wireless communications system comprising a UE and a network, the method comprising: receiving, at the UE from the network, information for identifying an AI / ML model or an AI / ML functionality; identifying, by the UE, at least one AI / ML model or AI / ML functionality based on the received information; performing, by the UE, an operation based on the identified at least one AI / ML model or AI / ML functionality, wherein the information for identifying an AI / ML model or an AI / ML functionality includes an ID associated with the AI / ML model or AI / ML functionality and meta data associated with the AI / ML model or AI / ML functionality.

[0013] In accordance with an example of the present disclosure, the operation includes reporting, by the UE to the network, information on the identified at least one AI / ML model or AI / ML functionality.

[0014] In accordance with an example of the present disclosure, the operation includes one or more of activating, deactivating, selecting, and / or switching the identified at least one AI / ML model or AI / ML functionality.

[0015] In accordance with an example of the present disclosure, the identifying includes identifying at least one AI / ML model or AI / ML functionality that belongs to a category indicated by the information for identifying an AI / ML model or an AI / ML functionality.

[0016] In accordance with an example of the present disclosure, the information for identifying an AI / ML model or an AI / ML functionality includes a first functionality and the identifying includes identifying at least one AI / ML model that provides the first functionality.

[0017] In accordance with an example of the present disclosure, the meta data includes information indicating or associated with one or more of: a category of the AI / ML model or AI / ML functionality; a similar AI / ML model or AI / ML functionality; a linked AI / ML model or AI / ML functionality; a feature of the AI / ML model or AI / ML functionality; a dataset associated with the AI / ML model or AI / ML functionality; a type of the AI / ML model or AI / ML functionality; an accuracy of the AI / ML model or AI / ML functionality; a reliability of the AI / ML model or AI / ML functionality; an age of the AI / ML model or AI / ML functionality; an expiry of the AI / ML model or AI / ML functionality; a security of the AI / ML model or AI / ML functionality; and an applicability of the AI / ML model or AI / ML functionality.

[0018] In accordance with an example of the present disclosure, the ID and meta data are included in one or more fields or headers appended to or included in a model ID data structure.

[0019] In accordance with an example of the present disclosure, the operation includes updating the ID and / or meta data associated with the AI / ML model or AI / ML functionality.

[0020] In accordance with an example of the present disclosure, the method further includes, receiving at the UE from the network, information configuring the mapping of the information for identifying an AI / ML model or an AI / ML functionality.

[0021] In accordance with an example of the present disclosure, the operation includes configuring, triggering, and / or performing data collection for the identified at least one AI / ML model or AI / ML functionality.

[0022] In accordance with an example of the present disclosure, the information for identifying an AI / ML model or an AI / ML functionality is received from a radio access network (RAN) of the wireless communications system, a network data analytics function (NWDAF) of the wireless communications systems, and / or a core network of the wireless communications system.

[0023] In accordance with an example of the present disclosure, the operation includes determining whether the identified at least one AI / ML model or an AI / ML functionality can be trained using similar data sets.

[0024] In accordance with an example of the present disclosure, the method further comprises updating by the network an ID and / or meta data associated with an AI / ML model or AI / ML functionality and providing an indication of the updated ID and / or meta data to the UE.

[0025] In accordance with an example of the present disclosure, the AI / ML models and AI / ML functionalities are network-side, UE-side, and / or two-sided models / functionalities.

[0026] In accordance with an example of the present disclosure, the wireless communications system is a 3GPP compliant wireless communications system.

[0027] In accordance with a second aspect of the present disclosure, there is provided a wireless communications system comprising a UE and a network, wherein the wireless communications system is configured to implement the first aspect of the present disclosure and any of the associated examples.

[0028] In accordance with a third aspect of the present disclosure, there is provided a method for a user equipment in a wireless communications system comprising the UE and a network, the method comprising: receiving, from the network, information for identifying an AI / ML model or an AI / ML functionality; identifying at least one AI / ML model or AI / ML functionality based on the received information; performing an operation based on the identified at least one AI / ML model or AI / ML functionality, wherein the information for identifying an AI / ML model or an AI / ML functionality includes an ID associated with the AI / ML model or AI / ML functionality and meta data associated with the AI / ML model or AI / ML functionality.

[0029] In accordance with a fourth aspect of the present disclosure, there is provided a user equipment (UE) for use in a wireless communication system comprises the UE and a network, wherein the UE comprises a transmitter, a receiver, and a processor, and the processor is configured in combination with the transmitter and / or receiver to: receive from the network, information for identifying an AI / ML model or an AI / ML functionality; identify at least one AI / ML model or AI / ML functionality based on the received information; perform an operation based on the identified at least one AI / ML model or AI / ML functionality, wherein the information for identifying an AI / ML model or an AI / ML functionality includes an ID associated with the AI / ML model or AI / ML functionality and meta data associated with the AI / ML model or AI / ML functionality.

[0030] Other aspects, advantages, and salient features of the disclosure will become apparent to those skilled in the art from the following detailed description taken in conjunction with the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Embodiments / examples of the present disclosure are further described hereinafter with reference to the accompanying drawings, in which: Figures 1a and 1b provide example configurations of an AI / ML model ID in accordance with the present disclosure; Figure 2 provides a block diagram of an exemplary network entity / function that may be used in certain examples of the present disclosure; and Figure 3 provides a flow diagram of an example method for AI / ML management in a wireless communications system in accordance with the present disclosure. DETAILED DESCRIPTION

[0032] The following description of examples of the present disclosure, with reference to the accompanying drawings, is provided to assist in a comprehensive understanding of certain examples of the present disclosure. The description includes various specific details to assist in that understanding but these are to be regarded as merely exemplary. Accordingly, those of ordinary skill in the art will recognize that various changes and modifications of the examples described herein can be made without departing from the scope of the disclosure.

[0033] The same or similar components may be designated by the same or similar reference numerals, although they may be illustrated in different drawings.

[0034] Detailed descriptions of techniques, structures, constructions, functions or processes known in the art may be omitted for clarity and conciseness, and to avoid obscuring the subject matter of the present disclosure.

[0035] The terms and words used herein are not limited to the bibliographical or standard meanings, but are merely used to enable a clear and consistent understanding of the disclosure.

[0036] Throughout the description of this specification, the words “comprise”, “include” and “contain” and variations of the words, for example “comprising” and “comprises”, means “including but not limited to”, and is not intended to (and does not) exclude other features, elements, components, integers, steps, processes, operations, functions, characteristics, properties and / or groups thereof.

[0037] Throughout the description of this specification, the singular form, for example “a”, “an” and “the”, encompasses the plural unless the context otherwise requires. For example, reference to “an object” includes reference to one or more of such objects.

[0038] Throughout the description, the expression “at least one of A, B and / or C” (or the like) and the expression “one or more of A, B and / or C” (or the like) should be seen to separately include all possible combinations, for example: A, B, C, A and B, A and C, A and B and C.

[0039] Throughout the description of this specification, language in the general form of “X for Y” (where Y is some action, process, operation, function, activity or step and X is some means for carrying out that action, process, operation, function, activity or step) encompasses means X adapted, configured or arranged specifically, but not necessarily exclusively, to do Y.

[0040] Features, elements, components, integers, steps, processes, operations, functions, characteristics, properties and / or groups thereof described or disclosed in conjunction with a particular aspect, embodiment or example are to be understood to be applicable to any other aspect, embodiment or example described herein unless incompatible therewith.

[0041] The following examples are applicable to, and use terminology associated with, 3GPP 4G (e.g., LTE) and / or 5G (e.g., NR). However, the skilled person will appreciate that the techniques disclosed herein are not limited to these examples or to 3GPP 4G (e.g., LTE) and / or 5G (e.g., NR), and may be applied in any suitable system or standard, for example one or more existing and / or future generation wireless communication systems or standards (e.g., B5G, 5G-Advanced, 6G etc.). The skilled person will appreciate that the techniques disclosed herein may be applied in any existing or future releases of 3GPP 4G (e.g., LTE) and / or 5G (e.g., NR) and / or 5G Advanced and / or 6G, and / or (3GPP Release 17, 18, 19, 20, etc.) or any other relevant standard. For example, the functionality of the various network entities and other features disclosed herein may be applied to corresponding or equivalent entities or features in other communication systems or standards. Corresponding or equivalent entities or features may be regarded as entities or features that perform the same or similar role, function, operation or purpose within the network.

[0042] Furthermore, the following also applies to the present disclosure: • The terms functionality / use-case / configuration / scenario / site may be used interchangeably. • The terms model and model functionality may be used interchangeably. • This disclosure also apply to non-3GPP entities. • The concepts, proposals, solutions, methods, embodiments, figures, and / or examples, presented in this disclosure, would apply to various type of communication systems, such as 4G, 4G-Advanced, 5G, 5G-Advanced, and 6G. Moreover, the above may also apply (in full or part or modified) to systems of Non-Terrestrial Networks (NR-NTN and / or loT-NTN and / or UAV, etc.), in addition to Terrestrial Networks (TN).

[0043] A particular network entity may be implemented as a network element on dedicated hardware, as a software instance running on a dedicated hardware, and / or as a virtualised function instantiated on an appropriate platform, e.g. on a cloud infrastructure.

[0044] The skilled person will appreciate that the present disclosure is not limited to the specific examples disclosed herein. For example: • The techniques disclosed herein are not limited to 3GPP 4G or 5G or 5G-Advanced and also apply to B5G and 6G systems. • One or more entities in the examples disclosed herein may be replaced with one or more alternative entities performing equivalent or corresponding functions, processes or operations. • One or more of the messages in the examples disclosed herein may be replaced with one or more alternative messages, signals or other type of information carriers that communicate equivalent or corresponding information. • One or more further elements, entities and / or messages may be added to the examples disclosed herein. • One or more non-essential elements, entities and / or messages may be omitted in certain examples. • The functions, processes or operations of a particular entity in one example may be divided between two or more separate entities in an alternative example. • The functions, processes or operations of two or more separate entities in one example may be performed by a single entity in an alternative example. • Information carried by a particular message in one example may be carried by two or more separate messages in an alternative example. • Information carried by two or more separate messages in one example may be carried by a single message in an alternative example. • The order in which operations are performed may be modified, if possible, in alternative examples. • The transmission of information between network entities is not limited to the specific form, type and / or order of messages described in relation to the examples disclosed herein.

[0045] Certain examples of the present disclosure may be provided in the form of an apparatus / device / network entity configured to perform one or more defined network functions and / or a method therefor. Such an apparatus / device / network entity may comprise one or more elements, for example one or more of receivers, transmitters, transceivers, processors, controllers, modules, units, and the like, each element configured to perform one or more corresponding processes, operations and / or method steps for implementing the techniques described herein. For example, an operation / function of X may be performed by a module configured to perform X (or an X-module). Certain examples of the present disclosure may be provided in the form of a system (e.g., a network) comprising one or more such apparatuses / devices / network entities, and / or a method therefor.

[0046] The number of AI / ML models for RAN optimisations is expected to grow rapidly in the next few years. In terms of managing these models in the network, the model ID can play a crucial role, by conveying meta data about the models. Without such meta data in the model ID, it would take additional resources to categorize and manage the vast number of AI / ML models expected in RAN in the near future. In the current RAN2 discussions, the AI / ML model ID is used only to uniquely identify a certain AI / ML model, which is a limitation on the potential capabilities the model ID can also be used for.

[0047] This disclosure provides a detailed derivation for the AI / ML model ID, which can reveal meta data about the AI / ML models, particularly for related or linked AI / ML Models

[0048] In particular, the present disclosure is centred around the use of a AI / ML model ID to convey meta data about the particular model or a group of similar models. This meta data can include (but not necessarily limited to): - Similar models: models that have (or share in part or all) certain features, properties, functionality, scenarios, use cases, configurations, data sets (e.g. data fortraining, monitoring, testing, etc), complexity (e.g. number of layers, parameters, other), size and / or other information related to an AI / ML model (or models), can be called ‘similar’ models. - Linked models: models that have (or share in part or all) certain relationships, interactions, collaborations, and / or any other actions related to those models, can be considered as ‘linked’ models.

[0049] Details and examples of the above two categories and how the AI / ML model IDs can be developed to showcase such meta data are given below.

[0050] Similarly, all solutions, examples, and figures, provided in relation to the AI / ML model ID, may also equally apply (or with some modifications) to the case / concept of AI / ML functionality / functionalities, AI / ML functionality identification (e.g. functionality ID), and / or functionality-based life-cycle-management (LCM) purposes. ‘Similar’ AI / ML models

[0051] The AI / ML models can be similar in many number of ways. For example, the following two broad types: 1) AI / ML models may be predicting (inferring) the future patterns of certain RAN parameters and reporting them, without doing any RAN parameter related modifications. For example, an AI / ML model can observe and then infer the power consumption in RAN and provide this as an input to a network management function. This data can be used to plan out and schedule the different power sources to the network (renewable, non-renewable) for example. 2) AI / ML models inferences can be used to directly update or change certain RAN parameter(s). An example would be to observe and then infer the RSRP reported values from a UE of the serving and neighbour cells and this inference to influence the handover preparations for the UE.

[0052] The main difference between these two types is that the first one (only) monitors RAN parameters and the second monitors and changes (other) RAN parameters.

[0053] Another ‘similarity’ between AI / ML models can be in the fact that they use the same (or similar, or related) input data sets. For example, RSRP / RSRQ data can be used as input to train an AI / ML model to infer the next Modulation and Coding Schemes (MCS) allocations. In a 5G beamforming network, the RSRP / RSRQ from the serving beam and the neighbouring beams can be used in another AI / ML model to facilitate beam switching through inference. The fact that these AI / ML models share the same or similar data sets can be very useful for the network to know early on - so that required resource allocations for data transfer and model transfer / delivery can be optimized, if both these AI / ML models are to be used in the network. The model ID may provide an easier and / or efficient way to convey this meta data.

[0054] In one example, the network, network entity, UE (or a group of UEs), server, OAM, operator, and / or application, can assign a given number of bits (or segment or part) of a given model ID (or models IDs) to indicate if an AI / ML model (or models) belong(s) to at least one category. In a related example, one or more models that belong to this category have the same pattern of bits in this segment of the model ID.

[0055] In some examples, the model (or models) may belong to more than one category.

[0056] In some examples, the network determines (or decides) whether at least one model belongs to at least one model category.

[0057] In some examples, the network entity (and / or function) and / or the UE (or a group of UEs) that assigns the model ID (or models IDs) also assigns the model category.

[0058] In some examples, the network indicates (or informs or reports) to at least one network entity (and / or function), the desired UE (or a group of UEs), external entity (or function), a server, application, OAM, the model (or a group of models) that belong to at least one model category.

[0059] In some examples, the network configures the UE (or a group of UEs) to be able to determine (or decide) whether at least one model belongs to at least one model category.

[0060] In some examples, the UE (or a group of UEs) may determine whether at least one model belongs to a given model category.

[0061] In some examples, the network and / or the UE may update the model (or models) category (-ies) and inform the other side of this update. In a related example, the entity that updates the category can update the model ID (e.g. segment) accordingly, and inform (or report) the other entity of the change of the model ID.

[0062] In some example, the UE indicates (or informs or reports) to at least one network entity (and / or function), the another UE (or a group of other UEs), external entity (or function), a server, application, OAM, that a given model (or models) belong(s) to at least one model category.

[0063] In some examples, the UE may send a request to the network asking if a given model (or models) belong(s) (or not belong anymore) to a given model category (or more than one category). In one example, the UE may send the request included in an existing or newly dedicated RRC and / or NAS signalling / messages and / or lEs. In a related example, the network may confirm (or acknowledge) to the UE if the model (or models) belong(s) to the model category, using a an existing or newly dedicated RRC and / or NAS signalling / messages and / or lEs. In an alternative example, the network may indicate to the UE (or group of UEs) that the model (or models) belong to a different category. This information may be also sent to the UE (or group of UEs) in the confirm (or acknowledge) message.

[0064] In some examples, the network and the UE (or group of UEs) may use the information included in the segments of the model ID, for example, information related to the data category to determine whether a given model (or models) can be trained using the same (or similar) data sets. For example, the network and / or the UE (or group of UEs) may not need to perform data collection (or acquire data) frequently to train models in the same category, if a model ID indicates that those models use same (or similar) data sets. That is, the UE can simply use the data sets (or slightly modified) to train all models in this category . This may avoid performing frequent data collection procedure for all models.

[0065] In another example, the information related to the model category may be included as part of the model meta data.

[0066] In another example, some bits in the model ID can convey information that describe the similarity features of those models included in the same category, for example, one bit to indicate if models use a common (or same or similar) input data set(s), one bit to indicate if models have similar complexity, one bit to indicate similar functionality / use case, etc. In a related example, a segment of 3 bits of the model ID, ‘100’, may indicate ‘1 ={similar data set}, 0 ={not similar / different complexity}, 0 ={not similar / different usecase}’.

[0067] In one example , the network reports to the desired UE (or a group UEs) the set (or group) of models included in the similar models category. In a related example, the network sends the mapping between models’ similarity features and bit order in the model ID (header).

[0068] In one example, the network may configure the mapping of a model ID (or bit or segments of the model ID) to models’ similarity features, for models stored in the network and / or UE (or group of UEs), using RRC and / or NAS signaling / messages and / or via system information broadcast (e.g. periodically, or on-demand). For example, the NG-RAN (or gNB) may configure the mapping by using newly defined and / or existing RRC signaling / messages.

[0069] In another example, the network may update (or modify) the mapping of a model ID (or bit or segments of the model ID) to models’ similarity features, for models stored in the network and / or UE (or group of UEs), using RRC and / or NAS signaling / messages messages and / or via system information broadcast (e.g. periodically, or on-demand). For example, the NG-RAN (or gNB) may update the mapping by using newly defined and / or existing RRC signaling / messages.

[0070] In another example, the UE (or group of UEs) may map (or update an existing mapping) of model ID (bit or segments) to models’ similarity features, for example, for models stored in the UE (or group of UEs). Additionally, the UE (or group of UEs) may inform the network of the mapping (or updated mapping) using RRC and / or NAS signaling / messages.

[0071] In another example, the network and / or the UE may share information related to their capability to support the mapping of models’ similarity features to a model ID. In an alternative example, the network and / or the UE may share information related to their capability to support model ID segmentation and / or mapping to indicate a set of models with similar features.

[0072] In one example, the network may use the information in the model ID, and / or any other information related to the model (or models) that belong(s) to a given category, to select and / or activate at least one model at the network-side, UE-side, and / or two-sided models, using the models category identification. In certain examples, the network may activate at least one model (or model part), at the UE (e.g. UE-side, or UE-part of two-sided models) using one indication (or report) message (or flag or bit in a message).

[0073] In another example, the network may use the information in the model ID, and / or any other information related to the model (or models) that belong(s) to a given category, to deactivate at least one model at the network-side, UE-side, and / or two-sided models, using the models’ category identification. In certain example, the network may deactivate at least one model (or model part), at the UE (e.g. UE-side, or UE-part of two-sided models), using one indication (or report) message (or flag or bit in a message).

[0074] In an example, the information related to the model (or models) category (or categories), model ID (or models IDs), or model meta data, and / or any other information (or data) related to the model (or models) and / or model operation (e.g. LCM procedures), may be exchanged with at least one network entity (and / or function). A certain example would be that the information may be exchanged with the NWDAF, and / or NWDAF and CN, and / or NWDAF and the UE (or a group of UEs), and / or NWDAF and the NG-RAN, and / or a mix of or all the previous entities.

[0075] In another example, the network and / or UE (or a group of UEs) may trigger data collection for a set (or group) of models that share, belong to, and / or have similar data sets for a given LCM purpose. ‘Linked’ AI / ML models

[0076] In this type, AI / ML models can be ‘linked’ such that the output(s) from one model can provide (one or more of the) inputs to another AI / ML model. One example is the RSRP / RSRQ input based MCS allocation model previously mentioned. This is linked to a model which optimizes the HARQ re-transmissions, taking RSRP / RSRQ values and any errors in MCS allocations as inputs. In this case, if both models are operating, the performance HARQ re-transmission model (A) will be influenced by the performance of the MCS allocation model (B). If B is operating optimally in terms of reducing errors, there will be very little or no corrective work that the model A has to perform. In another scenario, an operator may choose to ultilize both models so as to optimize the data throughput. In this case, model B will allocate MCS more aggressively, relying on model A to correct any errors through re-transmissions. Likewise, there are many implications and usages of linked AI / ML models and the network will benefit from knowing these relationships through the model ID meta data.

[0077] In one example, the model IDs can be configured to show the ‘links’ between or amongst 2 or more such AI / ML models. As in the above example, the models can be categorized as ‘parent’ or ‘child’ models and this can be reflected by a few bits (or segment or part) in the model ID.

[0078] In another example, information such as the level of dependency between models and how many network approved models are present in this ‘link’ can also be conveyed by few bits (or segment or part) in the model ID.

[0079] In one example, the network may select and / or activate a set (or group) of linked models at the network-side, UE-side, and / or two-sided models, using the information in the models’ category identification.

[0080] In another example, the network may deactivate a set (or group) of linked models at the network-side, UE-side, and / or two-sided models, using the information in the models’ category identification.

[0081] In another example, the network and / or UE (or a group of UEs) may trigger data collection for a set (or group) of linked models that share, belong to, and / or have similar data sets for a given LCM purpose.

[0082] In another example, the information related to the model (or models) category (or categories), model ID (or models IDs), or model meta data, and / or any other information (or data) related to the model (or models) and / or model operation (e.g. LCM procedures), may be exchanged with at least one network entity (and / or function). A certain example would be that the information may be exchanged with the NWDAF, and / or NWDAF and CN, and / or NWDAF and the UE (or a group of UEs), and / or NWDAF and the NG-RAN, and / or a mix or all the previous entities. Configuration of the AI / ML model ID

[0083] The model ID can be configured to include header(s) which indicate the features and attributes as discussed above. An example model ID configuration would be to concatenate these headers (e.g. Category ID 102, Feature ID 104, Dataset ID 106) in front of the unique model ID part (e.g. 108), which is unique to the particular AI / ML model. Such an example model ID 100 is depicted in the Figure 1a.

[0084] It should be stressed that this is only an example configuration - there can be many variations in arrangement of these headers and even splitting up of these headers into sub-fields. For example, the model ID may have only some of the headers / fields shown in Figure 1a and / or described above, and / or may include additional headers. The basic idea of appending the unique model ID with identifiable headers / fields related to some form of meta data is the key aspect of this example configuration.

[0085] Figure 1b provides another more general example of a model ID 110, which includes an “other ID” 114 to reflect any other type of common meta data that could be identified in a similar way to that described above. Further types of data may or may not be present in the model ID, as illustrated by the optional presence of 112. The sizes of the various fields in Figures 1a and 1b are for illustration purposes only and they may take any suitable bit-size. As for Figure 1a, the model ID of Figure 1b may have only some of the headers / fields shown and / or described above.

[0086] Although AI / ML models have mainly be discussed with respect to their use in the RAN, it should be noted that the use of the models in the RAN is merely an example and they may be applied to any suitable part of the network and network entities thereof. For example, the network may run AI / ML models from 3rd party clients and provide some outputs to these clients (e.g. location data). Here again, the AI / ML model IDs can be used to indicate meta data about the models. This way the model management becomes easier for the network, as in these RAN examples provided above.

[0087] Further examples in accordance with the present disclosure are set out below, where the examples may be combined in any appropriate form and also combined with any of the approaches set out above.

[0088] In accordance with an example of the present disclosure, there is provided an AI / ML model ID data format / structure that includes an indication of one of more types of meta data / or meta data itself in addition to a unique model ID. The types of meta data may include, but not limited to, one or more of including a category ID, a feature ID, and a dataset ID. In other words, the model ID may include information that provides or indicates a further description of the model.

[0089] The indication / meta data may be included in one or more headers or fields appended to or included in an enlarged model ID.

[0090] The category ID may indicate a predetermined category that a model is a member of or related to. The category types may take any suitable form, for example, models that predict behaviours and models that control behaviours.

[0091] The feature ID may indicate one or more features of the model and / or one or more features that are similar / common to other models and / or other models in the same category.

[0092] The dataset ID may indicate one or more datasets used by the model and / or datasets that are similar / common to other models and / or other models in the same category.

[0093] Information indicated by one or more of the category ID, feature ID, dataset ID or other meta data may include, but are not limited to, one or more of a model type, model complexity, model accuracy, model reliability, model dataset type, model age, model expiry, model security, model applicability, and model features.

[0094] In some examples, the model ID may include further headers / fields, such as a field explicitly indicating related models, using their unique model ID for instance.

[0095] The model ID may be configured when the model is first generated or configured. The model ID may also be updated if required, for example, due to new datasets, new models, or additional features being included in the model.

[0096] Model IDs may be distributed to any suitable network entity by any other suitable network entity, for example, any network entity that is required to provide or be provided with information on AI / ML models. For example, a gNB may transmit a model ID, a category ID or any other meta data indication to a UE. Alternatively, a AI / ML related network entity may transmit a model ID, a category ID or any other meta data indication to another network entity that requires or has requested information on AI / ML models.

[0097] The meta data of the model ID may also be leveraged to indicate one or more models without indicating individual unique model IDs, thus providing a more efficient way of communicating data concerning AI / ML models throughout a network.

[0098] In some examples, via the use of the meta data of the model ID, a network entity may be able to quickly and easily identify or indicate models that correspond to a request from another network entity, for example, models that utilise similar data sets and / or offer similar predictive or control functions.

[0099] It will be appreciated that examples of the present disclosure may be realized in the form of hardware, software or a combination of hardware and software. Certain examples of the present disclosure may provide a computer program comprising instructions or code which, when executed, implement a method, system and / or apparatus in accordance with any aspect, example and / or embodiment disclosed herein. Certain embodiments of the present disclosure provide a machine-readable storage storing such a program.

[00100] Figure 2 is a block diagram of an exemplary network entity / function that may be used in examples of the present disclosure, such as the techniques disclosed in relation to any of the figures. For example, any of the network entities, network function etc. may be provided in the form of the network entity illustrated in Figure 2. The skilled person will appreciate that a network entity / function may be implemented, for example, as a network element on a dedicated hardware, as a software instance running on a dedicated hardware, and / or as a virtualised function instantiated on an appropriate platform, e.g. on a cloud infrastructure.

[00101] The entity 200 comprises a processor (or controller) 201, a transmitter 203 and a receiver 205. The receiver 205 is configured for receiving one or more messages from one or more other network entities, for example as described above. The transmitter 203 is configured for transmitting one or more messages to one or more other network entities, for example as described above. The processor 201 is configured for performing one or more operations, for example according to the operations as described above.

[00102] For example, entity 200 may configure, update, generate, transmit, and / or receive the example model IDs described above or elements thereof. The entity may also be use the above-described model ID to identify and / or use AI / ML models suitable for a particular function.

[00103] Figure 3 provides a flow diagram of an example method for AI / ML management in a wireless communications system that utilises examples of the model ID described above, where the wireless communications system includes but is not limited to a UE and a network (e.g. one or more network entities) for communicating with the UE.

[00104] At step S302, the UE receives, from the network, information for identifying an AI / ML model or an AI / ML functionality. The information for identifying an AI / ML model or an AI / ML functionality may include any of the data / information described above as being part of the model ID e.g. unique IDs, meta data etc.

[00105] At step S304, the UE identifies at least one AI / ML model or AI / ML functionality based on the received information. For example, if the information for identifying an AI / ML model or an AI / ML functionality includes a category of an AI / ML model or AI / ML functionality, the UE may identify AI / ML models and / or AI / ML functionalities that belong to the category.

[00106] At step S306, the UE performs an operation based on the identified at least one AI / ML model or AI / ML functionality. For example, the UE may report to the network information on the identified at least one AI / ML model or AI / ML functionality, the UE may activate, deactivate, select, and / or switch the identified at least one AI / ML model or AI / ML functionality, or the UE may configure, trigger, and / or perform data collection for the identified at least one AI / ML model or AI / ML functionality.

[00107] For all of the examples / aspects / embodiments etc. described above / herein, it should be considered that the corresponding features / operations apply in any order or combination, and that furthermore there exists the possibility to omit one or more features / operations.

[00108] Moreover, for all of the examples, embodiments, aspects etc. above, these apply to at least LTE, NR, NR NTN or loT NTN (note this list is merely to give some examples and should not be seen as limiting), including any related signalling / messages on any of the inferences X2, Xn, NG, S1, F1, etc (again, this list is merely to give some examples and should not be seen as limiting).It will be appreciated that, in each example / embodiment / aspect etc. described above, one or more features or operations may be omitted, modified or moved (e.g., to change the order of the features or the operations), if desired and appropriate.

[00109] Additionally, where the figures illustrating example method flows include text in relation to a specific step / operation, it will be appreciated that this text is simply an example of the corresponding step / operation, where a more general definition (such as may be found in the description of the corresponding step) may apply for the step / operation.

[00110] Additionally, regarding all of the above, one or more features or operations etc. from any example / embodiment may be combined with features or operations from any other example / embodiment. That is, the present disclosure should be considered to include all combinations of examples / embodiments disclosed herein, as appropriate, as well as combinations of individual features within and between each example / embodiment, as appropriate.

[00111] The techniques described herein may be implemented using any suitably configured apparatus and / or system. Such an apparatus and / or system may be configured to perform a method according to any aspect, embodiment or example disclosed herein. Such an apparatus may comprise one or more elements, for example one or more of receivers, transmitters, transceivers, processors, controllers, modules, units, and the like, each element configured to perform one or more corresponding processes, operations and / or method steps for implementing the techniques described herein. For example, an operation / function of X may be performed by a module configured to perform X (or an X-module). The one or more elements may be implemented in the form of hardware, software, or any combination of hardware and software.

[00112] It will be appreciated that examples of the present disclosure may be implemented in the form of hardware, software or any combination of hardware and software. Any such software may be stored in the form of volatile or non-volatile storage, for example a storage device like a ROM, whether erasable or rewritable or not, or in the form of memory such as, for example, RAM, memory chips, device or integrated circuits or on an optically or magnetically readable medium such as, for example, a CD, DVD, magnetic disk or magnetic tape or the like.

[00113] It will be appreciated that the storage devices and storage media are embodiments of machine-readable storage that are suitable for storing a program or programs comprising instructions that, when executed, implement certain examples of the present disclosure. Accordingly, certain examples provide a program comprising code for implementing a method, apparatus or system according to any example, embodiment and / or aspect disclosed herein, and / or a machine-readable storage storing such a program. Still further, such programs may be conveyed electronically via any medium, for example a communication signal carried over a wired or wireless connection.

[00114] While the disclosure has been shown and described with reference to certain examples, it will be understood by those skilled in the art that various changes in form and detail may be made therein without departing from the scope of the disclosure.

[00115] The reader's attention is directed to all papers and documents which are filed concurrently with or previous to this specification in connection with this application and which are open to public inspection with this specification, and the contents of all such papers and documents are incorporated herein by reference. Acronyms and Definitions 3GPP 3rd Generation Partnership Project 5G 5th Generation 5GC 5G Core 5QI 5G QoS Identifier 5GS 5G System 5GSM 5G System Session Management 5GMM 5G System Mobility Management AF Application Function Al Artificial Intelligence AM Acknowledged Mode AMF Access and Mobility Management Function AS Application Server ASP Application Service Provider ATG Air-To-Ground AUSF Authentication Server Function CDN Content Delivery Network DCAF Data Collection Application Function DNAI Data Network Access Identifier DNN Data Network Name DNS Domain Name Server DRB Data Radio Bearer eNB Evolved Node B EPC Evolved Packet Core FEC Forward Error Correction FQDN Fully Qualified Domain Name GBR Guaranteed Bit Rate gNB Next generation Node B GPSI Generic Public Subscription Identifier GW Gateway HSS Home Subscriber Service IAB Integrated Access and Backhaul ID Identity / ldentifier lloT Industrial Internet of Things loT Internet of Things IMEI International Mobile Equipment Identities IP Internet Protocol l-SMF Intermediate SMF LADN Local Area Data Network LL SSM Lower Layer SSM MBMS Multimedia Broadcast / Multicast Service MBS Multicast / Broadcast Service MBSF Multicast / Broadcast Service Function MBSTF Multicast / Broadcast Service Transport Function MB-SMF Multicast / Broadcast Session Management Function MB-UPF Multicast / Broadcast User Plane Function ML Machine Learning MME Mobility Management Entity MN Master Node MNF Monitoring Network Function MNO Mobile Network Operator MT Mobile Termination NAS Non-Access Stratum NEF Network Exposure Function NR New Radio NRF Network Repository Function NG-RAN Next Generation Radio Access Network NG-eNB Next Generation eNB NSA Non-Standalone NSSF Network Slice Selection Function NTN Non-Terrestrial Networks NW Network NWDAF Network Data Analytics Function OS Operating System OSAPP OS Application PCF Policy Control Function PCO Protocol Configuration Options PDR Packet Detection Rule PDU Protocol Data Unit PTM Point To Multipoint PTP Point to Point QFI QoS Flow Identifier (ID) QoS Quality of Service RACH Random Access Channel RAN Radio Access Network RRC Radio Resource Control RSD Route Selection Descriptor RSRP Reference Signal Received Power SA Standalone SDAP Service Data Adaptation Protocol SDU Service Data Unit SGW Serving Gateway SIM Subscriber Identity Module SLA Service Level Agreement SM Session Management SMF Session Management Function SN Secondary Node S-NSSAI Single Network Slice Selection Assistance Information SSB Synchronization Signal Block SSM Source Specific IP Multicast address SSC Session and Service Continuity SRB Signaling Radio Bearer SUPI Subscription Permanent Identifier TA Tracking Area TAI Tracking Area Identity TE Terminal Equipment TM Transparent Mode TMGI Temporary Mobile Group Identity TS Technical Specification UAV Unmanned Aerial Vehicle UDM Unified Data Manager UDR Unified Data Repository UE User Equipment UL Uplink UM Unacknowledged Mode UP User Plane UPF User Plane Function URLLC Ultra-Reliable and Low-Latency Communication URSP UE Route Selection Policy

Claims

1. A method for AI / ML management in a wireless communications system comprising a UE and a network, the method comprising:receiving, at the UE from the network, information for identifying an AI / ML model or an AI / ML functionality;identifying, by the UE, at least one AI / ML model or AI / ML functionality based on the received information;performing, by the UE, an operation based on the identified at least one AI / ML model or AI / ML functionality,wherein the information for identifying an AI / ML model or an AI / ML functionality includes an ID associated with the AI / ML model or AI / ML functionality and meta data associated with the AI / ML model or AI / ML functionality.

2. The method of claim 1, wherein the operation includes reporting, by the UE to the network, information on the identified at least one AI / ML model or AI / ML functionality.

3. The method of claim 1 or 2, wherein the operation includes one or more of activating, deactivating, selecting, and / or switching the identified at least one AI / ML model or AI / ML functionality.

4. The method of any preceding claim, wherein the identifying includes identifying at least one AI / ML model or AI / ML functionality that belongs to a category indicated by the information for identifying an AI / ML model or an AI / ML functionality.

5. The method of any preceding claim, wherein the information for identifying an AI / ML model or an AI / ML functionality includes a first functionality and the identifying includes identifying at least one AI / ML model that provides the first functionality.

6. The method of any preceding claim, wherein the meta data includes information indicating or associated with one or more of:a category of the AI / ML model or AI / ML functionality;a similar AI / ML model or AI / ML functionality;a linked AI / ML model or AI / ML functionality;a feature of the AI / ML model or AI / ML functionality;a dataset associated with the AI / ML model or AI / ML functionality;a type of the AI / ML model or AI / ML functionality;an accuracy of the AI / ML model or AI / ML functionality;a reliability of the AI / ML model or AI / ML functionality;an age of the AI / ML model or AI / ML functionality;an expiry of the AI / ML model or AI / ML functionality;a security of the AI / ML model or AI / ML functionality; and an applicability of the AI / ML model or AI / ML functionality.

7. The method of any preceding claim, wherein the ID and meta data are included in one or more fields or headers appended to or included in a model ID data structure.

8. The method of any preceding claim, wherein the operation includes updating the ID and / or meta data associated with the AI / ML model or AI / ML functionality.

9. The method of any preceding claim, wherein the method further includes, receiving at the UE from the network, information configuring the mapping of the information for identifying an AI / ML model or an AI / ML functionality.

10. The method of any preceding claim, wherein the operation includes configuring, triggering, and / or performing data collection for the identified at least one AI / ML model or AI / ML functionality.

11. The method of any preceding claim, wherein the information for identifying an AI / ML model or an AI / ML functionality is received from a radio access network (RAN) of the wireless communications system, a network data analytics function (NWDAF) of the wireless communications systems, and / or a core network of the wireless communications system.

12. The method of any preceding claim, wherein the operation includes determining whether the identified at least one AI / ML model or an AI / ML functionality can be trained using similar data sets.

13. The method of any preceding claim, wherein the method further comprises updating by the network an ID and / or meta data associated with an AI / ML model or AI / ML functionality and providing an indication of the updated ID and / or meta data to the UE.

14. The method of any preceding claim, wherein the AI / ML models and AI / ML functionalities are network-side, UE-side, and / or two-sided models / functionalities.

15. The method of any preceding claim, wherein the wireless communications system is a 3GPP compliant wireless communications system.

16. A wireless communications system comprising a UE and a network, wherein the wireless communications system is configured to implement the method of any of claims 1 to 15.

17. A method for a user equipment in a wireless communications system comprising the UE and a network, the method comprising:receiving, from the network, information for identifying an AI / ML model or an AI / ML functionality;identifying at least one AI / ML model or AI / ML functionality based on the received information;performing an operation based on the identified at least one AI / ML model or AI / ML functionality,wherein the information for identifying an AI / ML model or an AI / ML functionality includes an ID associated with the AI / ML model or AI / ML functionality and meta data associated with the AI / ML model or AI / ML functionality.

18. A user equipment (UE) for use in a wireless communication system comprises the UE and a network, wherein the UE comprises a transmitter, a receiver, and a processor, and the processor is configured in combination with the transmitter and / or receiver to:receive from the network, information for identifying an AI / ML model or an AI / ML functionality;identify at least one AI / ML model or AI / ML functionality based on the received information;perform an operation based on the identified at least one AI / ML model or AI / ML functionality,wherein the information for identifying an AI / ML model or an AI / ML functionality includes an ID associated with the AI / ML model or AI / ML functionality and meta data associated with the AI / ML model or AI / ML functionality.26

Images