Grouping multiple entities to participate in machine learning operations

Abstract

Various aspects of this disclosure relate to a network entity for grouping multiple entities to participate in machine learning (ML) operations, wherein the multiple entities are coupled to a wireless communication system, the network entity comprising: at least one memory; and at least one processor coupled to the at least one memory and configured such that the network entity: obtains information related to the ML operation; creates a group of multiple entities to perform the ML operation based on the information related to the ML operation; determines a unique group identifier for the group; and transmits the unique group identifier to each of the multiple entities.

Description

Technical Field

[0001] This disclosure relates to wireless communications, and more specifically, to grouping multiple entities to participate in machine learning (ML) operations. Background Technology

[0002] A wireless communication system may include one or more network communication devices, such as base stations, that support wireless communication for one or more user communication devices, which may also be referred to as user equipment (UE) or other suitable terms. The wireless communication system may support wireless communication with one or more user communication devices by utilizing the resources of the wireless communication system (e.g., time resources (e.g., symbols, time slots, subframes, frames, etc.) or frequency resources (e.g., subcarriers, carriers, etc.)). Furthermore, the wireless communication system may support wireless communication across various radio access technologies, including third-generation (3G), fourth-generation (4G), fifth-generation (5G), and other suitable radio access technologies beyond 5G (e.g., sixth-generation (6G)).

[0003] It has identified use cases and requirements for supporting the distribution, delivery, and training of artificial intelligence (AI) / machine learning (ML) models for various applications (e.g., video / speech recognition, robot control, vehicle control), and also covers support for distributed AI training / inference based on direct device connectivity. Summary of the Invention

[0004] The article “a” preceding an element is unrestricted and should be understood to refer to “at least one” or “one or more” of those elements. The terms “a,” “at least one,” “one or more,” and “at least one of one or more” are interchangeable. As used herein, the word “or,” as used in a list of items (e.g., a list of items beginning with phrases such as “at least one,” “one or more,” or “one or two”) indicates an inclusive list, such that a list of at least one of, for example, A, B, or C means A or B or C or AB or AC or BC or ABC (i.e., A and B and C). Furthermore, as used herein, the phrase “based on” should not be construed as referring to a closed set of conditions. For example, without departing from the scope of this disclosure, an example step described as “based on condition A” may be based on both condition A and condition B. In other words, as used herein, the phrase “based on” should be interpreted in the same manner as the phrase “at least partially based on.” Furthermore, as used herein, the term “set” may comprise one or more elements.

[0005] Some embodiments of the methods and apparatus described herein may further include a network entity for grouping multiple entities to participate in machine learning (ML) operations, wherein the multiple entities are coupled to a wireless communication system, the network entity comprising: at least one memory; and at least one processor coupled to the at least one memory and configured such that the network entity: obtains information related to the ML operation; creates a group of multiple entities to perform the ML operation based on the information related to the ML operation; determines a unique group identifier for the group; and transmits the unique group identifier to each of the multiple entities.

[0006] The information associated with the ML operation may include one or any combination of the following: (i) an ML model identifier; (ii) a service identifier or service area or service profile; (iii) at least one target user equipment identifier; and (iv) an analysis identifier.

[0007] The information associated with the ML operation may include one or any combination of the following: (v) one or more KPIs of the ML operation; (vi) data requirements of the ML operation; (vii) one or more entity types; (viii) instructions on whether model training and / or inference are required; (ix) model vendor information; and (x) training job ID.

[0008] The at least one processor may be configured to cause the network entity to: send a request message to an ML model repository, the request message including the information related to the ML operation; and receive a response message from the ML model repository, the response message including a list of candidate entities to perform the ML operation.

[0009] The at least one processor may be configured to cause the network entity to select the plurality of entities from the list of candidate entities to create the group.

[0010] The at least one processor may be configured to cause the network entity to randomly determine the unique group identifier.

[0011] The at least one processor may be configured to: determine a mapping from the unique group identifier to a set of member identifiers of the plurality of entities in the group; and store the mapping in the at least one memory.

[0012] The at least one processor may be configured to: detect a triggering event for modifying the group; update the group based on the triggering event to create an updated group of multiple entities to perform the ML operation; and modify the mapping such that the unique group identifier is associated with a set of member identifiers of the multiple entities in the updated group.

[0013] The triggering event may be: an entity joining the group; one of the plurality of entities leaving the group; one of the plurality of entities becoming unavailable to perform the ML operation; an expected or predicted failure of one or more KPIs that satisfy the ML operation; one of the plurality of entities being unable to perform the ML operation; or an expected or predicted failure or congestion of the wireless interface coupled to at least one of the plurality of entities.

[0014] The at least one processor may be configured to monitor the state of the plurality of entities in the group to detect the triggering event.

[0015] The at least one processor may be configured to: periodically send a status request to each of the plurality of entities in the group; and receive a status response from each of the plurality of entities in the group.

[0016] The at least one processor can be configured to configure the plurality of entities in the group to periodically send state updates to the network entities.

[0017] The at least one processor may be configured to transmit the unique group identifier and the identity of the member of the updated group to deliver to at least one of the plurality of entities in the updated group.

[0018] The network entity may be a Service Enabler Architecture Layer (SEAL) group management server. The plurality of entities in the group may include one or any combination of the following: at least one FL client, at least one FL server, at least one FL aggregator or collaborator, at least one ML client, and at least one ML server.

[0019] The network entity may be a Service Enabler Architecture Layer (SEAL) group management server, an AI enabler server, an application data analytics enabler server, or an edge enabler server.

[0020] The plurality of entities in the group may include at least one edge application server and / or at least one UE application client.

[0021] The at least one processor may be configured to transmit the identities of the plurality of entities in the group to at least one of the plurality of entities in the group.

[0022] Some embodiments of the methods and apparatus described herein may further include a method for grouping multiple entities to participate in a machine learning (ML) operation, wherein the multiple entities are coupled to a wireless communication system, the method comprising: obtaining information related to the ML operation; creating a group of multiple entities to perform the ML operation based on the information related to the ML operation; determining a unique group identifier for the group; and transmitting the unique group identifier to deliver to each of the multiple entities.

[0023] Some embodiments of the methods and apparatus described herein may further include a processor for grouping multiple entities to participate in machine learning (ML) operations, the processor comprising: at least one controller coupled to at least one memory and configured such that the processor: obtains information related to the ML operation; creates a group of multiple entities to perform the ML operation based on the information related to the ML operation; determines a unique group identifier for the group; and outputs the unique group identifier to each of the multiple entities. Attached Figure Description

[0024] Figure 1 Examples of wireless communication systems according to aspects of this disclosure are described.

[0025] Figure 2 This describes the Advanced Service Enabler Architecture Layer (SEAL) architecture based on aspects of this disclosure.

[0026] Figure 3 This describes an architecture for enabling edge applications based on aspects of this disclosure.

[0027] Figure 4a and 4b Describe the use cases for Edge Application Server (EAS) bundled.

[0028] Figure 5 This describes an architecture for implementing group management of entities based on aspects of this disclosure.

[0029] Figure 6 This is a flowchart of a first embodiment for grouping multiple entities to participate in ML operations.

[0030] Figure 7 This is a flowchart of a second embodiment for grouping multiple entities to participate in ML operations.

[0031] Figure 8 An example of a processor 800 according to aspects of this disclosure is described.

[0032] Figure 9 An example of a network equipment (NE) 900 according to aspects of this disclosure is described.

[0033] Figure 10 A flowchart illustrating the method performed by NE according to aspects of this disclosure.

[0034] Figure 11 Explanation of aspects based on this disclosure Figure 5 How can the architecture be coupled to a wireless communication system? Detailed Implementation

[0035] 3GPP has identified use cases and requirements for supporting the distribution, delivery, and training of artificial intelligence (AI) / machine learning (ML) models for various applications (e.g., video / speech recognition, robot control, vehicle control), and also covers support for distributed AI training / inference based on direct device connectivity (in TS 22.261).

[0036] Regarding AI / ML support, 3GPP SA2 (System Architecture Group) has studied the following two aspects:

[0037] • Enhancements to the Network Data Analysis Function (NWDAF) to support analytics with Federation Learning (FL) enabled to improve analytics output. In this enhancement, NWDAFs acting as FL members (servers or clients) register with the Network Repository Function (NRF) and can be discovered by other Network Functions (NFs) (currently only by other NWDAFs).

[0038] • Enhancements to the 5G Core (5GC) (as specified in Clause 5.46 of TS 23.501) to assist AI / ML (including FL) operations in the application layer (between one or more AI / ML users and AI / ML servers). In this enhancement, the Network Exposure Function (NEF) assists the AI / ML application server in scheduling available UEs to participate in AI / ML federated learning. Furthermore, the 5GC can assist in selecting UEs to act as FL clients by providing a list of target member UEs to an Application Function (AF) that subscribes to the NEF to be notified of a subset of UEs (i.e., a list of candidate UEs that meet certain screening criteria).

[0039] Additionally, 3GPP SA6 has specified an application layer architecture in 3GPP TS 23.436 to implement data analytics as a new Service Enabler Architecture Layer (SEAL) service, also known as Application Data Analytics Enabled Services (ADAES). This type of architecture (in...) Figure 2 (As explained in the documentation) It provides an application-layer analytics framework that offers general analytics exposure and value-added services for vertical industries and ASPs. This includes application-layer analytics related to end-to-end application performance, edge load, service API availability, location accuracy, and slice-related performance and fault analysis.

[0040] Group management is defined in 3GPP SA6. One aspect of group management defined in 3GPP SA6 is the SEAL group management service. The group management (GM) service is a SEAL service that provides group management capabilities for one or more vertical applications. It consists of a GM server (GMS) and one or more GM clients (GMCs), and... Figure 2 This describes the network-internal functional model used for group management in the SEAL group management service.

[0041] The group management client communicates with the group management server via the GM-UU reference point. The group management client provides support for group management functions to the Vertical Application Layer (VAL) client via the GM-C reference point. The VAL server communicates with the group management server via the GM-S reference point. The group management server interacts with the NEF of the underlying 3GPP network system via the N33 reference point to execute the group management process for 5G Virtual Network (5GVN) groups.

[0042] TS 23.434 10.2 also specifies service-based representation and network-outside deployment models (GMC-to-GMC interactions). The capabilities provided in group management are presented in the process of TS 23.434 10.3, including:

[0043] - Group creation

[0044] - Group Information Query

[0045] - Group membership update performed by authorized user / UE / VAL server

[0046] - Group Configuration Management

[0047] - Location-based group creation

[0048] - Group member left

[0049] - Temporary group formation within the VAL system

[0050] These capabilities are general and designed to support vertical applications (e.g., automotive, factory, unmanned aerial systems (UAS)...); however, in general, group creation and management applies to groups for VAL UEs or VAL entities, and not to different types of enablers or groups for specific purposes (e.g., ML support).

[0051] Another aspect of group management defined in 3GPP SA6 is EDGEAPP, which specifies the application layer architecture for edge services. In this architecture, such as... Figure 3 As described above, the following entities can be defined:

[0052] - Edge Enabler Server (EES) provides the necessary support functions for edge application servers and edge enabler clients, such as:

[0053] a) Provide configuration information to the edge enabler client to enable application data service exchange with the edge application server;

[0054] c) The ability to interact with the 3GPP core network to access network functions directly (e.g., via PCF) or indirectly (e.g., via SCEF / NEF / SCEF+NEF);

[0055] e) Support 3GPP network capabilities to be exposed to the external edge application server via EDGE-3

[0056] - Edge Enabler Client (EEC) provides the support functions required by application clients, such as retrieving and providing configuration information to enable application data service exchange with edge application servers; and discovering available edge application servers in the edge data network.

[0057] The Edge Configuration Server (ECS) provides the support functions required for edge enabler clients to connect to the edge enabler server. These functions of the Edge Configuration Server are related to providing edge configuration information to the EEC, which is used to establish a connection with the EES.

[0058] Edge Application Server (EAS) is an application server that resides in an edge data network and performs server functions.

[0059] - Application Client (AC): An application client is an application that resides in the UE and performs client functions.

[0060] One of the features in EDGEAPP Rel-18 is support for EAS bundles. This support is primarily for situations where EAS are grouped together due to dependencies or the need to operate together, even if these EAS may belong to different edge / cloud platforms.

[0061] Two use cases have been identified in 3GPP TR 23.700-98, which are in Figure 4a and 4b The explanation is as follows.

[0062] Figure 4aThis section describes an example of an EAS bundle. To provide service to end users, a typical AC communicates with multiple endpoints (i.e., multiple EASs). In online games supporting a large number of users, different game functions are distributed across multiple servers: (i) a game engine for managing game state and user input (e.g., EAS1), (ii) an in-game chat server for communication between players (e.g., EAS2), and (iii) a capture server for capturing, rendering, encoding, and transmitting images to players' devices (e.g., EAS3). If each of these EASs is individually discovered, controlled, and relocated, this can impact the overall quality of service.

[0063] Figure 4b This section describes a composite EAS instance. To collaborate with other EASs to provide services (weather, traffic, maps, etc.), EAS context processing and composite EAS support may be required at the edge compatibility layer. When UE mobility occurs, methods for rearranging the composite EAS context may be needed to provide continuous composite EAS services. Additionally, a method may be needed to locate EASs that provide services to composite EASs within an EDN to which the UE has moved.

[0064] One aspect not covered in the 3GPP specifications is how to support the management of ML / FL groups during ML / FL operations (e.g., training) (e.g., how to create and manage a group of ML / FL members).

[0065] Federation learning (or more generally, ML workflows) involves multiple ML / FL members, which may act as one or more FL servers, FL aggregators, or FL clients to jointly or coordinately perform ML training and / or inference processes. Grouping these entities for specific ML tasks (e.g., training) can be important for performance optimization, especially where FL members reside on external systems and end devices (with varying availability and capabilities). Embodiments of this disclosure address how to support the grouping of application-layer entities to perform ML operations (e.g., acting as FL participants), and how to create and manage groups for specific ML / FL tasks involving multiple members. Specifically, embodiments of this disclosure provide a mechanism for creating and managing entity groups to act as ML / FL members for specific vertical / application service provider (ASP) requirements. Such grouping is supported by a new enabling function that generates group identities and can dynamically provide mappings to member IDs.

[0066] Federation learning (FL) can be defined as a machine learning technique that enables multiple FL clients to train a model by exchanging parameters rather than exchanging / sharing local datasets. The FL server manages FL operations by maintaining and updating a global ML model, selecting and managing FL clients, executing aggregation strategies, scheduling training in a federated manner, and communicating with FL clients. FL clients provide various aspects of data for FL operations, such as data collection, data preparation, and using the data for training and / or inference while simultaneously communicating updates to the ML model to the FL server.

[0067] The aspects of this disclosure are described in the context of wireless communication systems.

[0068] Figure 1 This describes an example of a wireless communication system 100 according to aspects of this disclosure. The wireless communication system 100 may include one or more NEs 102, one or more UEs 104, and a core network (CN) 106. The wireless communication system 100 may support various radio access technologies. In some embodiments, the wireless communication system 100 may be a 4G network, such as an LTE network or an LTE-A network. In some other embodiments, the wireless communication system 100 may be an NR network, such as a 5G network, a 5G-A network, or a 5G Ultra Wideband (5G-UWB) network. In other embodiments, the wireless communication system 100 may be a combination of 4G and 5G networks, or other suitable radio access technologies, including IEEE 802.11 (Wi-Fi), IEEE 802.16 (WiMAX), and IEEE 802.20. The wireless communication system 100 may support radio access technologies beyond 5G, such as 6G. In addition, the wireless communication system 100 can support technologies such as Time Division Multiple Access (TDMA), Frequency Division Multiple Access (FDMA), or Code Division Multiple Access (CDMA).

[0069] One or more NEs 102 may be distributed across a geographical area to form a wireless communication system 100. One or more of the NEs 102 described herein may be, include, or be referred to as a network node, base station, network element, network function, network entity, radio access network (RAN), NodeB, eNodeB (eNB), next-generation NodeB (gNB), or other suitable terms. NEs 102 and UEs 104 may communicate via a communication link, which may be a wireless or wired connection. For example, NEs 102 and UEs 104 may perform wireless communication (e.g., receive signaling, transmit signaling) via a Uu interface.

[0070] NE 102 can provide a geographic coverage area, for which NE 102 can support services for one or more UEs 104 within the geographic coverage area. For example, NE 102 and UE 104 can support wireless communication of signals related to services (e.g., voice, video, packet data, messaging, broadcasting, etc.) according to one or more radio access technologies. In some embodiments, NE 102 can be mobile, for example, a satellite associated with a non-terrestrial network (NTN). In some embodiments, different geographic coverage areas 112 associated with the same or different radio access technologies can overlap, but different geographic coverage areas can be associated with different NEs 102.

[0071] One or more UEs 104 may be distributed across a geographical area of ​​the wireless communication system 100. UE 104 may include or be referred to as a remote unit, mobile device, wireless device, remote device, subscriber device, transmitter device, receiver device, or some other suitable term. In some embodiments, UE 104 may be referred to as a unit, station, terminal, or client, and other instances thereof. Additionally or alternatively, UE 104 may be referred to as an Internet of Things (IoT) device, an Internet of Everything (IoE) device, or a Machine Type Communication (MTC) device, and other instances thereof.

[0072] UE 104 may be able to support direct wireless communication with other UE 104 via a communication link. For example, UE 104 may support direct wireless communication with another UE 104 via a device-to-device (D2D) communication link. In some implementations, such as vehicle-to-vehicle (V2V) deployments, vehicle-to-everything (V2X) deployments, or cellular-V2X deployments, the communication link 114 may be referred to as a side link. For example, UE 104 may support direct wireless communication with another UE 104 via a PC5 interface.

[0073] NE 102 may support communication with CN 106 or with another NE 102 or both. For example, NE 102 may interface with other NE 102 or CN 106 via one or more backhaul links (e.g., S1, N2, N2, or network interfaces). In some embodiments, NE 102 may communicate directly with each other. In some other embodiments, NE 102 may communicate with each other or indirectly (e.g., via CN 106). In some embodiments, one or more NE 102 may include sub-components, such as access network entities, which may be instances of access node controllers (ANCs). The ANC may communicate with one or more UE 104s via one or more other access network transmitting entities (which may be referred to as radio headends, smart radio headends, or transmit-receive points (TRPs)).

[0074] CN 106 can support user authentication, access authorization, tracking, connectivity, and other access, routing, or mobility functions. CN 106 can be an evolved packet core (EPC) or a 5G core (5GC), which may include control plane entities that manage access and mobility (e.g., a mobility management entity (MME), access and mobility management functions (AMF)) and user plane entities that route or interconnect packets to external networks (e.g., a serving gateway (S-GW), a packet data network (PDN) gateway (P-GW), or a user plane function (UPF)). In some implementations, the control plane entities may manage non-access stratum (NAS) functions of one or more UEs 104 served by one or more NEs 102 associated with CN 106, such as mobility, authentication, and bearer management (e.g., data bearers, signaling bearers, etc.).

[0075] CN 106 can communicate with the packet data network via one or more backhaul links (e.g., via S1, N2, N2, or another network interface). The packet data network may contain an application server. In some implementations, one or more UEs 104 can communicate with the application server. UE 104 can establish a session (e.g., a Protocol Data Unit (PDU) session, etc.) with CN 106 via NE 102. CN 106 can use the established session (e.g., an established PDU session) to route services (e.g., control information, data, etc.) between UE 104 and the application server. A PDU session may be an instance of a logical connection between UE 104 and CN 106 (e.g., one or more network functions of CN 106).

[0076] In the wireless communication system 100, NE 102 and UE 104 can use the resources of the wireless communication system 100 (e.g., time resources (e.g., symbols, time slots, subframes, frames, etc.) or frequency resources (e.g., subcarriers, carriers)) to perform various operations (e.g., wireless communication). In some embodiments, NE 102 and UE 104 may support different resource structures. For example, NE 102 and UE 104 may support different frame structures. In some embodiments, such as in 4G, NE 102 and UE 104 may support a single frame structure. In some other embodiments, such as in 5G and other suitable radio access technologies, NE 102 and UE 104 may support various frame structures (i.e., multiple frame structures). NE 102 and UE 104 may support various frame structures based on one or more parameter sets.

[0077] The wireless communication system 100 may support one or more parameter sets, and the parameter sets may include subcarrier spacing and cyclic prefixes. A first parameter set (e.g., μ=0) may be associated with a first subcarrier spacing (e.g., 15 kHz) and a regular cyclic prefix. In some embodiments, the first parameter set (e.g., μ=0) associated with the first subcarrier spacing (e.g., 15 kHz) may utilize one time slot per subframe. A second parameter set (e.g., μ=1) may be associated with a second subcarrier spacing (e.g., 30 kHz) and a regular cyclic prefix. A third parameter set (e.g., μ=2) may be associated with a third subcarrier spacing (e.g., 60 kHz) and a regular cyclic prefix or an extended cyclic prefix. A fourth parameter set (e.g., μ=3) may be associated with a fourth subcarrier spacing (e.g., 120 kHz) and a regular cyclic prefix. A fifth parameter set (e.g., μ=4) may be associated with a fifth subcarrier spacing (e.g., 240 kHz) and a regular cyclic prefix.

[0078] Time intervals for resources (e.g., communication resources) can be organized according to frames (also called radio frames). Each frame may have a duration, for example, 10 milliseconds (ms). In some embodiments, each frame may contain multiple subframes. For example, each frame may contain 10 subframes, and each subframe may have a duration, for example, 1 ms. In some embodiments, each frame may have the same duration. In some embodiments, each subframe of a frame may have the same duration.

[0079] Alternatively, the time intervals of resources (e.g., communication resources) can be organized according to time slots. For example, a subframe may contain a certain number (e.g., quantity) of time slots. The number of time slots in each subframe may also depend on one or more parameter sets supported in the wireless communication system 100. For example, the first, second, third, fourth, and fifth parameter sets (i.e., μ=0, μ=1, μ=2, μ=3, μ=4) associated with corresponding subcarrier intervals of 15 kHz, 30 kHz, 60 kHz, 120 kHz, and 240 kHz may utilize a single time slot per subframe, two time slots per subframe, four time slots per subframe, eight time slots per subframe, and 16 time slots per subframe, respectively. Each time slot may contain a certain number (e.g., quantity) of symbols (e.g., OFDM symbols). In some embodiments, the number (e.g., quantity) of time slots in a subframe may depend on the parameter set. For a conventional cyclic prefix, a time slot may contain 14 symbols. For an extended cyclic prefix (e.g., applicable to a 60 kHz subcarrier spacing), a time slot may contain 12 symbols. The relationships between the number of symbols per time slot, the number of time slots per subframe, and the number of time slots per frame for both the regular and extended cyclic prefixes may depend on the parameter set. It should be understood that references to the first parameter set (e.g., μ=0) associated with the first subcarrier spacing (e.g., 15 kHz) may be used interchangeably between subframes and time slots.

[0080] In the wireless communication system 100, the electromagnetic (EM) spectrum can be divided into various categories, frequency bands, channels, etc., based on frequency or wavelength. For example, the wireless communication system 100 may support one or more operating frequency bands, such as frequency range names FR1 (410 MHz to 7.125 GHz), FR2 (24.25 GHz to 52.6 GHz), FR3 (7.125 GHz to 24.25 GHz), FR4 (52.6 GHz to 114.25 GHz), FR4a or FR4-1 (52.6 GHz to 71 GHz), and FR5 (114.25 GHz to 300 GHz). In some embodiments, NE 102 and UE 104 may perform wireless communication on one or more of the operating frequency bands. In some embodiments, FR1 may be used by NE 102 and UE 104, as well as other equipment or devices, for cellular communication services (e.g., control information, data). In some implementations, FR2 can be used by NE 102 and UE 104, as well as other equipment or devices, for short-range, high data rate capabilities.

[0081] FR1 may be associated with one or more parameter sets (e.g., at least three parameter sets). For example, FR1 may be associated with a first parameter set containing a 15 kHz subcarrier spacing (e.g., μ=0); a second parameter set containing a 30 kHz subcarrier spacing (e.g., μ=1); and a third parameter set containing a 60 kHz subcarrier spacing (e.g., μ=2). FR2 may be associated with one or more parameter sets (e.g., at least two parameter sets). For example, FR2 may be associated with a third parameter set containing a 60 kHz subcarrier spacing (e.g., μ=2); and a fourth parameter set containing a 120 kHz subcarrier spacing (e.g., μ=3).

[0082] Figure 5 This describes an example of architecture 500 for implementing group management of entities (e.g., entities acting as ML entities or FL clients) at the application enablement layer, according to aspects of this disclosure. Such group management includes group creation, monitoring, and updating. This can be based on the AI / ML lifecycle / workflow / pipeline, which is based on 1) analytics events / services performed by ADAES or 2) VAL requirements of ML / FL support services. Multiple entities formed within a group according to embodiments of the invention are coupled to a wireless communication system, such as the wireless communication system 100 described above. This in… Figure 11 The explanation is as follows. That is to say, entities within the group can be, for example, UE 104, clients or applications running on UE 104, servers, or network elements, such as network elements in core network 106.

[0083] like Figure 5 As shown, architecture 500 includes an ML model producer 502, an ML model repository 504, an ML model consumer 506, an ML model management service consumer 508, and an ML model management service producer 510. Architecture 500 also includes at least one ML model training service producer 512 and / or at least one ML model inference service producer 514.

[0084] ML model producer 502 is an entity that provides ML models. Such an entity may be an application-specific client at a VAL server, EAS, or UE 104. Such an entity may belong to a non-3GPP system or edge / cloud platform provider. In some implementations, ML model producer 502 may be the mobile network operator (MNO) itself.

[0085] The ML model repository 504 may be a Common API Framework (CAPIF) core function (CCF) as defined in 3GPP TS 23.222, an edge or central platform repository, an Application Layer-Analysis Data Repository Function (A-ADRF), a VAL User Database (UDB) and application layer registry as defined in 3GPP TS 23.434, an embedded registry at ADAES, or a Network Repository Function (NRF). The ML model repository 504 is configured to store data. For example, the ML model repository 504 may store ML models, ML model IDs, and / or information about candidate participants (e.g., ML or FL candidate participants) capable of performing ML operations.

[0086] ML model consumer 506 is an entity that consumes ML models after negotiation with ML model producer 502. The negotiation can be indirect. For example, a third party registers an ML model at the MNO's ML model repository 504 (e.g., an edge or cloud registry), and another trusted third party consumes the ML model after negotiation with the MNO. ML model consumer 506 can be a VAL server. In embodiments of this disclosure, the VAL server can be a game server, an AI application server, an XR server, or a V2X server.

[0087] ML model training service producer 512 is the entity that undertakes the task of training the ML model of the identified model. In the case of federated learning (FL), multiple producers are being identified. In the context of SA6, such ML model training service producer 512 can be an enabling function on the server or client side. In some scenarios, after negotiation with the MNO or SA6 service provider, ML model training service producer 512 can be an external application or system (e.g., ETSI MEC, O-RAN, etc.).

[0088] ML Model Inference Service Producer 514 is an entity that provides ML model inference. Such entities can be those capable of obtaining, for example, real-time data, and can improve trained ML models with more accurate / fine-grained data. Instances of such entities can be VAL UEs (Enabler Clients, VAL Clients, or ACs).

[0089] Architecture 500 further includes the ML model management service consumer ML-MMSC 508 (which can be, for example, an FL aggregator, an application data analytics enabler server, or an edge enabler server (EES)), and the ML model management server producer ML-MMSP 510 (which can be, for example, a SEAL group management server, an AI / ML enabler server, an application data analytics enabler server, or an edge enabler server). While ML-MMSC 508 and ML-MMSP 510 are... Figure 5While shown as separate entities, in some embodiments of this disclosure, the functions of ML-MMSC 508 and ML-MMSP 510 may be performed by a single entity.

[0090] In architecture 500, entities act as consumers of ML model training services. An ML model training service consumer is an entity that consumes ML model training services provided by one or more producers (e.g., ML model training server producer 512). In the case of FL (Flexible Interchange), such a consumer can be an FL server / aggregator. Such a consumer can be a network function (e.g., ANLFNWDAF), or a VAL server or any other enabler (SEAL, EDGE). Such entities can be UE functions. ML model consumers 506 and / or ML-MMSC 508 can act as ML model training service consumers.

[0091] In Architecture 500, entities assume the role of consumers of ML model inference services. An ML model inference service consumer is an entity that consumes ML model inference services. Such entities can be producers of ML model training services at specific layers or even at network functions (such as NWDAF or Model Training Logic Function (MTLF)) 512.

[0092] In Architecture 500, the entity assumes the role of a producer of ML group membership management services. This service can be a standalone service or part of an ML model management service. In other words, ML-MMSP 510 can perform the role of a producer of ML group membership management services. Producers of this type of service provide group management for ML model participants (ML clients / ML servers or FL servers and a list of FL clients) as part of the group configuration for a given ML model and use cases. Use cases can be defined by application services, service regions, service types, slices, analysis events, or VAL requirements.

[0093] In Architecture 500, the entity assumes the role of a consumer of the ML group member management service: the consumer of the group management service can be the producer of the ML model training service 512 or any other entity involved in the ML model lifecycle. ML-MMSC508 can perform the role of a consumer of the ML group member management service.

[0094] In step S502, ML model producer 502 registers the ML model in ML model repository 504. In step S502, the ML model identifier (ID) associated with the ML model is stored in ML model repository 504. Additionally, ML model information, context, and / or parameters may be provided and stored in ML model repository 504. It should be understood that multiple models may be registered in repository 504 by one or more ML model producers 502.

[0095] In step S504, candidate participants are able to register for ML operations at the ML model repository 504, such that information identifying the candidate participants is stored at the ML model repository 504. Each candidate participant stores information at the ML model repository 504 indicating what ML operations the candidate participant may be involved in. For example, the ML model repository 504 may store each candidate participant in association with: (ii) an ML model identifier; (ii) a service identifier or service region or service profile (e.g., service type, such as URLLC); (iii) at least one target user equipment identifier; (iv) an analytics identifier; and / or the candidate participant's capabilities. The service identifier may be an application service ID, a vertical service ID, an application ID, or an application server ID. Candidate participants may include at least one ML model training service producer 512 and / or at least one ML model inference service producer 514.

[0096] At step S506, the ML model consumer 506 sends a request to the ML-MMSC 508 for providing AI / ML support services. When the ML model consumer 506 is a VAL server, the request can be a VAL request. The request sent at step S506 can be a request for ADAES analysis or a request for support for the training / inference of the ML model. This request may include the ML model ID, application service ID or profile, vertical service requirements, service area, a list of VAL UEs to which this request applies, PLMN / NPN ID, and / or data requirements for training (e.g., requirements for handling real-time data).

[0097] At step S508, based on the request, ML-MMSC 508 determines that consumer group management services are needed (e.g., multiple ML participants are needed for FL or ML model training partitions or ML training jobs).

[0098] At step S510, ML-MMSC 508 sends an ML member group management request to ML-MMSP 510 to create or configure a group (which may be, for example, by service or service region or model ID or analytics ID) including ML or FL participants for a given task. The ML member group management request includes information related to the ML operation (which may be an FL operation). The information related to the ML operation may include one or more of the following: (i) ML model identifier; (ii) service identifier (which may be application service ID, vertical service ID, application ID, or application server ID) or service region or service profile; (iii) at least one target user equipment identifier; (iv) analytics identifier; (v) one or more KPIs of the ML operation; (vi) data requirements of the ML operation; (vii) one or more entity types; (viii) instructions regarding whether model training and / or inference are required; (ix) model vendor information; and (x) training job ID.

[0099] ML-MMSP 510 is configured to create a set of multiple entities based on information related to an ML operation to perform the ML operation. This can be achieved by ML-MMSP 510 sending a request message to ML model repository 504 at step S512, the request message including information related to the ML operation. At step S512, ML-MMSP 510 receives a response message from ML model repository 504, the response message including a list of candidate entities to perform the ML operation. Specifically, the list of candidate entities is associated with information related to the ML operation in ML model repository 504. Specifically, the list of candidate entities may be associated with a given ML model ID, service ID, or analytics ID.

[0100] At step S514, ML-MMSP 510 is configured to use a list of candidate entities to determine a group of multiple entities for performing ML operations. ML-MMSP 510 may determine the members of the group from the list of candidate entities based on policy information or based on some predefined criteria. ML-MMSP 510 may determine the members of the group based on the capabilities of the candidate participants (computing resources, maximum load, thresholds), whether the candidate participants are fixed or mobile nodes and their availability, the proximity between the candidate participants and any interface constraints, the dependencies between the candidate participants, how close the candidate participants are to the data producer (based on data requirements), and / or the energy status / constraints of the candidate participants.

[0101] At step S514, ML-MMSP 510 is configured to determine a unique group identifier for the group.

[0102] At step S514, ML-MMSP 510 can determine a mapping from a unique group identifier to a set of member identifiers of multiple entities in the group, such as the mapping from group ID N to members #1, #2, and #3.

[0103] The ML-MMSP 510 may also obtain at least one of the following: ML-related ID, process ID, and ML training job ID. The ML-MMSP 510 can generate ML-related IDs / process IDs / ML training job IDs. Alternatively, the ML-related IDs / process IDs / ML training job IDs may be provided by the ML-MMSC 508. The ML-related ID identifies the joint learning process used to train the ML model; this parameter may be included when the service is used for joint learning. The ML training job or ML process ID identifies the job / process used to train the ML model. For the FL use case, the ML-related ID may include one or more ML training job IDs. The ML process ID may also identify the ML inference process.

[0104] At step S514, ML-MMSP 510 can also determine the mapping from unique group identifiers to ML-related IDs / process IDs / ML training job IDs. For example, the group ID of a group consisting of members #1, #2, and #3 can be mapped to job #x, job #y, or related ID #z.

[0105] An ML training job ID can be used for a given ML task, but a group can include entities that perform multiple ML tasks. For example, multiple training jobs may occur across two servers in the group for two application services (e.g., analytics service #1 and analytics service #2).

[0106] At step S514, ML-MMSP 510 may randomly generate a unique group identifier for the group, or the unique group identifier may encompass contextual information indicating one of the mappings referenced above.

[0107] The ML-MMSP 510 is configured to transmit a unique group identifier to each of several entities selected for the group. The transmission of the unique group identifier can be direct to multiple entities or via another entity (e.g., the ML-MMSC 508). Depending on the type of message transmitted, this may be followed by an ACK / NACK or a response.

[0108] For example, at step S516, ML-MMSP 510 may emit a unique group identifier to be delivered to one or more ML model training service producers 512, and / or at step S518, ML-MMSP 510 may emit a unique group identifier to be delivered to one or more ML model inference service producers 514.

[0109] The ML-MMSP 510 can be further configured to act as multiple entities in a launch group to deliver to at least one of the multiple entities in the group (e.g., an ML server or an FL server).

[0110] After including the ML model training service producer 512 into the determined group, at step S520, the ML model training service producer 512 sends the trained model with a unique group identifier to the ML-MMSC 508. Although in Figure 5 Not shown, but ML model inference service producer 514 can send inference results with unique group identifiers to ML training service producer 512 or to ML-MMSC 508 (depending on deployment).

[0111] The ML-MMSP 510 can additionally monitor the status of created groups, and specifically, periodically monitor the status of group members to detect triggering events for modifying the group. Triggering events can be: an entity joining a group; one of multiple entities leaving a group; one of multiple entities becoming unavailable to perform ML operations; anticipated or predicted failure to satisfy one or more KPIs of an ML operation; one of multiple entities being unable to perform an ML operation; or anticipated or predicted failure or congestion of the radio interface coupled to at least one of the multiple entities.

[0112] ML-MMSP 510 can monitor members by subscribing to the ML model repository 504 to receive notifications about updates to group members (such as entering / leaving the group).

[0113] The ML-MMSP 510 can monitor the members of a group by periodically sending status requests to each of the multiple entities in the group; and receive status responses from each of the multiple entities in the group.

[0114] The ML-MMSP 510 can monitor group members and configure them to periodically transmit state updates to network entities. Specifically, the message transmitted in steps S516 and S518, which includes a unique group identifier, may include information for configuring participants to periodically transmit state updates.

[0115] The ML-MMSP 510 can monitor group members by sending on-demand status requests to each of the multiple entities in the group; and receive status responses from each of the multiple entities in the group. On-demand status requests can be sent based on one or more of the following:

[0116] - When a new VAL request for debugging enable service or an ML member group management request is received;

[0117] - ML member state changes or predicted state changes (e.g., FL member moves out of coverage);

[0118] - Changes in UE mobility, energy, or capabilities;

[0119] - Find better groupings based on KPIs;

[0120] - If the KPIs used to enable the service cannot be met (e.g., confidence level not reached).

[0121] Monitoring performed by ML-MMSP 510 may also include continuously checking the status of group participants (e.g., FL clients) within the group and whether the local ML model is aligned. This step may involve direct interaction between ML-MMSP 510 and the ML training service producer 512 (via API) or interaction between ML-MMSP 510 and the FL server / aggregator (if alignment issues are captured in the FL server). Potential misalignment of the local ML model may trigger changes in group membership (the addition of new members or the departure of members with misaligned model outputs).

[0122] In response to the detection of a triggering event, the ML-MMSP 510 can update the group to create an updated group of multiple entities to perform ML operations, and modify the mapping so that a unique group identifier is associated with a set of member identifiers among the multiple entities in the updated group. That is, the unique group identifier remains unchanged, but the member list changes.

[0123] The ML-MMSP 510 can transmit a unique group identifier and the identity of the updated group members to be delivered to at least one of a plurality of entities in the updated group. In some embodiments, the ML-MMSP 510 transmits a unique group identifier and the identity of the updated group members to be delivered to each of the plurality of entities in the updated group. The transmission of the unique group identifier and the identity of the updated group members can be direct to the plurality of entities or via another entity (e.g., ML-MMSC 508).

[0124] Now refer to Figure 6The first embodiment describes a set of functions at the GMS (i.e., ML-MMSP 510 is SEALGMS), the ML model consumer 506 is a VAL server (operating as an ML / FL server), and the ML-MMSC 508 is an AI enabler server (which may also be an application data analytics enabler server). Specifically, the consumer can be an AI enabler server or a VAL server, which can be, for example, an FL server / aggregator. We refer to the AI ​​enabler server (or AI / ML enabler server) herein as a middleware server for supporting AI / ML services. Such servers provide middleware services as NetApp on top of the core network to allow AI / ML applications to integrate with 3GPP systems. The role of the AI ​​enabler server may include supporting the selection and grouping of FL members, or the discovery and registration of ML participants. The role of the AI ​​enabler server may include uniformly exposing AI / ML capabilities from the 3GPP core network 106, acting as an application function (AF), and configuring AI enabler client applications at mobile devices. In some implementations, the AI ​​enabler server may be deployed as the SEAL application data analytics enabler server as specified in TS 23.436.

[0125] At step S602, VAL server 506 sends a request to AI enabler server 508 to support the ML / FL workflow. The request may contain information related to the ML operation, including any of the instances described herein. For example, the request may include an ML model ID, a list of target UEs, data requirements, a description of the FL / ML training or inference job, a list of candidate ML / FL participants (if known), and / or a list of selected ML / FL participants (if VAL server 506 is an ML / FL aggregator), etc. Alternatively, AI enabler server 508 may receive a request from VAL server 506 to group FL / ML members.

[0126] At step S604, the AI ​​enabler server 508 determines, based on the received request, that a group consisting of the required ML / FL members needs to be created. In the case of FL, the FL aggregator can be either the VAL server 506 or the AI ​​enabler server 508, depending on the request.

[0127] At step S606, the AI ​​enabler server 508 sends a request to the FL group function at the group management server (GMS) 510 to assume group management of the ML / FL workflow / process based on the request in step S602. The request issued at step S606 may include one or more of the following: ML model ID, service ID or service region or service profile, one or more KPIs of the ML / FL process, data requirements, the type of ML / FL entity required, whether model training and / or inference is required, model vendor information, and training job ID (if known), etc. It should be noted that the FL group function may be a new service or function, or part of any existing enabler server. The request issued at step S606 may alternatively take the form of a subscription request to GMS 510 (e.g., a subscription-notification model).

[0128] At step S608, GMS 510 authorizes a request and, at step S610, retrieves ML / FL candidate or available ML / FL members from the ML model repository 504 for a given model ID / type / profile, or for a given service or service region, or for a given training / inference job ID, etc. GMS 510 can also request and receive the ability to act as an ML / FL member at a VALUE via GM clients in a given service region. For example, GMS 510 can query all GM clients in the service region for the status of the VALUE and its availability / ability to act as an ML / FL member.

[0129] At step S611, GMS 510 determines a group of multiple entities for performing ML operations by selecting ML / FL members or by configuring a group based on the selected ML / FL members by the ML / FL server / aggregator. GMS 510 generates a unique group identifier (group ID) and may also create and store a mapping of the generated group ID to a group of ML / FL member IDs within the group. The group determination may alternatively be jointly determined by GMS 510 and AI enabler server 508, or confirmed by AI enabler server 508.

[0130] GMS 510 transmits the group ID (and optionally the group membership identity) to members of the group. This transmission of the group ID can be direct to multiple entities within the group. Alternatively, the transmission can be indirect. For example, GMS 510 can notify the AI ​​enabler server 508, and the AI ​​enabler server 508 can notify the ML / FL group members of the group ID. Such transmission can take the form of a request or command to a candidate ML / FL member who will act as a selected ML / FL member for a given task (ML operation). Such transmission can contain a training job ID or related ID or FL process / workflow ID.

[0131] like Figure 6As shown, at step S612, GMS 510 transmits the group ID (and optionally group membership) to one or more candidate entities (candidate ML / FL members) selected for inclusion in the group (602). Furthermore, at step S614, GMS 510 may transmit the group ID (and optionally group membership) to one or more AI enabler clients (e.g., GM clients) on one or more UEs selected for inclusion in the group (606). The AI ​​enabler client is the counterpart of the AI ​​enabler server on the VAL UE side. The role of the AI ​​enabler client is configured by the AI ​​enabler server via a standardized API and interacts with the VAL application (or application or application client) to confirm the group. The AI ​​enabler client can be, for example, a SEAL client (as specified in TS 23.434) or an EEC capability (as specified in TS 23.558).

[0132] Although Figure 6 Not shown in the document, but notifications can also be sent (directly or indirectly via AI enabler server 508) to the ML model repository 504 to store group IDs and mappings to ML / FL members.

[0133] In an alternative embodiment, GMS 510 may send requests to candidate entities (e.g., clients) to join a group in the ML / FL process. Based on the response from each candidate, GMS 510 creates a group ID for the FL service and optionally creates training jobs and / or ML-related IDs.

[0134] At step S616, for the VAL client 604 (which can be an ML or FL client) on the UE 104 selected as an ML / FL member, the AI ​​enabler client 606 transmits information related to the ML operation to the VAL client 604 and may also request its consent / confirmation.

[0135] At step S618, the AI ​​enabler server 508 sends a response to the VAL request emitted at step S602, indicating information related to group creation and ML operations. For example, the response emitted at step S618 may include one or any combination of the following: ML model ID, analysis ID, service ID, group ID, bundle ID, ML-related ID, process ID, ML training job ID, member ID, and member type information.

[0136] In steps S620a and S620b, members of the group (e.g., ML / FL clients) use the group ID to send the output of the locally trained ML model to the VAL server 506. The group ID may be equivalent to an ML-related ID or a bundled training job ID.

[0137] At step S622, GMS 510 (FL group function) continuously monitors the group status and the availability / condition of its members. Furthermore, GMS 510 detects triggers for changing the group (a new member joins, an existing member leaves, or an existing member has a problem / is unavailable). Figure 6 As shown, the initial trigger detection can be performed by the VAL server 506 and transmitted to the GMS 510 at step S621, whereby the GMS 510 triggers based on the received communication detection. Similarly, the initial trigger detection can be performed by the AI ​​enabler server 508 and transmitted to the GMS 510, whereby the GMS 510 triggers based on the received communication detection.

[0138] Then, GMS 510 updates the group by changing the mapping from the group ID to the member list (the group ID remains the same, but the member list changes).

[0139] At steps S624 and S626, GMS 510 sends the updated group notification to the relevant entities in a manner similar to steps S612 and S614 (directly or via AI enabler server 508).

[0140] The GMS 510 or AI enabler server 508 can also notify the ML model consumer (VAL server) 506 and / or ML model repository 504 of changes in group membership.

[0141] Now refer to Figure 7 The second embodiment is described. In EDGEAPP, the EAS bundle scenario involves grouping EAS with dependencies. In the scenario of EAS grouping for ML / FL processes (e.g., training), taking into account EES or other enablers (e.g., EEC) and device applications (AC) within the group, the ASP may not provide grouping because group members can be modified as part of the AI / ML support service.

[0142] In this example embodiment, ML-MMSC 508 is an EES, ML-MMSP 510 is a SEAL server, and ML model consumer 506 is an EAS or AI enabler server (which can operate as an FL server / aggregator). EES or AI enabler 506 triggers EAS grouping and utilizes the logical functions of ML-MMSP 510 (which can be, for example, the logical functions of other enablers such as EES, ADAES, or AI enablers), and notifies the EAS 702 and AC 704 that will be grouped together. Therefore, the bundle ID is provided by the SA6 function, not by the ASP. This requirement is unique for AI / ML because grouping is based on a dynamically changeable selection of ML training / inference entities. ML-MMSP 510 with grouping functionality can be a new AI enabler server, an application data analytics enabler server, GMS, or even an EES.

[0143] At step S702, EES 508, together with EAS (FL / ML Server) 506, detects that 1) a bundle of EAS, or 2) a bundle of EAS and AC, or 3) a bundle of AC, is required to support the ML / FL workflow. Such detection may be based on an EAS request, may begin from EES 508, or may be triggered jointly.

[0144] At step S704, EES 508 determines that ML / FL members (candidates or selected based on the FL server) need to be grouped.

[0145] At step S706, EES 508 sends an ML group management request to the corresponding SEAL server 510 that has ML / FL group functionality. Such requests contain information related to ML operations (which may be FL operations), examples of which are described herein.

[0146] At step S708, the SEAL server 510 authorizes the request and at step S710 retrieves ML / FL candidate or available ML / FL members from the ML model repository 504 for a given model ID / type / configuration file, or for a given service or service region, or for a given training / inference job ID, etc.

[0147] The ML-MMSP 510 can also request and receive the ability of a VAL UE to act as an ML / FL member via a GM client in a given service area. For example, the GMS 510 can query all GMCs in the service area for the status of the VAL UE and its availability / ability to act as an ML / FL member.

[0148] At step S712, SEAL server 510 determines a group by selecting ML / FL members or by configuring a group based on the selected ML / FL members by ML / FL server / aggregator 506. SEAL server 510 generates a group identity and may also create / store a mapping of the generated group ID to a set of ML / FL member IDs within the group. Group determination may alternatively be determined or confirmed jointly with EES 508. The generated group ID may be a “bundle ID” as defined in 3GPP TS 23.558.

[0149] At step S712, SEAL server 510 transmits the group ID (and optionally the group membership identity) to the members of the group. This transmission of the group ID (and optionally the group membership identity) can be direct to multiple entities within the group. Alternatively, the transmission can be indirect. For example, at step S714, SEAL server 510 can transmit a notification to EES 508, and EES 508 can notify the group ID to the EAS / AC acting as an ML / FL member. Such a transmission can take the form of a request or command to a candidate ML / FL member who will act as a selected ML / FL member for a given task (ML operation). Such a transmission can contain a training job ID or a related ID or an FL process / workflow ID.

[0150] like Figure 7 As shown, at step S716, the SEAL server 510 may transmit a group ID (and optionally a group membership) to the EEC 706 on the UE 104, where the UE 104 has an AC 704 selected for inclusion in the group.

[0151] Although Figure 7 Not shown in the document, but notifications can also be sent (directly or indirectly via EES 508) to the ML model repository 504 to store group IDs and mappings to ML / FL members.

[0152] At step S718, for a UE selected as an ML / FL member, the EEC transmits information related to the ML operation to AC 704 and may also request its consent / confirmation.

[0153] At step S720, the SEAL server 510 sends a response to the EES 508, indicating group creation and information related to ML operations.

[0154] Alternatively, if the SEAL server 510 does not directly notify the EAS, then at step S722, the EES 508 may notify the EAS 702, which acts as an ML / FL member, of information related to the ML operation.

[0155] Therefore, in the first embodiment, the FL group function is an enhanced capability of SEAL GMS, and the consumer is an AI enabler server or a VAL server, which may be, for example, an FL server / aggregator. In the second embodiment, the FL group function is applicable to EDGE scenarios where groups include EAS and / or AC, and the role of EES is to coordinate group management (either by jointly supporting group management or by leveraging SEAL capabilities and exposing them to EAS).

[0156] In embodiments of this disclosure, the entity selected for the group may be, for example, an FL client, an FL server, an FL aggregator or collaborator, an ML client, or an ML server.

[0157] In this paper, we refer to the FL client as an entity that acts as an ML model training service producer when using FL. The FL client provides local ML model output to the FL server / aggregator. We refer to the FL server as an entity that configures and globally trains models based on the local output from the FL client. FL aggregators or collaborators may resemble FL servers, but in some scenarios (e.g., hierarchical FL), aggregators or collaborators may only aggregate local ML model training data from the FL client / FL server. We refer to the ML server as an AI / ML endpoint, which may be an application server that provides ML models or configures and trains ML models. We refer to the ML client as an AI / ML endpoint, which is configured by the AI / ML server to perform ML operations. For example, the ML client is located on the UE side and receives ML training output from the ML server, and based on this, the ML client uses the output as input for ML model inference (e.g., based on UE local data).

[0158] like Figure 6 As shown in the examples, ML clients or FL clients can be applications running on UE 104. Figure 7 As shown in the examples, the ML client or FL client can be an application client on UE 104 or Edge Application Server (EAS).

[0159] Figure 8An example of a processor 800 according to aspects of this disclosure is described. Processor 800 may be an example of a processor configured to perform various operations according to the examples described herein. Processor 800 may include a controller 802 configured to perform various operations according to the examples described herein. Processor 800 may optionally include at least one memory 804, which may be, for example, an L1 / L2 / L3 cache. Additionally or alternatively, processor 800 may optionally include one or more arithmetic logic units (ALUs) 806. One or more of these components may be electronically communicated or otherwise coupled (e.g., operatively, communicatively, functionally, electronically, electrically) via one or more interfaces (e.g., buses).

[0160] Processor 800 may be a processor chipset and includes a protocol stack (e.g., a software stack) executed by the processor chipset to perform various operations (e.g., receive, acquire, retrieve, transmit, output, forward, store, determine, identify, access, write, read) according to the examples described herein. The processor chipset may include one or more cores, one or more caches (e.g., memory local to the processor chipset (e.g., processor 800) or included in the processor chipset), or other memory (e.g., random access memory (RAM), read-only memory (ROM), dynamic RAM (DRAM), synchronous dynamic RAM (SDRAM), static RAM (SRAM), ferroelectric RAM (FeRAM), magnetic RAM (MRAM), resistive RAM (RRAM), flash memory, phase-change memory (PCM), and others).

[0161] Controller 802 can be configured to manage and coordinate various operations of processor 800 (e.g., signaling, receiving, acquiring, retrieving, transmitting, outputting, forwarding, storing, determining, identifying, accessing, writing, and reading) to enable processor 800 to support various operations according to the examples described herein. For example, controller 802 can operate as a control unit of processor 800, generating control signals that manage the operation of various components of processor 800. These control signals include enabling or disabling functional units, selecting data paths, initiating memory accesses, and coordinating operation timing.

[0162] Controller 802 may be configured to fetch (e.g., fetch, retrieve, receive) instructions from memory 804 and determine subsequent instructions to be executed to enable processor 800 to support various operations according to the examples described herein. Controller 802 may be configured to track the memory addresses of instructions associated with memory 804. Controller 802 may be configured to decode instructions to determine the operations to be performed and the operands involved. For example, controller 802 may be configured to interpret instructions and determine control signals to be output to other components of processor 800 to enable processor 800 to support various operations according to the examples described herein. Alternatively or additionally, controller 802 may be configured to manage data flow within processor 800. Controller 802 may be configured to control data transfers between registers, arithmetic logic unit (ALU), and other functional units of processor 800.

[0163] Memory 804 may include one or more caches (e.g., memory local to or included in processor 800), or other memories such as RAM, ROM, DRAM, SDRAM, SRAM, MRAM, flash memory, etc. In some embodiments, memory 804 may reside within or on the processor chipset (e.g., local to processor 800). In some other embodiments, memory 804 may reside outside the processor chipset (e.g., remotely from processor 800).

[0164] Memory 804 may store computer-readable, computer-executable code containing instructions that, when executed by processor 800, cause processor 800 to perform the various functions described herein. The code may be stored in a non-transitory computer-readable medium, such as system memory or another type of memory. Controller 802 and / or processor 800 may be configured to execute computer-readable instructions stored in memory 804 to cause processor 800 to perform various functions. For example, processor 800 and / or controller 802 may be coupled to or coupled to memory 804, and processor 800, controller 802, and memory 804 may be configured to perform the various functions described herein. In some instances, processor 800 may include multiple processors, and memory 804 may include multiple memories. One or more of the multiple processors may be coupled to one or more of the multiple memories, which may be individually or collectively configured to perform the various functions described herein.

[0165] One or more ALU 806s may be configured to support various operations according to the examples described herein. In some embodiments, one or more ALU 806s may reside within or on a processor chipset (e.g., processor 800). In some other embodiments, one or more ALU 806s may reside outside the processor chipset (e.g., processor 800). One or more ALU 806s may perform one or more calculations on data, such as addition, subtraction, multiplication, and division. For example, one or more ALU 806s may receive input operands and an opcode that determines the operation to be performed. One or more ALU 806s may be configured with various logic and arithmetic circuitry, including adders, subtractors, shifters, and logic gates, to process and manipulate data according to the operation. Alternatively, one or more ALU 806s may support logical operations such as AND, OR, XOR, NOR, and NAND, enabling one or more ALU 806s to handle conditional operations, comparisons, and bitwise operations.

[0166] Processor 800 may support wireless communication according to examples disclosed herein. Processor 800 may be configured or operable to support a component for: obtaining information related to ML operations; creating a group of multiple entities to perform ML operations based on the information related to ML operations; determining a unique group identifier for the group; and transmitting the unique group identifier to each of the multiple entities.

[0167] Figure 9 An example of NE 900 according to aspects of this disclosure is described. NE 900 may include a processor 902, a memory 904, a controller 906, and a transceiver 908. The processor 902, memory 904, controller 906, or transceiver 908, or various combinations thereof, or various components thereof, may be examples of components for performing the various aspects of this disclosure as described herein. These components may be coupled via one or more interfaces (e.g., operatively, communicatively, functionally, electronically, electrically). NE 900 may correspond to ML-MMSP 510 and perform the functions of ML-MMSP 510 as described herein.

[0168] Processor 902, memory 904, controller 906, or transceiver 908, or various combinations or components thereof, may be implemented in hardware (e.g., a circuit system). The hardware may include a processor, digital signal processor (DSP), application-specific integrated circuit (ASIC), or other programmable logic device, or any combination thereof configured or otherwise supporting components for performing the functions described in this disclosure.

[0169] Processor 902 may include intelligent hardware devices (e.g., a general-purpose processor, DSP, CPU, ASIC, FPGA, or any combination thereof). In some embodiments, processor 902 may be configured to operate memory 904. In some other embodiments, memory 904 may be integrated into processor 902. Processor 902 may be configured to execute computer-readable instructions stored in memory 904 to cause NE 900 to perform various functions of this disclosure.

[0170] Memory 904 may comprise volatile or non-volatile memory. Memory 904 may store computer-readable, computer-executable code containing instructions that, when executed by processor 902, cause NE 900 to perform the various functions described herein. The code may be stored in a non-transitory computer-readable medium, such as memory 904 or another type of memory. Computer-readable media includes both non-transitory computer storage media and communication media, wherein the communication media includes any medium that facilitates the transfer of a computer program from one place to another. Non-transitory storage media may be any available medium accessible by a general-purpose or special-purpose computer.

[0171] In some implementations, processor 902 and memory 904 coupled to processor 902 may be configured to cause NE 900 to perform one or more of the functions described herein (e.g., processor 902 executing instructions stored in memory 904). For example, processor 902 may support wireless communication at NE 900 according to an example disclosed herein. NE 900 may be configured to support a component for: obtaining information related to ML operations; creating a group of entities to perform ML operations based on the information related to ML operations; determining a unique group identifier for the group; and transmitting the unique group identifier to each of the entities.

[0172] Controller 906 manages the input and output signals of NE 900. Controller 906 can also manage peripheral devices not integrated into NE 900. In some embodiments, controller 906 may utilize an operating system such as iOS®, ANDROID®, WINDOWS®, or other operating systems. In some embodiments, controller 906 may be implemented as part of processor 902.

[0173] In some embodiments, the NE 900 may include at least one transceiver 908. In other embodiments, the NE 900 may have more than one transceiver 908. The transceiver 908 may represent a wireless transceiver. The transceiver 908 may include one or more receiver chains 910, one or more transmitter chains 912, or a combination thereof.

[0174] Receiver chain 910 may be configured to receive signals (e.g., control information, data, packets) via a wireless medium. For example, receiver chain 910 may include one or more antennas for receiving signals over the air or via a wireless medium. Receiver chain 910 may include at least one amplifier (e.g., a low-noise amplifier (LNA)) configured to amplify the received signal. Receiver chain 910 may include at least one demodulator configured to demodulate the received signal and obtain transmitted data by reversing the modulation technique applied during signal transmission. Receiver chain 910 may include at least one decoder for decoding the demodulated signal to receive the transmitted data.

[0175] Transmitter chain 912 can be configured to generate and transmit signals (e.g., control information, data, packets). Transmitter chain 912 may include at least one modulator for modulating data onto a carrier signal, preparing the signal for transmission over a wireless medium. The at least one modulator may be configured to support one or more techniques, such as amplitude modulation (AM), frequency modulation (FM), or digital modulation schemes like phase shift keying (PSK) or quadrature amplitude modulation (QAM). Transmitter chain 912 may also include at least one power amplifier configured to amplify the modulated signal to an appropriate power level suitable for transmission over a wireless medium. Transmitter chain 912 may also include one or more antennas for transmitting the amplified signal into the air or a wireless medium.

[0176] Figure 10 A flowchart illustrating a method according to an aspect of this disclosure is provided. The operation of the method can be implemented by a NE performing the functions of the ML-MMSP 510 as described herein. In some embodiments, the NE may execute a set of instructions to control the functional elements of the NE to perform the described functions.

[0177] At step 1002, the method may include obtaining information related to the ML operation. The operation at step 1002 may be performed according to the examples described herein. In some implementations, aspects of the operation at step 1002 may be determined by reference to [reference needed]. Figure 9 The described NE execution.

[0178] At step 1004, the method may include creating a set of multiple entities to perform the ML operation based on information related to the ML operation. The operation at step 1004 may be performed according to an example as described herein. In some implementations, aspects of the operation at step 1004 may be derived from, as referenced... Figure 9 The described NE execution.

[0179] At step 1006, the method may include a unique group identifier to determine the group. The operation of step 1006 may be performed according to the examples described herein. In some implementations, aspects of the operation of step 1006 may be determined by reference to [reference needed]. Figure 9 The described NE execution.

[0180] At 1008, the method may include issuing a unique group identifier to be delivered to each of the plurality of entities. The operation of 1006 may be performed according to the examples described herein. In some embodiments, aspects of the operation of step 1008 may be provided by reference to [reference needed]. Figure 9 The described NE execution.

[0181] Figure 11 This document explains how the components described herein can be coupled to a wireless communication system 100 that includes an SA6 enable layer.

[0182] Specifically, Figure 11 This describes an architecture for supporting vertical FL (Flexible Interaction) layers, which includes associated AI-enabled analytics entities (e.g., NWDAF) within the 5G core, an enabling layer (specifically the SEALADAE layer or AI enabler) on top of the core network 106, and operation and management (OAM) functions as the AF. This enabling layer spans both the server and device sides (referred to as "enabler clients") and has interfaces to application-specific layers at both the server and client sides. The AI ​​enabler server and AI enabler client may correspond to the clients described in 3GPP TS 23.434, TS 23.558, and TS 23.222.

[0183] It should be noted that the method described herein describes one possible implementation, and the operation and steps may be rearranged or modified in other ways, and other implementations are possible.

[0184] The description herein is provided to enable those skilled in the art to make or use this disclosure. Various modifications to this disclosure will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other variations without departing from the scope of this disclosure. Therefore, this disclosure is not limited to the examples and designs described herein, but should be given the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A network entity for grouping multiple entities to participate in machine learning (ML) operations, wherein the multiple entities are coupled to a wireless communication system, the network entity comprising: At least one memory; and At least one processor, coupled to the at least one memory and configured to enable the network entity to: Obtain information related to the ML operation; Based on the information related to the ML operation, a set of multiple entities is created to perform the ML operation; Determine a unique group identifier for the group; and The unique group identifier is emitted to be delivered to each of the plurality of entities.

2. The network entity of claim 1, wherein the information associated with the ML operation includes one or any combination of the following: (i) an ML model identifier; (ii) a service identifier or service area or service profile; (iii) at least one target user equipment identifier; and (iv) an analysis identifier.

3. The network entity according to claim 1 or 2, wherein the information related to the ML operation includes one or any combination of the following: (v) one or more KPIs of the ML operation; (vi) data requirements of the ML operation; (vii) one or more entity types; (viii) instructions on whether model training and / or inference are required; (ix) model vendor information; and (x) training job ID.

4. The network entity according to any of the preceding claims, wherein the at least one processor is configured to cause the network entity to: Send a request message to the ML model repository, the request message including the information related to the ML operation; and Receive a response message from the ML model repository, the response message including a list of candidate entities to perform the ML operation.

5. The network entity of claim 4, wherein the at least one processor is configured to cause the network entity to: Select the plurality of entities from the list of candidate entities to create the group.

6. The network entity according to any of the preceding claims, wherein the at least one processor is configured to cause the network entity to randomly determine the unique group identifier.

7. The network entity according to any of the preceding claims, wherein the at least one processor is configured to: Determine the mapping from the unique group identifier to a set of member identifiers of the plurality of entities in the group; and The mapping is stored in the at least one memory.

8. The network entity of claim 7, wherein the at least one processor is configured to: Detect trigger events used to modify the group; Based on the triggering event, update the group to create an updated group of multiple entities to perform the ML operation; and The mapping is modified such that the unique group identifier is associated with a set of member identifiers of the plurality of entities in the updated group.

9. The network entity according to claim 8, wherein the triggering event is: The entity joins the group; One of the plurality of entities leaves the group; One of the plurality of entities cannot be used to perform the ML operation; Failure to meet the expected or predicted one or more KPIs of the ML operation; One of the multiple entities is unable to perform the ML operation; or The expected or predicted failure or congestion of the wireless interface coupled to at least one of the plurality of entities.

10. The network entity of claim 8 or 9, wherein the at least one processor is configured to monitor the state of the plurality of entities in the group to detect the triggering event.

11. The network entity of claim 10, wherein the at least one processor is configured to: Periodically send status requests to each of the plurality of entities in the group; and Receive a status response from each of the plurality of entities in the group.

12. The network entity of claim 10, wherein the at least one processor is configured to configure the plurality of entities in the group to periodically transmit state updates to the network entity.

13. The network entity according to any one of claims 8 to 12, wherein the at least one processor is configured to transmit the unique group identifier and the identity of the member of the updated group to deliver to at least one of the plurality of entities in the updated group.

14. The network entity according to any of the preceding claims, wherein the network entity is a Service Enabler Architecture Layer SEAL Group Management Server.

15. The network entity of claim 14, wherein the plurality of entities in the group comprises one or any combination of the following: at least one FL client, at least one FL server, at least one FL aggregator or collaborator, at least one ML client and at least one ML server.

16. The network entity according to any one of claims 1 to 13, wherein the network entity is a Service Enabler Architecture Layer (SEAL) group management server, an AI enabler server, an application data analytics enabler server, or an edge enabler server.

17. The network entity of claim 14, wherein the plurality of entities in the group comprises at least one edge application server and / or at least one UE application client.

18. The network entity according to any of the preceding claims, wherein the at least one processor is configured to transmit the identity of the plurality of entities in the group to be delivered to at least one of the plurality of entities in the group.

19. A method for grouping multiple entities to participate in machine learning (ML) operations, wherein the multiple entities are coupled to a wireless communication system, the method comprising: Obtain information related to the ML operation; Based on the information related to the ML operation, a set of multiple entities is created to perform the ML operation; Determine a unique group identifier for the group; and The unique group identifier is emitted to be delivered to each of the plurality of entities.

20. A processor for grouping multiple entities to participate in machine learning (ML) operations, the processor comprising: At least one controller, coupled to at least one memory and configured to enable the processor to: Obtain information related to the ML operation; Based on the information related to the ML operation, a set of multiple entities is created to perform the ML operation; Determine a unique group identifier for the group; and Output the unique group identifier to each of the plurality of entities.

Images