Support for machine learning-enabled analysis

Abstract

Various embodiments relate to an analytics enabler entity or method for selecting one or more application entities to perform machine learning model processing for application layer data analysis. The analytics enabler entity comprises memory and a memory-coupled processor. The processor is configured to cause the analytics enabler entity to receive service requirements for an application layer data analysis task from an analytics consumer, wherein the service requirements comprise analytics parameters for the analysis task; identify the requirements for providing machine learning-enabled analytics for the analysis task; and, based on the identified requirements for providing machine learning-enabled analytics and the analytics parameters and application entity parameters, select an application entity from candidate entities to perform machine learning model processing.

Description

Technical Field

[0001] The present disclosure relates to wireless communication, and more particularly, to data analysis services.

Background Art

[0002] A wireless communication system may include one or more network communication devices, such as a base station, that can support wireless communication for one or more user communication devices, sometimes known 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 resources of the wireless communication system (e.g., time resources (e.g., symbols, slots, subframes, frames, etc.) or frequency resources (e.g., subcarriers, carriers, etc.)). Additionally, the wireless communication system may support wireless communication across various wireless access technologies, including third-generation (3G) wireless access technology, fourth-generation (4G) wireless access technology, fifth-generation (5G) wireless access technology, particularly suitable wireless access technologies beyond 5G (e.g., sixth-generation (6G)).

[0003] Additionally, the system comprises one or more wireless communication platforms in edge and / or cloud data networks. Such platforms may include applications and / or edge enabling services.

Summary of the Invention

Means for Solving the Problems

[0004] The article “a” preceding an element is unrestricted and is 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” may be interchangeable. When used herein, including in the claims, “or” as used in an enumeration of items (for example, an enumeration of items followed by a phrase such as “at least one of,” “one or more of,” or “one or both of”) indicates an inclusive enumeration, such as an enumeration of at least one of A, B, or C meaning A or B or C or AB or AC or BC or ABC (i.e., A and B and C). Also, when used herein, the phrase “based on” should not be interpreted as a reference to a closed set of conditions. For example, an exemplary step described as “based on condition A” may be based on both condition A and condition B without departing from the scope of this disclosure. In other words, when used herein, the phrase “based on” should be interpreted similarly to the phrase “at least partially based on.” Furthermore, when used herein, including in the claims, “set” may include one or more elements.

[0005] Some implementations of the methods and apparatus described herein may include an analysis enabler entity for selecting one or more application entities to perform machine learning model processing for application layer data analysis, the analysis enabler entity comprising at least one memory and at least one processor coupled to at least one memory, the at least one processor configured to cause the analysis enabler entity to receive service requirements for an application layer data analysis task from an analysis consumer, wherein the service requirements comprise at least one analysis parameter for the application layer data analysis task; identify the requirements for providing machine learning-enabled analysis for the application layer data analysis task; and, upon identifying the requirements for providing machine learning-enabled analysis, retrieve machine learning model information from a machine learning model repository; and select at least one application entity from one or more candidate entities to perform machine learning model processing based on at least one analysis parameter and at least one application entity parameter.

[0006] In one embodiment, the analysis enabler entity is further configured to send machine learning model information to at least one selected application entity, to configure at least one selected application entity using the machine learning model information, and to receive derived application data analysis output from at least one selected application entity based on trained and / or inferred machine learning model data.

[0007] In one embodiment, the analysis enabler entity is further configured to process the received derived application data analysis output based on at least one of the analysis parameters of the service requirement, wherein processing comprises one or more of aggregating or filtering the data from the received derived analysis output, and to send the processed derived analysis output to the analysis consumer.

[0008] In one embodiment, the machine learning model processing comprises at least one ML model lifecycle operation, the model lifecycle operation being one of either a machine learning model inference operation or a machine learning training operation.

[0009] In one embodiment, the analysis enabler entity is further configured to identify at least one candidate entity, optionally at least one of which is an application enabler server and / or client.

[0010] In one embodiment, the analysis enabler entity is further configured to receive analysis parameters in a service requirement that have a representation of a consumer type, and optionally receive that the consumer type is one of the following: a vertical application layer server, an edge application server, an edge enabler server application client, an edge enabler client, a vertical application layer client, a network function, an application function, a management function or server, or an external application, and to select at least one application entity based at least in part on the consumer type.

[0011] In one embodiment, the analysis enabler entity is further configured to identify requirements for machine learning-enabled analysis for an application layer data analysis task by either receiving indications from an analysis consumer or determining, based on at least one analysis parameter, that machine learning-enabled analysis is required.

[0012] In one embodiment, at least one application entity parameter is one of the following: proximity to data producer, signaling cost, latency, data accessibility, credibility level, reliability, energy consumption, cost, or edge network location and applications supported within the edge network.

[0013] In one embodiment, the analytical enabler entity is further configured to select at least one application entity for performing machine learning model processing, wherein the selection of at least one application entity comprises selecting either or both of the following: application layer machine model training functionality or application layer machine learning model inference functionality.

[0014] Some implementations of the methods and apparatus described herein may further include a method for selecting one or more application entities to perform machine learning model processing for application layer data analysis, the method comprising: receiving a service requirement for an application layer data analysis task from an analysis consumer, wherein the service requirement comprises at least one analysis parameter for the application layer data analysis task; identifying a requirement for providing machine learning-enabled analysis for the application layer data analysis task; having identified a requirement for providing machine learning-enabled analysis, obtaining machine learning model information from an application layer machine learning model repository; and selecting at least one application entity from one or more candidate entities to perform machine learning model processing based on at least one analysis parameter and at least one application entity parameter.

[0015] In one embodiment, the method may further comprise the steps of: configuring at least one selected application entity using machine learning model information; and receiving derived analytical output from the at least one selected application entity based on trained and / or inferred machine learning model data.

[0016] In one embodiment, the method may further comprise the steps of processing the received derived analytical output by aggregating or filtering data based on vertical service requirements, and sending the processed derived analytical output to an analytical consumer.

[0017] In one embodiment, the machine learning model processing comprises at least one ML model lifecycle operation, the model lifecycle operation being one of either a machine learning model inference operation or a machine learning training operation.

[0018] In one embodiment, the method may further comprise the step of identifying at least one candidate entity, optionally at least one of the candidate entities being an application enabler server and / or client.

[0019] In one embodiment, the method may further comprise the steps of receiving an analysis parameter in a service requirement that includes a consumer type representation, wherein the consumer type is optionally one of a vertical application layer server, an edge application server, an edge enabler server application client, an edge enabler client, a vertical application layer client, a network function, an application function, a management function or server, or an external application, and selecting at least one application entity based at least partially on the consumer type.

[0020] In one embodiment, the method may further include the step of identifying requirements for machine learning-enabled analysis for an application layer data analysis task, which includes either receiving a display from an analysis consumer, or determining, based on at least one analysis parameter, that machine learning-enabled analysis is required.

[0021] In one embodiment, the method may further comprise at least one application entity parameter being one of the following: proximity to data producer, signaling cost, latency, data accessibility, credibility level, reliability, energy consumption, cost, or location of edge network and applications supported within the edge network.

[0022] In one embodiment, the method may further comprise the step of selecting at least one application entity, which includes the step of selecting either or both of the following: application layer machine model training functionality or application layer machine learning model inference functionality.

[0023] Some implementations of the methods and apparatus described herein may further include a processor for selecting one or more application entities to perform machine learning model processing for application layer data analysis, the processor comprising at least one controller coupled to at least one memory, the at least one controller configured to cause the processor to: obtain service requirements for an application layer data analysis task from an analysis consumer, wherein the service requirements comprise at least one analysis parameter for the application layer data analysis task; identify requirements for machine learning-enabled analysis for the application layer data analysis task;, having identified the requirements for machine learning-enabled analysis, obtain machine learning model information from an application layer machine learning model repository; select at least one application entity from one or more candidate entities to perform machine learning model processing based on at least one analysis parameter and at least one application entity parameter; and output a display of at least one application entity.

[0024] In one embodiment, the processor is further configured to use machine learning model information to configure at least one selected application entity and to receive derived analytical outputs from at least one selected entity based on trained and / or inferred machine learning model data.

[0025] In one embodiment, the processor processes the received derived analysis output based on at least one of at least one analysis parameter of service requirements, and the processing comprises one or more of aggregating data from the received derived analysis or filtering the data, and the processor is further configured to output the processed derived analysis to an analysis consumer.

[0026] A network entity comprising an application entity for performing machine learning model processing for application layer data analysis, the network entity comprising at least one memory and at least one processor coupled to the at least one memory, the at least one processor configured to cause the network entity to receive machine learning model information from an analysis enablement entity, configure based on the machine learning model information to perform one or more of a machine learning training task, a machine learning inference task, or a data collection task, and send the derived analysis output and / or the collected data to the analysis enablement entity.

Brief Description of the Drawings

[0027] [Figure 1] A diagram showing an example of a wireless communication system according to an aspect of the present disclosure. [Figure 2] A schematic diagram showing a high-level architecture for an ADAE service. [Figure 3] A diagram showing an example of a collaborative deployment according to an aspect of the present disclosure. [Figure 4] A diagram showing an example of a machine learning life cycle building block according to an aspect of the present disclosure. [Figure 5] A diagram showing high-level steps of a process for configuring machine learning analysis according to an embodiment of the present disclosure. [Figure 6] An example of a signaling and data flow diagram according to an aspect of the present disclosure. [Figure 7]This is an example of signaling and data flow diagrams according to the aspects of this disclosure. [Figure 8] This figure shows an example of a user device (UE) according to the aspects of this disclosure. [Figure 9] This figure shows an example of a processor according to the embodiments of this disclosure. [Figure 10] This figure shows an example of a network device (NE) according to the embodiments of this disclosure. [Figure 11] This is a flowchart of a method performed by NE according to the aspects of this disclosure. [Figure 12] This is a flowchart of a method performed by NE according to one embodiment. [Modes for carrying out the invention]

[0028] 3GPP® aims to support network data analytics services in 5G core networks, with data analytics services provided by the Network Data Analytics Function (NWDAF) in 3GPP standard TS23.288. The application data analytics enablement server may support the use of machine learning-enabled analytics to predict application layer performance and edge load. However, there is still no mechanism for how artificial intelligence / machine learning lifecycle aspects can be supported in collaborative ADAES deployments. These aspects include, for example, where and how ML models are trained internally or externally from the ADAE layer to derive the analysis, whether machine model inference is required to extend the analysis using real-time application data, and what impact this has on the signaling and capabilities required within the SEAL / ADAE layer.

[0029] The present invention provides an apparatus and method for selecting one or more application entities for performing machine learning model processing for application layer data analysis. The selection may be based on several criteria for both the entities to be selected and the consumer types. This has the advantage of providing an efficient and optimal selection of entities for which machine learning data analysis should be performed.

[0030] The aspects of this disclosure will be described in the context of wireless communication systems.

[0031] Figure 1 shows an example of a wireless communication system 100 according to an aspect of the present disclosure. The wireless communication system 100 may comprise one or more wireless communication platforms. The wireless communication system 100 may also comprise one or more NEs 102, one or more UEs 104, and a core network (CN) 106. The CN 106 may be implemented via at least one NE 102. The NE 102 may comprise a base station, for example, a gNB. One or more NEs 102 may embody at least one of the following: an application enablement layer, an edge enablement layer, a service, or a functionality. Such a layer is part of the system and may be implemented together with the CN 106.

[0032] The wireless communication system 100 may support various radio access technologies. In some implementations, the wireless communication system 100 may be a 4G network such as an LTE network or an LTE Advanced (LTE-A) network. In some other implementations, the wireless communication system 100 may be an NR network such as a 5G network, a 5G Advanced (5G-A) network, or a 5G Ultra Wideband (5G-UWB) network. In other implementations, 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. Additionally, the wireless communication system 100 may support technologies such as Time Division Multiple Access (TDMA), Frequency Division Multiple Access (FDMA), or Code Division Multiple Access (CDMA).

[0033] One or more NE102s may be distributed across a geographical area to form a wireless communication system 100. One or more of the NE102s described herein may be, include, or be referred to as network nodes, base stations, network elements, network functions, network entities, radio access networks (RANs), node B, e-node B (eNB), next-generation node B (gNB), or other preferred terms. The NE102s and UE104s may communicate via a communication link that may be a wireless or wired connection. For example, the NE102s and UE104s may perform wireless communication (e.g., receive signaling, transmit signaling) via a Uu interface.

[0034] NE102 may provide a geographic coverage area to which NE102 can support services for one or more UE104 within the geographic coverage area. For example, NE102 and UE104 may support wireless communication of signals relating to services (e.g., voice, video, packet data, messaging, broadcast, etc.) according to one or more radio access technologies. In some implementations, NE102 may be mobile, for example, a satellite associated with a non-terrestrial network (NTN). In some implementations, different geographic coverage areas 112 associated with the same or different radio access technologies may overlap, but different geographic coverage areas may be associated with different NE102s.

[0035] One or more UE104 may be distributed across the entire geographical area of ​​the wireless communication system 100. UE104 may include, or be referred to as, a remote unit, mobile device, wireless device, remote device, subscriber device, transmitter device, receiver device, or several other preferred terms. In some implementations, UE104 may be referred to as a unit, station, terminal, or client, among other examples. Additionally or alternatively, UE104 may be referred to as an Internet of Things (IoT) device, an Internet of Everything (IoE) device, or a machine-type communications (MTC) device, among other examples.

[0036] UE104 may support direct wireless communication with other UE104s via a communication link. For example, UE104 may support direct wireless communication with another UE104 via a device-to-device (D2D) communication link. In some implementations, such as vehicle-to-vehicle (V2V), vehicle-to-everything (V2X), or cellular V2X deployments, the communication link 114 may be called a side link. For example, UE104 may support direct wireless communication with another UE104 via the PC5 interface.

[0037] An NE102 may support communication with a CN106, or with another NE102, or both. For example, an NE102 may interface with other NE102s or CN106s through one or more backhaul links (e.g., S1, N2, N3, or network interfaces). In some implementations, NE102s may communicate directly with each other. In some other implementations, NE102s may communicate indirectly with each other (e.g., via a CN106). In some implementations, one or more NE102s may include sub-components such as an access network entity, which may be an example of an access node controller (ANC). An ANC may communicate with one or more UE104s through one or more other access network transmitting entities, which may be called radio heads, smart radio heads, or transmit / receive points (TRPs).

[0038] CN106 may support user authentication, access permission, tracking, connectivity, and other access, routing, or mobility functions. CN106 may be an advanced packet core (EPC) or 5G core (5GC) that includes control plane entities managing access and mobility (e.g., Mobility Management Entity (MME), Access and Mobility Management Function (AMF)) and user plane entities routing packets or interconnecting to external networks (e.g., Serving Gateway (S-GW), Packet Data Network (PDN) Gateway (P-GW), or User Plane Function (UPF)). In some implementations, the control plane entities may manage non-access layer (NAS) functions, such as mobility, authentication, and bearer management (e.g., data bearers, signal bearers, etc.), for one or more UE104 serviced by one or more NE102 associated with CN106.

[0039] CN106 may communicate with the packet data network via one or more backhaul links (e.g., via S1, N2, N3, or another network interface). The packet data network may include an application server. In some implementations, one or more UE104s may communicate with the application server. UE104s may establish a session (e.g., a protocol data unit (PDU) session) with CN106 via NE102. CN106 may use the established session (e.g., an established PDU session) to route traffic (e.g., control information, data, etc.) between UE104 and the application server. A PDU session may be an example of a logical connection between UE104 and CN106 (e.g., one or more network functions of CN106).

[0040] In the wireless communication system 100, NE102 and UE104 may use the resources of the wireless communication system 100 (e.g., time resources (e.g., symbols, slots, subframes, frames, etc.) or frequency resources (e.g., subcarriers, carriers)) to perform various operations (e.g., wireless communication). In some implementations, NE102 and UE104 may support different resource structures. For example, NE102 and UE104 may support different frame structures. In some implementations, such as in 4G, NE102 and UE104 may support a single frame structure. In some other implementations, such as in 5G and particularly preferred radio access technologies, NE102 and UE104 may support various frame structures (i.e., multiple frame structures). NE102 and UE104 may support various frame structures based on one or more numerologies.

[0041] One or more numerologies may be supported in the wireless communication system 100, and the numerology may include a subcarrier interval and a cyclic prefix. A first numerology (e.g., μ=0) may be associated with a first subcarrier interval (e.g., 15 kHz) and a normal cyclic prefix. In some implementations, the first numerology (e.g., μ=0) associated with the first subcarrier interval (e.g., 15 kHz) may utilize one slot per subframe. A second numerology (e.g., μ=1) may be associated with a second subcarrier interval (e.g., 30 kHz) and a normal cyclic prefix. A third numerology (e.g., μ=2) may be associated with a third subcarrier interval (e.g., 60 kHz) and a normal cyclic prefix or an extended cyclic prefix. A fourth numerology (e.g., μ=3) may be associated with a fourth subcarrier interval (e.g., 120 kHz) and a normal cyclic prefix. A fifth numerology (e.g., μ=4) may be associated with a fifth subcarrier interval (e.g., 240 kHz) and a normal cyclic prefix.

[0042] The time intervals of resources (for example, communication resources) may be organized according to frames (also called wireless frames). Each frame may have a duration, for example, 10 milliseconds (ms). In some implementations, each frame may contain multiple subframes. For example, each frame may contain 10 subframes, each subframe may have a duration, for example, 1 ms. In some implementations, each frame may have the same duration. In some implementations, each subframe of a frame may have the same duration.

[0043] As an addition or alternative, the time intervals of resources (e.g., communication resources) may be organized according to slots. For example, a subframe may contain a certain number (e.g., a quantity) of slots. The number of slots in each subframe may also depend on one or more numerologies supported in the wireless communication system 100. For example, the first, second, third, fourth, and fifth numerologies (i.e., μ=0, μ=1, μ=2, μ=3, μ=4) associated with the respective subcarrier intervals of 15kHz, 30kHz, 60kHz, 120kHz, and 240kHz may utilize a single slot per subframe, two slots per subframe, four slots per subframe, eight slots per subframe, and sixteen slots per subframe, respectively. Each slot may contain a certain number (e.g., a quantity) of symbols (e.g., OFDM symbols). In some implementations, the number (e.g., a quantity) of slots for a subframe may depend on the numerology. For a normal cyclic prefix, a slot may contain 14 symbols. For an extended cyclic prefix (applicable, for example, to a 60 kHz subcarrier interval), a slot may contain 12 symbols. The relationship between the number of symbols per slot, the number of slots per subframe, and the number of slots per frame for normal and extended cyclic prefixes may depend on the numerology. It should be understood that a reference to a first numerology (e.g., μ=0) associated with a first subcarrier interval (e.g., 15 kHz) can be used interchangeably between subframes and slots.

[0044] In the wireless communication system 100, the electromagnetic (EM) spectrum may be divided into various classes, frequency bands, frequency 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 designations 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 implementations, the NE102 and UE104 may perform wireless communication over one or more of these operating frequency bands. In some implementations, FR1 may also be used by the NE102 and UE104 in equipment or devices for cellular communication traffic (e.g., control information, data). In some implementations, FR2 may be used by NE102 and UE104 among instruments or devices for short-range high data rate capabilities.

[0045] FR1 may be associated with one or more numerologies (for example, at least three numerologies). For example, FR1 may be associated with a first numerology including a 15 kHz subcarrier interval (e.g., μ=0), a second numerology including a 30 kHz subcarrier interval (e.g., μ=1), and a third numerology including a 60 kHz subcarrier interval (e.g., μ=2). FR2 may be associated with one or more numerologies (for example, at least two numerologies). For example, FR2 may be associated with a third numerology including a 60 kHz subcarrier interval (e.g., μ=2), and a fourth numerology including a 120 kHz subcarrier interval (e.g., μ=3).

[0046] In the Third Generation Partnership Project (3GPP), data analytics services are provided by Network Data Analytics Functions (NWDAF) as standardized in 3GPP standard TS23.288, aiming to support network data analytics services in fifth-generation (5G) core networks. Such analytics can collect data from other network functions (NF), application functions (AF), and / or operation and maintenance (OAM), and the analysis results can be provided to third-party application functions to provide statistics and predictions related to slice load levels, observed service experience, NF load, network performance, user equipment (UE) related analytics (mobility, communication), user data congestion, quality of service (QoS) sustainability, data network DN performance, etc. Furthermore, in 3GPP standard SA5 (TS28.104), Management Data Analytics Service (MDAS) provides data analytics for the network. MDAS can be deployed at different levels, for example, at the network element level, e.g., at the gNB level, at the domain level (e.g., Radio Access Network (RAN), Core Network (CN), Network Slice Subnet), or centrally (e.g., at the Public Land Mobile Network (PLMN) level). The purpose of MDAS is to provide root cause analysis for complex problems and to optimize network resource allocation (at the network / domain level and at the slice / slice subnet level).

[0047] Additional analytical capabilities in 3GPP are described in the 3GPP standard SA6 (TS23.436), which specifies the Application Data Analytics Enablement Service (ADAES) for performing application layer and edge / cloud analytics outside the 3GPP domain. ADAES can be viewed as an AF that possesses analytical capabilities and also has interfaces to the UE side (ADAE client) and OAM. Figure 2 is a schematic diagram showing the high-level architecture for the ADAE service 200. This architecture comprises a client side 201 with a vertical application layer VAL client 202 and an Application Data Analytics Enablement Client (ADAE-C) 203, a server side 204 with one or more of a VAL server 205 and an Application Data Analytics Enablement Server (ADAE-S) 206, and a 3GPP network 207.

[0048] In some embodiments, application layer data analysis refers to the analysis defined in the 3GPP standard TS23.436, which is incorporated herein by reference and referred to hereafter as TS23.436. In other embodiments, the term “application layer data analysis” may be more broad and refer to analysis that provides statistics, predictions, or prescriptions for at least one application parameter used for an application service or session, where the application service or session is communicated over a wireless communication system. Examples of such parameters are application service performance indicators, such as round-trip time (RTT), throughput, latency, jitter, application service quality (QoS) or experience quality (QoE), application or UE mobility-related parameters, server availability and reliability-related parameters, edge load, or performance parameters.

[0049] Referring to Figure 2, the VAL server 205 may communicate with ADAES 206 via the ADAE-S reference point 208. ADAES, acting as the application function AF, may communicate with the 5G core network function 207. This communication may involve one or more of the following, which are within the 3GPP network but not shown in Figure 2: the network exposure function NEF, the user plane function UPF, or the operation and maintenance OAM. The communication may be via the N33 reference point 209 when communicating with the NEF, via the N6 reference point 210 when communicating with the UPF, or via the ADAE-OAM interface when communicating with the OAM.

[0050] ADAES supports analytics (e.g., VAL server performance, edge load analysis, location analysis, etc.), and such analytics methods may also be machine learning-enabled. There are different deployments and business models for ADAES supported by TS23.436. Examples of different deployments include ADAES within a public land mobile network (PLMN), in an edge computing service provider (ECSP), or in a vertical domain. Regarding deployment models, there are three possible scenarios: centralized, distributed, and collaborative.

[0051] Figure 3 shows an example of a collaborative deployment 300 according to the embodiments of this disclosure. As described in TS23.436, multiple ADAES can be deployed in different edge data networks (EDNs) or data networks (DNs) and can be deployed by the same ADAE provider. Such collaborative deployments enable local to global analytical derivation, which may be necessary to improve the level of analytical reliability. A centrally deployed ADAES can also act as an ADAE analytical aggregator entity, which can configure edge-deployed ADAES to derive analysis for different sub-areas.

[0052] In the example in Figure 3, there are two EDNs, namely EDN#1 301 and EDN#2 302, and a centralized DN 303. ADAE servers, namely ADAE#1.1 304, ADAE#1.2 305, and ADAE#1 306, are located within each network. Each EDN has Edge Application Servers (EAS) 307 and 308, and Edge Enabler Servers (EES) 309 and 310, while the centralized network has a VAL server 311.

[0053] One example of the use of analysis that can be performed within the exemplary network in Figure 3 is the use of EDN#1 or EDN#2 load values, which can be used to help predict the performance of VAL server 311 within the centrally located network. Such deployments are also applicable to machine learning-based analysis methods such as supervised learning, where the centrally located ADAES can act as machine learning model training entities, and the edge-located ADAES can act as machine learning model inference entities (using edge data to improve prediction accuracy).

[0054] The statistics / forecasts indicating that the edge has deployed ADAES correspond to ADAES service areas 316 and 317, which are equal in size to the EES / EAS service areas. A central ADAE server covers all PLMN areas 318 and is used to coordinate or jointly perform analytics with distributed ADAES. Such analytics services may be provided to consumers at a central DN such as a VAL server or SEAL service, or even on the PLMN side (e.g., NWDAF consuming service experience analytics). ADAES may support the use of machine learning-enabled analytics to predict application layer performance and edge load. However, as described above, there is still no mechanism for how artificial intelligence / machine learning lifecycle aspects can be supported in collaborative ADAES deployments. These aspects include, for example, where and how machine learning models are trained internally or externally to derive analytics, whether machine model inference is required to extend analytics using real-time application data, and what impact this has on the signaling and capabilities required within the SEAL / ADAE layer.

[0055] Figure 4 shows an example of machine learning lifecycle building blocks 400 according to the embodiments of this disclosure. These building blocks comprise data acquisition 401, data preparation 402, machine learning model selection 403, machine learning model deployment 404, machine learning model training 405 and machine learning model inference 406, and prediction / prescription 407 based on training / inference. All of these building blocks may be in one location, for example, in a centralized deployment such as a data center, or alternatively, distributed across edge / cloud and network domains (i.e., core network, management system, and radio access network (RAN)). Furthermore, different variants may be possible based on the objectives and requirements for different vertical deployments. Referring to Figure 4, three configurable options are shown for Vehicle to Anything (V2X) 408, Future Factory (FF) 409, and Unified Access Service (UAS) 410. The interaction of the building blocks of the machine learning lifecycle with the application-specific layer 411, the application enabler layer 412, the edge enabler layer 413, and the NWDAF / MDAS 414 is illustrated.

[0056] One example of ML analysis use is in the case of Automotive / Vehicle-to-Anything (V2X) Vertical 408, where different ML algorithms and ML model types may be used to meet more critical traffic efficiency requirements. Another example is in the case of Gaming Vertical, where different ML model lifecycles may be used to ensure the best quality of experience (QoE) for multiplayer games. For the reasons mentioned above, it is important to configure the following: (i) AI / ML parameters based on service requirements and analytical services (e.g., ML model type and algorithm to be used). (ii) AI / ML lifecycle operational deployments to support different service requirements and network deployments. (iii) Whether and how some AI / ML lifecycle operations are exposed to external entities or entities within 5GS / UE. (iv) How should the processing of different vertical AI / ML lifecycle operations be redundant to minimize signaling and complexity loads while ensuring that the requirements of all vertical customers are met?

[0057] As described above, in order to efficiently perform AI / ML analysis, the embodiment includes an analysis enabler entity whose role is to select one or more application entities for performing machine learning model processing for application layer data analysis. This is done in response to receiving service requirements for application layer data analysis tasks from analysis consumers. The analysis enabler entity identifies the requirements for providing machine learning-enabled analysis, then retrieves machine learning model information from a machine learning model repository, and selects at least one application entity from one or more candidate entities for performing machine learning model processing based on at least one analysis parameter, the retrieved machine learning model information, and at least one application entity parameter.

[0058] The Analytical Enabler entity may be an Application Data Analysis Enabler (ADAE) server or client, as specified in the 3GPP standard TS23.434, incorporated herein by reference and hereafter referred to as TS23.434. Alternatively, the Analytical Enabler entity may be any other SEAL defined entity, as specified in TS23.434, or EDGEAPP, as specified in the 3GPP standard TS23.558, incorporated herein by reference and hereafter referred to as TS23.558. Alternatively, the Analytical Enabler entity may be a new AI / ML support enabler entity dedicated to ML model configuration and processing.

[0059] The analytical enabler entity is configured to provide support for AI / ML lifecycle control aspects, such as controlling ML model processing entities and associated parameters.

[0060] In some implementations, the analytical enabler entity can be deployed as part of the core network and / or as part of OAM (as an extension of MDAS or new AI / ML support functionality).

[0061] The application entity may be located within the network entity and may have the role of performing machine learning processing in response to selections and configurations by the analytics enabler entity. In embodiments, the application entity may be an application enabler entity or edge enabler entity, as defined in 3GPP SA6 incorporated herein by reference, or an application function provided to the network operator by the network operator or a trusted third party, as defined in 3GPP SA2 (3GPP TS23.501), also incorporated herein by reference. In embodiments, the application entity may also comprise an application enabler client in the user equipment UE.

[0062] The service requirements include at least one analytical parameter for the application layer data analysis task, which identifies the requirements for machine learning-enabled analysis for the application layer data analysis task.

[0063] In some embodiments, the application entity parameter is either a current status parameter based on measured and / or historical data related to that parameter, but it may also be an expected or predicted or estimated parameter.

[0064] Figure 5 shows a high-level step 500 of the process for configuring machine learning analysis according to one embodiment of the present disclosure. The architecture comprises a VAL customer 501, a central ADAES 502, a machine learning training entity 503, (A-ADRF) 504, an edge ADAES 505, a central ADAES 506, a 3GPP network 507, and multiple ADAECs 508, 509, 510, 511. Figure 5 further shows instances of ML model inference 512 and analysis output 513. In some implementation forms, the functionality illustrated in ADAES / ADAEC may potentially be deployed as new enabling capabilities in a standalone AI / ML enabler server / client.

[0065] According to one embodiment of this disclosure, a method that can be implemented in an architecture such as that shown in Figure 5 enables the selection of ML model entities and configuration of model lifecycle parameters based on the required analysis and, in one embodiment, also on the customer type. This solution results in extensions of the analysis services provided by the analysis enablement functionality, and, in embodiments, analysis related to VAL server or session performance, as well as analysis related to edge load, as specified in the 3GPP standard TS23.436 incorporated by reference. Those skilled in the art will understand that the architecture is merely an example, that many different configurations of application entities and different numbers of different types of application entities are possible, and that this disclosure is not limited to any one configuration. Network entities capable of performing machine learning processing are sometimes referred to as application entities.

[0066] An analytical process or service selected by an application entity for it may be called an application layer data analysis task. The terms analytical service, analytical process, or analytical task may be used interchangeably. An entity requesting an analytical task may be called an analytical customer, and a request for a task may be called a service request. A service request may comprise one or more of vertical requirements and application requirements. A service request may comprise one or more of the following: consumer ID, analytical ID, analytical filter information, analytical type (forecast, statistical), VAL service ID, target data producer profile criteria, preferred confidence level, relevant area, time validity, destination EAS ID, destination EES ID, or DNN / DNAI. A consumer may be a VAL server as defined in TS23.434. A VAL server may be a generalized server applicable to different verticals, and therefore, in the standard, a VAL server may be a V2X server (as defined in TS23.286, TS23.287), a UAS server (as defined in TS23.255), or an IIOT server. Consumers can also be edge application servers (EAS) or edge enabler servers (EES), or application clients or edge enabler clients (as defined in TS23.558), VAL clients (as defined in TS23.434), NFs (such as NWDAF) or AFs (as defined in TS23.501), management functions / servers, or external applications (e.g., MEC servers or MEC applications).

[0067] In the example in Figure 5, the analyzed customer is VAL customer 501, and the analyzed enabler entity is central ADAES 502. However, those skilled in the art will understand that this disclosure is not limited to this configuration.

[0068] According to one embodiment, the following steps are performed. In the first step, a VAL server, which may be a vertical-specific server or an edge / cloud server, subscribes to C-ADAES (anchor analytics server) (514) to request an analytics service (for example, for analytics ID = "VAL server #1 perf analytics"). This request may also indicate the use of AI / ML-enabled analytics, and optionally, ML model information or profiles and IDs. This request may be called a service requirement.

[0069] In the second step, C-ADAES identifies, based on the deployment, the need for application entities best suited to perform ML-enabled analysis, and in particular model training and inference. C-ADAES also identifies ML model information / profiles for a given analysis ID / event ID (if not provided). C-ADAES may also configure ML model training parameters (algorithm, ...) on request.

[0070] In the third step, C-ADAES requests ML models from an ML model repository based on identification / profile or analysis ID (515). Such a repository may be an A-ADRF or an external registry or a registry in the core network (e.g., NRF, ADRF). C-ADAES receives / fetches the ML models based on the request. Figure 5 shows an A-ADRF, but those skilled in the art will recognize that this is not the only option and this disclosure is not limited to a specific source of model information.

[0071] In the fourth step, C-ADAES determines which one or more entities, i.e., application entities, will undertake ML model training (referred to as the "Selected ML Model Training Entity") based on the requirements / criteria for data collection and model processing. This is selected from a list of candidate ML model training entities, which may include the following: a. C-ADAES, b. Other locally / edge-expanded ADAES, c. One or more VALUEs that support ADAEC and have the capability to perform ML model training (capabilities related to processing power, energy constraints, and latency constraints), d. External entities (for example, ML forecasting services provided by third-party providers).

[0072] The criteria for selecting application entities by the analysis enabler entity may be one or more of the following: (i) Model training should be performed near the data producer. This criterion applies when it is preferable that training be performed locally in an entity close to the source of the data that will be used as input for training. One example might be when training an ML model for edge load analysis on the destination edge platform rather than in a central cloud, when the data source is expected to be provided by the edge platform itself (e.g., real-time and historical load data from the edge platform, such as the number of connections per EAS / EES). (ii) Signaling costs and latency to be minimized for collecting data from distributed data producers. Under this criterion, the primary determinants of the selection are based on the cost and potential delays of collecting data from multiple sources. For example, if ML model training is expected to be performed in a central cloud (C-ADAES) or at the edge / VALUE, and training inputs are expected to be required from OAM, edge platforms, and VALUE, then the cost of selecting an ML model training entity in either the central cloud or at the edge or VALUE should be evaluated, taking into account the time required to collect data and whether this may impact the efficiency of the analytics service. (iii) Trustfulness / rating or reliability or availability of candidate ML model training entities. In some entities, the reliability / availability of these entities may be considered untrustworthy by network operators (for example, an application enabler server may be in a trustworthy operator domain, but VALUE may be untrustworthy), or their availability may change frequently or unpredictably (for example, VALUE may go from connected mode to idle mode, or go out of coverage for some time, or have poor channel quality in some areas), and one criterion for selection may be the reliability / availability of these entities, and selection may be based on this factor. (iv) Energy and processing power constraints on ML model candidates. ML model training requires significant processing power, and some devices (primarily UEs) may have energy constraints. It may be necessary to constrain the use of these entities based on their capabilities and, if available, their power levels (e.g., battery status). Such criteria may be evaluated based on input from VALUEs, or based on profiling entities that should only be used when other entities are unavailable and act as model training entities that are "power-constrained," or based on analytical services. (v) Data accessibility based on different stakeholders involved. ML model training requires significant processing power, and some devices (primarily UEs) may have energy constraints, and the use of these entities may need to be restricted based on their capabilities and, if available, their power levels (e.g., battery status). Such criteria may be evaluated based on input from VALUEs, or based on profiling entities that should only be used when other entities are unavailable and act as model training entities that are "power-constrained," or based on analytical services. (vi) Price / cost of model training by external entities, and the trade-off between cost and benefit. This criterion applies to cases where training may be external based on an agreement with an IT company specializing in ML model training. This may be a practice to offload power-consuming training, but this can be costly as the external entity may charge a lot to achieve it. Therefore, one determinant is whether or not training should be offloaded if the ADAE layer has the capability to do so internally, given the accessibility of the data. (vii) Location of the edge network (for example, based on DNAI). In this case, the selection of the ML model training entity can be based on the location of the edge network and its area coverage, so that for edge analysis, if training is not possible on the destination edge network, training must be performed on another entity that is (topologically or physically) close to it. (viii) Application services supported by the edge network. This criterion concerns selecting ML model training entities based on the application services supported at the destination edge. For example, if the analysis is on EAS performance or load for several EAS (e.g., gaming application services), or for a V2X service, or for a group of different application services, for a target vertical such as a stadium provider, then the edge network undertaking ML model training must support these services, or if there are two or more candidate entities that should act as ML model training entities for selection, one of which supports more services (as a possible criterion). (ix) Requirements related to the analysis task (time granularity, confidence level requirements, analysis service area). Such requirements are primarily performance and service requirements related to the analysis events to which the ML model training will be applied. For example, given a high confidence level or granularity (how quickly the output is expected to be delivered), different choices may be made to select different entities (closer to the edge or in a central location), as these choices may have an impact on the efficiency of the analysis.

[0073] When an application entity is instantiated and connected to an analytics enabler entity, the analytics enabler entity may know a list of available entities that should act as ML model training or inference entities, i.e., application entities. Alternatively, a list of candidate application entities may be pre-configured, or such information may be provided by the ML model repository in response to requests for ML model information. At the time of initial registration / instance creation in each edge cloud, application entities may register their capabilities and data availability with the ML model repository or with the analytics enabler entity.

[0074] If different combinations of criteria are to be used, the decision may be based on a pre-configured policy for each analysis event and / or for each consumer type. Such a policy may be executed as “weights” or “ratings” coefficients used to enable ADAES to select the best entities for performing ML model processing. In some embodiments, the selection decision may be based on solving a weighted utility optimization problem to find the best entities to act as ML training entities based on a given network utility target / objective (e.g., based on a pre-configured policy such as energy efficiency or minimum latency or high reliability level), and on constraints relating to the use of some entities (e.g., relating to credit or data accessibility). Such a problem may be solved by various heuristic / metahouristic methods (e.g., based on graph theory, game theory), or by approximate solutions.

[0075] In the fifth step of the embodiment in Figure 5, C-ADAES joins the selected ML model training entity 503 to begin ML model training (516), and when requesting to join, it may also provide an ML model ID or address for fetching directly from A-ADRF, or it may provide local model parameters. The join response is followed by a positive or negative result (517). If the selected ML model training entity is outside the ADAE layer, this may involve more interactions to set up a service agreement and interactions related to billing, which are outside the scope of the present invention.

[0076] In the sixth step, C-ADAES receives trained ML model data from the selected ML training entities (518).

[0077] In step 7, C-ADAES may also determine a list of entities (local ADAES and / or ADAEC) that should perform ML model inference to support real-time analysis if they themselves do not perform inference. This may be done, for example, to increase the reliability level of the analysis using real-time or near-real-time data, or when the analysis event requires edge / local real-time data (e.g., edge load analysis).

[0078] In one embodiment, the consumer may simply request C-ADAES to train an ML model, or (if the consumer runs the ML model itself) it may provide analytical output. In this embodiment, C-ADAES sends analytical output based on the trained ML model to the consumer, and steps 8, 9, and 10 described below may be omitted.

[0079] In some embodiments, such a decision in the seventh step may occur either with or only after the fourth step. In this instance, the ML training entity sends the trained ML model directly to ADAES / ADAEC, which is expected to perform ML model inference to derive online analysis (in which case the fifth step also includes a list of entities and their addresses that will act as ML model inference entities).

[0080] In step 8, if C-ADAES delegates model inference to a local / edge or UE entity, C-ADAES distributes the ML model to the selected ML model inference entity (local ADAES or ADAEC) (519).

[0081] In step 9, the C-ADAES or one or more ADAE layer entities performing ML model inference derive the analysis output (513).

[0082] In step 10, if C-ADAES delegates model inference to a local / edge or UE entity, C-ADAES collects the analysis output (520) and decides whether to further aggregate or process the output, or simply store it in A-ADRF.

[0083] In the 11th step, C-ADAES sends the derived analysis output (which may be processed, aggregated, or original) to the VAL server / consumer (521).

[0084] Figure 6 shows an example of a signaling and data flow diagram 600 according to an embodiment of the present disclosure. This embodiment is directed towards improving VAL performance analysis services using ML-enabled training and inference at the ADAE layer or via an external entity. In this embodiment, the analysis enabler entity is a centralized ADAE server (C-ADAES) 601 that receives commands from the analysis consumer 602. An application model training entity 603, a 5G network 604, an EDN 605 with ADAE-C 607 accompanied by a VAL server 606, and an A-ADRF registry 608 are provided. Signaling and data transfer comprises the following steps:

[0085] Step 1: A consumer of the ADAES analysis service sends a VAL performance analysis subscription request (including ML model inference capabilities) to ADAES (609). This request is sometimes referred to as a service requirement. In embodiments, this request may include one or more of the following: - Consumer ID, Analysis ID, Analysis Filter Information, Analysis Type (Predictive, Statistical), VAL Service ID, Target - VALUE ID, target VAL server ID, target data producer profile criteria, preferred confidence level, area, and time validity. - ML-enabled analysis requirements / flags, ML model profiles / context to assist in selecting ML models by ADAE layer

[0086] Step 2: ADAES sends a subscription response to the consumer of the analysis service as an affirmative or negative response (610).

[0087] Step 3: In one embodiment, ADAES determines the need for ML-enabled analysis (611). In an alternative embodiment, the request for ML analysis may be sent by the analysis consumer. ADAES then determines at least one entity to be considered for the ML model training entity (internal or external to SEAL), as well as at least one entity for performing ML model inference on the analysis ID and / or consumer ID. In embodiments, the decision may be based on the capabilities of the application entity (i.e., one or more of the entity's energy constraints, processing capacity, availability / reliability), as well as the granularity of the analysis event's expected output (based on, for example, expected analysis output and whether it is an offline or online analysis), preferences for entities that should act as ML model training / inference entities near the relevant data producers, and interoperability constraints assuming a multi-vendor model and data producers.

[0088] One example in one embodiment is training an ML model to predict the RTT deviation for a V2X application session in a target area and time. While ML model training may be performed offline in a central or edge ADAES (given the location of the V2X server), ML model inference may be chosen to be performed in a V2X group leading UEs, or in V2X-UEs acting as roadside units (RSUs), or in a selected set of V2X UEs selected based on their capabilities.

[0089] Step 4a: ADAES (central or anchor) sends a subscription request to the ML model registry (which in one embodiment may be the Application layer-Analytics Data Repository Function (A-ADRF)) to receive one or more ML model identification information / information (e.g., V2X server #1) for the analysis ID and / or consumer type (612). The ML model registry is assumed to have a pre-configured mapping between the analysis ID and the ML model, or this may be determined and provided by ADAES or, in Step 1, by the analytics customer.

[0090] Step 4b: ADAES (Central or Anchor) receives a subscription response (613) which includes ML model information and mapping to analysis ID / consumer type, ML model file address for the initial model, permission to update the ML model (related to who the consumer is), ML model vendor ID, a list of permitted / preferred ML model training entities to train the model, and a billing model for using the ML model.

[0091] Step 5: The ADAES (central or anchor) joins the selected (local / edge / central) function (Application Layer ML Model Training Function) to perform ML model training (614), and provides ML model information and the file address of the initial ML model or the ML model itself. The ADAES may also provide requirements regarding the training algorithm to be used, the type of training, whether it is online or offline training, and the time to provide the trained model and address / entity ID for sending the trained model (if the ML model inference entity is not C-ADAES by another entity, e.g., ADAEC or edge ADAES).

[0092] Step 6: An ADAES may choose to derive an analysis for VAL performance using ML model inference on its own, or it may delegate ML model inference to one or more ADAECs or edge ADAESs and act as an aggregator (615). In the latter case, the ADAES joins an entity that acts as an ML model inference entity and sends the requirements for performing the analysis and the configuration parameters for performing ML model inference. The ADAES may also provide the ML model training entity, the ML model aggregator, and the IDs and addresses of the ML models that report configuration information. As Step X below, the application layer ML model training function (local / edge or global) collects data from data producers to train the model (based on assumptions about the data producers, as in the procedures in Sections 8.2.2 and 8.2.3 of TS23.436).

[0093] Step 7. The ML model training entity trains the ML model (616) (how the collection of data for model training occurs is outside the scope of this solution) and sends the trained model to the ML model inference entity either indirectly (617) or directly (618) via C-ADAES. If the ML model inference entity is VALUE, this occurs via the ADAE-UU interface.

[0094] Step 8. The ML model inference entity (ADAES or ADAEC) performs ML model inference (618) and then derives the analysis (prediction) based on the procedures in Sections 8.2.2 and 8.2.3 of TS23.436 for VAL performance analysis.

[0095] Step 9. The ML model inference entity sends its analysis output to C-ADAES.

[0096] Step 10. C-ADAES processes / filters or aggregates the analysis output from one or more ML model inference entities based on the analysis ID and consumer type (620). Different aggregations or processing may be possible depending on the service requirements.

[0097] Step 11. C-ADAES sends the processed / aggregated or original analytical output to the consumer as a notification (621).

[0098] In the above, the entities selected for performing ML training or inference are sometimes referred to as application entities.

[0099] Figure 7 shows an example of a signaling and data flow diagram 700 according to an embodiment of this disclosure. This embodiment targets improving edge load analysis services using ML-enabled training and inference at the ADAE layer or via an external entity. This procedure is an extension / supplement to the procedure in Section 8.8 of TS23.436 (applying both the request response model and the join notification model).

[0100] In this embodiment, two or more edge ADAES may be used to provide any of the ML model inferences, but the ML model training occurs in the cloud. The scenario for having multiple edge ADAES to provide ML model inferences may be based on the following assumptions: - If the EAS / EES service area is covered by two or more EDNs, which may partially overlap, edge load analysis may be performed on an EAS / EES basis, and therefore, edge load analysis should take into account the analysis outputs from two or more ADAESs among the two or more EDNs. - Edge load analysis per EES / EAS / EDN / DNAI when the analysis request covers a list of destination EES / EAS / EDN / DNAIs (and the edge service areas do not necessarily overlap). - When the processing load or energy consumption in the destination EDN (where the EES / EAS is hosted) is high, edge load analysis and ML model training / inference per EDN / EES / EAS are handled in a different central DN or another ADAES within the EDN where the expected load is lower.

[0101] In this embodiment, the analytics enabler entity is a centralized ADAE server (C-ADAES) 702 that receives commands from the analytics consumer 701. An application model training entity 703, an edge ADAES 704, a 5G network 705, an EDN 706 with EAS / EES 707, and an A-ADRF registry 708 are provided. Signaling and data transfer consist of the following steps:

[0102] Step 1: A consumer of the ADAES analytics service submits an edge analytics enrollment request to ADAES (709). This request is sometimes referred to as a service requirement. In one embodiment, this request may include one or more of the following: - Consumer ID, Analysis ID, Analysis Filter Information, Analysis Type (Predictive, Statistical), VAL Service ID, Target Data Producer Profile Criteria, Preferred Confidence Level, Area, Time Validity, Destination EAS ID, Destination EES ID, DNN / DNAI, - ML-enabled analysis requirements / flags, ML model information including ML model profiles / context to support ML model selection by ADAE layer, ML model identification information (if known by the VAL customer), and an optional list of initial ML models. - A service area that identifies a single EDN service area or a list of EDN service areas.

[0103] Step 2: ADAES sends a subscription response to the consumer of the edge analytics service as an affirmative or negative response (710).

[0104] Step 3: In one embodiment, ADAES determines the need for ML-enabled analysis (711). In an alternative embodiment, a request for ML analysis may be sent by the analysis consumer. ADAES then determines at least one entity to be considered for the ML model training entity (internal or external to SEAL), as well as at least one entity for performing ML model inference on the analysis ID, DNN / DNAI, and / or consumer ID. The determination may, in embodiments, be based on one or more capabilities (energy constraints, processing power, availability / reliability of the entity on the target edge platform), as well as the predicted output granularity of the analysis event (based on, for example, the expected analysis output and whether it is an offline or online analysis), preferences for entities that should act as ML model training / inference entities near the relevant data producers / EDNs, and interoperability constraints assuming a multi-vendor model and data producers. Entities selected to perform ML training or inference may be referred to as application entities.

[0105] The entity determination may not only be for an instantiated ADAES, but may also indicate a target EDN where the ADAES may be instantiable and need to be instantiated to undertake ML training and / or inference. In some embodiments, such a determination may trigger the instantiation of an ADAES in the destination EDN.

[0106] Step 4a: ADAES (Central or Anchor) sends a join request to the ML Model Registry (which may be A-ADRF) to receive one or more ML Model Identifiers / Information (e.g., EAS#1) for the Analysis ID and / or Consumer Type, and based on the Destination EDN / DNAI (712). The ML Model Registry is assumed to have a pre-configured mapping between Analysis IDs / EDNs and ML Models, or this may be determined and provided by ADAES or, in Step 1, by the VAL Customer.

[0107] Step 4b: ADAES (Central or Anchor) receives a subscription response (713) which includes ML model information and mapping to analysis ID / consumer type / EDN, ML model file address for the initial model, permission to update the ML model (related to who the consumer is), ML model vendor ID, ECSP identifying who is expected to provide the ML model, a list of permitted / preferred ML model training entities for training the model in one or more EDNs, and a billing model for using the ML model provided by the vendor / ECSP.

[0108] Step 5: If the ADAES is not an ML model training entity, it joins the entity selected to perform ML model training (714) and provides ML model information and the file address of the initial ML model or the ML model itself. The ADAES may also provide requirements regarding the training algorithm to be used, the type of training, whether it is online or offline training, and the time to provide the trained model and address / entity ID for sending the trained model (if the ML model inference entity is not a C-ADAES by one or more edge ADAES).

[0109] Step 6: An ADAES may choose to use ML model inference on its own to derive an analysis for VAL performance, or it may delegate ML model inference to one or more ADAECs or edge ADAESs and act as an aggregator (715). In the latter case, the ADAES joins an entity that acts as an ML model inference entity and sends the requirements for performing the analysis and the configuration parameters for performing the ML model inference. The ADAES may also provide the ML model training entity, the ML model aggregator, and the IDs and addresses of the ML models that report configuration information.

[0110] Step 7. The ML model training entity trains the ML model (716) (collection of data for model training is outside the scope of this solution) and sends the trained model to the ML model inference entity either indirectly (717) or directly (718) via C-ADAES.

[0111] Step 8. The ML model inference entity (one or more edge ADAES in one or more EDNs) performs ML model inference (719) and then derives an analysis (prediction) based on the procedure in Section 8.8.2 of TS23.436 for edge load performance analysis.

[0112] Step 9. The ML model inference entity sends the analysis output to C-ADAES (720).

[0113] Step 10. C-ADAES processes / filters or aggregates the analysis output from one or more ML model inference entities based on the analysis ID and consumer type (721). Different aggregations or processing may be possible based on service requirements.

[0114] Step 11. C-ADAES sends the processed / aggregated or original analytical output to the consumer as a notification (722).

[0115] Figure 8 shows an example of a UE800 according to an aspect of the present disclosure. The UE800 may include a processor 802, memory 804, controller 806, and transceiver 808. The processor 802, memory 804, controller 806, or transceiver 808, or various combinations thereof, or various components thereof, may be examples of means for performing various aspects of the present disclosure as described herein. These components may be coupled via one or more interfaces (for example, operationally, communicatively, functionally, electronically, or electrically).

[0116] The processor 802, memory 804, controller 806, or transceiver 808, or any combination or component thereof, may be implemented in hardware (e.g., circuit configuration). 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 as a means for performing or otherwise supporting the functions described herein.

[0117] The processor 802 may include an intelligent hardware device (for example, a general-purpose processor, DSP, CPU, ASIC, FPGA, or any combination thereof). In some implementations, the processor 802 may be configured to operate memory 804. In some other implementations, memory 804 may be integrated into the processor 802. The processor 802 may be configured to execute computer-readable instructions stored in memory 804 in order to cause the UE800 to perform various functions of this disclosure.

[0118] Memory 804 may include volatile or non-volatile memory. Memory 804 may store computer-readable computer-executable code, which, when executed by processor 802, causes UE 800 to perform various functions described herein. The code may be stored in memory 804 or in a non-temporary computer-readable medium such as another type of memory. The computer-readable medium includes both non-temporary computer storage medium and communication medium, including any medium that facilitates the transfer of computer programs from one place to another. The non-temporary storage medium may be any available medium that can be accessed by a general-purpose or dedicated computer.

[0119] In some implementations, the processor 802 and the memory 804 coupled to the processor 802 may be configured to cause the UE 800 to perform one or more of the functions described herein (for example, the processor 802 to execute instructions stored in memory 804). For example, the processor 802 may support wireless communication in the UE 800 in accordance with examples such as those disclosed herein. The UE 800 may be configured to support means for receiving machine learning model information from an analytical enabler entity, means for configuring based on the machine learning model information to perform one or more of the machine learning training tasks, machine learning inference tasks, or data acquisition tasks, and means for sending derived analytical outputs and / or collected data to the analytical enabler entity. These features are illustrated and described with respect to user equipment, but this is only an example of the types of entities that can perform these tasks. Any of the entities described above or in the claims may perform these tasks as an application entity.

[0120] The controller 806 may manage input and output signals for the UE800. The controller 806 may also manage peripherals not integrated into the UE800. In some implementations, the controller 806 may utilize an operating system such as iOS®, ANDROID®, WINDOWS®, or other operating systems. In some implementations, the controller 806 may be implemented as part of the processor 802.

[0121] In some implementations, the UE800 may include at least one transceiver 808. In some other implementations, the UE800 may have two or more transceivers 808. Transceiver 808 may represent a wireless transceiver. Transceiver 808 may include one or more receiver chains 810, one or more transmitter chains 812, or a combination thereof.

[0122] The receiver chain 810 may be configured to receive signals (e.g., control information, data, packets) via a wireless medium. For example, the receiver chain 810 may include one or more antennas for receiving signals in the air or via a wireless medium. The receiver chain 810 may include at least one amplifier (e.g., a low-noise amplifier (LNA)) configured to amplify the received signal. The receiver chain 810 may include at least one demodulator configured to demodulate the received signal and obtain the transmitted data by reversing the modulation technique applied during the transmission of the signal. The receiver chain 810 may include at least one decoder for decoding the demodulated signal to receive the transmitted data.

[0123] The transmitter chain 812 may be configured to generate and transmit signals (e.g., control information, data, packets). The transmitter chain 812 may include at least one modulator for modulating data over a carrier signal to prepare a 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 such as phase-shift keying (PSK) or quadrature amplitude modulation (QAM). The transmitter chain 812 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. The transmitter chain 812 may also include one or more antennas for transmitting the amplified signal into the air or into the wireless medium.

[0124] Figure 9 shows an example of a processor 900 according to an aspect of the present disclosure. The processor 900 may be an example of a processor configured to perform various operations as illustrated herein. The processor 900 may include a controller 902 configured to perform various operations as illustrated herein. The processor 900 may optionally include at least one memory 904, which may be, for example, an L1 / L2 / L3 cache. In addition or alternatively, the processor 900 may optionally include one or more arithmetic logic units (ALUs) 906. One or more of these components may communicate electronically or otherwise (for example, operably, communicatively, functionally, electronically, electrically) via one or more interfaces (e.g., buses).

[0125] The processor 900 may be a processor chipset and may include a protocol stack (e.g., a software stack) executed by the processor chipset to perform various operations (e.g., receiving, acquiring, retrieving, transmitting, outputting, transferring, storing, determining, identifying, accessing, writing, reading) in accordance with 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 900), or memory contained therein), 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).

[0126] The controller 902 may be configured to manage and coordinate various operations of the processor 900 (e.g., signaling, receiving, acquiring, retrieving, transmitting, outputting, transferring, storing, determining, identifying, accessing, writing, and reading) in order to enable the processor 900 to support various operations as illustrated in the examples described herein. For example, the controller 902 may act as a control unit of the processor 900, generating control signals that manage the operations of various components of the processor 900. These control signals include enabling or disabling functional units, selecting data paths, initiating memory access, and coordinating the timing of operations.

[0127] The controller 902 may be configured to fetch instructions from memory 904 (e.g., acquire, retrieve, receive) and determine subsequent instructions to be executed to support various operations, such as those described herein. The controller 902 may be configured to track the memory addresses of instructions associated with memory 904. The controller 902 may be configured to decode instructions and the operands involved to determine the operations to be executed. For example, the controller 902 may be configured to interpret instructions and determine control signals to be output to other components of the processor 900 to support various operations, such as those described herein. Additionally or alternatively, the controller 902 may be configured to manage the flow of data within the processor 900. The controller 902 may be configured to control the transfer of data between registers, arithmetic logic units (ALUs), and other functional units of the processor 900.

[0128] Memory 904 may include one or more caches (for example, memory local to or contained within the processor 900), or other memory such as RAM, ROM, DRAM, SDRAM, SRAM, MRAM, or flash memory. In some implementations, memory 904 may reside within or on the processor chipset (for example, locally to the processor 900). In some other implementations, memory 904 may reside outside the processor chipset (for example, remotely from the processor 900).

[0129] Memory 904 may store computer-readable computer-executable code, which, when executed by the processor 900, causes the processor 900 to perform various functions described herein. The code may be stored in a non-temporary computer-readable medium, such as system memory or another type of memory. The controller 902 and / or the processor 900 may be configured to execute computer-readable instructions stored in memory 904 to cause the processor 900 to perform various functions. For example, the processor 900 and / or the controller 902 may be coupled to memory 904, and the processor 900, controller 902, and memory 904 may be configured to perform various functions described herein. In some examples, the processor 900 may include multiple processors, and memory 904 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 various functions described herein.

[0130] One or more ALU906s may be configured to support various operations, such as those described herein. In some implementations, one or more ALU906s may reside within or on a processor chipset (e.g., processor 900). In some other implementations, one or more ALU906s may reside outside of a processor chipset (e.g., processor 900). One or more ALU906s may perform one or more calculations on data, such as addition, subtraction, multiplication, and division. For example, one or more ALU906s may receive input operands and operation codes that determine the operation to be performed. One or more ALU906s may be configured with various logic and arithmetic circuits, including adders, subtractors, shifters, and logic gates, to process and manipulate data according to the operation. As an addition or alternative, one or more ALU906s may support logical operations such as AND, OR, exclusive OR (XOR), inverted OR (NOR), and inverted AND (NAND), enabling one or more ALU906s to handle conditional operations, comparisons, and bitwise operations.

[0131] The processor 900 may support wireless communication in accordance with examples such as those disclosed herein. The processor 900 may be configured or operable to support means for obtaining service requirements for an application layer data analysis task from an analysis consumer, wherein the service requirements comprise at least one analysis parameter for the application layer data analysis task; means for identifying requirements for machine learning-enabled analysis for the application layer data analysis task; means for obtaining machine learning model information from an application layer machine learning model repository once the requirements for machine learning-enabled analysis have been identified; means for selecting at least one application entity from one or more candidate entities for performing machine learning model processing based on at least one analysis parameter and at least one application entity parameter; and means for outputting a display of at least one application entity.

[0132] Figure 10 shows an example of an NE1000 according to an aspect of the present disclosure. The NE1000 may include a processor 1002, a memory 1004, a controller 1006, and a transceiver 1008. The processor 1002, the memory 1004, the controller 1006, or the transceiver 1008, or various combinations thereof, or various components thereof, may be examples of means for performing various aspects of the present disclosure as described herein. These components may be coupled via one or more interfaces (for example, operationally, communicatively, functionally, electronically, or electrically).

[0133] The processor 1002, memory 1004, controller 1006, or transceiver 1008, or any combination or component thereof, may be implemented in hardware (e.g., circuit configuration). 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 as a means for performing or otherwise supporting the functions described herein.

[0134] The processor 1002 may include an intelligent hardware device (for example, a general-purpose processor, a DSP, a CPU, an ASIC, an FPGA, or any combination thereof). In some implementations, the processor 1002 may be configured to operate the memory 1004. In some other implementations, the memory 1004 may be integrated into the processor 1002. The processor 1002 may be configured to execute computer-readable instructions stored in the memory 1004 in order to cause the NE 1000 to perform various functions of this disclosure.

[0135] Memory 1004 may include volatile or non-volatile memory. Memory 1004 may store computer-readable computer-executable code, which, when executed by processor 1002, causes NE 1000 to perform various functions described herein. The code may be stored in memory 1004 or in a non-temporary computer-readable medium such as another type of memory. The computer-readable medium includes both non-temporary computer storage media and communication media, including any medium that facilitates the transfer of computer programs from one location to another. The non-temporary storage medium may be any available medium that can be accessed by a general-purpose or dedicated computer.

[0136] In some implementations, the processor 1002 and the memory 1004 coupled to the processor 1002 may be configured to cause the NE1000 to perform one or more of the functions described herein (for example, the processor 1002 to execute instructions stored in the memory 1004). For example, the processor 1002 may support wireless communication in the NE1000 in accordance with examples such as those disclosed herein. The NE1000 may be configured to support means for receiving service requirements for application layer data analysis tasks from an analysis consumer, wherein the service requirements comprise at least one analysis parameter for the application layer data analysis task; means for identifying requirements for providing machine learning-enabled analysis for the application layer data analysis task; means for retrieving machine learning model information from an application layer machine learning model repository once requirements for providing machine learning-enabled analysis have been identified; and means for selecting at least one application entity from one or more candidate entities for performing machine learning model processing based on at least one analysis parameter and at least one application entity parameter. The controller 1006 may manage input and output signals for the NE1000. The controller 1006 may also manage peripherals not integrated into the NE1000. In some implementations, the controller 1006 may utilize an operating system such as iOS®, ANDROID®, WINDOWS®, or other operating systems. In some implementations, the controller 1006 may be implemented as part of the processor 1002.

[0137] In some implementations, the NE1000 may include at least one transceiver 1008. In some other implementations, the NE1000 may have two or more transceivers 1008. A transceiver 1008 may represent a wireless transceiver. A transceiver 1008 may include one or more receiver chains 1010, one or more transmitter chains 1012, or a combination thereof.

[0138] The receiver chain 1010 may be configured to receive signals (e.g., control information, data, packets) via a wireless medium. For example, the receiver chain 1010 may include one or more antennas for receiving signals in the air or via a wireless medium. The receiver chain 1010 may include at least one amplifier (e.g., a low-noise amplifier (LNA)) configured to amplify the received signal. The receiver chain 1010 may include at least one demodulator configured to demodulate the received signal and obtain the transmitted data by reversing the modulation technique applied during the transmission of the signal. The receiver chain 1010 may include at least one decoder for decoding the demodulated signal to receive the transmitted data.

[0139] The transmitter chain 1012 may be configured to generate and transmit signals (e.g., control information, data, packets). The transmitter chain 1012 may include at least one modulator for modulating data over a carrier signal to prepare a 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 such as phase-shift keying (PSK) or quadrature amplitude modulation (QAM). The transmitter chain 1012 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. The transmitter chain 1012 may also include one or more antennas for transmitting the amplified signal into the air or into the wireless medium.

[0140] Figure 11 shows a flowchart of a method according to an aspect of the present disclosure. The operation of the method may be carried out by a UE as described herein. In some implementations, the UE may execute a set of instructions to control functional elements of the UE to perform the functions described herein.

[0141] Figure 11 shows a flowchart of a method according to an aspect of the present disclosure. The operation of the method may be carried out by an NE as described herein. In some implementations, the NE may execute a set of instructions for controlling the functional elements of the NE to perform the functions described herein.

[0142] In 1102, the method may include receiving service requirements for an application layer data analysis task from an analysis consumer, the service requirements comprising at least one analysis parameter for the application layer data analysis task. The operation of 1102 may be performed according to examples such as those described herein. In some implementations, the operation of 1102 may be performed by an NE as described with reference to Figure 10.

[0143] In 1104, the method may include identifying requirements for providing machine learning-enabled analytics for application layer data analysis tasks. The operation of 1104 may be performed according to examples such as those described herein. In some implementations, the operation of 1104 may be performed by a NE as described with reference to Figure 10.

[0144] In 1106, the method may, after identifying the requirements for providing machine learning-enabled analysis, include in 1108 retrieving machine learning model information from an application layer machine learning model repository, and in 1110 selecting at least one application entity from one or more candidate entities for performing machine learning model processing based on at least one analysis parameter and at least one application entity parameter. The operations of 1106, 1108, and 1110 may be performed according to examples such as those described herein. In some implementations, the operations of 1106, 1108, and 1110 may be performed by an NE as described with reference to Figure 10.

[0145] Figure 12 shows a flowchart of the method according to an aspect of the present disclosure, as a further feature of the embodiment of Figure 11. The operation of the method may be carried out by an NE as described herein. In some implementations, the NE may execute a set of instructions for controlling the functional elements of the NE to perform the functions described herein.

[0146] In 1202, the method shown in Figure 11 may include configuring at least one selected application entity using machine learning model information. The operation of 1202 may be performed according to examples such as those described herein. In some implementations, the operation of 1202 may be performed by an NE as described with reference to Figure 10.

[0147] In 1204, the method may further include, in one embodiment, receiving derived analytical outputs based on trained and / or inferred machine learning model data from at least one selected application. The operation of 1204 may be performed according to examples such as those described herein. In some implementations, the operation of 1204 may be performed by an NE as described with reference to Figure 10.

[0148] In 1206, the method may optionally further include processing the received derived analytical output by, in one embodiment, specifically by aggregating or filtering the data based on vertical service requirements. The operation of 1206 may be performed according to examples such as those described herein. In some implementations, the operation of 1206 may be performed by an NE as described with reference to Figure 10.

[0149] In 1208, the method may optionally further include, in one embodiment, sending the processed derived analytical output to the analytical consumer. The operation of 1208 may be performed according to examples such as those described herein. In some implementations, the operation of 1208 may be performed by an NE as described with reference to Figure 10.

[0150] It should be noted that the methods described herein represent possible implementations, and that the operations and steps may be rearranged or modified in other ways, and that other implementations are possible.

[0151] The descriptions herein are provided to enable those skilled in the art to construct or use the disclosure. Various modifications of the disclosure will become apparent to those skilled in the art, and the general principles set forth herein may be applied to other variations without departing from the scope of the disclosure. Accordingly, the disclosure is not limited to the examples and designs described herein, and should be given the broadest scope that corresponds to the principles and novel features disclosed herein.

[0152] Abbreviation AF Application Function NF Network Function NWDAF Network Data Analysis Function OAM Operation and Maintenance UE User Equipment MDAS Managed Domain Analysis Service C-ADAES Centralized Application Data Analytics Enabler Service / Server ANLF Analysis Logic Function MTLF Model Training Logic Function DNAI (Data Network Access Identifier) ADAEC Application Data Analytics Enabler Client TRLF Trustworthy Rating Logic Function ML (Machine Learning) ADAE Application Data Analytics Activation VAL Vertical Application Layer A-ADRF Application Layer Analysis Data Repository Function RTT (Round Trip Time) EES Edge Enabler Server EAS Edge Application Server EDN (Edge Data Network) DNN Data Network Name [Explanation of Symbols]

[0153] 100 Wireless Communication Systems 102 Network Equipment (NE) 104 User Equipment (UE) 106 Core Network (CN) 200 ADAE Services 201 Client side 202 Vertical Application Layer VAL Client 203 Application Data Analysis Activation Client (ADAE-C) 204 Server side 205 VAL Server 206 Application Data Analysis Activation Server (ADAE-S) 207 3GPP Network, 5G Core Network Function 208 ADAE-S reference point 209 N33 reference point 210 N6 reference point 300 Cooperative deployment 301 EDN#1 302 EDN#2 303 Centralized DN 304 ADAE#1.1 305 ADAE#1.2 306 ADAE#1 307, 308 Edge Application Server (EAS) 309, 310 Edge Enabler Servers (EES) 311 VAL Server 316, 317 ADAES Service Area 318 PLMN area 400 Machine Learning Lifecycle Building Blocks 401 Data Collection 402 Data Preparation 403 Machine Learning Model Selection 404 Machine Learning Model Deployment 405 Machine Learning Model Training 406 Machine Learning Model Inference 407 Prediction / Prescription 408 Vehicle-to-Anything (V2X) 409 Future Factory (FF) 410 Unified Access Service (UAS) 411 Application-Specific Layers 412 Application Enabler Layer 413 Edge Enabler Layer 414 NWDAF / MDAS 501 VAL Customer 502 Central ADAES 503 Machine Learning Training Entities 504 A-ADRF 505 Edge ADAES 506 Central ADAES 507 3GPP Network 508, 509, 510, 511 ADAEC 512 ML Model Inference 513 Analysis Output 601 Centralized ADAE Server (C-ADAES) 602 Analysis Consumer 603 Application Model Training Entity 604 5G network 605 EDN 606 VAL Server 607 ADAE-C 608 A-ADRF Registry 701 Analysis Consumer 702 Centralized ADAE Server (C-ADAES) 703 Application Model Training Entity 704 Edge ADAES 705 5G network 706 EDN 707 EAS / EES 708 A-ADRF Registry 800 UE 802 Processor 804 memory 806 Controller 808 Transceiver 810 Receiver Chain 812 Transmitter Chain 900 processor 902 Controller 904 memory 906 Arithmetic Logic Unit (ALU) 1000 NE 1002 Processor 1004 memory 1006 Controller 1008 Transceiver 1010 Receiver Chain 1012 Transmitter Chain

Claims

1. An analysis enabler entity for selecting one or more application entities to perform machine learning model processing for application layer data analysis, At least one memory, The at least one processor coupled to the at least one memory and The at least one processor provides the analysis enabler entity, Receiving service requirements for an application layer data analysis task from an analysis consumer, wherein the service requirements comprise at least one analysis parameter for the application layer data analysis task. To identify the requirements for providing machine learning-enabled analytics for the aforementioned application layer data analysis task, Identifying the requirements for providing machine learning-enabled analytics, Retrieving machine learning model information from a machine learning model repository, Based on the at least one analysis parameter, the acquired machine learning model information, and the at least one application entity parameter, at least one application entity for executing the machine learning model processing is selected from one or more candidate entities. An analysis enabler entity configured to perform the following actions.

2. Sending the machine learning model information to at least one selected application entity, Using the aforementioned machine learning model information, configure the at least one selected application entity, The recipient receives derived application data analysis output based on trained and / or inferred machine learning model data from at least one selected application entity. The analytical enabler entity according to claim 1, further configured to perform the following:

3. Processing the received derived application data analysis output based on at least one of the at least one analysis parameters of the service requirements, wherein the processing comprises one or more of aggregating or filtering data from the received derived analysis output. Sending the processed derived analytical output to the analytical consumer. The analysis enabler entity according to claim 2, further configured to perform the following:

4. The analytical enabler entity according to any one of claims 1 to 3, wherein the machine learning model processing comprises at least one ML model lifecycle operation, the model lifecycle operation being one of a machine learning model inference operation or a machine learning training operation.

5. An analysis enabler entity according to any one of claims 1 to 4, configured to identify at least one candidate entity, optionally, at least one of the at least one candidate entity being an application enabler server and / or client.

6. Receiving analytical parameters in the service requirements that include a consumer type indication, optionally, the consumer type being one of the following: vertical application layer server, edge application server, edge enabler server application client, edge enabler client, vertical application layer client, network function, application function, management function or server, or external application. Selecting the at least one application entity based at least partially on the aforementioned consumer type and It is further configured to do the following: The analysis enabler entity according to any one of claims 1 to 5, optionally further configured to identify the requirements for machine learning-enabled analysis for the application layer data analysis task by either receiving a display from the analysis consumer or determining, based on the at least one analysis parameter, that machine learning-enabled analysis is required.

7. The analytics enabler entity according to any one of claims 1 to 6, wherein the at least one application entity parameter is one of proximity to data producers, signaling cost, latency, data accessibility, credibility level, reliability, energy consumption, cost, or location of edge network and applications supported within the edge network.

8. The analytical enabler entity according to any one of claims 1 to 7, wherein the analytical enabler entity is configured to select at least one application entity for performing the machine learning model processing, and the selection of at least one application entity comprises selecting either or both of the following: application layer machine model training functionality or application layer machine learning model inference functionality.

9. A method for selecting one or more application entities to perform machine learning model processing for application layer data analysis, A step of receiving service requirements for an application layer data analysis task from an analysis consumer, wherein the service requirements comprise at least one analysis parameter for the application layer data analysis task. The steps include identifying requirements for providing machine learning-enabled analytics for the aforementioned application layer data analysis task, Identifying the requirements for providing machine learning-enabled analytics, Steps include: obtaining machine learning model information from the application layer machine learning model repository, A step of selecting at least one application entity from one or more candidate entities for executing the machine learning model processing based on the at least one analysis parameter and at least one application entity parameter. A method that includes [a certain feature].

10. The steps include: configuring the at least one selected application entity using the machine learning model information; The steps include receiving derived analytical output based on trained and / or inferred machine learning model data from at least one selected application entity, and The method according to claim 9, further comprising:

11. The steps include processing the received derived analysis output by aggregating or filtering the data, particularly based on vertical service requirements, The steps include sending the processed derived analytical output to the analytical consumer. The method according to claim 10, further comprising:

12. The method according to any one of claims 9 to 11, wherein the machine learning model processing comprises at least one ML model lifecycle operation, the model lifecycle operation being one of a machine learning model inference operation or a machine learning training operation.

13. The method according to any one of claims 9 to 12, further comprising the step of identifying at least one candidate entity, wherein at least one of the at least one candidate entity is an application enabler server and / or client.

14. A step of receiving an analysis parameter in the service requirements that includes a consumer type representation, wherein the consumer type is optionally one of a vertical application layer server, an edge application server, an edge enabler server application client, an edge enabler client, a vertical application layer client, a network function, an application function, a management function or server, or an external application. The steps of selecting the at least one application entity based at least partially on the consumer type and Furthermore, The method according to any one of claims 9 to 13, optionally comprising the step of identifying the requirements for machine learning-enabled analysis for the application layer data analysis task, the step of receiving a display from the analysis consumer, or the step of determining that machine learning-enabled analysis is required based on the at least one analysis parameter.

15. The method according to any one of claims 9 to 14, wherein the at least one application entity parameter is one of proximity to a data producer, signaling cost, latency, data accessibility, credibility level, reliability, energy consumption, cost, or location of an edge network and applications supported within the edge network.

16. The method according to any one of claims 9 to 15, wherein the step of selecting at least one application entity comprises the step of selecting one or both of the following: application layer machine model training functionality or application layer machine learning model inference functionality.

17. A processor for selecting one or more application entities to perform machine learning model processing for application layer data analysis, It comprises at least one controller coupled to at least one memory, and the at least one controller provides the processor, Obtaining service requirements for an application layer data analysis task from an analysis consumer, wherein the service requirements comprise at least one analysis parameter for the application layer data analysis task. To identify the requirements for machine learning-enabled analysis for the aforementioned application layer data analysis task, Identifying the requirements for machine learning-enabled analysis, Retrieving machine learning model information from the application layer machine learning model repository, Based on the at least one analysis parameter and the at least one application entity parameter, select at least one application entity from one or more candidate entities for executing the machine learning model processing. Outputting the display of at least one application entity A processor configured to perform the following task.

18. Using the aforementioned machine learning model information, configure the at least one selected application entity, The system receives derived analytical output based on trained and / or inferred machine learning model data from at least one selected entity. The processor according to claim 17, further configured to perform the following:

19. Processing the received derived analysis output based on at least one of the at least one analysis parameters of the service requirements, wherein the processing comprises one or more of aggregating or filtering data from the received derived analysis. Outputting the processed derived analysis to the analysis consumer. The processor according to claim 18, further configured to perform the following:

20. A network entity comprising an application entity for performing machine learning model processing for application layer data analysis, wherein the application entity is At least one memory, The at least one processor coupled to the at least one memory and The system includes, and the at least one processor provides the network entity, Receiving machine learning model information from an enabler entity, The machine learning model information is configured to perform one or more of the following tasks: machine learning training, machine learning inference, or data collection. Send the derived analysis output and / or collected data to the analysis enabler entity. A network entity configured to perform the following actions.

Images