Method, apparatus and system for model selection in communication networks
Patent Information
- Application Number
- PCT/CN2025/100117
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2025-02-26
- Filing Date
- 2025-06-10
- Publication Date
- 2026-09-03
Smart Images

Figure CN2025100117_03092026_PF_FP_ABST
Abstract
Description
METHOD, APPARATUS AND SYSTEM FOR MODEL SELECTION IN COMMUNICATION NETWORKSCROSS-REFERENCE TO RELATED APPLICATIONS
[0001] This application claims priority to US provisional patent application No. 63 / 763,636, filed on February 26, 2025, which is hereby incorporated by reference in its entirety.TECHNICAL FIELD
[0002] The present disclosure relates to the field of communication technologies, and in particular, to a method, apparatus and system for model selection and incorporation of customer feedback in communication networks.BACKGROUND
[0003] In future generation wireless communication systems (e.g., 6G systems) , it is envisioned that customers can request a communication system to perform a variety of tasks. Within the future generation wireless communication system, such tasks are performed by a single Network Capability (NC) or multiple NCs. A customer initiates the process by providing an “intent” to the system. The intent is a natural language description of the requirement of the customer. The system breaks down the intent into a set of specific tasks, and delegates each task to an NC that specializes on performing that task. The system also determines how the NCs interact and exchange data while performing their tasks. The overall program for NC interaction is referred to as the “mission” and performing those tasks is referred to as the “mission execution” . When performing a task, an NC accepts an input and produces an output. The input to an NC may be provided by the customer of a future wireless communication system (e.g., 6G system) , or by other NCs. Similarly, the output of an NC may be provided back to the customer, or to another NC. The NCs themselves may be provided by a third party. This provider of the NC is referred to as the NC Provider (NCP) . Along with the NC, the NCP provides information such as the list of tasks the NC can perform, and the formats of the input and output parameters for each of those tasks. Examples of input and output data may include, but are not limited to, multimedia types, data matrices, vectors or any other types that embody data. A wireless communication system may utilize such information for resolving the mission from the intent.
[0004] This background information is provided to reveal information believed by the applicant to be of possible relevance to the present disclosure. No admission is necessarily intended, nor should be construed, that any one of the preceding information constitutes prior art against the present disclosure.SUMMARY
[0005] In a first aspect, a method implemented by a first service provider node in a network is provided by an embodiment of the present disclosure, and the method includes: receiving information related to a mission and information related to at least one second service provider node, where the information related to the mission is generated based on a customer’s intent and the at least one second service provider node is identified based on the mission; communicating with the at least one second service provider node based on the information related to the mission. The communicating includes: transmitting sample data to the at least one second service provider node; and receiving information related to at least one model from each of the at least one second service provider node, where the at least one model is selected based on the sample data.
[0006] The first service provider receives mission-related information generated based on the customer’s intent and identifies the necessary second service provider nodes to execute the mission. The system can be allowed to break down complex tasks into subtasks and assign them to suitable nodes, thereby increasing the overall accuracy and adaptability of task execution. By using sample data to select models, the second service provider nodes can select the model suited for the current task data dynamically. The efficiency of the second service provider node can be improved by enabling automatic model selection based on the sample data.
[0007] In an implementation, the first service provider node includes a first network element and a second network element, where the first network element performs the method of the first aspect. The first network element may be an execution agent (EA) . The second network element may be a planning agent (PA) .
[0008] In an implementation, receiving the information related to the mission and the information related to the at least one second service provider node includes receiving, by the first network element of the first service provider node, the information related to the mission and the information related to the at least one second service provider node from a second network element of the first service provider node. The first network element is responsible for receiving the mission, coordinating with the components in the second service provider node to execute the mission, providing the second service provider node with service requirement parameters (SRPs) , and updating an SRP policy. The SRPs are parameters that configure the operational aspects of the second service provider node. The SRP policy is a mechanism that generates SRPs according to the operational requirements of the mission. The second network element is responsible for interpreting the intent provided by the customer and generating the mission that accomplishes the customer’s requirement.
[0009] In an implementation, transmitting the sample data includes: receiving, from the at least one second service provider node, a notification indicating that sample data is needed; and receiving the sample data from at least one of a network entity (NE) and a previous second service provider node. By specifying that sample data is required, the system can configure itself to better suit the mission, ahead of receiving the actual data.
[0010] In an implementation, the NE is associated with a customer, and the customer’s intent is received by the second network element from the NE.
[0011] In an implementation, the method further includes: generating respective one or more SRPs for each of the at least one second service provider node, where the one or more SRPs are generated based on the information related to the at least one model; and transmitting the respective one or more SRPs to each of the at least one second service provider node. By generating and transmitting SRPs, the model in the second service provider node can be configured adaptively for the task to be performed, leading to better performance and resource utilization.
[0012] In an implementation, generating the one or more SRPs further includes: receiving network state information; and generating the one or more SRPs based on at least one SRP policy using the network state information. The SRPs can be dynamically adjusted based on real-time network conditions. Therefore, the tasks can be executed efficiently under various conditions.
[0013] In an implementation, the method further includes: receiving feedback from the customer; and updating one or more SRP policies based on the feedback using one or more learning algorithms. By updating the SRP policies according to the feedback from the customer, task executions can be adjusted to better meet customer expectations. The performance and resource utilization of the system can be improved, which leads to higher customer satisfaction.
[0014] In a second aspect, a method implemented by a second service provider node in a network is provided by an embodiment of the present disclosure, and the method includes: receiving sample data from a first service provider node; selecting at least one model from a plurality of models based on the sample data; obtaining information related to the at least one model; and transmitting the information related to the at least one model to the first service provider node.
[0015] In an implementation, receiving the sample data includes: receiving a request related to at least one task of a mission from the first service provider node; and transmitting a response to the first service provider node indicating whether the sample data is needed. Some second service provider nodes may have only one model or a single algorithm designed to handle a specific type of data or task. In such cases, there is no need for sample data to perform model selection because there is only one model to choose from. The requirement for sample data can be specified within the second service provider node itself. By specifying whether sample data is required, the system avoids unnecessary data processing and transmission. This leads to more efficient operation, especially for the second service provider nodes that does not need sample data for model selection.
[0016] In an implementation, the request includes a request to notify if the sample data is needed.
[0017] In an implementation, the sample data is received by the first service provider node from at least one of a network entity (NE) and a previous second service provider node.
[0018] In an implementation, the NE is associated with a customer, and a customer’s intent is received from the NE. The NE may act as an interface between the customer and the network. The NE can provide sample data that represents the type of data the customer wants to process.
[0019] In an implementation, the information related to the at least one model includes information to identify the at least one model.
[0020] In an implementation, the method includes: receiving one or more service requirement parameters (SRPs) from the first service provider node, where the one or more SRPs are generated based on the information related to the at least one model; and converting, the one or more SRPs to respective one or more configuration parameters (CPs) based on the at least one model. The CPs are the second service provider node’s implementation-dependent parameters that are not exposed to the EA.The SRP is a parameter that specifies the requirements for performing a task. The SRPs can include details such as processing delay, resource usage, hardware requirements (e.g., CPU or GPU) , and other operational constraints. Having an SRP to CP conversion mechanism allows the SRPs to be used more efficiently within a second service provider node.
[0021] In an implementation, the second service provider node includes a first network component, a second network component and a third network component and the at least one model includes a first sub-model corresponding to the first network component, a second sub-model corresponding to the second network component and a third sub-model corresponding to the third network component.
[0022] In an implementation, the first network component includes a plurality of likelihood estimators. The first network component may use the plurality of likelihood estimators to evaluate which model is most suitable for processing the sample data. Each LE computes a likelihood value indicating how well its corresponding model matches the sample data. The first component selects the model with the highest likelihood value.
[0023] In an implementation, the information related to the at least one model includes information related to the first sub-model corresponding to the at least one model.
[0024] In an implementation, the method further includes: transmitting, by the first network component to the second network component and the third network component, the information related to the first sub-model corresponding to the at least one model; obtaining, by the second network component, the second sub-model corresponding to the at least one model based on the information related to the first sub-model; and obtaining, by the third network component, the third sub-model corresponding to the at least one model based on the information related to the first sub-model.
[0025] In an implementation, the method further includes: receiving, by the second network component, one or more service requirement parameters (SRPs) from the first service provider node, where the one or more SRPs are generated based on the information related to the at least one model; converting, by the second sub-model of the second network component, the one or more SRPs to respective one or more configuration parameters (CPs) based on the at least one model; and transmitting, by the second network component, the one or more CPs to the third network component.
[0026] In a third aspect, a system is provided according to an embodiment of the present disclosure. The system includes: a network entity (NE) ; an autonomous programming component; a network state observer (NSO) ; and a plurality of second service provider nodes. The autonomous programming component is configured to: receive a customer’s intent from the NE; and receive network state information from the NSO. The autonomous programming component can be the aforementioned first service provider node. The autonomous programming component is further configured to perform the method according to the first aspect or any implementations of the first aspect. Each of the plurality of second service provider nodes can be the aforementioned second service provider node. Each of the plurality of second service provider nodes is configured to perform the method according to the second aspect or any implementations of the second aspect.
[0027] In a fourth aspect, a first service provider node in a network is provided according to an embodiment of the present disclosure. The first service provider node may include various modules configured to execute the method according to the first aspect or any implementations of the first aspect.
[0028] In a fifth aspect, a second service provider node in a network is provided according to an embodiment of the present disclosure. The second service provider node may include various modules configured to execute the method according to the second aspect or any implementations of the second aspect.
[0029] In a sixth aspect, a first service provider node in a network is provided according to an embodiment of the present disclosure. The first service provider node may include at least one processor, where the at least one processor is configured to execute the method according to the first aspect or any implementations of the first aspect.
[0030] In a seventh aspect, a second service provider node in a network is provided according to an embodiment of the present disclosure. The second service provider node may include at least one processor, where the at least one processor is configured to execute the method according to the second aspect or any implementations of the second aspect.
[0031] In an eighth aspect, a computing device cluster is provided according to an embodiment of the present disclosure. The computing device cluster may include a processing circuitry for performing the method according to the first aspect or any implementations of the first aspect, or the method according to the second aspect or any implementations of the second aspect.
[0032] In a ninth aspect, a computer program product is provided according to an embodiment of the present disclosure. The computer program product may include computer-executable instructions which, when executed by a processor, cause the processor to execute the method according to the first aspect or any implementations of the first aspect, or the method according to the second aspect or any implementations of the second aspect.
[0033] In a tenth aspect, a computer program is provided according to an embodiment of the present disclosure. The computer program may include computer-executable instructions which, when executed by a processor, cause the processor to execute the method according to the first aspect or any implementations of the first aspect, or the method according to the second aspect or any implementations of the second aspect.
[0034] In an eleventh aspect, a non-transitory computer-readable storage medium is provided according to an embodiment of the present disclosure. The non-transitory computer-readable storage medium may include computer-executable instructions which, when executed by a processor, cause the processor to execute the method according to the first aspect or any implementations of the first aspect, or the method according to the second aspect or any implementations of the second aspect.
[0035] In a twelfth aspect, according to an embodiment of the present disclosure, a chip is provided. The chip may include an input / output (I / O) interface and a processor, where the processor is configured to call and run computer-executable instructions stored in a memory, to enable a device, in which the chip is present, to execute the method according to the first aspect or any implementations of the first aspect, or the method according to the second aspect or any implementations of the second aspect.BRIEF DESCRIPTION OF THE DRAWINGS
[0036] The accompanying drawings are used to provide a further understanding of the present disclosure, constitute a part of the specification, and are used to explain the present disclosure together with the following specific embodiments, but should not be construed as limiting the present disclosure.
[0037] FIG. 1 is a schematic illustration of a 6G system conceptual structure.
[0038] FIG. 2 illustrates an example deployment of a 6G system according to an evolutionary solution.
[0039] FIG. 3 illustrates an example of an apparatus in a communication system.
[0040] FIG. 4 illustrates a network architecture according to one or more embodiments of the present disclosure.
[0041] FIG. 5 illustrates a network architecture that employs enhanced 5G NFs according to one or more embodiments of the present disclosure.
[0042] FIG. 6 illustrates a system architecture and connectivity between components of the system according to one or more embodiments of the present disclosure.
[0043] FIG. 7A is a schematic flowchart of a method for model selection according to one or more embodiments of the present disclosure.
[0044] FIG. 7B is a schematic flowchart of a communication procedure for model selection according to one or more embodiments of the present disclosure.
[0045] FIG. 8 shows an example of components of a first network element in a first service provider node according to one or more embodiments of the present disclosure.
[0046] FIG. 9A illustrates policy execution in the first network element of the first service provider node according to one or more embodiments of the present disclosure.
[0047] FIG. 9B illustrates policy training in the first network element of the first service provider node according to one or more embodiments of the present disclosure.
[0048] FIG. 10 is a schematic flowchart of a method for model selection according to one or more embodiments of the present disclosure.
[0049] FIG. 11 illustrates an organization of models in a second service provider node according to one or more embodiments of the present disclosure.
[0050] FIG. 12A and FIG. 12B illustrate a model selection mechanism within a first network component (e.g., Service Control Function (SCF) ) of the second service provider node according to one or more embodiments of the present disclosure.
[0051] FIG. 13A, FIG. 13B and FIG. 13C illustrate example methods of training a Likelihood Estimator (LE) according to one or more embodiments of the present disclosure.
[0052] FIG. 14 illustrates an example procedure for training a sub-model of a second network component (e.g., Task Control Function (TCF) ) according to one or more embodiments of the present disclosure.
[0053] FIG. 15 illustrates an example procedure for model submission according to one or more embodiments of the present disclosure.
[0054] FIG. 16 is a call flow diagram illustrating a procedure for a system initialization stage according to one or more embodiments of the present disclosure.
[0055] FIG. 17 is a call flow diagram illustrating a procedure for an initialization stage at a second service provider node according to one or more embodiments of the present disclosure.
[0056] FIG. 18 is a call flow diagram illustrating a procedure for a processing stage at a second service provider node according to one or more embodiments of the present disclosure.
[0057] FIG. 19 illustrates a call flow diagram for a customer feedback stage according to one or more embodiments of the present disclosure.
[0058] FIG. 20 illustrates a procedure for model submission to a second service provider node by a model provider (MP) according to one or more embodiments of the present disclosure.
[0059] FIG. 21 is a schematic structural diagram of a first service provider node according to one or more embodiments of the present disclosure.
[0060] FIG. 22 is a schematic structural diagram of a second service provider node according to one or more example embodiments of the present disclosure.
[0061] FIG. 23 is a schematic structural diagram of an apparatus according to one or more implementations of the present disclosure.DETAILED DESCRIPTION
[0062] In the following description, reference is made to the accompanying figures, which form part of the present disclosure, and which show, by way of illustration, specific aspects of embodiments of the present disclosure or specific aspects in which embodiments of the present disclosure may be used. It is understood that embodiments of the present disclosure may be used in other aspects and include structural or logical changes not depicted in the figures. The following detailed description, therefore, is not to be taken in a limiting sense, and the scope of the present disclosure is defined by the appended claims.
[0063] An evolutionary solution of a 6G system architecture design and procedure design are described in the present application. The evolutionary solution is designed by enhancement of 5G system.
[0064] The proposed 6G network architecture has been designed with a few important principles and requirements: openness, trustworthiness, simplicity in standardization, scalability, rapid deployment of 6G networks and future-proofing.
[0065] The proposed 6G network architecture design applies modularization strategy, utilizes service-based X-as-a-service or anything-as-a-service (XaaS) concepts and network virtualization techniques.
[0066] For all of the procedure designs, modularization of procedures may be implemented. A procedure of the 6G System may include some procedures that can be reused by other procedures. Such a reusable procedure is defined as a basic procedure.
[0067] A complex procedure can, thus, include multiple sequential or parallel basic procedures. It is expected that such methodology can simplify designs of procedures.
[0068] The 6G System leverages service-based architecture and XaaS concept. XaaS services in the 6G System are categorized into three layers. The 6G System conceptual structure is shown in FIG. 1.
[0069] Infrastructure Layer includes infrastructures supporting 6G services. Among them are wireless networks (Radio Access Network (RAN) , Core Network (CN) ) infrastructures, Cloud / data center infrastructures, satellite networks, storage / database infrastructures, and sensing networks, and etc. These infrastructures can be provided by a single provider or by multiple providers.
[0070] In FIG. 1, each XaaS service is provided by identified 5G logical functions. In the evolutionary solution, a XaaS service can be provided with 5G enhancement by more than one approaches. FIG. 1 is only an example.
[0071] In the 6G System conceptual structure: -Network for AI (NET4AI) is a new type of service in 6G CN / RAN which enables network with the capability to conduct / execute AI training / inferencing task (s) . i.e., AI task (s) , by network-based computing and communication resources. In this application, the evolutionary solution to support NET4AI service by enhancing the network data analytics function (NWDAF) in 5G system are described. -A NET4Data service provides a decentralized architecture for data stakeholders to collaboratively manage data lifecycle events. These data lifecycle events include data storage and data sharing. The data could be public, private, sensitive, confidential. In the present application, the NET4Data service could be integrated into the 5GS, or could be enhanced by the 5GS. -Data analysis and management (DAM) focus on different types of data: network data (e.g., data collected from network functions, XaaS service) , ISAC data (3GPP-based sensing data (e.g., from UE and RAN) , Non-3GPP-based sensing data (e.g., from Radar, LiDAR, Wi-Fi Sensing) ) , sensor data (e.g., data from camera sensor, video sensor) , and other data (e.g., Digital user data, 3rd party data, synthetization data, and AI data) . DAM provides services for a variety of data consumers, e.g., XaaS service, 3rd party, NF, UE, etc. 5G system logical functions for example: NWDAF, DCCF, and MFAF of control plane can be enhanced to support DAM service in an evolutionary solution. -Network for Digital World (NET4DW) as a service provides the capability of intelligent integration / synthesis of information from the physical world and digital world (DW) . Customers of NET4DW can be individuals, industries, governments. The customers can have the capability of creation, control, and management of a variety of applications running in the DW such as virtual reality applications. DW services can be supported by enhancing 5G functions and adding new functions (e.g., an evolutionary solution) where necessary. -Network for connectivity (NET4CON) as a service provides a capability to support exchange of messages and data among new 6G services. The basic capabilities of NET4CON include to manage logical topology among XaaS services and between 6G XaaS services and all types of 6G system customers, to introduce intelligent GWs for controlling dynamic forwarding based on configured procedure principle and to support anonymous interactions among these XaaS services and customers by the introduced intelligent GWs. The NET4CON service is provided by enhancement of 5G system. -Mission Management (MM) as a Service provides a capability to program provisioning of XaaS services at Service Layer to provide mission services. A mission is to achieve a designated goal, known as mission goal, which includes providing PDU connectivity and optionally providing data processing. The MM services include the following: mission information management service, mission session management service, mission execution and access management service. -Resource Management (RM) as a Service provides a capability of life-cycle management of a variety of slices and over-the-air resource assignment to wireless devices. -Service Provisioning Management (SPM) as a Service provides a capability of control and management of 6G service access by customers and provisioning of requested services. The capability is provided by ID management, unified authentication, anonymous service authorization and key management. -Connectivity Management (CM) as a service provides a capability of reachability management of 6G wireless devices and D-users in NET4DW in order to support connectivity establishment between wireless devices / D-Users and XaaS services of 6G System. Note that physical locations of D-Users can be changed. A CM service can be deployed across multiple Basic Architecture Structure (BAS) domains. -Protocol as a Service provides a capability to design service customized protocol stacks for identified interfaces.
[0072] FIG. 2 illustrates an example deployment of a 6G system according to an evolutionary solution.
[0073] The “+” represents “enhanced” , for example, the 5G AMF-Mobility function is enhanced, denoted as AMF-Mobility+, the 5G RRC function is enhanced, denoted as RRC+, the 5G Network Repository Function (NRF) is enhanced, denoted as NRF+, the 5G Session Management Function (SMF) is enhanced, denoted as SMF+, the 5G Network Exposure Function (NEF) is enhanced, denoted as NEF+, the 5G Authentication Server Function (AUSF) is enhanced, denoted as AUSF+, other enhanced functions are not described in detail herein.
[0074] C / M Radio Bearer (C / M RB) of a 6G device: over-the-air connection for carrying control signaling for over-the-air interface management and C / M plane messages. A 6G device can have multiple C / M RBs.
[0075] Data Radio Bearer (Data RB) of a 6G device: over-the-air connection for carrying Data plane traffic. A 6G device can have multiple Data RBs.
[0076] RB endpoint: endpoint of an RB at network side. An endpoint of an RB protocol stack (e.g., PDCP) can be in, e.g., a RAN BAS domain, but not limited to. In other words, an RB endpoint can be flexibly deployed / selected for a device.
[0077] RB handler: over-the-air interface protocol stack handler. An RB handler is defined as a logical function which perform RB protocol stack operations after getting configurations. A protocol handler is PDCP-only handler or whole protocol stack handler. An RB handler accepts RB configuration from Connectivity Management (CM) service. An RB handler also accepts security configuration, e.g., keying material, from Service Provisioning Management (SPM) service.
[0078] The NET4CON service which is main service impacting on 6G system architecture is implemented by enhanced 5G Service Communication Proxy (SCP+) as C / M plane GW and enhanced 5G User Plane Function (UPF+) as data plane GW. Proposed per device / D-User C / M session and data session are defined as logical connection between a device / D-User and its serving SCP+ (C / M-TW-GW) and serving UPF+ (Data-TW-GW) . All XaaS services are deployed across multiple BAS / clouds.
[0079] The 6G customer can be of various types, including a device (e.g., electronic device ED, terminal device) , apparatus, a chip, an equipment (e.g., user equipment) etc. For example, the customer may be an individual customer, a business customer, etc. The 6G customer is used to connect persons, objects, machines, etc. The 6G customer may be widely used in various scenarios including, for example, cellular communications, device-to-device (D2D) , vehicle to everything (V2X) , peer-to-peer (P2P) , machine-to-machine (M2M) , MTC, internet of things (IoT) , virtual reality (VR) , augmented reality (AR) , mixed reality (MR) , metaverse, digital twin, industrial control, self-driving, remote medical, smart grid, smart furniture, smart office, smart wearable, smart transportation, smart city, drones, robots, remote sensing, passive sensing, positioning, navigation and tracking, autonomous delivery and mobility, etc.
[0080] Each 6G customer represents any suitable end user device for wireless operation and may include such devices (or may be referred to but not limited to) as a user equipment (UE) or a user device or a terminal device, a wireless transmit / receive unit (WTRU) , a mobile station, a fixed or mobile subscriber unit, a cellular telephone, a station (STA) , a MTC device, a personal digital assistant (PDA) , a smartphone, a laptop, a computer, a tablet, a wireless sensor, a consumer electronics device, a smart book, a vehicle, a car, a truck, a bus, a train, or an IoT device, wearable devices (such as a watch, a pair of glasses, head mounted equipment, etc. ) , an industrial device, or an apparatus in (e.g. module, modem, or chip) or including the forgoing devices, among other possibilities. Future generation 6G customer may be referred to using other terms. When a 6G customer performs (or is configured to perform) a method described herein, it may be interpreted as the ED, one or more module (or units) in the ED, a circuit or chip, or a combination thereof, may perform the method. For example, the circuit or chip may include a modem chip, also referred to as a baseband chip, a system on chip (SoC) including a modem core, or system in package (SIP) ) , and the like, and may be responsible for one or more communication functions in the ED.
[0081] FIG. 3 illustrates an example of an apparatus 320 in a communication system (e.g., a future generation network architecture illustrated in FIG. 2) . The apparatus 320 may be a UE, a network node such as the AN, any components in the AN, the CN or any Network Function of the CN (AMF+, SMF+ or any other network functions illustrated in FIG. 2) . As shown in FIG. 3, the apparatus 320 may include at least one processor 260. Only one processor 260 is illustrated to avoid congestion in the drawing. The processor 260 may perform (or control the apparatus 320 to perform) operations (or methods) described herein as being performed by the apparatus 320.
[0082] When the apparatus is the AN, components of the AN or the apparatus is the UE, the apparatus 320 may further include a transmitter 252 and a receiver 254 coupled to one or more antennas. One, some, or all of the antennas may alternatively be panels. The transmitter 201 and the receiver 203 may be integrated, e.g. as a transceiver. The transceiver is configured to modulate data or other content for transmission by at least one antenna or a network interface controller (NIC) . The transceiver is also configured to demodulate data or other content received by the at least one antenna. Each transceiver includes any suitable structure for generating signals for wireless or wired transmission and / or processing signals received wirelessly or by wire. Each antenna includes any suitable structure for transmitting and / or receiving wireless or wired signals. In present disclosure, the transceiver (or transmitter 252 and / or receiver 254) may be viewed as an interface circuit.
[0083] The apparatus 320 may include at least one memory 258. The memory 258 stores instructions used to perform operations described herein. The memory 258 may also store data used, generated, or collected by the apparatus 320. For example, the memory 258 could store software instructions or modules configured to implement some or all of the functionality and / or embodiments described herein and that are executed by the one or more processors 260.
[0084] A person skilled in the art should understand that embodiments of this application may be provided as a method, an apparatus or system, computer-readable storage medium, or a computer program product. Therefore, this application may use a form of a hardware-only embodiment, a software-only embodiment, or an embodiment with a combination of software and hardware. Moreover, this application may use a form of a computer program product that is implemented on one or more computer-usable storage media (including, but not limited to, a disk memory, an optical memory, and the like) that include computer-usable program code.
[0085] FIG. 4 illustrates a network architecture according to one or more embodiments of the present disclosure. As shown in FIG. 4, a user equipment (UE) is an extended reality (XR) device. The UE may be equipped with communication units to connect with one or more communication networks.
[0086] The mobile network may have an Access Network (AN) that may provide wireless or wired interfaces, or both, for the UE to connect with a data network (DN) , and NFs of the Access Network (AN) and Core Network (CN) . The AN may include a radio management unit, and transmit (Tx) and receive (Rx) points to support radio transmission and reception, and sensing functionalities.
[0087] FIG. 4 illustrates a core network architecture. The CN may include one or more of the following network functions (NFs) : -Autonomous Network Capability Programming (A-CAP) function: This function provides autonomous programming capabilities or services. The A-CAP function knows the capabilities of other NFs, and selects different NFs to perform one or more tasks requested by a network service consumer, such as a UE, an AF, or NFs. -Connection Management Function (CMF) : The CMF may provide functionalities to support control plane (CP) signaling between the electronic devices (EDs) , the UE and the NFs in the CN. The CMF may also manage the mobility of the electronic device (ED) and the UE. -Session Management Function (SMF) : The SMF may provide control plane functionalities to create and manage user plane or data plane connections between the ED, the UE and the NFs, and between the ED, the UE and the data network (DN) . -Data Storage Function (DSF) : The DSF may provide functionalities to store data of one or more of UE data, user data, NF data, application data, and network operation data, and any other types of data. The DSF may be a unified data repository (UDR) function in a 5G system. -Data Management Function (DMF) : The DMF may provide functionalities to manage one or more DSFs. For example, some NF may send a data record of a data type to the DMF, then the DMF may select a DSF instance to store certain types of data. The DMF may be a Unified Data Management (UDM) function in a 5G system.
[0088] In some embodiments, the DMF may manage UE and ED subscription data. The ED and UE data subscription data may be provided by one or more of the following methods: -The OAM (Operation, Administration, and Maintenance) function of the network may provide or configure one or more parameters of user subscription data. -A network entity (NE) , such as ED, DWCF, may provide one or more of parameters of user subscription data. The DWCF may manage some real-time digital twin (DT) information of the UE and ED, e.g., real-time location, or assigned / expected service location of an ED. The DWCF may send the real-time DT information of the ED to the DMF and the DMF may store the DT information of ED in a DSF. -Policy Function (PF) : The PF may create policies for different operations of the network and may provide policies to NFs, EDs, UEs, AN, DN. -Security Function (SF) : The SF may provide one or more authorization functions, authentication function, and data security protection for one or more EDs, UEs, NFs in the AN, NFs in the CN, AN, and NFs in the DN. -Location Management Function (LMF) : The LMF may provide one or more functionalities: detect the UE and ED location, estimate the location of the UE and ED, track the mobility of the UE and ED. -Network Entity Repository (NER) : The NEF may provide functionalities for a network entity (NE) to register its NE profile so that other NEs can discover, select, and use the services of this NE. -Control Plane Gateway (CP GW) : The CP GW may provide an interface for NFs in the DN to access the services provided by NFs of the mobile network. -Data Plane Function (DPF) : The DPF may provide one or more services: receiving data of NDT and NFs; processing the received data; forwarding the received data; sending processed data. -Data Plane Gateway (DP GW) : The DP GW may provide an interface to send or receive data between the mobile network and other entities in the DN.
[0089] The mobile network may provide NFs to host or support digital world (DW) applications. Some example of DW applications may include digital twin applications, metaverse applications, and other applications.
[0090] The following NFs may support DW applications: -Data Collection and Distribution Function (DCDF) : The DCDF may provide one or more of the following functionalities: Data collection from NEs, such as sensors, UE, NF in the mobile network, NF in the DN; data storage management for the collected data stored in one or more Sensor Data Storage Functions (SDSF) ; data distribution to other NFs that request the data. -DW Control Function (DWCF) : The DWCF may perform one or more tasks to create and manage DW applications which include managing the operation of DW applications. -DW artificial intelligence and machine learning (AIML) model training function (MTF) : The MTF may use the collected sensor data, or any other types of data such as partially developed AI or ML models developed by other NEs during a federated learning process, to derive an AI or ML model to support DW applications. -DW artificial intelligence and machine learning (AIML) model repository function (MRF) : The MRF may provide one or more of the following services: store the AIML models derived by the MTF, and distribute AIML models to other NFs and ED. An AIML model may be complete or incomplete. If the AIML model is complete, other NEs can use the complete AIML model to infer the data. If the AIML model is incomplete, other NEs may use the incomplete AIML model to further develop the incomplete AIML independently or jointly to create a complete AIML model. -Object Context Repository function (OCRF) : The OCRF may provide one or more services: store object context in real-time, and distribute object contexts to subscribed NFs. -Object context: e.g., UE context, ED context, NF context. -DW Data Processing Function (DW DPF) : The DW DPF may provide one or more services: -Get one or more AIML models from the MRF; -Get the sensor data from the UE, ED and NFs; -use one or more of AIML models or other methods to process the collected sensor data to detect the real world (RW) objects; convert the detected RW objects into one or more virtual world (VW) objects that can be used by one or more DW applications; -Run application software of DW applications; -Generate actuator data for actuator devices: For example, video data for video games, patient monitoring videos in hospitals, robot monitoring in smart factories, vehicle monitoring for intelligent transport system operator, lighting control in smart city or performance data. -Send actuator control command and actuator data to actuator devices.
[0091] The DN may host one or more applications, e.g., DW applications. The DW applications may be implemented by having a DW Controller (DWC) hosted in an Application Function (AF) , and a DW application hosted in an Application Server (AS) . The DWC may provide control functionalities. The DW AS may host application software of DW applications.
[0092] In some examples, as illustrated in FIG. 5, the functionalities of the above NFs may be implemented by modifying NFs of the 5G network as follows:
[0093] FIG. 5 is an illustration of a network architecture that employs enhanced 5G NFs according to one or more embodiments of the present disclosure.
[0094] The 5G access and mobility management function may be enhanced (5G AMF+) to provide functionalities of the CMF.
[0095] The 5G session management function may be enhanced (5G SMF+) may be enhanced to provide functionalities of the SMF.
[0096] The 5G policy control function may be enhanced (5G PCF+) to provide functionalities of the PF.
[0097] The 5G network exposure function may be enhanced (5G NEF+) to provide functionalities of the CP GW.
[0098] The 5G network repository function may be enhanced (5G NRF+) to provide functionalities of the NER.
[0099] The 5G unified data management function may be enhanced (5G UDM+) to provide functionalities of the DMF.
[0100] The 5G unified data repository may be enhanced (5G UDR+) to provide functionalities of the SDRF.
[0101] The 5G Authentication Server Function may be enhanced (5G AUSF+) to provide functionalities of the SF.
[0102] The 5G Data Collection Coordination Function may be enhanced (5G DCCF+) to provide functionalities of the DCDF.
[0103] The 5G Location Management Function may be enhanced (5G LMF+) to provide functionalities of the LMF.
[0104] The 5G Service Communication Proxy may be enhanced (5G SCP+) to support the DWCF indirect communications with other control plane functions.
[0105] Data plane functions:
[0106] The 5G User Plane Function may be enhanced (5G UPF+) to provide functionalities of the DPF, DWDPF, and DP GW.
[0107] The 5G Analytics Data Repository Function may be enhanced (5G ADRF+) to provide functionalities of the OCRF, SDSF, MRF.
[0108] The 5G Network Data Analytics Function (NWDAF) Model Training Logical Function (MTLF) may be enhanced (5G NWDAF-MTLF+) to provide functionalities of the MTF.
[0109] The DW may provide services to mobile users, UE, NF, and AF. The services of DW may be accessible by using natural languages. In this disclosure, the services of DW are described by a Network Capability Description Language (NCDL) . The capabilities or services of DW are described in a friendly format for artificial intelligence (AI) and / or large language model (LLM) to enable full automation of network operation.
[0110] In future generation wireless communication systems (e.g., 6G systems) , it is envisioned that customers can request the communication system to perform a variety of tasks. Within the future generation wireless communication system, such tasks are performed by a single Network Capability (NC) or multiple NCs. A customer initiates the process by providing an “intent” to the system. The intent is a natural language description of the requirement of the customer. The system breaks down the intent into a set of specific tasks, and delegates each task to an NC that specializes on performing that task. The system also determines how the NCs should interact and exchange data while performing their tasks. The overall program for NC interaction is referred to as the “mission” and performing those tasks is referred to as the “mission execution” .
[0111] When performing a task, an NC accepts an input and produces an output. The input to an NC may be provided by the customer of a future wireless communication system (e.g., 6G system) , or by other NCs. Similarly, the output of an NC may be provided back to the customer, or to another NC. For instance, consider a mission that includes two NCs. The first NC retrieves data from the customer, processes the data, and then passes the processed data to the second NC. The second NC performs further operations on the data and sends the final result back to the customer. In this example, the two NCs create a sequential processing system. However, a mission may include more complex processing topologies.
[0112] The NCs themselves may be provided by a third party. This provider of the NC is referred to as the NC Provider (NCP) . Along with the NC, the NCP provides information such as the list of tasks the NC can perform, and the formats of the input and output parameters for each of those tasks. Examples of input and output data include, but are not limited to, multimedia types, data matrices, vectors or any other types that embody data. The 6G system makes use of such information for resolving the mission from the intent.
[0113] Firstly, some NCs may employ a variety of algorithms and models to perform a task. For example, consider the task of upscaling images with super-resolution imaging. An NC may provide super-resolution imaging capability as a part of its services. Upon providing a low-resolution image, an algorithm in the NC converts the low-resolution image into a realistic high-resolution image. This conversion may be accomplished by an Artificial Intelligence (AI) model or any other type of algorithm designed or trained to perform the conversion.
[0114] Secondly, prior to performing a task, an NC can be configured using a set of Service Requirement Parameters (SRP) . An SRP can be in any data format that represents a parameter. For example, an NC can be configured to either use or not use a hardware accelerator (e.g., a graphics processing unit (GPU) or tensor processing unit (TPU) ) instead of the central processing unit (CPU) . This configuration can be provided with a binary SRP that selects one of the two available choices. There may exist other types of parameters as well. For example, a 3-ary parameter may configure whether the computations are required to be performed using 32-bit, 16-bit, or 8-bit floating point numbers, and a continuous parameter may configure the target maximum delay for a task. When the NC is provided, the NCP specifies the default SRPs that are suitable for the average use of the NC.
[0115] Aspects of the present disclosure may address one or more of the following issues: The NCs may lack the ability to support inputs from different datasets. For example, considering the super-resolution imaging example, it is assumed that the NC uses an AI model and the model is trained on a dataset consisting of only landscape sceneries (e.g., scenes of waterfalls, mountains etc. ) . This model may not be suitable to perform super-resolution imaging on the images from a different dataset, say a dataset consisting of images of wildlife (e.g., images of zebras and lions) . If the customer’s input to an NC is from the dataset that is considerably different from the dataset used to train the model, the mismatch between the training data and the input data may yield an inferior performance in the output.
[0116] One solution to address the dataset mismatch issue is to train an AI model using many possible training datasets. However, the complexity of the model that is required to capture details of all training datasets might be quite large to a point where such an approach may not be feasible. Even if such a large model is trained, the computation requirement for a single use of the model may be significantly higher than that for a (smaller) model trained on a single dataset. In addition, the model needs to be retrained every time a new dataset is available, which may pose significant scalability and maintainability issues.
[0117] Furthermore, NCs may lack the ability for the customers to provide feedback based on the services they receive. As a result, the NCs may miss the opportunity to improve themselves based on the customer’s feedback. For example, consider a situation where an NC provides a service for generating artificial but realistic images. The NC may use a CPU or a GPU for the image generation. This can be configured by providing a binary SRP. Using the CPU may be slower but comes with low cost (cost for resource use) , and using the GPU may be faster, however, it comes with high cost. When the customer sends the request to generate an image, the system may be able to find a combination of SRPs that can balance the time and cost. After providing the final output to the customer, the customer may be able to provide feedback of whether the actual time and cost struck the right balance. This process becomes more complex when multiple NCs interact to perform a task, and each NC is configurable with a set of SRPs. Existing solutions may lack such a mechanism for selecting SRPs and the means for incorporating the user feedback.
[0118] Aspects of the present disclosure resolve the aforementioned drawbacks which may include the inability of a model to support inputs from multiple datasets, and a lack of a mechanism for generating SRPs for NCs and providing feedback to customers.
[0119] According to a first aspect of the present disclosure, a mechanism is introduced for supporting multiple models in an NC, and a method is provided for selecting the most suitable model based on the user’s input. This system is referred to as the Model Selection Mechanism (MSM) . All models perform the same task. However, each model specializes in one particular dataset. When the NC is provided with sample data, it selects the model that works best for the sample data, and uses that model for processing the actual input from the customer. The sample data may be provided by the customer or other NCs. A few examples of this mechanism are as follows.
[0120] Example 1: Consider an NC that performs super-resolution imaging. The customer provides a set of low-resolution images, and the NC produces and converts the low-resolution images to high-resolution images. For this task, the NC may use two models, where model 1 is specifically trained for landscape images and model 2 is specifically trained for images of wildlife. Given some sample data consisting of a few low-resolution images, the NC may decide which models to use for processing the whole dataset.
[0121] Example 2: Consider an NC that performs audio transcribing. The customer provides an audio file containing a spoken conversation or speech, and the NC processes this audio input to produce a text transcription of the spoken words. For this task, the NC may use two models, where model 1 is specifically trained for transcribing English conversations and model 2 is specifically trained for transcribing conversations in Spanish. The customer provides a few seconds of the audio clip as sample data, based on which the NC decides which model to use.
[0122] This method of selecting models scales easily as the number of training datasets and the models grow, and a single use of the system does not incur more computational cost than that of a single model. For example, it is easy to add a third model trained on French.
[0123] According to a second aspect of the present disclosure, a mechanism is introduced for customers to provide feedback based on the outputs they receive. The feedback is then incorporated into the system for providing better SRPs. This mechanism is referred to as the Self-Improvement Mechanism (SIM) . While a mission is executed, the overall coordination of NCs is conducted by the Autonomous Network Capability Programming (A-CAP) function. The A-CAP is responsible for providing the NCs with SRPs. The SRPs are generated by an artificial neural network referred to as the SRP policy, or policy in short. The SRPs configure the NCs prior to the execution of the mission. A few examples where SRPs configure an NC are as follows.
[0124] Example 1: Consider an NC consisting of a hardware accelerator (e.g., a GPU or a TPU) in addition to the CPU. The A-CAP may produce a binary SRP that selects whether or not to use the hardware accelerator. Specifically, the A-CAP outputs an integer where the value is 0 if only CPU should be used, and 1 if the hardware accelerator may be used.
[0125] Example 2: The NC may have the option of using 32-bit, 16-bit, or 8-bit floating point numbers in its computations. The A-CAP may output an integer SRP where the value is 0, 1, or 2 for each of the corresponding options.
[0126] Example 3: The NC may be configured to operate with a maximum delay for a given task. The A-CAP may output a continuous SRP within a range (e.g., 0.5 seconds to 3 seconds) to configure the maximum delay.
[0127] When the mission execution is concluded, the customer provides feedback to the A-CAP. For example, the feedback may in the form of a scalar, where the value is positive if the output is satisfactory, and negative if the output is not satisfactory. When the feedback is received, the A-CAP adjusts the SRP policy in accordance with the feedback. The updated criteria take effect in the next time the service is used.
[0128] FIG. 6 illustrates a system architecture and connectivity between components of the system according to one or more embodiments of the present disclosure.
[0129] Referring to FIG. 6, the system model of the present invention includes a Network Entity (NE) 600, an Autonomous Network Capability Programming (A-CAP) function 603, a Network State Observer (NSO) 606, and multiple Network Capabilities (NCs) 608 and 609. The A-CAP 603 consists of a Gateway (GW) 605, a Planning Agent (PA) 604, and an Execution Agent (EA) 607. The A-CAP 603 can also be referred to as a first service provider node in the present disclosure. The NE 600 consists of a Control Plane 601 and a Data Plane 602. Each NC 608 or 609 consists of a Service Control Function (SCF) 6081 or 6091 and a Task Control Function (TCF) 6082 or 6092 in its control plane 601, and a Processing Service Function (PSF) 6083 or 6093 in its data plane 602. Each NC 608 or 609 has an NC Identifier (NCID) that uniquely identifies that NC. The NC 608 or 609 can also be referred to as a second service provider node in the present disclosure.
[0130] The functioning of the components in FIG. 6 are summarized below.
[0131] Network Entity (NE) Control Plane 601: This component represents the customer of the 6G system and it implements the following control plane functions, for example, but not limited to: initiating contact with the GW 605 and sending the intent to GW 605, providing sample data through the GW 605 upon request, generating feedback and providing feedback through the GW 605. The intent is a natural language description of the requirement of the customer. Sample data is a small sample of data similar to the input data. Examples of input and output data include multimedia types, data matrices, vectors or any other types that embody data.
[0132] Network Entity (NE) Data Plane 602: This component implements the following data plane functions, for example, but not limited to: providing input data to the PSFs 6083 and 6093, and receiving output data from PSFs 6083 and 6093, if applicable.
[0133] Gateway (GW) 605: This component implements following functions, for example, but not limited to: retrieving customer’s intent from the NE 600, providing the intent to the PA 604 and retrieving the mission from the PA 604, selecting the EA 607 according to the mission, and relaying information between the NE control plane 601 and the EA 607.
[0134] When the GW 605 receives the intent from the customer, the GW 605 generates a unique ID for the intent. All subsequent communications between any two entities (including the customer) contain this ID that can uniquely identify the intent and the mission. This ID is referred to as the mission ID.
[0135] Planning Agent (PA) 604: This component is responsible for interpreting the intent provided by the customer and generating the mission that accomplishes the customer’s requirement. The PA 604 is aware of the capabilities of each NC 608 or 609. Specifically, the PA 604 is aware of which tasks an NC 608 or 609 can perform, what inputs required for each task, and what outputs are produced by each task. The mission generated by the PA 604 includes a structured description of what NCs 608 and 609 are required to execute the mission, what tasks each NC 608 or 609 is required to perform, and how the NCs 608 and 609 are required to interact to exchange inputs and outputs. The PA can also be referred to as a second network element in the present disclosure.
[0136] Network State Observer (NSO) 606: The NSO 606 is responsible for collecting data related to the state of the system (e.g., delay and the load of the network connecting the NCs 608 and 609) . This component implements the following functions, for example, but not limited to: monitoring the state of the network that connects the NCs 608 and 609 (e.g., network speed of the links connecting NCs 608 and 609 and their delay etc. ) , and providing such information to the EA 607 upon request.
[0137] Execution Agent (EA) 607: The EA 607 implements functions for receiving the mission from the GW 605, coordinating with the components in the NCs 608 and 609 such as SCF 6081 or 6091 and TCF 6082 or 6092 to execute the mission, providing the NCs 608 and 609 with the SRPs, and updating the SRP policy based on feedback from the customer. Details of the operation within the EA 607 are further described in conjunction with FIG. 8, FIG. 9A and FIG. 9B. The EA can also be referred to as a first network element in the present disclosure.
[0138] Service Control Function (SCF) 6081 or 6091: This component initializes the NC 608 or 609 for mission execution. It is responsible for receiving sample data provided by the customer or other NCs 608 and 609, selecting the most suitable model that matches the sample data, and providing the information about the selected model to rest of the components in the NC 608 or 609. In addition, SCF 6081 or 6091 shares the sample data with the PSF 6083 or 6093. Details of the model selection operation are further described in conjunction with FIG. 11, FIG. 12A and FIG. 12B. The SCF can also be referred to as a first network component in the present disclosure.
[0139] Task Control Function (TCF) 6082 or 6092: This component implements the functions for receiving the SRPs from the EA 607, converting the SRPs to Configuration Parameters (CP) . The CPs are the NC’s 608 or 609 implementation-dependent parameters that are not exposed to the EA 607. The CPs are sent to the PSF 6083 or 6093. Details of the SRP to CP conversion are further described in conjunction with FIG. 14. The TCF can also be referred to as a second network component in the present disclosure.
[0140] Processing Service Function (PSF) 6083 or 6093: This component implements the functions for receiving the information from the SCP about the selected model, receiving the CPs from the TCF 6082 or 6092, receiving the input data from the customer or the PSF 6083 or 6093 of another NC, processing the input data, and producing an output. In addition, the PSF 6083 or 6093 is responsible for receiving sample data from the SCF 6081 or 6091, processing the sample data, and sending the processed sample data to another NC, upon request. The PSF can also be referred to as a third network component in the present disclosure.
[0141] Aspects of the present disclosure introduce the Model Selection Mechanism (MSM) and the Self-Improvement Mechanism (SIM) . The MSM supports multiple AI models in an NC and provides a method for selecting the most suitable AI model based on the user’s input. The SIM provides a method for customers to provide feedback based on the outputs they receive. The feedback is then incorporated into the system for providing better SRPs.
[0142] FIG. 7A is a schematic flowchart of a method for model selection according to one or more embodiments of the present disclosure. This method can be implemented by one or more network components or network functions. For example, the network components may include a first service provider node, which may be an autonomous programming component, such as an A-CAP. As shown in FIG. 7A, the method can include the following steps.
[0143] At S701, receive information related to a mission and information related to at least one second service provider node. The information related to the mission is generated based on a customer’s intent and the at least one second service provider node is identified based on the mission.
[0144] At S702, communicate with the at least one second service provider node based on the information related to the mission.
[0145] FIG. 7B is a schematic flowchart of a communication procedure for model selection according to one or more embodiments of the present disclosure. The communication procedure involves the first service provider node and the at least one second service provider node. The at least one second service provider node may be an NC. Referring to FIG. 7B, the method may include the following steps.
[0146] At S7021, the first service provider node transmits sample data to the at least one second service provider node, and the at least one second service provider node receives the sample data.
[0147] At S7022, each of the at least one second service provider node transmits information related to at least one model to the first service provider node, and the first service provider node receives the information related to at least one model from each of the at least one second service provider node. The at least one model is selected based on the sample data.
[0148] In the embodiment, the first service provider node receives information related to a mission. The mission is derived from the customer’s intent, which is a natural language description of the requirement of the customer. In an implementation, the first service provider node may include a first network element and a second network element. The first network element performs the steps of S701 and S702. The first network element may be a network component responsible for executing the mission. The first network element may be an EA. The first network element may transmit sample data to the at least one second service provider node, and receive the information related to at least one model from each of the at least one second service provider node. The intent may be interpreted, for example, by using natural language processing (NLP) techniques, and a mission may be generated based on the interpreted intent. The mission is used to offer a service through a collection of tasks. In an implementation, the first service provider node may receive the information related to the mission and the information related to the at least one second service provider node from a second network element in the network. The second network element may be a PA. The second network element generates the mission based on the intent, and transmits the information related to the mission and the information related to the at least one second service provider node to the first network element. The information related to the mission may include a structured description of what second service provider node is required to execute the mission, what task the second service provider node is required to perform, and how the second service provider node is required interact to exchange inputs and outputs. The information related to at least one second service provider node may include information to identify the at least one second service provider node.
[0149] For example, the intent may be “Enhance the resolution of a low-resolution image and analyze the content of the image” . In this case, the mission generated based on the intent may include two tasks, i.e., Task 1: super-resolution image processing, and Task 2: image content analysis. For a second service provider node responsible for Task 1, the inputs are low-resolution images, outputs are high-resolution images, and its outputs are required to pass to a second service provider node responsible for Task 2. For the second service provider node responsible for Task 2, the inputs are high-resolution images from the second service provider node responsible for Task 1, the outputs are analysis results.
[0150] In the embodiment, the first service provider node communicates with the identified second service provider node (s) to coordinate the execution of the mission. The communication between the first service provider node and the at least one second service provider node includes transmission of sample data. The sample data is a small sample of data similar to the input data. The sample data may be representative of the actual input data that will be processed. The sample data is used to help the second service provider node (s) select the most appropriate model for processing the actual input data. Each second service provider node processes the sample data and selects the most suitable model (s) based on the sample data. After determining the at least one model, the second service provider node receives information related to the at least one model, and transmits information related to the at least one model to the first service provider node. The information related to the at least one model may include information to identify the at least one model. For example, the information to identify the at least one model may be indices related to the at least one model, which can also be referred to model identity (ID) .
[0151] In an implementation, the second service provider node may receive a request related to at least one task of a mission from the first service provider node, and transmit a response to the first service provider node indicating whether the sample data is needed. The request may include a request to notify if the sample data is needed. Some second service provider nodes may have only one model or a single algorithm designed to handle a specific type of data or task. In such cases, there is no need for sample data to perform model selection because there is only one model to choose from. The requirement for sample data can be specified within the second service provider node itself. By specifying whether sample data is required, the system avoids unnecessary data processing and transmission. This leads to more efficient operation, especially for the second service provider nodes that does not need sample data for model selection. The second service provider node may transmit a notification indicating that sample data is needed to the first service provider node, and the first service provider node receives the notification indicating that sample data is needed from the second service provider node. In this case, the first service provider node may transmit a request for the sample data to at least one of a corresponding network entity (NE) and a corresponding previous second service provider node. Sample data may be provided by the customer or another second service provider node. Examples of sample data include multimedia types, data matrices, vectors or any other types that embody data. The first service provider node may receive the sample data from at least one of an NE and a previous second service provider node. In an implementation, the NE may be associated with a customer, and the customer’s intent is received by the second network element from the NE.
[0152] In the embodiment, the first service provider node receives mission-related information generated based on customer intent and identifies the necessary second service provider nodes to execute the mission. The system can be allowed to break down complex tasks into subtasks and assign them to the most suitable nodes, thereby increasing the overall accuracy and adaptability of task execution. By using sample data to select models, the second service provider nodes can select the model best suited for the current task data dynamically. The efficiency of the second service provider node can be improved by enabling automatic model selection based on the sample data.
[0153] In an implementation, prior to performing a task, the second service provider node can be configured using one or more service requirement parameters (SRPs) . The first service provider node may generate respective one or more service requirement parameters for each of the at least one second service provider node, and transmit the respective one or more SRPs to each of the at least one second service provider node. The one or more SRPs are generated based on the information related to the at least one model. The SRP is a parameter that specifies the requirements for performing a task. The SRPs can include details such as processing delay, resource usage, hardware requirements (e.g., CPU or GPU) , and other operational constraints. The first service provider node (e.g., a first network element of the first service provider node) may generate one or more SRPs based on the information related to the selected model (s) for the task. By generating and transmitting SRPs, the model in the NC can be configured adaptively for the task to be performed, leading to better performance and resource utilization. The second service provider node may receive one or more service requirement parameters (SRPs) from the first service provider node, and may convert the one or more SRPs to respective one or more configuration parameters (CPs) based on the at least one model. An example of an SRP is the target delay for processing. An NC (which is an example of the aforementioned second service provider node) may use a multitude of implementation-dependent methods for achieving a specified target delay. Depending whether the delay is small or large, the NC may use the GPU or the CPU for its processing. In that case the CP is a binary parameter that indicates which hardware to use. Two NCs that provide the same service may have different implementations. Both NCs accept identical SRPs, and a sub-model (such as, a TCF model) corresponding to a network component in each NC converts SRPs to implementation-dependent CPs. Having a SRP to CP conversion mechanism allows the SRPs to be used more efficiently within an NC.
[0154] In an implementation, the one or more SRPs may be generated based on network state information and an SRP policy. The first service provider node may receive network state information, and generate the one or more SRPs based on at least one SRP policy using the network state information. The network state information can include metrics such as network latency, bandwidth availability, load on network nodes, and other relevant data that can affect task execution. The network state information may come in a form of a vector of real values. Each component in the vector represents one aspect of the network state. The network state information may be provided by another network component, for example, an NSO, which is responsible for collecting data related to the state of the system (e.g., delay and the load of the network connecting the NCs) . An SRP policy is a predefined strategy that dictates how SRPs should be generated based on various factors, including network state information. The SRP policy provides a framework for dynamically adjusting SRPs to optimize task execution under different network conditions. The SRP policy, or policy in short, may be an artificial neural network. An SRP policy takes network state information from the NSO as input, and produces SRPs as outputs. The SRPs can be dynamically adjusted based on real-time network conditions. Therefore, the tasks can be executed efficiently under different conditions.
[0155] In an implementation, a self-improvement mechanism is also introduced in the present disclosure. The first service provider node may receive feedback from the customer, and update one or more SRP policies based on the feedback using one or more learning algorithms. After the execution of the mission, the customer can provide feedback to the first service provider node. A component (e.g., a policy updater (PU) ) in the first service provider node can be responsible for receiving customer feedback, and updating the policy. The first service provider node may employ machine learning techniques, such as reinforcement learning, to analyze the feedback and update the SRP policies. The feedback may be in the form of a vector. Each component of the vector represents one aspect of the system’s performance as perceived by the customer. For example, a two-dimensional vector can be considered as the form of the feedback. The first dimension may represent the delay of the output, and the second dimension may represent the accuracy of the output. A positive value for the first component expresses the system delay is acceptable, and a negative value expresses the delay is too high. The feedback can be converted to a scalar reward value by computing the average across the dimensions. To update the policy, the PU employs standard reinforcement learning techniques such as policy gradient and deep Q learning. The policy updater may map the feedback vector to a scalar by averaging, and use the scalar as a reward signal used in reinforcement learning. By updating the SRP policies according to the feedback from the customer, the task executions can be adjusted to better meet customer expectations. The performance and resource utilization of the system can be improved, which leads to higher customer satisfaction.
[0156] FIG. 8 shows an example of components of first network element in a first service provider node according to one or more embodiments of the present disclosure. An A-CAP is taken as an example of the first service provider node and an EA is taken as an example of the first network element for illustration. An EA of the A-CAP is responsible for providing the second service provider nodes (e.g., NCs) with SRPs and updating the SRP policy based on the feedback from the customer. As shown in FIG. 8, the EA consists of a Policy Database (PD) , a Policy Executor (PE) , and a Policy Updater (PU) . PD is a database consisting of multiple policies. PE is responsible for generating SRPs using a policy. PU is responsible for receiving customer feedback, and updating the policy. The function of each component is described next.
[0157] Each policy stored in PD is tied to a particular NC model combination. For example, consider a mission that involves two NCs where the first and second NC consists of 3 and 4 models (as illustrated in FIG. 11) , respectively. In total, there exists 3x4=12 model combinations. In this case, the EA holds 12 SRP policies corresponding to each of the model combinations.
[0158] PE is responsible for retrieving a policy from the PD and generating SRPs. Once all NCs have selected the indices of the at least one models that they are going to use (as explained in conjunction with FIG. 11) , the NCs report that information back to the EA (see FIG. 8) . The PE retrieves from the PD the SRP policy that corresponds to the model combination.
[0159] FIG. 9A illustrates policy execution in a first network element of a first service provider node according to one or more embodiments of the present disclosure. An A-CAP is taken as an example of the first service provider node and an EA is taken as an example of the first network element for illustration purposes. As shown in FIG. 9A, a PE in an EA of the A-CAP uses a selected policy to generate SRPs. The SRP policy, or policy in short, is an artificial neural network. An SRP policy takes network state information from an NSO as input, and produces SRPs as outputs. The SRPs are provided to NCs (specific examples of the aforementioned second service provider nodes) in the mission when the mission is executed. Network state information may come in a form of a vector of real values. Each component in the vector represents one aspect of the network state. For example, the NSO may provide a 2d vector, where the first and second components respectively represent the average delay and the throughput of the network links connecting the NCs.
[0160] FIG. 9B illustrates policy training in a first network element of the first service provider node according to one or more embodiments of the present disclosure. After the execution of the mission, the customer can provide feedback to the EA. The PU in EA is responsible for incorporating the customer feedback to the system. As shown in FIG. 8, the PU receives customer feedback along with information for identifying the mission (e.g., the mission ID) , and the PU retrieves the policy relevant to the mission. As shown in FIG. 9B, the PU updates the retrieved policy as described next.
[0161] In an implementation, new policies may be added to the PD. Whenever an MP submits a new model to an NC, EA creates a new set of SRP policies corresponding to each model combination that includes the newly submitted model. For example, consider two NCs where the first and second NC consists of 3 and 4 models, respectively. In total, there exists 3x4=12 model combinations and the EA holds 12 policies. If an MP adds a new model to the second NC, the number of model combinations rise to 3x5=15. In this case EA creates 15-12=3 new policies. The procedure for model submission is described later in conjunction with FIG. 20.
[0162] The new policies have not been trained when they are first created, and there does not exist a separate training process for newly created policies. The training of the new policies is conducted as they are used, using the customer feedback (same process outlined above) . Due to lack of training, the new policies may yield inferior performance at the beginning. To compensate for this lack of performance, the services that use new policies may be provided to the customer at a discounted fee.
[0163] Details of the second service provider node and model selection by the second service provider node will be described below.
[0164] FIG. 10 is a schematic flowchart of a method for model selection according to one or more embodiments of the present disclosure. This method can be implemented by a second service provider node, such as an NC. As shown in FIG. 10, the method can include the following steps.
[0165] At S1001, receive sample data from a first service provider node.
[0166] At S1002, select at least one model from a plurality of models based on the sample data.
[0167] At S1003, obtain information related to the at least one model.
[0168] At S1004, transmit the information related to the at least one model to the first service provider node.
[0169] In the embodiment, the second service provider node receives sample data, which is used to evaluate and select the most suitable model for processing the actual input data. The sample data may be received from a first network element of the first service provider node, such as an EA of an A-CAP. The second service provider node selects at least one model from a plurality of models based on the sample data for subsequent processing. The selection process may involve using LEs to determine which model best matches the sample data. The second service provider node receives the information related to the selected model, and then transmit the information related to the selected model to the first service provider node. The information related to the at least one model is used to identify the at least one model. By using the sample data, the second service provider node can select the most suitable model for processing the actual input data, which can improve performance of the system. The efficiency of the system can be improved by enabling automatic model selection based on the sample data. Moreover, this approach enhances the flexibility and adaptability of the system, making it well-suited for future-generation wireless communication systems like 6G.
[0170] In an implementation, the second service provider node may include a first network component, a second network component and a third network component. Each model of the at least one model may include a first sub-model corresponding to the first network component, a second sub-model corresponding to the second network component and a third sub-model corresponding to the third network component. The first network component may be an SCF. The first network component is used to initialize the second service provider node, receive the sample data, select at least one model, provide model information to rest components in the second service provider node. The first sub-model may be a likelihood estimator.
[0171] In an implementation, the information related to the at least one model may include information related to the first sub-model corresponding to the at least one model. The information related to the first sub-model corresponding to the at least one model is information identifying the first sub-model. The first network component may select the at least one model based on the sample data. For example, an index for the at least one selected model is n, and the information related to the first sub-model is n for the model-n-SCF.
[0172] In an implementation, the first network component may include a plurality of likelihood estimators. The second service provider node may hold a plurality of models and each model of the plurality of models may include a likelihood estimator of the plurality of likelihood estimators. The first network component may use the plurality of likelihood estimators to evaluate which model is most suitable for processing the sample data. Each LE computes a likelihood value indicating how well its corresponding model matches the sample data. The first component selects the model with the highest likelihood value.
[0173] The second network component may be used to receive SRPs and convert the SRPs to CPs. The second network component may be a TCF. The third network component may be used to process the input data, and produce an output. The third network component may receive the information about the selected model from the first network component, and receive the CPs from the second network component. The third network component then may configure the third sub-model corresponding to the selected model, and process the input data using the configured third sub-model and produce an output.
[0174] In an implementation, the first network component may transmit the information related to the first sub-model corresponding to the at least one model to the second network component and the third network component. The second network component may obtain the second sub-model corresponding to the at least one model based on the information related to the first sub-model. The third network component may obtain the third sub-model corresponding to the at least one model based on the information related to the first sub-model. The information related to the first sub-model corresponding to the at least one model may be information used for identifying the first sub-model. For example, the most suitable model that matches the sample data selected by the first network component may be LE-n. In this case, the information related to the first sub-model corresponding to the at least one model may be n. The first network component transmits n to the second network component and the third network component. The second network component can obtain the second sub-model based on the information n. For example, the second network component can retrieve a model-n-TCF. The third network component can obtain the third sub-model based on the information n. For example, the third network component can retrieve a model-n-PSF.
[0175] In an implementation, the second network component may receive one or more SRPs from the first service provider node, where the one or more SRPs are generated based on the information related to the at least one model. The second sub-model of the second network component may convert the one or more SRPs to respective one or more CPs based on the at least one model. The second network component may transmit the one or more CPs to the third network component.
[0176] Specific examples of the method for model selection performed by the second service provider node will be described in the following. An NC is taken as an example of the second service provider node, an SCF is taken as an example of the first network component in the second service provider node, a TCF is taken as an example of the second network component in the second service provider node, and a PSF is taken as an example of the third network component in the second service provider node for illustration.
[0177] The NC consists of an SCF and a TCF in its control plane, and a PSF in its data plane. According to an implementation of the present disclosure, an NC can hold multiple models that are designed for or trained on different datasets. These models are either provided along with the NC, or added to the NC after provisioning the NC. An entity that provides new models is referred to as a Model Provider (MP) . The models in an NC are already trained by their MPs, and the NC does not conduct any form of training on the models.
[0178] Assume that an NC holds N models. The models are labeled model-1, …, and model-N. Each model is a collection of three sub-models. The model-n consists of three sub-models labeled LE-n (an example of the aforementioned first sub-model) , model-n-TCF (an example of the aforementioned second sub-model) , and model-n-PSF (an example of the aforementioned third sub-model) . These three sub-models are stored in and operated by SCF, TCF and the PSF, respectively. All sub-models in model-n are trained using the same dataset. A model may be supplied along with an NC, or they can be submitted to the system at a later time. The mechanism for model submission is described later in conjunction with FIG. 20.
[0179] FIG. 11 illustrates an organization of models in a second service provider node according to one or more embodiments of the present disclosure. FIG. 11 illustrates how the multiple AI models are stored and organized in an NC 1100.
[0180] Service Control Function (SCF) 1101: The SCF 1101 holds LE-1, …, and LE-N. These sub-models are referred to as the Likelihood Estimators (LE) . The SCF 1101 uses LEs and sample data to implement a model selection process. This process is described in conjunction with FIG. 12A and FIG. 12B. Sample data may be provided by the customer or another NC. Examples of sample data include multimedia types, data matrices, vectors or any other types that embody data. The SCF 1101 receives sample data, selects the most suitable model that matches the sample data using LEs, and notifies TCF 1102, PSF 1103, and EA of the selected model. Assuming the index picked out and communicated by SCF 1101 is n, TCF 1102 and PSF 1103 proceed to use model-n-TCF and model-n-PSF, respectively for their subsequent processing.
[0181] Task Control Function (TCF) 1102: The TCF 1102 holds model-1-TCF, …, and model-N-TCF. Assume the TCF 1102 uses model-n-TCF for its processing. The TCF model converts the SRPs received from the EA to Configuration Parameters (CP) . I. e., model-n-TCF takes SRPs as inputs, and produces CPs as outputs. Both SRPs and CPs may be vectors of real values or any other data format that may be used to represent a parameter. The SRPs are known to both the EA and the NC 1100, whereas CPs are known only to the NC.
[0182] Processing Service Function (PSF) 1103: The PSF 1103 holds model-1-PSF, …, and model-N-PSF. Assuming the PSF 1103 uses model-n-TCF for the mission. This component is responsible for processing the input data using model-n-TCF and producing an output. The input data may be provided by the customer or another NC. Prior to processing, the PSF 1103 configures itself per the CPs provided by the TCF 1102. Examples of input and output data include multimedia types, data matrices, vectors or any other types that embody data.
[0183] The SCF 1101 takes sample data as input and outputs the index of the model that best matches the sample data.
[0184] FIG. 12A and FIG. 12B illustrate a model selection mechanism within a first network component (e.g., SCF) of the second service provider node according to one or more embodiments.
[0185] The SCF consists of N Likelihood Estimators (LE) . The arrangement of LEs within an SCF is shown in FIG. 12A. Note that LE-n is a sub-model in model-n as presented in FIG. 11. The LE-n measures how likely the model-n is a good model for processing the sample data. The LE-n takes sample data as its input and outputs a likelihood value, a scalar. The larger the likelihood, the better the model-n is for processing sample data. According to FIG. 12A, the SCF outputs the index of the LE that produces the largest likelihood. This index is shared with the TCF, the PSF and the EA.
[0186] FIG. 12B presents one possible implementation of an LE 1200 which works best for multimedia data types such as images. The LE 1200 takes in sample data as input, and outputs a scalar value. The LE 1200 consists of three sub-components. The procedures of training these components are presented later in conjunction with FIG. 13A and FIG. 13B. The operation of the three components are as follows:
[0187] Component 1 is an autoencoder 1201 trained to produce a reconstruction of the input. The autoencoder 1201 is a neural network that accepts sample data as input, and outputs a reconstruction of the sample data.
[0188] Component 2 is an error computer 1202 which computes the error of the reconstruction and outputs the error as a scalar. As shown in FIG. 12B, component 2 is provided both the sample data and the reconstruction. The method of error computation may vary from one input type to the other. For example, if the input is an image, component 2 may compute the mean squared error (MSE) . If the input is a sequence of characters, component 2 may compute the Levenshtein distance. In any case, the output of component 2 is a measure of the performance of component 1 on sample data. If sample data is similar to the dataset used to train the autoencoder, the reconstruction error is expected to be small. Conversely, if sample data is considerably different from the dataset used to train the autoencoder, the reconstruction error is expected to be large.
[0189] Component 3 is an error to likelihood converter 1203 which maps the reconstruction error to a likelihood value. Component 3 consists of a probability mass function (PMF) , where x-axis is the reconstruction error, and the y-axis is the probability mass. As illustrated in FIG. 12B, given a reconstruction error e, component 3 computes the area of the PMF to the right of the reconstruction error e, and outputs this value as the likelihood.
[0190] In an implementation, NC model training is provided. Assume an MP wants to submit model-n including of LE-n, model-n-TCF, and model-n-PSF. First, the MP trains the sub-models in model-n using a single dataset. The MP then submits the three sub-models to the system. The training procedures of sub-models are outlined below. The model submission procedure is presented later with FIG. 20.
[0191] Training LE-n: FIG. 12B presents one possible method of implementing the LE. Out of the three components in LE-n shown in FIG. 12B, only component 1 and 3 require training. Component 2 is a function that does not require any training. The training of component 1 and 3 takes place as follows.
[0192] FIG. 13A and FIG. 13B and FIG. 13C illustrate example methods of training an LE according to one or more embodiments of the present disclosure
[0193] Prior to training component 1 and 3, the dataset is partitioned into two sets: the training set, and the test set. The partitioning may follow current best practices (e.g., 70%of data samples in training set, 30%in test set) .
[0194] Component 1, the autoencoder 1301 (see FIG. 13A) is trained using the training set. This component is trained using standard techniques for training autoencoders 1301 (e.g., stochastic gradient descent, dropout, batch normalization etc. ) .
[0195] After training component 1, the mechanism (i.e. the error computer 1302) in FIG. 13B is used to compute the reconstruction errors for all the samples in the test set. I. e., each sample in the test set is passed through the mechanism in FIG. 13B, computing the reconstruction errors for all samples. The histogram of the set of reconstruction errors is computed, and normalized so that the area under the histogram is 1. At this point the component 1 and component 3 (i.e. the error to likelihood converter 1303) have finished their training.
[0196] When a trained LE is submitted to the SCF, the SCF’s process for computing the likelihood of sample data is as follows. Refer to FIG. 13B and FIG. 13C, given the sample data, component 1 computes the reconstruction of the sample data, component 2 computes the reconstruction error (say e) , and finally, component 3 outputs the likelihood by computing area under the histogram, to the right of e.
[0197] Training model-n-PSF: The PSF model is a generic model or an algorithm that takes input data and produces an output. An example is an image super-resolution model where the input is an image, and the output is a high resolution of the image. The training of the PSF is conducted using the dataset and standard techniques (e.g., stochastic gradient descent, if an AI model) .
[0198] Training model-n-TCF: The TCF model takes Service Requirement Parameters (SRPs) as inputs and maps them to Configuration Parameters (CPs) . The CPs are implementation-dependent details that configure the NC, more specifically, the PSF. The TCF provides CPs to PSF, and the PSF configures itself per the CPs. If the TCF model is trained well, the actual performance of PSF will match with what’s outlined with SRPs. For example, assuming the processing delay of the PSF is an SRP, the actual processing time taken by the PSF will be in-line with the SRP. The training of model-n-TCF may be conducted using standard reinforcement learning (RL) techniques such as policy gradient and deep Q networks.
[0199] FIG. 14 illustrates an example procedure for training a sub-model of a second network component (e.g., Task Control Function (TCF) ) according to one or more embodiments of the present disclosure.
[0200] Referring to FIG. 14, the training steps are as follows:
[0201] The model trainer is responsible for training and updating model-n-TCF. The inputs of model-n-TCF are SRPs, and the outputs are CPs. The trainer generates a target SRP, e.g., a target processing delay, and the model-n-TCF maps the SRP to a CP, e.g., whether to use the CPU or the GPU.
[0202] The PSF configures itself according to the CPs, and processes input data using model-n-PSF. The outputs and the processing statistics are sent to the Service Parameter Estimator (SPE) .
[0203] The SPE is responsible for estimating actual service parameters, e.g., measuring the actual processing delay of the PSF. This information is provided to the Reward Generator.
[0204] The Reward Generator is responsible for quantifying the difference between target SRPs and actual service parameters, and generating an appropriate reward. For example, if the target processing delay and the actual processing delay are considerably different, the reward is a large negative number. The model trainer repeats this process of generating SRPs and computing the reward. The model trainer employs standard reinforcement learning methods for tuning model-n-TCF to maximize the reward. At the end of the training, the model-n-TCF will have learned to map SRPs to CPs in a way that the PSF performs in the way expressed through the SRPs.
[0205] The Model Provider (MP) may train the three sub-models LE-n, model-n-TCF, and model-n-PSF using the methods described above. The MP then submits the three sub-models to the system, after which the system adds the new models to the existing model collection. This process is further described.
[0206] FIG. 15 illustrates an example procedure for model submission according to one or more embodiments of the present disclosure. FIG. 15 presents the connectivity between the components involved in the model submission procedure.
[0207] The Model Provider (MP) 1501 is the entity providing with a new model to the system, and the Model Handler (MH) 1502 is the entity that coordinates with system components for storing the model.
[0208] Assume that the MP 1501 submits model-n, which consists of the sub-models LE-n, model-n-TCF, and model-n-PSF. These sub-models have already been trained per procedures described in conjunction with FIG. 13A, FIG. 13B, FIG. 13C and FIG. 14. The MP 1501 sends the three sub-models to MH 1502, and MH 1502 directs SCF 1505, TCF 1506, and PSF 1507 to store LE-n, model-n-TCF, and model-n-PSF, respectively. Finally, MH 1502 notifies A-CAP 1503 that a new model was added to an NC 1504, and provides to A-CAP 1503 the NCID along with the index of the added model. The call flow procedure involved in model submission is described in conjunction with FIG. 20.
[0209] An embodiment of the present disclosure provides a system. The system includes a network entity (NE) , an autonomous programming component, a network state observer (NSO) and a plurality of second service provider nodes. The autonomous programming component is configured to: receive a customer’s intent from the NE, and receive network state information from the NSO. The autonomous programming component is configured to perform the aforementioned method applied to the first service provider node. Each of the plurality of second service provider nodes is configured to perform the aforementioned method applied to the second service provider node.
[0210] Specific examples of operations of the system will be described below.
[0211] The system presented in FIG. 6 operates in four stages: system initialization, NC (an example of the aforementioned second service provider node) initialization, NC processing, and customer feedback. When more than one NC is involved in the mission, the order in which the NCs operate is specified in the mission. This is important in the NC initialization and NC processing stages. In the NC initialization stage, each NC initializes itself in the order specified by an EA (an example of the aforementioned first network element of the first service provider node) . After all the NCs are initialized, the system moves to the NC processing stages. In the NC processing stage, each NC processes input data in the order specified by the EA. A brief summary of each stage is provided below.
[0212] System initialization: in this stage, a customer initiates the process by sending an intent to the system, and a second network element (e.g., a PA) of the first service provider node and a first network element (e.g., an EA) of the first service provider node prepare the system for the executing the mission. This includes the PA generating mission, and the EA generating SRPs for the NCs. The procedures involved in this stage are outlined in FIG. 16.
[0213] NC initialization: in this stage, all NCs prepare themselves for the mission by selecting the best model for the sample data. Each NC is provided with sample data from the customer or other NCs. The model selection is performed by an SCF (an example of the aforementioned first network component) of an NC. The procedures involved in this stage are outlined in FIG. 17.
[0214] NC processing: after NCs complete the initialization, they move to the NC processing stage. In the processing stage, an NC is provided with SRPs and input data from the customer or other NCs. In each NC, a TCF (an example of the aforementioned second network component) converts SRPs to CPs, and a PSF (an example of the aforementioned third network component) processes the input. The procedures involved in this stage are outlined in FIG. 18.
[0215] Customer feedback: in this stage, the mission has already been full executed. The customer provides the feedback to the system, and the EA incorporates the feedback by updating the relevant SRP policy. The procedures involved in this stage are outlined in FIG. 19.
[0216] In addition to the listed four stages, FIG. 20 illustrates a procedure for model submission by the model provider. The subsequent figures detail the system operation in each stage.
[0217] System initialization stage
[0218] FIG. 16 is a call flow diagram illustrating a procedure for a system initialization stage according to one or more embodiments of the present disclosure.
[0219] The details of calls and actions are as follows:
[0220] 1. NE Control Plane 1601 sends an intent to a GW 1602. The intent consists of a natural language description of the requirement of the customer.
[0221] 2. The GW 1602 assigns a unique ID for the intent. The ID may be a string or a number. All subsequent communications related to this intent (and the corresponding mission) contains this ID for identification purposes. The ID is referred to as a mission ID.
[0222] 3. The GW 1602 sends the mission ID to the NE Control Plane 1601. The NE Control Plane 1601 stamps the mission ID to all future communications related to this intent.
[0223] 4. The GW 1602 sends the intent (along with the mission ID) to a PA 1603.
[0224] 5. The PA 1603 generates the mission based on the intent. The mission consists of a structured description of which NCs are required to perform the customer’s requirement and how these NCs communicate and interact. Each NC in the mission is identified by the NCID.
[0225] 6. The PA 1603 sends the mission to an EA 1604 via the GW 1602. The EA 1604 is responsible for executing the mission. The mission execution entails instructing the NCs to perform actions and communicate as outlined in the mission.
[0226] Having initialized in accordance with FIG. 16, the system moves to the NC initialization stage.
[0227] NC initialization stage
[0228] FIG. 17 is a call flow diagram illustrating a procedure for an initialization stage at a second service provider node according to one or more embodiments of the present disclosure.
[0229] There may be missions where more than one NC is involved. In such cases, the order of NC operation is provided in the mission. The procedure in FIG. 17 presents the call flow for initialization of one of the NCs. This NC is referred to as the current NC. If there exist NCs that operated previously, they are referred to as the previous NCs (no previous NCs exist if the current NC is the first NC to operate) .
[0230] 1. Based on the mission, the EA 1703 selects an NC (referred to as the current NC) and establishes connection with the SCF 1704 of the current NC.
[0231] 2. The EA 1703 sends to the SCF 1704 the details of the task the NC is required to perform. Also, the EA 1703 requests the SCF 1704 to notify whether sample data is needed.
[0232] 3. The SCF 1704 checks whether sample data is needed from the customer for the model selection. Not all NCs may require sample data from the customer. For example, an NC may have only one model or an algorithm, and therefore may not require sample data to perform a model selection. Whether the sample data is required or not is specified in the NC itself.
[0233] 4. If sample data is required from the customer, the SCF 1704 requests and receives it from the customer through the EA 1703, the GW 1702 and the NE Control Plane 1701. Sample data is a small sample of data similar to the input data. Examples of sample data include, but are not limited to, multimedia types, data matrices, vectors or any other types that embody data. After the SCF 1704 receives the sample data, and if the sample data is not sufficient, the SCF 1704 may re-request the NE Control Plane 1701 to send more sample data.
[0234] 5. The SCF 1704 evaluates whether sample data is needed from any of the previous NCs (similar to step 3) .
[0235] 6. If the sample data is required from previous NCs, the SCF 1704 requests and receives it through the EA 1703 and the TCF 1707 of each of the previous NCs and the PSF 1708 of previous NC (similar to step 4) . After the SCF 1704 receives the sample data, and if the sample data is not sufficient, the SCF 1704 may re-request other NCs to send more sample data.
[0236] 7. The SCF 1704 uses the sample data to select the best model that suits the sample data. This selected model is referred to as model-n, where ‘n’ is the index of the model. The model selection procedure is outlined in FIG. 11, FIG. 12A and FIG. 12B.
[0237] 8. The SCF 1704 communicates ‘n’ to the TCF 1705, the PSF 1706, and the EA 1703. The EA 1703 stores ‘n’ for SRP policy selection, as described in conjunction with FIG. 8, FIG. 9A and FIG. 9B.
[0238] 9. After receiving index ‘n’ , the TCF 1705 retrieves model-n-TCF and uses that model for subsequent processing. The organization of models within the TCF 1705 is illustrated in FIG. 11.
[0239] 10. After receiving index ‘n’ , the PSF 1706 retrieves model-n-PSF and uses that model for subsequent processing. The organization of models within the PSF 1706 is illustrated in FIG. 11.
[0240] 11. The SCF 1704 sends the sample data (received in step 4 and 6) to the PSF 1706.
[0241] 12. The PSF 1706 processes the sample data using model-n-PSF and stores the output results. The output may be considered sample data for another NC. If another NC requires sample data from the current NC, that NC will request sample data as described in step 6.
[0242] 13. When sample data processing is finished, the PSF 1706, the SCF 1704 and the TCF 1705 coordinate among themselves and notify the EA 1703 that NC initialization is complete.
[0243] 14. After receiving the initialization completion notification, the EA1703 checks whether there exist any NCs remaining in the mission waiting to initialize. If there exists more NCs, the EA 1703 selects an NC according to the order defined in the mission, and repeats steps 1 to 13. If no NCs remain, proceed to step 15.
[0244] 15. At this point, the initialization stage is complete for all the NCs. The EA 1703 has received the indices of the models selected by all the NCs (as depicted in step 8) . The EA 1703 retrieves the SRP policy corresponding to the selected model combination (as described in conjunction with FIG. 8) . The EA 1703 requests and receives network state information from the NSO 1709, generates and stores SRPs for all the NCs, as outlined in FIG. 8.
[0245] NC processing stage
[0246] FIG. 18 is a call flow diagram illustrating a procedure for a processing stage at a second service provider node according to one or more embodiments of the present disclosure. The NC processing stage takes place after the NC initialization stage.
[0247] The procedure in FIG. 18 outlines the call flow for one of the NCs. This NC is referred to as the current NC. The NCs that operated previously are referred to as the previous NCs (no previous NCs exist if the current NC is the first NC to operate) .
[0248] 1. The EA 1804 selects an NC (referred to as the current NC) and retrieves the SRPs relevant to the current NC (SRPs are stored as depicted in step 15 of FIG. 17) .
[0249] 2. The EA 1804 sends SRPs to the TCF 1805 of the current NC.
[0250] 3. The TCF 1805 receives the SRPs and uses the model-n-TCF to generate the CPs, according to the procedure described in conjunction with FIG. 11. The CPs are NC’s implementation-dependent configurations as described in conjunction with FIG. 11 and FIG. 14.
[0251] 4. The TCF 1805 sends the CPs to the PSF 1806.
[0252] 5. The PSF 1806 evaluates whether input data is required from the customer. Not all NCs may require input data from the customer. For example, some NCs may require data only from other NCs. Whether input data is required from the customer or not is specified in the NC itself.
[0253] 6. If input data is required from the customer, the PSF 1806 requests the input data from the customer through the TCF 1805, the EA 1804, the GW 1803 and the NE Control Plane 1802. The PSF 1806 receives the input data from the NE Data Plane 1801. Examples of input data include multimedia types, data matrices, vectors or any other types that embody data.
[0254] 7. The PSF 1806 evaluates whether input data is required from one or more other NCs. Some NCs may not require any data from other NCs. Whether input data is required or not from other NCs is specified in the NC itself.
[0255] 8. If input data is required from previous NC (s) , the PSF 1806 requests data from each of the previous NC (s) through the TCF 1805, the EA 1804and the TCF 1807 of the previous NC. The PSF 1806 receives the input data from the PSF 1808 of the previous NC.
[0256] 9. The PSF 1806 configures itself according to the CPs received in step 4, and the PSF 1806 processes input data using model-n-PSF (the model selected in step 10 of FIG. 17) . The processing entails providing input data to the PSF model and producing the outputs.
[0257] 10. When processing is complete, the PSF 1806 notifies that to the EA 1804 through the TCF1805.
[0258] 11. After receiving the processing complete notification, the EA 1804 checks whether there exist any remaining NCs according to the mission. If there exists more NCs, the EA 1804 selects an NC according to the order defined in the mission, and repeats steps 1 to 10. If no NCs remain, proceed to step 12.
[0259] 12. Optionally, the EA 1804 notifies that processing is complete to the Network Entity (NE) Control plane 1802.
[0260] 13. The NE Control Plane 1802 stores information needed to provide feedback. This information consists of the mission ID (stored as depicted in step 3 of FIG. 16) for identifying the mission. The format of the feedback is described in conjunction with FIG. 15. In summary, the feedback may be in the form of a vector. Each component of the vector represents one aspect of the system’s performance as perceived by the customer.
[0261] Customer feedback stage
[0262] FIG. 19 illustrates a call flow diagram for a customer feedback stage according to one or more embodiments of the present disclosure. This procedure takes place any time after completion of the mission execution.
[0263] 1. The NE Control Plane 1901 retrieves the feedback generated for the mission (according to step 13 of FIG. 18) . The format of the feedback is specified in conjunction with FIG. 15.
[0264] 2. The feedback is sent to the EA 1903 via the GW 1902 along with the mission ID.
[0265] 3. The EA 1903 retrieves the SRP policy corresponding to the mission ID and updates the policy according to the procedure described in conjunction with FIG. 15.
[0266] Model submission by model provider
[0267] In addition to the main four stages related to mission execution, FIG. 20 illustrates a procedure for providing a model to an NC by a Model Provider (MP) according to one or more embodiments of the present disclosure. This call flow may take place independently of any mission or any of the four stages outlined above.
[0268] FIG. 20 shows a call-flow diagram for new model submission. FIG. 20 illustrates a procedure for model submission to a second service provider node by a MP according to one or more embodiments of the present disclosure.
[0269] 1. An MP 2001 trains model-n (including all its sub-models LE-n, model-n-TCF, and model-n-PSF) using a dataset. The training procedure for sub-models is illustrated in FIG. 13A, FIG. 13B, FIG. 13C and FIG. 14.
[0270] 2. The MP 2001 sends the sub-models. The MP 2001 also sends the NCID of the NC 2004 the models should be added to.
[0271] 3. The MH 2002 coordinates with the relevant NC 2004 (the one with the NCID) to store sub-models. The SCF 2005, the TCF 2006 and the PSF 2007 store LE-n, model-n-TCF, and model-n-PSF, respectively. The NC 2004 notifies the MH 2002 of the index of the newly added model.
[0272] 4. The MH 2002 notifies the A-CAP 2003 the index of the newly added model along with the NCID. The EA in the A-CAP 2003 follows the procedure described in conjunction with FIG. 9A and FIG. 9B, to create a new set of policies that involve the newly added model.
[0273] A model selection mechanism (MSM) is provided in the present disclosure. The MSM allows an NC to expand its AI model portfolio by adding new models trained on new datasets. Moreover, the MSM increases NC efficiency by enabling automatic model selection based on customer’s sample data.
[0274] A self-improvement mechanism (SIM) is also provided in the present disclosure. The SIM allows customers to provide feedback based on the services they receive, leading to higher customer satisfaction. Moreover, the SIM allows the system to improve itself by incorporating the customer feedback, and increasing the quality of services offered by the system.
[0275] Although this disclosure refers to illustrative embodiments, this is not intended to be construed in a limiting sense. Various modifications and combinations of the illustrative embodiments, as well as other embodiments of the disclosure, will be apparent to persons skilled in the art upon reference to the description.
[0276] Features disclosed herein in the context of any particular embodiments may also or instead be implemented in other embodiments. Method embodiments, for example, may also or instead be implemented in apparatus, system, and / or computer program product embodiments. In addition, although embodiments are described primarily in the context of methods and apparatus, other implementations are also contemplated, as instructions stored on one or more non-transitory computer-readable media, for example. Such media could store programming or instructions to perform any of the various methods consistent with the present disclosure.
[0277] Embodiments of products related to the communication methods are further described.
[0278] FIG. 21 is a schematic structural diagram of a first service provider node according to one or more embodiments of the present disclosure. As shown in FIG. 21, the first service provider node 2100 may include: a communicating module 2101, configured to: receive information related to a mission and information related to at least one second service provider node, where the information related to the mission is generated based on a customer’s intent and the at least one second service provider node is identified based on the mission; and communicate with the at least one second service provider node based on the information related to the mission. The communicating module 2101 is configured to transmit sample data to the at least one second service provider node; and receive indices related to at least one model from each of the at least one second service provider node, where the at least one model is selected based on the sample data.
[0279] In an implementation, the first service provider node comprises a first network element and a second network element, wherein the aforementioned communicating module 2101 is implemented via the first network element.
[0280] In an implementation, the communicating module 2101 is configured to receive the information related to the mission and the information related to the at least one second service provider node from a second network element in the network.
[0281] In an implementation, the communicating module 2101 is configured to receive, from the at least one second service provider node, a notification indicating that sample data is needed; and receive the sample data from at least one of a network entity (NE) and a previous second service provider node.
[0282] In an implementation, the NE is associated with a customer, and the customer’s intent is received by the second network element from the NE.
[0283] In an implementation, the first service provider node 2100 further includes a generating module 2102, configured to generate respective one or more service requirement parameters (SRPs) for each of the at least one second service provider node, where the one or more SRPs are generated based on the information related to the at least one model. The communicating module 2101 is configured to transmit the respective one or more SRPs to each of the at least one second service provider node.
[0284] In an implementation, the communicating module 2101 is configured to receive network state information. The generating module 2102 is configured to generate the one or more SRPs based on at least one SRP policy using the network state information.
[0285] In an implementation, the communicating module 2101 is configured to receive feedback from the customer. The first service provider node 2100 further includes an updating module 2103, configured to update one or more SRP policies based on the feedback using one or more learning algorithms.
[0286] FIG. 22 is a schematic structural diagram of a second service provider node according to one or more example embodiments of the present disclosure. As shown in FIG. 22, the second service provider node 2200 may include: a communicating module 2201, configured to: receive sample data from a first service provider node; and select at least one model from a plurality of models based on the sample data. The communicating module 2201 is configured to: obtain information related to the at least one model; and transmit the information related to the at least one model to the first service provider node.
[0287] In an implementation, the communicating module 2201 is configured to receive a request related to at least one task of a mission from the first service provider node; and transmit a response to the first service provider node indicating whether the sample data is needed.
[0288] In an implementation, the request includes a request to notify if the sample data is needed.
[0289] In an implementation, the sample data is received by the first service provider node from at least one of a network entity (NE) and a previous second service provider node.
[0290] In an implementation, the NE is associated with a customer, and a customer’s intent is received from the NE.
[0291] In an implementation, the information related to the at least one model includes information to identify the at least one model.
[0292] In an implementation, the second service provider node 2200 further includes a control module 2202, configured to receive one or more service requirement parameters (SRPs) from the first service provider node, where the one or more SRPs are generated based on the information related to the at least one model. The control module 2202 is configured to convert the one or more SRPs to respective one or more configuration parameters (CPs) based on the at least one model.
[0293] In an implementation, the second service provider node includes a first network component, a second network component and a third network component and the at least one model includes a first sub-model corresponding to the first network component, a second sub-model corresponding to the second network component and a third sub-model corresponding to the third network component.
[0294] In an implementation, the first network component includes a plurality of likelihood estimators.
[0295] In an implementation, the information related to the at least one model includes information related to the first sub-model corresponding to the at least one model.
[0296] In an implementation, the second service provider node 2200 further includes a processing module 2203. The communicating module 2201 is configured to transmit the information related to the first sub-model corresponding to the at least one model to the control module 2202 and the processing module 2203. The control module 2202 is configured to obtain the second sub-model corresponding to the at least one model based on the information related to the first sub-model. The processing module 2203 obtains the third sub-model corresponding to the at least one model based on the information related to the first sub-model. The communicating module 2201 may be implemented by the first network component. The control module 2202 may be implemented by the second network component. The processing module 2203 may be implemented by the third network component.
[0297] In an implementation, the control module 2202 is configured to receive one or more service requirement parameters (SRPs) from the first service provider node, where the one or more SRPs are generated based on the information related to the at least one model. The control module 2202 is configured to convert the one or more SRPs to respective one or more configuration parameters (CPs) based on the at least one model by the second sub-model of the second network component. The control module 2202 is configured to transmit the one or more CPs to the processing module 2203.
[0298] It should be understood by a person skilled in the art that, the relevant description of the above modules in the embodiments of the present disclosure may be understood with reference to the relevant description of the method in the embodiments of the present disclosure.
[0299] FIG. 23 is a schematic structural diagram of an apparatus according to one or more implementations of the present disclosure. As shown in FIG. 23, the apparatus 2300 includes a processor 2301, an interface 2302 for communicating with other devices, a memory 2303 is coupled to the processor 2301. The memory 2303 may be stored with computer execution instructions, and the processor 2301 executes computer execution instructions stored in the memory 2303 to enable the apparatus to execute any of the above methods. In some implementations, the memory 2303 may be included or may not be included in the apparatus.
[0300] It should be noted that the apparatus in the present disclosure may also be implemented as a device, or one or more components included in a device, such as, a processor or a chip. The device may be user equipment, a terminal, a network device, a network function, a network node, or another network element, which is not limited in the present disclosure.
[0301] An embodiment of the present disclosure provides a system, including: the apparatus executing any of the above methods.
[0302] An embodiment of the present disclosure provides a chip, including an input / output (I / O) interface and a processor, where the processor is configured to call and run a computer program stored in a memory, to enable a device installing with the chip to perform any of the above methods.
[0303] It should be noted that the memory in the systems and the methods described in this specification includes but is not limited to these memories and a memory of any other appropriate type.
[0304] An embodiment of the present disclosure provides a computer-readable medium carrying a program code which, when executed by a processor, any of the above methods is performed.
[0305] Optionally, the computer-readable medium may be specifically a memory.
[0306] An embodiment of the present disclosure provides a computer program product storing instructions which, when executed, cause an apparatus to perform any of the above methods.
[0307] An embodiment of the present disclosure provides a computer program storing instructions which, when executed, cause an apparatus to perform any of the above methods.
[0308] Note that when the request or the response mentioned above includes multiple different contents for indicating multiple different pieces of information, the multiple contents can be indicated separately in multiple request / response messages or together in a request / response message.
[0309] Note that the network elements mentioned in the present disclosure are all logical network elements, which can be implemented as individual devices, or can be implemented as chips or modules that could be integrated into a certain device.
[0310] Although the present disclosure describes methods and processes with steps in a certain order, one or more steps of the methods and processes may be omitted or altered as appropriate. One or more steps may take place in an order other than that in which they are described, as appropriate.
[0311] Note that the expression “at least one of A or B” , as used herein, is interchangeable with the expression “A and / or B” . It refers to a list in which you may select A or B or both A and B. Similarly, “at least one of A, B, or C” , as used herein, is interchangeable with “A and / or B and / or C” or “A, B, and / or C” . It refers to a list in which you may select: A or B or C, or both A and B, or both A and C, or both B and C, or all of A, B and C. The same principle applies for longer lists having a same format.
[0312] Although the present disclosure is described, at least in part, in terms of methods, a person of ordinary skill in the art will understand that the present disclosure is also directed to the various components for performing at least some of the aspects and features of the described methods, be it by way of hardware components, software or any combination of the two. Accordingly, the technical solution of the present disclosure may be embodied in the form of a software product. A suitable software product may be stored in a pre-recorded storage device or other similar non-volatile or non-transitory computer readable medium, including DVDs, CD-ROMs, USB flash disk, a removable hard disk, or other storage medium, for example. The software product includes instructions tangibly stored thereon that enable a processing device (e.g., a personal computer, a server, or a network device) to execute examples of the methods disclosed herein. The machine-executable instructions may be in the form of code sequences, configuration information, or other data, which, when executed, cause a machine (e.g., a processor or other processing device) to perform steps in a method according to examples of the present disclosure.
[0313] The present disclosure may be embodied in other specific forms without departing from the subject matter of the claims. The described example embodiments are to be considered in all respects as being only illustrative and not restrictive. Selected features from one or more of the above-described embodiments may be combined to create alternative embodiments not explicitly described, features suitable for such combinations being understood within the scope of this disclosure.
[0314] All values and sub-ranges within disclosed ranges are also disclosed. Also, although the systems, devices and processes disclosed and shown herein may include a specific number of elements / components, the systems, devices and assemblies could be modified to include additional or fewer of such elements / components. For example, although any of the elements / components disclosed may be referenced as being singular, the embodiments disclosed herein could be modified to include a plurality of such elements / components. The subject matter described herein intends to cover and embrace all suitable changes in technology.
[0315] Although embodiments have been described above with reference to the accompanying drawings, those of skill in the art will appreciate that variations and modifications may be made without departing from the scope thereof as defined by the appended claims.
[0316] Please note that the different examples may be implemented separately or combined. Although a combination of features is shown in the illustrated embodiments, not all of them need to be combined to realize the benefits of various examples of the present disclosure. In other words, a system or method designed according to an embodiment of the present disclosure will not necessarily include all of the features shown in any one of the figures or all of the portions schematically shown in the figures. Moreover, selected features of one example embodiment may be combined with selected features of other example embodiments.
[0317] Although this disclosure has been described with reference to illustrative embodiments, the description is not intended to be construed in a limiting sense. Various modifications and combinations of the illustrative embodiments, as well as other examples of the disclosure, will be apparent to persons skilled in the art upon reference to the description. It is therefore intended that the appended claims encompass any such modifications or embodiments.
Claims
1.A method performed by a first service provider node in a network, the method comprising:receiving information related to a mission and information related to at least one second service provider node, wherein the information related to the mission is generated based on a customer’s intent and the at least one second service provider node is identified based on the mission;communicating with the at least one second service provider node based on the information related to the mission, wherein the communicating comprises:transmitting sample data to the at least one second service provider node; andreceiving information related to at least one model from each of the at least one second service provider node, wherein the at least one model is selected based on the sample data.2.The method of claim 1, wherein the first service provider node comprises a first network element and a second network element, wherein the first network element performs the method of claim 1.3.The method of claim 1 or 2, wherein receiving the information related to the mission and the information related to the at least one second service provider node comprises receiving, by the first network element of the first service provider node, the information related to the mission and the information related to the at least one second service provider node from the second network element of the first service provider node.4.The method of any one of claims 1 to 3, wherein transmitting the sample data comprises:receiving, from the at least one second service provider node, a notification indicating that sample data is needed; andreceiving the sample data from at least one of a network entity (NE) and a previous second service provider node.5.The method of claim 3 or 4, wherein the NE is associated with a customer, and the customer’s intent is received by the second network element from the NE.6.The method of claim 1, further comprising:generating respective one or more service requirement parameters (SRPs) for each of the at least one second service provider node, wherein the one or more SRPs are generated based on the information related to the at least one model; andtransmitting the respective one or more SRPs to each of the at least one second service provider node.7.The method of claim 6, wherein generating the one or more SRPs further comprises:receiving network state information; andgenerating the one or more SRPs based on at least one SRP policy using the network state information.8.The method of claim 1, further comprising:receiving feedback from the customer; andupdating one or more SRP policies based on the feedback using one or more learning algorithms.9.A method performed by a second service provider node in a network, the method comprising:receiving sample data from a first service provider node;selecting at least one model from a plurality of models based on the sample data;obtaining information related to the at least one model; andtransmitting the information related to the at least one model to the first service provider node.10.The method of claim 9, wherein receiving the sample data comprises:receiving a request related to at least one task of a mission from the first service provider node; andtransmitting a response to the first service provider node indicating whether the sample data is needed.11.The method of claim 10, wherein the request comprises a request to notify if the sample data is needed.12.The method of any one of claims 9 to 11, wherein the sample data is received by the first service provider node from at least one of a network entity (NE) and a previous second service provider node.13.The method of claim 12, wherein the NE is associated with a customer, and a customer’s intent is received from the NE.14.The method of claim 13, wherein the information related to the at least one model comprises information to identify the at least one model.15.The method of any one of claims 9 to 14, further comprising:receiving one or more service requirement parameters (SRPs) from the first service provider node, wherein the one or more SRPs are generated based on the information related to the at least one model; andconverting, the one or more SRPs to respective one or more configuration parameters (CPs) based on the at least one model.16.The method of any one of claims 9 to 14, wherein the second service provider node comprises a first network component, a second network component and a third network component and wherein the at least one model includes a first sub-model corresponding to the first network component, a second sub-model corresponding to the second network component and a third sub-model corresponding to the third network component.17.The method of claim 16, wherein the first network component comprises a plurality of likelihood estimators.18.The method of claim 16 or 17, wherein the information related to the at least one model comprises information related to the first sub-model corresponding to the at least one model.19.The method of claim 18, comprising:transmitting, by the first network component to the second network component and the third network component, the information related to the first sub-model corresponding to the at least one model; andobtaining, by the second network component, the second sub-model corresponding to the at least one model based on the information related to the first sub-model; andobtaining, by the third network component, the third sub-model corresponding to the at least one model based on the information related to the first sub-model.20.The method of any one of claims 16 to 19, further comprising:receiving, by the second network component, one or more service requirement parameters (SRPs) from the first service provider node, wherein the one or more SRPs are generated based on the information related to the at least one model;converting, by the second sub-model of the second network component, the one or more SRPs to respective one or more configuration parameters (CPs) based on the at least one model; andtransmitting, by the second network component, the one or more CPs to the third network component.21.A system comprising:a network entity (NE) ;an autonomous programming component;a network state observer (NSO) ; anda plurality of second service provider nodes, wherein:the autonomous programming component is configured to:receive a customer’s intent from the NE; andreceive network state information from the NSO,wherein the autonomous programming component is configured to perform the method of any one of claims 1 to 8; andwherein each of the plurality of second service provider nodes is configured to perform the method of any one of claims 9 to 20.22.A first service provider node in a network, comprising:a communicating module, configured to:receive information related to a mission and information related to at least one second service provider node, wherein the information related to the mission is generated based on a customer’s intent and the at least one second service provider node is identified based on the mission;communicate with the at least one service provider node based on the information related to the mission, wherein the communicating module is configured to:transmit sample data to the at least one second service provider node; andreceive indices related to at least one model from each of the at least one second service provider node, wherein the at least one model is selected based on the sample data.23.A second service provider node in a network, comprising:a communicating module, configured to:receive sample data from a first service provider node; andselect at least one model from a plurality of models based on the sample data;wherein the communicating module is configured to:obtain information related to the at least one model; andtransmit the information related to the at least one model to the first service provider node.24.A first service provider node in a network, comprising at least one processor coupled to a memory storing a set of instructions;wherein the at least one processor is configured to read the set of instructions in the memory and execute the method of any one of claims 1 to 8.25.A second service provider node in a network, comprising at least one processor coupled to a memory storing a set of instructions;wherein the at least one processor is configured to read the set of instructions in the memory and execute the method of any one of claims 9 to 20.26.A computer-readable storage medium having instructions stored thereon which, when executed by an apparatus, cause the apparatus to perform the method of any one of claims 1 to 20.27.A computer program product storing instructions which, when executed, cause an apparatus to perform the method of any one of claims 1 to 20.