Intelligent agents in a communication network

Abstract

Example embodiments of this disclosure relate to intelligent agents (IAs) in a communication network. One method includes: determining, based on a determination that an intelligent agent (IA) is required to respond to a request, an execution plan to be performed by at least one IA; and sending the execution task of the execution plan to a second device hosting a target IA among the at least one IA.

Description

Technical Field

[0001] Various exemplary embodiments of this disclosure generally relate to the field of telecommunications, and more particularly to methods, apparatuses, devices, and computer-readable storage media for intelligent agents (IA) in communication networks. Background Technology

[0002] A communication network can be used as a facility to enable communication between two or more communication devices or to provide communication devices with access to a data network. A mobile or wireless communication network is an example of a communication network. Communication devices may be served by an application server.

[0003] Communication networks can operate according to standards provided by organizations such as the 3rd Generation Partnership Project (3GPP) or the European Telecommunications Standards Institute (ETSI). Examples of standards provided by 3GPP are the so-called 3GPP standards for cellular technology generations, such as the 3GPP standards for 4G, 5G, 6G, and so on. Summary of the Invention

[0004] In a first aspect of this disclosure, a first apparatus is provided. The first apparatus includes: at least one processor; and at least one memory storing instructions, which, when executed by the at least one processor, cause the first apparatus to: determine, based on a request, an execution plan to be executed by the at least one IA, and send the execution plan to a second apparatus hosting a target IA in the at least one IA.

[0005] In a second aspect of this disclosure, a method is provided. The method includes: determining, based on a determination that an intelligent agent (IA) is required to respond to a request, an execution plan to be executed by at least one IA; and sending the execution task of the execution plan to a second device hosting a target IA among the at least one IA.

[0006] In a third aspect of this disclosure, a first apparatus is provided. The first apparatus includes: components for determining an execution plan to be executed by at least one IA based on a determination that an intelligent agent (IA) is needed to respond to a request; and components for sending the execution task of the execution plan to a second apparatus hosting a target IA in at least one IA.

[0007] In a fourth aspect of this disclosure, a computer-readable medium is provided. The computer-readable medium includes instructions stored thereon for causing a device to at least execute the method according to the second aspect.

[0008] It should be understood that the summary portion is not intended to identify key or essential features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0009] Some exemplary embodiments will now be described with reference to the accompanying drawings, in which: Figure 1 An example communication environment in which example embodiments of the present disclosure may be implemented is shown; Figure 2 The signaling flow for an example procedure for IA orchestration according to some example embodiments of this disclosure is shown; Figure 3 A schematic diagram of an intelligent agent framework according to some example embodiments of the present disclosure is shown; Figure 4 An example Unified Modeling Language (UML) diagram of an information model for IA orchestration according to some example embodiments of the present disclosure is shown; Figure 5 The following illustrates example signaling flows between a consumer and a producer regarding the lifecycle of a task or intent, according to some example embodiments of this disclosure; Figure 6 The signaling flow for an example process for agent discovery and configuration according to some example embodiments of this disclosure is shown; Figure 7 Example diagrams are shown illustrating examples of use cases in which intelligent agents are assisting in network management and optimization, according to some example embodiments of this disclosure; Figure 8 Example diagrams are shown illustrating examples of use cases in which an intelligent agent is customized for consumer-oriented tasks according to some example embodiments of the present disclosure; Figure 9 A flowchart is shown illustrating a method implemented at a first device according to some exemplary embodiments of the present disclosure; Figure 10 A simplified block diagram of a device suitable for implementing example embodiments of the present disclosure is shown; and Figure 11 A block diagram of an example computer-readable medium according to some example embodiments of the present disclosure is shown.

[0010] Throughout the accompanying drawings, the same or similar reference numerals denote the same or similar elements. Detailed Implementation

[0011] The principles of this disclosure will now be described with reference to some exemplary embodiments. It should be understood that these embodiments are described for illustrative purposes only and to assist those skilled in the art in understanding and implementing this disclosure, without imposing any limitation on the scope of this disclosure. The embodiments described herein can be implemented in various ways other than those described below.

[0012] In the following description and claims, unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure pertains.

[0013] References to "an embodiment," "embodiment," "example embodiment," etc., in this disclosure indicate that the described embodiment may include a particular feature, structure, or characteristic, but not every embodiment needs to include that particular feature, structure, or characteristic. Furthermore, such phrases do not necessarily refer to the same embodiment. Moreover, when a particular feature, structure, or characteristic is described in connection with an embodiment, whether explicitly described or not, it is believed that its influence on such feature, structure, or characteristic in conjunction with other embodiments is within the knowledge of those skilled in the art.

[0014] It should be understood that although various elements may be described herein using prefixes such as "first," "second," etc., these elements should not be limited by these terms. These terms are used only to distinguish one element from another, and they do not restrict the order of the terms. For example, without departing from the scope of the exemplary embodiments, a first element may be referred to as a second element, and similarly, a second element may be referred to as a first element. As used herein, the term "and / or" includes any and all combinations of one or more of the listed terms.

[0015] As used herein, “at least one of the following: ” and “at least one of ” and similar wording, where the list of two or more elements is connected by “and” or “or”, means at least any one of the elements, or at least any two or more of the elements, or at least all of the elements.

[0016] As used herein, unless explicitly stated otherwise, the action “in response to A” does not indicate that the action is performed immediately after “A” occurs and may include one or more intervention steps.

[0017] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments. As used herein, the singular forms “a,” “an,” and “the” are also intended to include the plural forms unless the context clearly indicates otherwise. It will be further understood that the terms “comprising,” “including,” “having,” “containing,” and / or “including” as used herein specify the presence of the stated features, elements, and / or components, but do not exclude the presence or addition of one or more other features, elements, components, and / or combinations thereof.

[0018] As used in this application, the term "circuit system" may refer to one or more of the following: (a) Hardware circuit implementation only (such as implementation in analog and / or digital circuits only), and (b) A combination of hardware circuitry and software, such as (if applicable): (i) A combination of (multiple) analog and / or digital hardware circuits and software / firmware, and (ii) Any part of the (multiple) hardware processors having software (including (multiple) digital signal processors working together to enable a device (such as a mobile phone or server) to perform various functions), software, and (multiple) memory), and (c) Multiple hardware circuits and / or multiple processors, such as multiple microprocessors or a portion thereof, that require software (e.g., firmware) to operate, but the software may not be present when operation is not required.

[0019] This definition of circuit system applies to all uses of the term in this application, including in any claim. As another example, as used in this application, the term circuit system also covers implementations of hardware circuitry or processors (or processors in general) or a portion thereof and their accompanying software and / or firmware. For example, if applicable to a particular claim element, the term circuit system also covers baseband integrated circuits or processor integrated circuits used in mobile devices or servers, cellular network devices, or other computing or networking devices.

[0020] As used herein, the term "communication network" refers to a network that conforms to any suitable communication standard, such as New Radio (NR), Long Term Evolution (LTE), LTE-A Advanced (LTE-A), Wideband Code Division Multiple Access (WCDMA), High-Speed ​​Packet Access (HSPA), Narrowband Internet of Things (NB-IoT), etc. Furthermore, communication between terminal devices and network devices in a communication network can be performed according to any suitable generated communication protocol, including but not limited to first-generation (1G), second-generation (2G), 2.5G, 2.75G, third-generation (3G), fourth-generation (4G), 4.5G, fifth-generation (5G), 5.5G, sixth-generation (6G) communication protocols and / or any other currently known or future-developed protocols. Embodiments of this disclosure can be applied to a variety of communication systems. Given the rapid development in communications, there will naturally be future types of communication technologies and systems that can implement this disclosure. The scope of this disclosure should not be limited to the aforementioned systems only.

[0021] As used herein, the term "network device" refers to a node in a communications network through which terminal devices access the network and receive services. Network devices can refer to base stations (BS) or access points (APs), such as Node B (NodeB or NB), evolved Node B (eNodeB or eNB), NR NB (also known as gNB), Remote Radio Unit (RRU), Radio Head (RH), Remote Radio Head (RRH), repeater, Integrated Access and Backhaul (IAB) node, low-power node (such as femtoseconds, picoseconds), non-terrestrial network (NTN) or non-terrestrial network equipment (such as satellite network equipment, low Earth orbit (LEO) satellites, and geostationary Earth orbit (GEO) satellites), spacecraft network equipment, etc., depending on the terminology and technology applied. In some example embodiments, the radio access network (RAN) split architecture includes a centralized unit (CU) and a distributed unit (DU) at the IAB donor node. An IAB node includes a mobile terminal (IAB-MT) portion that behaves like a UE toward its parent node, and a DU portion of the IAB node that behaves like a base station toward the next-hop IAB node.

[0022] The term "terminal device" refers to any terminal device capable of wireless communication. As an example and not a limitation, a terminal device may also be referred to as a communication device, user equipment (UE), subscriber station (SS), portable subscriber station, mobile station (MS), or access terminal (AT). Terminal devices can include, but are not limited to, mobile phones, cellular phones, smartphones, Voice over IP (VoIP) phones, wireless local loop phones, tablets, wearable terminal devices, personal digital assistants (PDAs), portable computers, desktop computers, image capture terminal devices (such as digital cameras), gaming terminal devices, music storage and playback devices, in-vehicle wireless terminal devices, wireless endpoints, mobile stations, laptop embedded devices (LEEs), laptop devices (LMEs), USB dongles, smart devices, wireless customer premises equipment (CPEs), Internet of Things (IoT) devices, watches or other wearable devices, head-mounted displays (HMDs), vehicles, drones, medical devices and applications (e.g., remote surgery), industrial devices and applications (e.g., robots and / or other wireless devices operating in the context of industrial and / or automated processing chains), consumer electronics devices, devices operating on commercial and / or industrial wireless networks, etc. The terminal device may also correspond to the mobile terminal (MT) portion of an IAB node (e.g., a relay node). In the following description, the terms "terminal device," "communication device," "terminal," "user equipment," and "UE" are used interchangeably.

[0023] As used herein, the terms “resource,” “transmission resource,” “resource block,” “physical resource block” (PRB), “uplink resource,” or “downlink resource” can refer to any resource used to perform communication, such as communication between a terminal device and a network device, including resources in the time domain, frequency domain, spatial domain, code domain, or any other combination of time, frequency, spatial, and / or code domain resources used to implement communication. In the following, unless explicitly stated otherwise, resources in both the frequency and time domains will be used as examples of transmission resources used to describe some exemplary embodiments of this disclosure. Note that the exemplary embodiments of this disclosure are equally applicable to other resources in other domains.

[0024] As used herein, the term "intelligent agent" (IA) (also simply agent) can refer to an entity that makes decisions and takes actions based on its perception of its environment. An IA can be implemented based on any suitable algorithm, such as rule-based, classical machine learning models, or deep learning models. An intelligent agent can act as an autonomous system capable of perceiving its environment using different data modalities and leveraging various tools to derive decisions and take actions. In the telecommunications field, an agent can correspond to a product or solution capable of performing specific telecommunications tasks (e.g., adjusting the power level of a gNB).

[0025] Generative artificial intelligence (AI) is a method for creating new content of different types that follows the characteristics of the training data. One approach to generative AI is the Large Language Model (LLM) for generating plausible text / language based on input queries. Currently, there are many examples of proprietary and open-source models available and used in various applications. A general-purpose LLM (base model) is typically obtained through extensive training on large amounts of data to capture relationships between words and acquire general capabilities for text understanding, processing, and generation. This process is called pre-training. Fine-tuning is the process of adapting the general-purpose model to domain-specific tasks, such as understanding technical text and recommending administrative actions in the telecommunications domain. This can be done by selectively tuning / training a subset of model parameters or a set of newly added parameters.

[0026] LLMs use a single input and output data modality, and that modality is language. Large multimodal models (LMMs) combine various data modalities, such as text, audio, vision, sensor data, etc., to capture the correlations between different modalities. This approach can be applied to any kind of data, including web data. Finally, small language models (SLMs) are less computationally intensive than LLMs in terms of training and inference, and have good performance, especially if they are trained and used for specific problems, thus showing high relevance in telecommunications applications in addition to LLMs and LMMs.

[0027] AI agents are generally defined as entities that make decisions and act based on their perception of their environment. Different types of AI agents can exist, such as rule-based, reinforcement learning-based, and LLM / LMM / SLM-based agents. LLM / LMM / SLM-based agents utilize LLM / LMM / SLM as a "brain" to extend their ability to perceive their environment and take actions through multimodal perception strategies that utilize external tools (such as internet searches) or consider different types / modalities of input data (such as audio, vision, text, web KPIs and metrics).

[0028] Chain-of-Thought (CoT) is another strategy for reasoning and planning by enabling LLMs to provide outputs and stepwise descriptions (i.e., "chains of thought," while breaking down high-level tasks into smaller ones). LLM-based agents exhibit performance improvements due to their ability to capture knowledge, interpret instructions, causes, etc. Furthermore, LLM-based agents can understand language and multimodal inputs, such as customer documents, ticket resolution reports, and user feedback.

[0029] LLM has demonstrated significant reasoning capabilities. Furthermore, LLM can be extended with other components, such as memory, utilization of different tools, environmental awareness, and critique or implementation—that is, the development of LLM agents that interact with the environment and perform autonomous actions, based on output generation of relevant information, such as use-case specific information that cannot be used as part of training data.

[0030] Such agents can interact with each other to accomplish high-level or complex tasks. In such a multi-agent system, agent collaboration is crucial for completing high-level tasks by performing a set of smaller tasks, such as searching, optimizing, allocating resources, etc.

[0031] An agent typically has several components: memory, planning, and action execution. It may also have a policy component or a critique component. In a network, each functional entity can have at least one agent supported by at least one LLM / LMM / SLM.

[0032] Each agent can be aware of the network state between the functional entity itself and its directly related functional entities, for example, by using tools / APIs for externally related network environment data, consumer services from other network functions for acquiring, for example, FCAP data and performing analysis on the latest collected data for network state analysis, or by using the capabilities of other agents in the related network entities.

[0033] Each agent can process and store collected, generated analytics data and other task-related context in memory, which can be retrieved throughout the lifecycle of the request execution. Each agent can break down requests from consumers (e.g., users, other agents) into tasks or action plans, execute the action plans, and then generate the final result. Each agent can have policy controls for planning, action execution, and working memory capabilities.

[0034] In the case of intelligent agents, enabling controlled use of agents in telecommunications systems and activating and controlling their collaboration are fundamental to unlocking the full potential of intelligent agents.

[0035] According to some example embodiments of this disclosure, a solution for an IA (Integrated Automation) system in a communication network is provided. In the solution, a first device receiving a request determines whether an IA is needed to respond to the request. If an IA is needed, the first device determines an execution plan to be performed by at least one IA based on the request. The first device then sends the execution task of the execution plan to a second device hosting a target IA among the at least one IA.

[0036] Instead of targeting a single (simple) intelligent agent, the proposed solution enables collaboration between different intelligent agents. In this way, improvements in network performance and automation are achieved, allowing the full potential of intelligent agents to solve complex tasks. Therefore, network performance and automation can be enhanced.

[0037] The exemplary embodiments of this disclosure will now be described in detail with reference to the accompanying drawings.

[0038] First refer to Figure 1 . Figure 1 An example communication environment 100 in which exemplary embodiments of the present disclosure may be implemented is illustrated. The communication environment 100 may include a first device 110 and a second device 120 for communication. The second device 120 hosts an intelligent agent 130. The first device 110 and the second device 120 may be any suitable device in a communication network. For example, the first device 110 may be or can be included in a terminal device, radio access network (RAN) equipment, core network (CN) function / element, management device / function, or third-party application / service. Similarly, the second device 120 may be or can be included in a terminal device, radio access network (RAN) equipment, core network (CN) function / element, management device / function, or third-party application / service.

[0039] Communication in communication environment 100 can be implemented according to any suitable communication protocol, including but not limited to cellular communication protocols, wireless local area network communication protocols (such as IEEE 802.11, etc.), and / or any other currently known or future-developed protocols. Furthermore, communication can utilize any suitable wireless communication technology, including but not limited to: Code Division Multiple Access (CDMA), Frequency Division Multiple Access (FDMA), Time Division Multiple Access (TDMA), Frequency Division Duplex (FDD), Time Division Duplex (TDD), Multiple Input Multiple Output (MIMO), Orthogonal Frequency Division Multiple Access (OFDM), Discrete Fourier Transform Extended OFDM (DFT-s-OFDM), and / or any other currently known or future-developed technologies.

[0040] It should be understood that Figure 1 The number of devices and their connections shown are for illustrative purposes only and do not impose any limitations. Communication environment 100 may include any suitable number of devices configured to implement the exemplary embodiments of this disclosure. Although not shown, it should be understood that one or more additional devices may be deployed in communication environment 100. For example, communication environment 100 may include additional devices communicating with first device 110, such as a third device sending a request to first device 110.

[0041] Figure 2 This section describes the signaling flow of an example process 200 for IA orchestration according to some example embodiments of this disclosure. Reference will be made to this section for discussion purposes. Figure 1 The process 200 is described by the first device 110 and the second device 120.

[0042] In the following text, the first device 110 may also be referred to as an IA orchestration device configured to perform IA orchestration or management. IA can be based on any suitable artificial intelligence / machine learning (AI / ML) model, such as LLM, LMM, or SLM. The scope of this disclosure is not limited in this respect.

[0043] In some embodiments, the first device 110 may be a newly defined core network function named Intelligent Agent Orchestration Function (IAOF). Alternatively, the first device 110 may be implemented in an existing core network function (e.g., AI / ML inference function). In some embodiments, the first device 110 may be implemented in an end device, management node / function, RAN node, or third-party device or application.

[0044] like Figure 2As shown, in process 200, based on the determination that an IA is needed to respond to a request, a first device 110 determines an execution plan 212 to be performed by at least one IA. This request may be received by the first device 110 from another device or generated by the first device 110 itself. The first device 110 also sends an execution task 215 of the execution plan to a second device 120 hosting the target IA in the at least one IA. The second device 120 (i.e., the target IA) receives the execution task 218 and can execute (not shown) the received execution task.

[0045] An execution plan can refer to a list of actions, operable (sub)tasks, or execution (sub)tasks with certain specific dependencies and / or constraints. The actions or (sub)tasks in the list can be executed directly or via an agent. Depending on the request, an execution plan can include a variety of execution tasks. For example, if the request is an intent such as "I want the RAN energy consumption in area X to be lower than Y kWh," the execution plan can include specific tasks such as transmit power adjustment, cell on / off handover, and selection of the most energy-efficient UPF. Specific tasks in the execution plan can be executed by the corresponding IA determined based on the request.

[0046] Figure 3 A schematic diagram of an intelligent agent framework 300 according to some example embodiments of the present disclosure is shown. The intelligent agents work together for specific tasks, such as fulfilling a network operator's intent to reduce energy consumption by a certain percentage. The intelligent agents involved may not have special relationships, or they may be constructed or organized in a hierarchical manner.

[0047] Intelligent agent 301 (referred to as the planning agent) receives a task (referred to as task A, such as a network operator intent) from consumer 330. Intelligent agent 301 needs to break down task A into smaller subtasks (e.g., task N1, task N2, ..., task Nm). As an example, Figure 3 Tasks N1, N2, and N3 are shown. Each subtask can be received by a smart agent. For example, smart agents 311, 312, and 313 receive tasks N1, N2, and N3, respectively. Task Nm can be further decomposed by another receiving smart agent (e.g., agent N1…agent Nn), however, this decomposition level is not visible to smart agent A. It is assumed that the subtask smart agents will be domain-specific agents supported by domain-specific LLM / LMM / SLM. For example, agent 311 can perform gNB power level adjustment, for which executor 321 could be as follows: Figure 3 The gNB 1 shown. Agent 312 can perform energy-aware UPF selection in the core network, and the executor 322 for it can be either an SMF or a UPF. Agent 313 can perform any action, and the executor 322 for it can be gNB N. Note that regarding Figure 3 The actions described are examples and have no limitations. For example... Figure 3 The challenge in the scenario illustrated is how intelligent agent 301 can know which intelligent agents are performing which tasks, their capabilities, and requirements. Intelligent agent 301 can also monitor and control the completion of task A by tracking the progress of all involved intelligent agents and evaluating the results of the actions of intelligent agent Nx. Some example embodiments for addressing these challenges are described below.

[0048] In some embodiments, based on the request, the first device 110 may first determine whether an IA is needed to respond to the request. If the first device 110 determines that an IA is needed, then the first device 110 may determine an execution plan and at least one IA for executing the execution plan. In some embodiments, the first device 110 may determine whether the request can be mapped to a set of create, read, update, delete (CRUD) operations. If so, then the first device 110 may determine that the request can be responded to by a simple CRUD operation, so that an IA is not needed. Conversely, if the first device 110 determines that the request cannot be mapped to a simple CRUD operation, then the first device 110 may determine that an IA is needed to respond to the request.

[0049] In some embodiments, the first device 110 may determine at least one IA for executing the execution plan by using a first IA hosted in the first device 110. The first IA hosted in the first device 110 may also be referred to hereinafter as the primary agent or planning agent. The planning agent may select a list of IAs corresponding to the execution plan and / or configure collaboration between the selected IAs.

[0050] In some embodiments, based on determining that an IA is needed to respond to a received request, the first device 110 may instantiate a first IA for planning an execution schedule, that is, determining an execution schedule and a plurality of corresponding IAs for executing the execution tasks(s) in the execution schedule(s). Alternatively, the first IA hosted in the first device 110 may be installed by a telecommunications system.

[0051] Alternatively, the first device 110 may determine, based on a request, at least one IA including a first IA hosted in the first device. That is, the first device 110 may use components other than the IA to determine an execution plan, and then instantiate the first IA in the device 110 to execute the execution plan. For example, a predetermined mapping from request to execution plan may be used.

[0052] In some embodiments, the first device 110 may determine an execution plan to be executed by at least one IA by decomposing the request into multiple execution tasks having execution dependencies and / or execution constraints. For example, the first device 110 may decompose or separate the request into execution plans by taking into account task dependencies, task constraints, proxy dependencies, and / or proxy constraints.

[0053] In some embodiments, the first device 110 may determine at least one IA for executing the execution plan based on the discovery and / or registration of IAs (and their capabilities). For example, multiple IAs may be registered to core network elements, and their corresponding capability information may be stored. The first device 110 may use any suitable means to discover IAs and their capability information, and select at least one IA from the discovered IAs for executing the determined execution plan based on the request and the IA capability information.

[0054] In some embodiments, the first device 110 may receive IA capability information from a third device that sends a request to the first device. The third device may also be referred to as a consumer requesting services from the first device 110 (also referred to as a producer in this case). In some embodiments, the first device 110 may receive IA capability information from the consumer as at least part of IA provisioning information. IA provisioning information may refer to provisioning / requirements or configurations determined by the consumer for the IA(x) ...(x)(x))(x)(x)(x)(x)(x)(x)(x)(x)(x)(x)(x)(x)(x)(x))(x)(x)(x

[0055] Alternatively or additionally, the first device 110 may receive IA capability information from a fourth device (e.g., a core network element to which IA is registered) responsible for the exposure or discovery service of IA capability information. For example, the first device 110 may receive IA capability information during the discovery process.

[0056] In some embodiments, IA capability information may include IA planning capabilities associated with the planning execution plan as described above. For example, the first device 110 may select a planning IA based on the planning capabilities exposed by different IAs in the network (and optional IA provisioning information sent from the consumer).

[0057] Alternatively or additionally, IA capability information may include information about at least one task supported by the agent, such as a list of tasks that the IA can perform, or a description of the tasks(s) that the IA can perform.

[0058] Alternatively or additionally, IA capability information may include capabilities for the agent's underlying model. For example, model-related capabilities may include the type of (generative) model used by the IA, such as LLM, LMM, etc. Model-related capabilities may include the size of the model powering the agent (e.g., large if the number of parameters exceeds one billion, otherwise small). Model-related capabilities may include the training methods used to obtain the model powering the IA. For example, pre-training can be used to obtain a general model, enabling the corresponding IA to have general capabilities, such as recognizing relationships between input data. Fine-tuning can be used to obtain a domain- or task-specific model, enabling the corresponding IA to have specific capabilities for handling specific tasks within a specific network domain. Capturing internal knowledge of a specific operational domain (e.g., in the RAN or core) can be used for fine-tuning.

[0059] Alternatively or additionally, IA capability information may include memory / context length capabilities related to the length of memory / context supported by the agent. For example, an LLM model supports a limited memory length (context length), and the agent's memory length is no longer than the model's context length.

[0060] Alternatively or additionally, IA capability information may include the ability to use external knowledge (i.e., knowledge outside the model) or external tools, such as internet searches. Alternatively or additionally, IA capability information may include the ability to collaborate with another agent. For example, an independent IA may be designated to perform a single task, rather than being designed to receive input from other IA(s). Collaborating IAs may require input from other IA(s).

[0061] Alternatively or additionally, IA capability information may include critical capabilities for validating and refining the output from another agent. For example, an IA may be able to validate the output of other IAs(s) and provide recommendations on their refinement. Alternatively or additionally, IA capability information may include fallback capabilities. It can indicate fallback options for resolving IA tasks in the event of task failure.

[0062] Alternative or additional locations, IA capability information may include responsibility and reliability capabilities. It may indicate, for example, IA availability in space (e.g., an area of ​​interest represented by geographic coordinates or a list of RAN nodes), IA availability in time (e.g., time of day), task completion reliability (e.g., “timely” task fulfillment), and / or maintainability / fault recovery.

[0063] Alternatively or additionally, IA capability information may include data aspects of IA. Examples may include the amount of data used to train models powering IA (e.g., the number of labels, data size) and the data types used for training and expected as input to the models. Examples of data types may include a list of languages ​​or applicable languages. Examples of data types may also include mobile network data, such as network metrics, key performance indicators (KPIs), and fault, configuration, billing, and performance (FCAP) management data. Examples of data types may also include other relevant data, such as data obtained from simulations, lab tests, field tests; customer specification documents related to products and features; and historical records of network operations, such as troubleshooting ticket solutions, service provision, customer satisfaction reports, etc. Examples of data types may also include audio data, video data, sensor data, etc.

[0064] In some embodiments, information elements (IEs) can be used to indicate IA capability information. (See now for reference.) Figure 4 . Figure 4 Example diagrams of an information model 400 for IA orchestration according to some example embodiments of the present disclosure are shown. Information model 400 includes information object classes (IOCs) and data types required to implement IA orchestration.

[0065] like Figure 4 As shown, it is named proxy capability ( AgentCapability The data type of 404 can be used to indicate IA capability information. Data type AgentCapability A 404 error can include the following attributes as shown in Table 1. In Table 1 and the following tables, M indicates mandatory, O indicates optional, and C indicates conditional.

[0066] Table 1 shows the attributes of IA capability information.

[0067] In Table 1, the attribute "supportedTaskTypeList" indicates the list of tasks the agent can perform along with the task description. The attribute "baseModelCapabilities" indicates the capabilities of the large model that supports the agent, or the types of models the agent connects to. These capabilities may include, for example, domain-specific strength, capturing internal knowledge of a specific operational domain (such as RAN or core); data aspects of the model (e.g., the amount or type of data used for model training); model capabilities derived from training; and fine-tuning (e.g., the ability to solve general or specific tasks).

[0068] The attribute "useExternalKnowledge" indicates the ability to directly use external knowledge other than that from the model. The attribute "externalKnowledgeTypeList" indicates the types of external knowledge.

[0069] The attribute "collaborationWithOtherAgent" indicates the ability to collaborate with another agent. The attribute "usingTool" indicates the use of external tools. The attribute "memoryLength" indicates the length of memory supported by the agent. The attribute "criticsCapabilities" indicates the ability to validate the output of other intelligent agents and provide recommendations on their refinement.

[0070] In some embodiments, the first device 110 may receive IA provisioning information and determine an execution plan to be performed by at least one IA based on the IA provisioning information. In some embodiments, the first device 110 may also receive updates to the IA provisioning information during the execution of the execution plan. As described above, the IA provisioning information may include IA capability information transmitted from the consumer.

[0071] In some embodiments, the IA provisioning information may further include IA agent information. Agent information may refer to general information about the IA. For example, agent information may include agent capability information as described above, agent type, a list of potential collaborating agents or conditions for filtering collaborating agents, agent responsibility score, and / or policies for the agent. Agent responsibility score may indicate the agent's responsibility. It may be a floating-point number in the range [0, 1], with higher values ​​implying higher responsibility. It may be determined based on evaluations from previous task executions through self-evaluation, feedback from the consumer, or overall evaluation. Policies for the agent may indicate control policies used to control agent capabilities, and details of the provisioning policies will be described below.

[0072] Figure 4 The information model 400 shown includes an IOC “Intelligent Agents” 403 to indicate the agent information. Table 2 shows example properties of the agent information.

[0073] Table 2. Example attributes of agent information

[0074] The attribute "agentType" indicates the type of agent. The attribute "agentCapabilities," of type "agentCapabilities," indicates the agent's capabilities. These capabilities can be derived from pre-training, fine-tuning a large model, or using external knowledge. It can be represented as: a list of tasks the agent can perform, along with task descriptions; the capabilities of the large model supporting the agent, or the type of model the agent connects to; the use of external knowledge, and the type of external knowledge; the ability to collaborate with other agents(s); the use of external tools; infrastructure capabilities (optional); and critical capabilities. Capabilities can include domain-specific strength, internal knowledge of capturing a specific operational domain (such as RAN or core); the agent's area of ​​focus (represented, for example, a list of geographic coordinates or RAN nodes); data aspects of the model (e.g., the amount or type of data used for model training); capabilities derived from training the model; and fine-tuning (e.g., the ability to solve general or specific tasks).

[0075] The attribute "agentAccountabilityScore" indicates the agent's accountability. It can be a floating-point number ranging from 0 to 1, with higher values ​​indicating greater accountability. It can be an evaluation of previous task performance through self-assessment, feedback from the consumer, or an overall assessment.

[0076] The "collocationAgentSelectionOptions" property specifies the criteria for filtering the agents used for collaboration within a subtask. Example filters can include location or data modality filters.

[0077] The attribute "policyForAgent" indicates the control policy used to control agent capabilities. It can include: the ability to activate / deactivate the agent; "ExecutionStrategy," indicating the priority, best performance, balance, and bestEnergySaving (thus minimizing interaction with other agents) for task execution; "ScopeOfExternalInteraction," a filter for the scope of each geographic location, each network function list, or agent type; "MaxNumberOfExternalAgent," indicating the maximum number of external agents for a given request; and "fallbackCapabilities," indicating fallback options for resolving subtasks when a subtask request fails.

[0078] In some example embodiments, IA policy information may include at least one of the following: managed activation / deactivation scope for agent capabilities, control over agent capabilities or behavior, task execution policies, scope of external interactions, or fallback options for task failures.

[0079] Policy or IA policy information for agents can be represented by the attribute "PolicyForAgent" in Table 2. Example attributes for agent policies are shown in Table 3.

[0080] Table 3. Example properties of strategies used for agents

[0081] For example, a policy for an agent may include a managed activation scope, which indicates the range of capabilities that can be activated or deactivated based on factors such as: on / off control, conditional activation / deactivation based on location, time, or specific task type. Managed capabilities may include IA capabilities as described above, such as supported consumer-side services (e.g., analytics output) from other network functions, or other services from the same network function, the ability to use external tools / services (APIs), the ability to interact with other agents, and the ability to use multiple different backend LLMs or LLM-based models. The managed activation scope may be represented by the attribute “managedActivationScope” shown in Table 3.

[0082] The "Execution Strategy" attribute indicates the priority of task execution. It can be a list of ENUM's best performance, balanced, and most energy-efficient (therefore, less interaction with other agents) options. The "scopeOfExternalInteraction" attribute indicates filters for the scope of each geographic location, each network function list, or agent type, etc. The "fallbackCapabilities" attribute indicates fallback options used to resolve subtasks when their requests fail.

[0083] In some embodiments, the agent-specific policy, i.e., the attribute "PolicyForAgent", may be part of the IA provisioning information.

[0084] In some embodiments, the first device 110 may receive IA policy information from a third device or mobile network operator (MNO) that sends a request to the first device. In some embodiments, the IA provisioning information may further include IA activation or deactivation configuration information. The IA activation or deactivation configuration information may include activation or deactivation of at least one of the following: agent, agent capabilities, network function (NF) service, or agent model.

[0085] In the example, the data type "Managed Activation" can represent the ability to activate / deactivate a smart agent, the managed activation / deactivation scope for the agent's capabilities (e.g., on / off control, condition-based activation / deactivation based on location, time, or specific task type, etc.). The data type "Managed Activation" can include example attributes of the data type "Managed Activation" shown in Table 4.

[0086] Table 4. Example properties of the data type "Managed Activation"

[0087] The attribute "activationNFservices" indicates that the agent can consume services from other network functions or other services from the same network function. It can be a list of services, or "ALL" to indicate all NF services.

[0088] The attribute "activationAsPerLocation" indicates a location constraint, meaning the agent can only be activated if the task execution involves a specific geographic area. It can be a list of location types (e.g., polygon, center and radius, cell ID).

[0089] The attribute "activationAsPerNF" indicates the NF on which the agent can be activated. It can be a list of NF instances of type NF. The attribute "activationAsPerTime" indicates a time constraint, meaning the agent can only be activated during the indicated time frame. It can be a list of time periods. The attribute "activatedToolService" indicates the use of external tools or services. It can be a list of tools or services activated for the agent.

[0090] The attribute "Maximum Number of External Agents (maxNumberOfExternalAgent)" indicates the maximum number of external smart agents for a given request. It can be an integer. Zero indicates that interaction with external agents is deactivated.

[0091] The attribute "activatedModel" indicates the use of multiple different backend LLMs or LLM-based models. It can be a list of LLM-based or LLM-based AIML models that are activated to support proxies. The attribute "deactivatedModel" indicates the prevention of the use of different backend LLMs or LLM-based models. It can be a list of LLM-based or LLM-based AIML models that are prevented from supporting proxies.

[0092] In some embodiments, the first device 110 may determine an agent job corresponding to the execution plan. For example, the agent job may represent a request. The first device 110 may maintain agent job information. The agent job information may be used to manage the execution plan.

[0093] In some embodiments, the agent job information may include at least one of the following: job identifier, job status, execution progress, job request, job context, or job reporting control. In some embodiments, at least a portion of the job identifier is configured as a public identifier of at least one IA.

[0094] In the example, such as Figure 4 As shown, information model 400 may include an IOC "AgenticJob" 405 for agent job information. Table 5 shows example attributes in the agent job information, such as in IOC "AgenticJob" 405.

[0095] Table 5. Example attributes of agent job information

[0096] The attributes "administrativeState" and "operationalState" are used to maintain the status of jobs. The attribute "jobId" can be used to associate tasks from multiple "Smart Agent" instances. "JobId" can be included when reporting job execution status allows MnS consumers to associate received status reports for the same request.

[0097] The "executionProgress" attribute indicates the current state and control information element, i.e., whether the job should be started, maintained, resumed, or canceled. The "jobRequest" attribute indicates a request from the consumer. It can be a task or subtask from the consumer.

[0098] The "jobContext" property indicates the job context during job execution. It can indicate whether it's a subtask as part of another task, or job memory information, etc. The "jobReportCtrl" property indicates how job reports are generated. It can specify job execution details at the report level, such as verbose, brief (default), or normal.

[0099] In some example embodiments, the first device 110 may send a response to the request to a third device, which in turn sends the request to the first device 110. The response may include a proxy job report. The proxy job report may be used to instruct the third device to execute the request.

[0100] In some embodiments, the agent job report may include at least one of the following: job identifier, job status, job result, or job failure reason. As an example, information model 400 may include an IOC “AgentJobReport” 402 representing the agent job report. Table 6 shows example attributes of the agent job report.

[0101] Table 6. Sample Attributes of Agent Job Reports

[0102] The "Job ID (jobId)" attribute can be used to associate a report with a task. The "Job Status (jobStatus)" attribute indicates the final status of the job execution, which could be success, failure, or partial success. The "Job Results (jobResults)" attribute indicates the detailed results of the report based on the report's configuration. It can include different levels of detail. When the status is failure or partial success, the "Job Failure Cause (jobFailureCause)" attribute indicates the possible reasons for failure. Reasons can be enumerated: 1) the job was executed too late, 2) insufficient data, etc.

[0103] In some embodiments, after receiving a performance report from each of at least one IA, the first device 110 may store task planning or execution metadata corresponding to the request and the identifier of at least one IA. For example, once the last IA provides its task performance report, the IAOF stores the task metadata, task planning metadata, and the IA's identifier for future use.

[0104] As described above, in some example embodiments, the first device 110 can be implemented in an AI / ML inference function. For this purpose, a new attribute "capabilityActivation" can be added to the IOC "AIMLInferenceFunction" 401, such as... Figure 4 As shown in Table 7, example properties are shown in IOC “AIMLI InferenceFunction” 401. The new property “capabilityActivation” indicates whether the agent capability is enabled or disabled. It can be a boolean list when a list of agents supported by the inference function exists.

[0105] Table 7. Example Attributes of Agency Work Reports

[0106] The above describes some example implementations. The general process of these examples will now be described. In the example, the IAOF receives consumer / operator prompts and translates them into executable network tasks using a model (e.g., LLM / LMM / SLM). If the task can be mapped to CRUD operations, the IAOF selects and applies these CRUD operations. If the task cannot be mapped to CRUD operations, the IAOF breaks the task down into subtasks. The IAOF discovers and selects IAs for subtask fulfillment and plans task fulfillment, including but not limited to the order of subtask fulfillment, deadlines, etc. The IAOF shares the subtask fulfillment instructions with each network entity of the selected IA(s) in the managed agent framework. The IA selected for subtask fulfillment executes actions according to each task fulfillment plan and reports to the IAOF upon task completion. Once the IA finally provides its task fulfillment report, the IAOF stores task metadata, task planning metadata, and the IA's identifier for future use.

[0107] To better understand the IA orchestration solution, some example processes are now described.

[0108] Figure 5 The signaling flow of an example process 500 between a consumer and a producer regarding the lifecycle of a task or intent, according to some example embodiments of this disclosure, is illustrated.

[0109] The consumer 501 of the MnS for AI / ML inference (also known as the inference MnS consumer 501) initiates a network request 505 for generative inference. This can be used for the intent of a specific network task or network request. The producer 502 of the MnS for AI / ML inference (also known as the inference MnS producer 502) checks 508 whether an agent is needed, for example, by evaluating the raw input / hints from the consumer. For example, if the task or intent can be accomplished via mapping to a direct CRUD operation or a set of CRUD operations without interaction with the relevant network entity or service, then no intelligent agent needs to be initiated. If no agent is needed, the inference MnS producer 502 can instruct the inference MnS consumer 501 to generate results directly, for example, using external tools.

[0110] If the evaluation determines that at least one smart agent is required, then the Inference MnS producer 502 instantiates a smart agent instance. For example, the Inference MnS producer 502 creates a 515 Smart Agent Management Object (MOI) instance. Alternatively, the smart agent instance can be installed by the system.

[0111] The inference MnS producer 502 (with the help of an LLM-based or LLM-based model) decomposes the 520 request and separates the request into execution plans. The execution plan can include details such as whether an external intelligent agent is needed, what kind of intelligent agent is needed, how many intelligent agents are needed, which service(s)(s) to consume, what tools to use, and how to access them.

[0112] The inference MnS producer 502 sends an initial response 525 back to the inference MnS consumer 501 to indicate that the request has been accepted and may further indicate whether more time is needed. Optionally, the inference MnS producer 502 may indicate (e.g., in natural language) subtasks related to the received prompts.

[0113] Inference MnS producer 502 can use 530 existing smart agent instances from other producers. Alternatively, inference MnS producer 502 can request the direct creation of smart agent instances for subtasks. Or, inference MnS producer 502 can send normal generative requests for subtasks. It is the relevant MnS producer that determines whether an agent is needed and, if so, creates a smart agent and negotiates with the primary agent (also known as the planning agent).

[0114] The Inference MnS producer 502 negotiates with other intelligent agent service producers 535 for subtask assignment. The Inference MnS producer 502 orchestrates 540 the interaction between the primary agent and the assigned agent to execute the assigned subtasks through interaction 545 with the assigned agent for subtask execution.

[0115] The inference MnS consumer 501 can receive 550 notifications during job execution. When the job execution is completed or fails, the inference MnS producer 502 can send a 555 job report to the inference MnS consumer 501.

[0116] Figure 6 The signaling flow of an example process 600 for agent discovery and configuration according to some example embodiments of the present disclosure is shown.

[0117] In the post-deployment debugging phase 601, the Inference MnS consumer 501 provisiones 605 general capabilities (e.g., the agent capabilities mentioned above) to the Inference MnS producer 502. The Inference MnS consumer 501 provisiones 610 general activation / deactivation configurations (e.g., the management activation mentioned above) to the Inference MnS producer 502. The Inference MnS consumer 501 provisiones 615 tool-related capabilities to the Inference MnS producer 502. Tool-related capabilities may include, for example, tool type, tool level, and tool policy. The Inference MnS consumer 501 provisiones 620 agent-specific policies to the Inference MnS producer 502. The Inference MnS consumer 501 provisiones 625 collaboration policies for using other agents to the Inference MnS producer 502. Collaboration policies may include, for example, agent-specific policies, context, and conditions. The Inference MnS consumer 501 provisiones 630 LLM or LLM-based models to the Inference MnS producer 502.

[0118] In runtime pre-configured phase 602, the inference MnS consumer 501 updates the decomposition strategy (e.g., agent capabilities) to the inference MnS producer 502. The inference MnS consumer 501 updates the tools (e.g., tool type, tool level, tool policy) to the inference MnS producer 502. The inference MnS consumer 501 updates the general activation / deactivation configuration (e.g., the aforementioned management activation) to the inference MnS producer 502. The inference MnS consumer 501 updates the collaboration strategy using other agents to the inference MnS producer 502. The collaboration strategy may include, for example, policies, context, and conditions specific to the agents. The inference MnS consumer 501 updates the LLM-based or LLM-based model to the inference MnS producer 502.

[0119] Now, let's describe some use cases. In telecommunications systems, simple service requests that can be mapped to direct CRUD operations or sets of CRUD operations and do not require bidirectional interaction do not require a proxy. For more complex service requests or tasks, an LLM (or LLM-based) proxy may be needed. The proxy can break down the request or task into action plans or execution plans. Each item can be a simpler task that can be mapped to a direct CRUD operation or a set of CRUD operations, and other items may need to interact with / consume services from other network functions, use external tools, or interact with proxies in other network functions.

[0120] The following sections describe two example use cases for the purpose of illustrating the application of a proxy framework in the telecommunications field. These use cases benefit from the solutions presented in this disclosure for managing different proxies, their capabilities, and interactions.

[0121] Figure 7Example Figure 700 illustrates an example use case where an intelligent agent is assisting in network management and optimization according to some example embodiments of this disclosure. In this example, the agent assists in network management and optimization, which is intent-based management.

[0122] In this scenario, human operator 701 can express an intent, such as "I want to achieve RAN energy consumption below YkWh in region X." This natural language intent is given to the planning intelligent agent 702. The planning intelligent agent 702 utilizes LLM (Limited Language Management) to understand the given human intent. It deduces that different energy-saving tasks need to be performed across several regions of interest. Based on the registered intelligent agents' capabilities or those performing specific tasks (e.g., power level optimization, RAN and core network energy consumption savings), the planning intelligent agent 702 determines which intelligent agents are most suitable and should participate in intent implementation.

[0123] In this example, the planning smart agent 702 can select three smart agents, each responsible for a specific task. For example, smart agent 703 is responsible for transmit power adjustment, smart agent 704 is responsible for cell on / off handover, and smart agent 703 is responsible for selecting the most energy-efficient UPF and routing according to service requirements. The planning smart agent 702 can monitor the execution of other smart agents. Along with further input and interaction with the environment, the planning smart agent 702 can determine whether the operator 701's intent has been fulfilled. Feedback can be sent to the operator, notifying whether the intent was fulfilled or failed.

[0124] Figure 8 Example diagrams are shown illustrating examples of use cases in which an intelligent agent is customized for consumer-oriented tasks, according to some example embodiments of this disclosure. Figure 8 An example of consumer-based assistance in everyday activities is shown. In this example, the agent can perform specific tasks to help a human complete his / her activities (e.g., a fitness instructor, cooking teacher, sightseeing guide, etc.).

[0125] In this use case, human user 801 can express a task using UE 802. The task could include, for example, creating a vacation itinerary at a specific location and requesting assistance with all relevant facility selection and booking, such as travel tickets, accommodation, restaurants, and landmark visits. To accomplish such a complex task, a planning intelligent agent 803 can determine that different agents might be needed, each dedicated to a part of the overall task. For example, an intelligent agent could be dedicated to accommodation and restaurant selection based on information about availability, ratings, customer budget, target quality level, etc. Another intelligent agent could be dedicated to selecting the most suitable landmarks and attractions to visit based on customer preferences, age, interests, budget, etc. All such intelligent agents need to collaborate with each other to complete the complex task of vacation planning and execution.

[0126] Given a consumer's task description and the capabilities of available agents, including their interaction abilities, the planning smart agent 803 selects the most suitable smart agents 804, 805, and 806. For example, in this case, smart agents capable only of assisting with the desired vacation location are also suitable, interacting with each other to synchronize booking times and restaurant locations with booking times for the attractions to be visited. If the required task can be satisfied within the constraints and availability of the given smart agents, the planning smart agent 803 can provide feedback to the user.

[0127] In summary, this disclosure proposes management mechanisms to enable control over the decomposition of complex tasks into smaller tasks and the execution of complex tasks, i.e., control over planning performed by intelligent agents. This includes providing planning strategies for controlling the planning process. The proposed management mechanisms may include the exposure and registration of descriptions of complex tasks that the intelligent agent can decompose into simpler tasks.

[0128] The proposed management approach may include the exposure and registration of task decomposition metadata. Examples of such exposure and registration include: previous complex task decompositions, the smaller tasks involved, task feasibility indicators (i.e., based on CRUD operations and / or smart agents), task performance level indicators when applying plans derived by (multiple) smart agents, and the costs of different task fulfillment plans in terms of monetary costs, energy expenditures, interface signaling, etc.

[0129] The proposed management approach may include defining planning strategies that manipulate and control the planning process to achieve task completion. For example, aspects such as: including or excluding which intelligent agents with which capabilities in the plan (e.g., filtering by type, geographic location, etc.), the maximum number of agents involved in the plan, which LLM / LMM / SLM models will be included as part of the intelligent agents in the plan, the priority of tasks within the plan, and how the mapping between tasks, agents, and models should look, such as one-to-one, one-to-many, etc.

[0130] In telecommunications systems, intelligent agents from different vendors can not only perform different tasks or task nuances, but also be based on entirely different models with varying capabilities and requirements related to the required inputs and the provision of different outputs. This disclosure proposes a mechanism for managing agent frameworks in communication networks. In this way, intelligent agents from different vendors can collaborate to perform tasks.

[0131] Figure 9 A flowchart of an example method 900 implemented at a first device according to some example embodiments of the present disclosure is shown. For the purposes of discussion, [the following will be discussed]. Figure 1The method 900 is described by the angle of the first device 110 in the middle.

[0132] At box 910, based on the determination that an intelligent agent (IA) is needed to respond to a request, the first device 110 determines an execution plan to be performed by at least one IA.

[0133] At frame 920, the first device 110 sends the execution task of the execution plan to the second device that hosts the target IA in at least one IA.

[0134] In some example embodiments, method 900 further includes: determining at least one IA by using a first IA hosted in the first device, or determining at least one IA including the first IA hosted in the first device based on a request.

[0135] In some example embodiments, determining at least one IA includes: discovering multiple IAs and IA capability information of the multiple IAs; and selecting at least one IA from the multiple IAs based on the request and the IA capability information.

[0136] In some example embodiments, method 900 further includes: receiving IA provisioning information, wherein the execution plan to be executed by at least one IA is also based on the IA provisioning information.

[0137] In some example embodiments, the IA provisioning information includes IA capability information, which includes at least one of the following: information on at least one task supported by the agent, the ability to use the agent's underlying model, the ability to use external knowledge or tools, the ability to collaborate with other agents, the context length of the IA, or the ability to validate and refine the output from other agents.

[0138] In some example embodiments, receiving IA provisioning information includes receiving IA capability information from at least one of a third device that sends a request to the first device or a fourth device that is responsible for the exposure or discovery service of IA capability information.

[0139] In some example embodiments, the IA provisioning information includes IA agent information, which includes at least one of the following: agent capabilities, policies for agents, agent type, agent accountability score, or a list of potential collaborating agents.

[0140] In some example embodiments, the IA provisioning information includes IA policy information, which includes at least one of the following: the managed activation / deactivation scope of agent capabilities, control over agent capabilities or behavior, task execution policies, the scope of external interactions, or fallback options for task failures.

[0141] In some example embodiments, receiving IA preset information includes receiving IA policy information from a mobile network operator (MNO) or a third device that transmits a request to the first device.

[0142] In some example embodiments, IA provisioning information includes IA activation or deactivation configuration information, which includes activation or deactivation of at least one of the following: agent, agent capabilities, network function (NF) service, or agent model.

[0143] In some example embodiments, method 900 further includes receiving updates to IA provisioning information during the execution of the execution plan.

[0144] In some example embodiments, method 900 further includes: determining a proxy job corresponding to the execution plan; and maintaining proxy job information for the proxy job.

[0145] In some example embodiments, the agent job information includes at least one of the following: job identifier, job status, execution progress, job request, job context, or job report control.

[0146] In some example embodiments, at least a portion of the job identifier is configured as a public identifier of at least one IA.

[0147] In some example embodiments, method 900 further includes sending a response to the request to a third device that sent the request to the first device, the response including a proxy job report.

[0148] In some example embodiments, the agent job report includes at least one of the following: job identifier, job status, job result, or job failure reason.

[0149] In some example embodiments, the execution task is one of a plurality of execution tasks in the execution plan, and determining the execution plan to be executed by at least one IA based on the request includes: decomposing the request into a plurality of execution tasks having at least one of execution dependencies or execution constraints.

[0150] In some example embodiments, method 900 further includes: determining whether an IA is needed to respond to the request based on the request.

[0151] In some example implementations, determining whether an IA is needed to respond to a request includes determining whether the request can be mapped to at least one of the following: a set of create, read, update, delete (CRUD) operations, a direct NF service, or an external tool call.

[0152] In some example embodiments, method 900 further includes: after receiving a fulfillment report from each of at least one IA, storing task planning or execution metadata corresponding to the request and an identifier of at least one IA.

[0153] In some example embodiments, method 900 further includes sending task planning or execution metadata to another device that is responsible for the exposure or discovery service of task planning or execution metadata.

[0154] In some example embodiments, at least one IA is based on at least one of the following: a large language model (LLM), a large multimodal model (LMM), or a small language model (SLM).

[0155] In some example embodiments, the first device is or is included in at least one of a terminal device, a radio access network device, a core network function, a management function, or a third-party application or service.

[0156] In some example embodiments, a first device capable of performing any method 900 (e.g., Figure 1 The first device 110 may include components for performing the corresponding operation of method 900. The device may be implemented in any suitable form. For example, the device may be implemented in a circuit or software module. The first device may be implemented as or included in... Figure 1 In the first device 110.

[0157] In some example embodiments, the first device includes: components for determining an execution plan to be executed by at least one IA based on a determination that an intelligent agent (IA) is required to respond to a request; and components for sending the execution task of the execution plan to a second device hosting a target IA in at least one IA.

[0158] In some example embodiments, the first device further includes: a component for determining at least one IA by using a first IA hosted in the first device, or a component for determining, based on a request, at least one IA including the first IA hosted in the first device.

[0159] In some example embodiments, the components for determining at least one IA include: components for discovering multiple IAs and IA capability information of the multiple IAs; and components for selecting at least one IA from the multiple IAs based on a request and the IA capability information.

[0160] In some example embodiments, the first apparatus further includes a component for receiving IA preset information, and wherein the determination of an execution plan to be performed by at least one IA is further based on the IA preset information.

[0161] In some example embodiments, the IA provisioning information includes IA capability information, which includes at least one of the following: information on at least one task supported by the agent, the ability to use the agent's underlying model, the ability to use external knowledge or tools, the ability to collaborate with other agents, the context length of the IA, or the ability to validate and refine the output from other agents.

[0162] In some example embodiments, the device component for receiving IA preset information is a component for receiving IA capability information from at least one of a third device that transmits a request to a first device or a fourth device that is responsible for the exposure or discovery service of IA capability information.

[0163] In some example embodiments, the IA provisioning information includes IA agent information, which includes at least one of the following: agent capabilities, policies for the agent, agent type, agent responsibility score, or a list of potential collaborating agents.

[0164] In some example embodiments, the IA provisioning information includes IA policy information, which includes at least one of the following: the managed activation / deactivation scope of agent capabilities, control over agent capabilities or behavior, task execution policies, the scope of external interactions, or fallback options for task failures.

[0165] In some example embodiments, the component for receiving IA preset information includes: a component for receiving IA policy information from a mobile network operator (MNO) or a third device that transmits a request to the first device.

[0166] In some example embodiments, IA provisioning information includes IA activation or deactivation configuration information, which includes activation or deactivation of at least one of the following: agent, agent capabilities, network function (NF) service, or agent model.

[0167] In some example embodiments, the first device further includes a component for receiving updates to IA preset information during the execution of the execution plan.

[0168] In some example embodiments, the first apparatus further includes: a unit for determining a proxy job corresponding to an execution plan; and a component for maintaining proxy job information of the proxy job.

[0169] In some example embodiments, the agent job information includes at least one of the following: job identifier, job status, execution progress, job request, job context, or job report control.

[0170] In some example embodiments, at least a portion of the job identifier is configured as a public identifier of at least one IA.

[0171] In some example embodiments, the first device further includes a component for sending a response to the request to a third device, the third device sending the request to the first device, the response including a proxy job report.

[0172] In some example embodiments, the agent job report includes at least one of the following: job identifier, job status, job result, or job failure reason.

[0173] In some example embodiments, the execution task is one of a plurality of execution tasks in the execution plan, and the component for determining the execution plan to be executed by at least one IA based on the request includes: a component for decomposing the request into a plurality of execution tasks having at least one of execution dependencies or execution constraints.

[0174] In some example embodiments, the first device further includes a component for determining, based on the request, whether an IA is needed to respond to the request.

[0175] In some example embodiments, the component for determining whether an IA is needed to respond to a request includes: a component for determining whether a request can be mapped to at least one of the following: a set of create, read, update, delete (CRUD) operations, a direct NF service, or an external tool call.

[0176] In some example embodiments, the first apparatus further includes a component for storing task planning or execution metadata corresponding to the request and an identifier of at least one IA after receiving a fulfillment report from each of at least one IA.

[0177] In some example embodiments, the first device further includes a component for sending task planning or execution metadata to another device responsible for the exposure or discovery service of task planning or execution metadata.

[0178] In some example embodiments, at least one IA is based on at least one of the following: a large language model (LLM), a large multimodal model (LMM), or a small language model (SLM).

[0179] In some example embodiments, the first device is or is included in at least one of a terminal device, a radio access network device, a core network function, a management function, or a third-party application or service.

[0180] Figure 10 This is a simplified block diagram of a device 1000 suitable for implementing exemplary embodiments of the present disclosure. The device 1000 can be provided to implement a communication device, for example, Figure 1The device 1000 includes one or more processors 1010, one or more memories 1020 coupled to the processors 1010, and one or more communication modules 1040 coupled to the processors 1010.

[0181] Communication module 1040 is used for bidirectional communication. Communication module 1040 has one or more communication interfaces to facilitate communication with one or more other modules or devices. The communication interface can represent any interface necessary for communication with other network elements. In some example embodiments, communication module 1040 may include at least one antenna.

[0182] As a non-limiting example, processor 1010 can be any type suitable for a local technology network and can include one or more of the following: general-purpose computer, special-purpose computer, microprocessor, digital signal processor (DSP), and processor based on a multi-core processor architecture. Device 1000 can have multiple processors, such as application-specific integrated circuit chips that are time-dependent on a clock that synchronizes with the main processor.

[0183] Memory 1020 may include one or more non-volatile memories and one or more volatile memories. Examples of non-volatile memories include, but are not limited to, read-only memory (ROM) 1024, electrically programmable read-only memory (EPROM), flash memory, hard disk, compact disc (CD), digital video disc (DVD), optical disc, laser disc, and other magnetic and / or optical storage. Examples of volatile memories include, but are not limited to, random access memory (RAM) 1022 and other volatile memories that will not persist for the duration of a power outage.

[0184] Computer program 1030 includes computer-executable instructions that are executed by an associated processor 1010. The instructions of program 1030 may include instructions for performing operations / actions of some example embodiments of this disclosure. Program 1030 may be stored in memory (e.g., ROM 1024). Processor 1010 can perform any suitable actions and processes by loading program 1030 into RAM 1022.

[0185] The exemplary embodiments of this disclosure can be implemented by program 1030, enabling device 1000 to execute as described in the reference. Figures 2 to 9 Any process discussed in this disclosure. Exemplary embodiments of this disclosure may also be implemented by hardware or a combination of software and hardware.

[0186] In some example embodiments, program 1030 may be tangibly contained in a computer-readable medium, which may be included in device 1000 (such as in memory 1020) or in other storage devices accessible by device 1000. Device 1000 may load program 1030 from the computer-readable medium into RAM 1022 for execution. In some example embodiments, the computer-readable medium may include any type of non-transitory storage medium, such as ROM, EPROM, flash memory, hard disk, CD, DVD, etc. As used herein, the term "non-transitory" is a limitation of the medium itself (i.e., tangible, not tactile), rather than a limitation of the persistence of data storage (e.g., RAM versus ROM).

[0187] Figure 11 An example of a computer-readable medium 1100 is shown, which may be in the form of a CD, DVD, or other optical storage disc. The computer-readable medium 1100 has a program 1030 stored thereon.

[0188] Generally, the various embodiments of this disclosure can be implemented in hardware or dedicated circuitry, software, logic, or any combination thereof. Some aspects can be implemented in hardware, and others can be implemented in firmware or software that can be executed by a controller, microprocessor, or other computing device. Although various aspects of the embodiments of this disclosure are shown and described as block diagrams, flowcharts, or using some other graphical representation, it should be understood that the blocks, apparatuses, systems, techniques, or methods described herein can be implemented in hardware, software, firmware, dedicated circuitry or logic, general-purpose hardware, or controllers or other computing devices, or some combination thereof, as non-limiting examples.

[0189] Some exemplary embodiments of this disclosure also provide at least one computer program product tangibly stored on a computer-readable medium, such as a non-transitory computer-readable medium. The computer program product includes computer-executable instructions that execute in a device on a target physical or virtual processor, such as those included in a program module, to perform any of the methods described above. Typically, a program module includes routines, programs, libraries, objects, classes, components, data structures, etc., that perform a particular task or implement a particular abstract data type. In various embodiments, the functionality of a program module can be combined or split among program modules as needed. The machine-executable instructions for a program module can execute within a local or distributed device. In a distributed device, a program module can reside on both local and remote storage media.

[0190] Program code used to perform the methods of this disclosure may be written in any combination of one or more programming languages. The program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that, when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a stand-alone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0191] In the context of this disclosure, computer program code or related data may be carried by any suitable carrier wave to enable a device, apparatus, or processor to perform the various processes and operations described above. Examples of carrier waves include signals, computer-readable media, etc.

[0192] Computer-readable media can be computer-readable signal media or computer-readable storage media. Computer-readable media can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatuses, or devices, or any suitable combination thereof. More specific examples of computer-readable storage media will include electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable optical disc read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0193] Furthermore, although operations are described in a specific order, this should not be construed as requiring that such operations be performed in the specific order shown or sequentially, or requiring that all shown operations be performed to achieve the desired result. In some cases, multitasking and parallel processing may be advantageous. Similarly, although several specific implementation details are included in the above discussion, these should not be construed as limiting the scope of this disclosure, but rather as a description of features that may be specific to particular embodiments. Unless explicitly stated otherwise, certain features described in the context of a single embodiment may also be implemented in combination in a single embodiment. Conversely, unless explicitly stated otherwise, various features described in the context of a single embodiment may also be implemented individually or in any suitable sub-combination in multiple embodiments.

[0194] Although this disclosure has been described in language specific to structural features and / or methodological actions, it should be understood that the disclosure as defined in the appended claims is not necessarily limited to the specific features or actions described above. Rather, the specific features and actions described above are disclosed as exemplary forms for implementing the claims.

Claims

1. A first device, comprising: At least one processor; as well as At least one memory storing instructions that, when executed by the at least one processor, cause the first device to: Based on the determination that an intelligent agent (IA) is needed to respond to the request, an execution plan is determined to be executed by at least one IA; as well as The execution task of the execution plan is sent to a second device that hosts the target IA in at least one of the IAs.

2. The first device according to claim 1, wherein the first device is further configured to: The at least one IA is determined by using a first IA hosted in the first device, or Based on the request, at least one IA is determined, including the first IA hosted in the first device.

3. The first apparatus according to claim 2, wherein determining the at least one IA comprises: Discover multiple IAs and their IA capability information; as well as Based on the request and the IA capability information, at least one IA is selected from the plurality of IAs.

4. The first device according to any one of claims 1 to 3, wherein the first device is further configured to: Receive IA pre-configured information, and The execution plan, which determines that will be executed by the at least one IA, is also based on the IA's pre-configured information.

5. The first device according to claim 4, wherein the IA preset information includes IA capability information, and the IA capability information includes at least one of the following: Information for at least one task supported by the agent, Capabilities of the underlying model for agents The ability to use external knowledge or tools The ability to collaborate with another agent IA context length, The ability to validate and refine the output from another agent.

6. The first apparatus according to claim 4, wherein receiving the IA preset information includes: The IA capability information is received from at least one of a third device that sends the request to the first device or a fourth device that is responsible for the exposure or discovery service of the IA capability information.

7. The first apparatus according to any one of claims 4 to 6, wherein the IA preset information includes IA proxy information, and the IA proxy information includes at least one of the following: Agency capabilities Strategies for using proxies, Agent type Agency responsibility score A list of potential collaborating agents.

8. The first apparatus according to any one of claims 4 to 7, wherein the IA preset information includes IA policy information, and the IA policy information includes at least one of the following: Managed activation / deactivation scope for proxy capabilities Control over the agent's capabilities or behavior Task execution strategy Scope of external interaction A rollback option for task failure.

9. The first apparatus according to claim 8, wherein receiving the IA preset information includes: The IA policy information is received from either the mobile network operator (MNO) or a third device that sends the request to the first device.

10. The first apparatus according to any one of claims 4 to 9, wherein the IA preset information includes IA activation or deactivation configuration information, the IA activation or deactivation configuration information including activation or deactivation of at least one of the following: acting, The ability of an agent Network Functions (NF) services The agent model.

11. The first device according to any one of claims 4 to 10, wherein the first device is further configured to: Updates to the IA preset information are received during the execution of the execution plan.

12. The first device according to any one of claims 1 to 11, wherein the first device is further configured to: Determine the proxy job corresponding to the execution plan; and Maintain the proxy job information of the proxy job.

13. The first apparatus according to claim 12, wherein the agent operation information includes at least one of the following: Job identification, job status, execution progress, job request, job context, or job report control.

14. The first apparatus of claim 13, wherein at least a portion of the job identifier is configured as a common identifier of the at least one IA.

15. The first device according to any one of claims 4 to 14, wherein the first device is further configured to: A response to the request is sent to a third device, which in turn sends the request to the first device. The response includes a proxy job report.

16. The first apparatus of claim 15, wherein the agent operation report comprises at least one of the following: Task identifier, task status, task result, or reason for task failure.

17. The first apparatus according to any one of claims 1 to 16, wherein the execution task is one of a plurality of execution tasks of the execution plan, and determining the execution plan to be executed by the at least one IA based on the request comprises: The request is decomposed into a plurality of execution tasks, wherein the plurality of execution tasks have at least one of execution dependency or execution constraint.

18. The first device according to any one of claims 1 to 17, wherein the first device is further configured to: Based on the request, determine whether an IA is needed to respond to the request.

19. The first apparatus of claim 18, wherein determining whether the IA is required to respond to the request based on the request comprises: Determine whether the request can be mapped to at least one of the following: a set of create, read, update, delete (CRUD) operations, a direct NF service, or an external tool call.

20. The first device according to any one of claims 1 to 19, wherein the first device is further configured to: After receiving a fulfillment report from each of the at least one IA, the task planning or execution metadata corresponding to the request and the identifier of the at least one IA are stored.

21. The first device according to claim 20, wherein the first device is further configured to: The task planning or execution metadata is sent to another device responsible for the exposure or discovery service of the task planning or execution metadata.

22. The first device according to any one of claims 1 to 21, wherein the at least one IA is based on at least one of the following: Large Language Model (LLM) Large multimodal model (LMM) Small Language Model (SLM).

23. The first apparatus according to any one of claims 1 to 22, wherein the first apparatus is or is included in at least one of a terminal device, a radio access network device, a core network function, a management function, or a third-party application or service.

24. A method comprising: Based on the determination that an intelligent agent (IA) is needed to respond to the request, an execution plan is determined to be executed by at least one IA; as well as The execution task of the execution plan is sent to a second device that hosts the target IA in at least one of the IAs.

25. A first device, comprising: A component for determining an execution plan to be performed by at least one IA based on a request that requires an intelligent agent (IA) to respond to the request; as well as A component for sending the execution task of the execution plan to a second device that hosts the target IA in at least one of the IAs.

26. A computer-readable medium comprising instructions stored thereon for causing a device to perform at least the method according to claim 24.

Images